Double Perovskite Cobaltites Integrated in a Monolithic and Noble Metal-Free Photoelectrochemical Device for Efficient Water Splitting

Water photoelectrolysis has the potential to produce renewable hydrogen fuel, therefore addressing the intermittent nature of sunlight. Herein, a monolithic, photovoltaic (PV)-assisted water electrolysis device of minimal engineering and of low (in the μg range) noble-metal-free catalysts loading is presented for unassisted water splitting in alkaline media. An efficient double perovskite cobaltite catalyst, originally developed for high-temperature proton-conducting ceramic electrolyzers, possesses high activity for the oxygen evolution reaction in alkaline media at room temperatures too. Ba1–xGd1–yLax+yCo2O6−δ (BGLC) is combined with a NiMo cathode, and a solar-to-hydrogen efficiency of 6.6% in 1.0 M NaOH, under 1 sun simulated illumination for 71 h, is demonstrated. This work highlights how readily available earth-abundant materials and established PV methods can achieve high performance and stable and monolithic photoelectrolysis devices with potential for full-scale applications.

j (mA cm Figure S9: Equivalent circuit used to fit the raw EIS data. All fittings were done in Zview (Scribner) and all χ 2 values were of the order of 10 -4 . Rs accounts for the solution resistance. Rd and CPEd account for the resistance and in parallel constant phase element for the diffusion/migration of ionic species in porous electrodes, respectively. This is a simplified approach and is proposed instead of the transmission line model. Rint and CPEint account for the interfacial charge transfer resistance and in parallel constant phase element, respectively (a). Nyquist plots of BGLC587, BGLC82, BGLC37 and IrO2 at potential for 10 mA cm -2 (b), onset potential (c) and non-faradaic region (d).       Transmittance (%)     The exponential value of the CPEd parameter is approx. 0.5, which clearly indicates this constant phase element has a Warburg behaviour, consistent with a purely diffusive process. Therefore, the Cdl parameter in the onset potentials was found by the interfacial CPE element alone. In fact, even if the effective capacitance is extracted, it is so large that it does not contribute to the total effective capacitance of the double layer.     Figure S7 shows XPS spectra for BGLC587 pre-and post-operation. The pre-operation survey spectrum shows the presence of Gd, La, Co and Ba together with C and O, while the post-operation survey is dominated by Na, F, O and C ( Figure S7a). The La 3d and O 1s spectra pre-operation ( Figure  S7b and c) are similar to spectra reported for the LaCoO 3 phase. 29 The IO component of O 1s can unambiguously be assigned to perovskite lattice oxygen, e.g., LaCoO 3 , while the IIO component can be assigned to other oxide phases, under-coordinated oxygen and/or adsorbed O 2 or OH species. 30,31 The

List of Tables
Gd 4d spectrum (Figure S7e), which contains multiple components due to final state interactions with the 4f 7 valence band electrons, was fitted by the procedure used by Thiede et al. 32 Although only one set of Gd components was necessary to obtain a good fit, the presence of more than one Gd phase cannot be completely ruled out.
The Ba 4d spectrum Figure S7f shows clear indications of two different Ba 4d states. This was also seen in the Ba 3d spectrum (Figure S7d), where the two components labelled I Ba and II Ba are consistent with previous reports for similar materials. 30,31 The fact that the I Ba and II Ba components in the Ba 3d 5/2 region are repeated with 2/3 of the peak area in the Ba 3d3/2 region verifies the identification as Ba 3d components.
The Co 2p spectrum is complicated, as the main peak is accompanied by several plasmon loss peaks, which vary in separation and intensity depending on the chemical state of Co. 33 To obtain a good fit of the spectrum in Figure S7d, the two components I Co and II Co had to be added. These components are repeated with 1/2 of the peak area in the Co 2p 1/2 region, demonstrating the presence of both Ba and Co in the sample. Due to the overlap between Ba 3d and Co 2p, however, a detailed description of the Co phase(s) was not possible. The Co 3p region is less described in the literature, but a doublet splitting of 1.1 eV is assumed. 34 The Co 3p spectrum in Figure S7g was fitted with 3p doublet components at a fixed distance of 1.1 eV. In addition, a second doublet set is added to the high BE region. As for Co 2p, it is unclear whether these secondary components are an indication of Co in a different state, or if it is just part of the satellite structure of the main component.
A complete determination of the composition of the post-operation powder based on the XPS spectra was challenging due to low sample amounts. However, certain observations can be made: Post-operation, the Co 2p/Ba 3d region is dominated by a component that is repeated with a 2:1 ratio between the 3/2 and 1/2 level, indicative of Co and thus suggesting a loss of Ba. The Co 2p component corresponds with the position of the weakest Co component in the pristine powder (IICo). The component labelled as IBa in pre-operation is still present, repeating with a 3:2 ratio between the 5/2 and 3/2 levels.
In the Ba 4d spectrum, the IBa components were also present, while the IIBa region appeared broader, possibly indicating a mix of several Ba phases. To estimate the relative content of Ba before and after cycling, quantification was done based on two regions not affected by peak overlap: Ba 4d and Co 3d, which confirms that the Ba content decreases relative to Co (Table S3). For the post-operation powder the La 3d and Gd 4d peaks overlapped with F KLL and Si 2s, which means that a reliable quantification of the relative content of Co, Ba, Gd and La could not be obtained.
The Gd 4d peak is broadened in a way that can be fitted by assuming Gd in two different chemical states, with identical multiplet splitting. Co 3p is also broadened, and can be fitted by increasing the intensity of the weak peak found in the pristine powder. This could be consistent with the relative change in peak intensity observed for Co 2p, but a more extensive XPS investigation, including pure reference samples of the different candidate compositions, is needed to fully understand the chemical decomposition of the BGLC system upon cycling.

Performance of the PV-PEC with IrO2 as the anode electrode
The STH begins at 8.4% (Figure S25), which is lower than the expected efficiency of 9.3% according to the fitted j-V curve of the mini module and the j-E curves of IrO2 and NiMo ( Figure S26a). This discrepancy is again attributed to ohmic losses in the electrolyte, but possibly due to the series connected potentiostat. In this case the evaporation of the electrolyte reduces the photocurrent density, but unlike the ea-PV-PEC, the efficiency of the device was constantly decreasing. After 24 h of continuous operation the efficiency was reduced to approx. 7.5% ( Figure S25). The loss is mainly attributed to dissolution of IrO2 as preliminary monolithic devices retrieved their efficiency after IrO2 is electrodeposited again (see Figure S27). We took this device for outdoor operation as well and as expected by the laboratory performance, as well as the j-E curves of the HER and OER catalysts, an STH fluctuating around 7% was achieved ( Figure S28). It can be seen that the sunlight intensity was significantly fluctuating due to passing clouds during the first hour of operation. It is also interesting to notice that under light intensities of around 0.4 and 0.2 Suns, the STH was well above 8%, reaching even 9%. Fittings of the j-V curves of the mini-PV module under different light intensities (0.2 to 0.8 Suns) to the j-E curves of IrO2 and NiMo further confirm the performance range of the PV-PEC device (Figure S26b-e). The discrepancies observed may be attributed to different performance of the solar cells under natural sunlight and laboratory emulation. Finally, the IrO2-coated FTO shows over 75% of the transmittance in the visible region, where the EQE has an excellent response ( Figure S29).