Chromium doped copper vanadates photoanodes for water splitting

Solar hydrogen obtained from photoelectrochemical water splitting offers a versatile approach towards the substitution of fossil fuels by decentralized and sustainable resources, like water and sun. In the present study we have investigated the Chromium doped Copper Vanadate (Cr:Cu3V2O8) as a candidate photoanode for photoelectrochemical water splitting. We have synthetized this material through a simple aqueous precipitation reaction, which easily allows compositional modifications. We have studied the effect of extrinsic doping with substitutional atoms like Chromium on the optical and photoelectrochemical properties. The main limiting factor for performance is related to the high bulk recombination, which is partially overcome by 0.75 at. % Chromium doping, with a five-fold enhancement of the charge separation efficiency at 1.23 V vs RHE. Despite this remarkable milestone, significant further improvement is needed for the technological exploitation of this material.


Introduction
The sun feeds the planet with an enormous amount of power, ~120 000 TW, that considerably exceeds the projected global energy demand for the next decades (~30 TW by 2050) in a moderate scenario. [1] Therefore, the conversion of sunlight into a profitable form of energy is almost mandatory to meet the needs of a growing world population with increasing living standards. Photoelectrochemical (PEC) water splitting offers an attractive technology for the conversion and storage of solar energy as chemical energy in the form of molecular bonds. In this process, the water molecule is decomposed into H 2 and O 2 under solar illumination. The O 2 evolution reaction (OER) is thermodynamically and kinetically more demanding, since four holes per molecule of produced oxygen are required. The strong oxidizing conditions needed for this reaction, severely limit the choice of adequate materials for viable operation. Consequently, the development of efficient and stable oxygen evolving photoanodes is one of the key challenges to success of this technology. [2,3] Since the pioneering study of Fujishima and Honda demonstrating UV light-assisted water splitting with TiO 2 , [4] intensive research efforts have been carried out to find earthabundant, efficient, stable and cost-effective materials that can absorb a significant fraction of the solar spectrum to split water. Metal oxides or oxo-metalates based on abundant materials, such as TiO 2 , [4,5] Fe 2 O 3 ,[6] WO 3 , [7] ZnO, [8] and BiVO 4 , [9,10] have been explored as candidate photoanode materials, satisfying both stability and cost requirements. Nevertheless, the low performance of these photoanodes attributed to poor electronic properties (Fe 2 O 3 and BiVO 4 ) and/or large bandgaps (TiO 2 , WO 3 and ZnO) [11] hinder their large-scale use in PEC systems. In order to overcome these problems, different strategies have been accomplished, involving combinatorial synthesis, [12] tuning the band structure of semiconductor materials by doping with different elements, [13,14] surface state passivation, [15,16] surface activation with OER catalysts, [17,18] and nanostructuring. [19][20][21][22] At present, metal vanadates, mainly BiVO 4 , are at the forefront of the research for PEC water splitting photoanodes. [13,23,24] A record photocurrent of 6.72 mA·cm -2 at 1.23 V vs RHE was obtained for BiVO 4 , close to its theoretical maximum value (7.5 mA·cm -2 ), by means of combining nanostructuring with adequate OER catalysts. [25] Triclinic NiV 2 O 6 films, fabricated by a vacuum deposition technique, have been recently tested for the first time as photoanodes for water oxidation. Although the observed photocurrent is relatively low (ca. 0.25 mA·cm -2 at 1.23 V vs RHE) according to the band gap (~2.4 eV) both the wide availability of the semiconductor components (Ni and V) and its stability in alkaline conditions postulate this material as a promising anode for PEC systems. [26] Similar to NiV 2 O 6 , Cu 3 V 2 O 8 is an n-type semiconductor composed only of first-row transition metals. Although it has been previously examined for different applications (Li-ion batteries and degradation of organic pollutants), [27,28] both, its bandgap near 2 eV and the adequate position of the valence band maximum make this material suitable for water photo-oxidation. To the best of our knowledge, there is only one recent study reporting the PEC performance of Cu 3 V 2 O 8 photoanodes. Seabold  copper vanadate, the V corrosion is mitigated through a self-passivation process in which V corrodes from the film, leaving behind a Cu-rich oxide surface layer that prevents further V corrosion. [31] This investigation confirms that copper vanadate has indeed emerged as a promising photoanode for water splitting due to its stability, particularly in weakly alkaline borate electrolytes.
In the present work, we report our efforts aiming at improving the charge transport and charge separation efficiency of this semiconductor material. We have employed a synthetic method inspired on that previously reported. [29] This material was doped with chromium (Cr: Cu 3 V 2 O 8 ), which has an atomic radii of 0.74 Å, close to that of Cu 2+ (0.73 Å), making feasible the exchange of both atoms in the CuO 6 octahedra of its crystalline structure, enhancing the extrinsic n-type doping of the semiconductor oxide. A detailed optoelectronic and photoelectrochemical characterization has been performed to quantitatively assess the contribution of the three fundamental processes involved in PEC, i.e. charge carrier generation, charge transport to the semiconductor-liquid interface and interfacial charge transfer, to the obtained photocurrent.

Materials and synthesis
Preparation of nanoparticles of Cu 3 V 2 O 7 (OH) 2 ·2H 2 O was carried out following the procedure described in reference [29], with slight modifications, using as reagents:

Structural, optical and photoelectrochemical characterizations
X-ray diffraction (XRD) data were obtained employing Cu Kα radiation at room temperature, scanning the samples from 10º to 70º (2θ) with a step of 0.02º. The morphology and thickness of both the precursor and oxide films were determined by scanning electron microscopy (SEM) using a JEOL JEM-3100F field emission scanning electron microscope. UV-Vis absorption spectra were recovered with a Cary 300 UV−Vis Varian spectrophotometer, between 300 and 800 nm. The absorbance (A) was estimated from transmittance (T) and diffuse reflectance (R) measurements as: . The indirect optical bandgap was estimated by the Tauc plot as: . In this expression, the absorption coefficient (α) was calculated by , where l is the thickness of the electrode.

Results and discussion
In order to optimize both the homogeneity and reproducibility for the preparation of the oxide films, the deposition of the precursor solution was rigorously controlled. In first place, the synthesis of the precursor by a precipitation reaction provided a very easy and  The morphology and particle size was determined by SEM (Fig. 2). The films of the precursor were composed of nanoflakes with around 70-80 nm size (Fig. 2a). In contrast, after annealing, both undoped and Cr-doped nanoparticles showed a globular morphology ( Fig. 2b and c), with lower particle size for the doped nanoparticles (approximately 40-100 nm vs 20-80 nm, respectively) at the optimum Cr concentration (0.75%).  Fig. 4. Fig. 4.a shows that the 800 nm thick samples yield the best performance, which can be attributed to the inhomogeneous material distribution as the number of spin coating cycles increases. With respect to the Cr content, the optimum doping concentration appears at 0.75%, which means a three-fold enhancement at 1.5 V vs RHE compared to pristine Cu 3 V 2 O 8 (Fig. 4b).
Dopant concentrations higher than 0.75% (i.e. 1% and 1.5%) did not improve the obtained photocurrent, as showed in the inset on Fig. 4b.  properties of the photoanode, as also illustrated by the similar charge injection yield obtained for both undoped and Cr doped materials (Fig. 6b).
The spectral signature of the photocurrent was characterized by incident photon to current conversion efficiency (IPCE) for both undoped and Cr-doped films without and with the addition of the hole scavenger (Fig. 5). Insets in this Fig. display the magnification of the onset region, showing a good correspondence between the onset wavelength for the IPCE and the absorbance of the films (Fig. 3), around 600 nm. There is also an excellent agreement between the integrated photocurrent extracted from the IPCE spectra and that obtained from the j-V measurements ( Fig. 5c and d), as illustrated in Supplementary   Information, Table SI1. . The calculated theoretical photocurrent was 12.90 mA·cm -2 and 11.21 mA·cm -2 for 800 nm thick undoped and Cr doped films, respectively. In aqueous electrolyte, the photocurrent density could be affected mainly by the charge separation efficiency on the bulk material and the charge injection efficiency related with the surface kinetic reaction or catalytic efficiency [33], so that: In presence of the hole scavenger, we assume that the catalytic efficiency is close to unity, so the obtained photocurrent density will be affected only by the charge separation efficiency:  .
The comparative behavior of both calculated yields is showed in Fig. 6  Impedance spectroscopy (IS) measurements were carried out to assess the electronic properties of the Cu 3 V 2 O 8 photoanodes, i.e., doping density (N D ) and flatband potential (V fb ), by means of Mott-Schottky (MS) plots. The value for the relative dielectric permittivity () was estimated as 44. [34] This study was performed in dark conditions covering a wide potential window (from 0.5 to 1.5 V vs RHE) and at single frequencies (10 Hz, 50 Hz and 100 Hz, respectively). The frequencies were selected from the region at which the real part of the capacitance remained constant (10 Hz -100 Hz) in preliminary multi-frequency tests (10 MHz -100 mHz) at a constant applied bias (see Supplementary   Information, Fig. SI5). Fig. 7 compares the MS plots obtained at a frequency of 10 Hz for doped and undoped Cu 3 V 2 O 8 films and with these plots, the values for N D and V fb were estimated. Note that identical MS plots were obtained for the measurements at 50 Hz and 100 Hz (see Supplementary Information, Fig. SI3). The lower slope reported for the Cr:Cu 3 V 2 O 8 samples with 0.75% and more significantly with 1% of chromiun content is connected to an increase in the doping density (Table 1)  substitution by Mo(VI) reported in other vanadates. [29,35] The statistical significance of these results is illustrated in Supplementary Information, Fig. SI6, where four identical samples were measured at each condition. The flatband potential, V fb , for Cu 3 V 2 O 8 photoanode is 0.79 V vs RHE, which is slightly more positive than the frequency-dependent values previously reported (between 0.63 and 0.69 V vs RHE) [29].
In addition, V fb for the optimum Cr-doped film was around 120 mV cathodically shifted (0.67 V vs RHE), although this beneficial displacement is not reflected on the onset potential for the photocurrent, see Fig. 5a, probably due to the excessive bulk recombination losses. It is important to highlight that these very positive flatband potentials are detrimental for technological applications and further efforts should be conducted to shift cathodically this potential. By substitutional Cr doping, replacing part of the Cu sites, the Fermi level of copper vanadate shifts towards the conduction band increasing the band bending at the semiconductor liquid junction and consequently enhancing charge separation (Fig. 6a).
This has a beneficial effect on the photoelectrochemical performance. On the other hand, there is an anomaly on the donor density for 0.5% and 1.5% Cr additions. EDS and XPS experiments were carried out to understand the correlation between added Cr during the synthetic process and incorporated Cr into the specimens. Unfortunately the Cr contents employed are below the detection limit of these techniques and further information could not be obtained. XRD analysis did not show the presence of secondary phases at any Cr concentration tested, but a significant shift of the maximum on the [012] direction ( Fig. 1) was registered at the highest concentration (1.5%), which could be related to a degradation of the film integrity as result of the strain induced by Cr replacing Cu positions into the periodic crystal lattice, with the consequent progressive decrease of the photocurrent showed in Fig. 4b for high Cr additions. In any case, further work is needed to understand the anomaly at the lowest Cr addition (0.5%). The morphological modification of the electrodes with Cr addition, illustrated by Supplementary Information, Fig. SI7 could slightly affect the donor densities measured, but it is not believed to the main reason explaining the observed behavior. In order to evaluate the stability of these films, chronoamperometric measurements were performed at 1.23 V vs RHE for 1h (Fig. 8), finding a very stable behavior of photocurrent response, with an overall loss of around the 14 % at the end of the measurement with respect the five initial minutes for the measures with hole scavenger. The higher losses observed in the measurements performed in presence of the hole scavenger is related with the fact that the rapid removal of the photogenerated holes at the surface by the sulfite can compete with photocorrosion process, which was reported indeed in Mo-doped copper vanadates as well [29]. This remarks the fact that, for better performance of this material, an appropriated surface modification with a suitable oxygen reaction catalyst that can also avoid direct contact of the semiconductor surface with the electrolyte to prevent photocorrossion, is needed.

Conclusions
In summary, we have studied the optical and photoelectrochemical properties of