Photoelectrochemical properties of BiOCl microplatelets ☆

The photoelectrochemical properties of highly crystalline and phase-pure BiOCl microplatelets synthesised via a room temperature ionic liquid method are reported. X-ray crystallography reveals a tetragonal BiOCl phase, while high resolution electron microscopy shows sheet-like structures with a cross section of approximately 5 μ m and thickness in the range of 500 nm. Di ﬀ use re ﬂ ectance spectroscopy shows a direct bandgap transition at 3.34 eV. Electrochemical measurements of as-prepared BiOCl powders deposited onto ﬂ uorine-doped tin oxide electrodes show a sharp cathodic current at − 0.10 V vs RHE at pH 10, which is linked to electron injection into the conduction band edge. Photoelectrochemical measurements in the presence of Na 2 SO 3 as hole-acceptor in solution exhibit a strong potential dependence, switching from cathodic to anodic photocurrents at potentials around 0.70 V vs RHE. The positive photocurrent is associated with SO 32 − oxidation, while the unexpected negative photocurrents are linked to cathodic material decomposition.


Introduction
The development of stable photoelectrodes for water splitting remains of one of the key challenges in material science; an area in which Roger Parsons made valuable contributions in the early 80's [1][2][3][4]. Bismuth oxychloride (BiOCl) is a V-VI-VII main group ternary wideband gap semiconductor oxyhalide composed of Earth abundant materials which has been considered in the context of solar fuels and photocatalysis [5][6][7][8][9]. BiOCl has a tetragonal structure (space group P4/ nmm) [6,10,11], comprising slabs of [Bi 2 O 2 ] 2 + interleaved with double layers of chloride ions [12], held together by non-bonding interactions. This structure also gives rise to an internal static electric field, running perpendicular to the layers, which may play a role in the separation of photogenerated charge carriers [13,14]. The conduction band (CB) of BiOCl is dominated by Bi 6p orbitals, while the valence band (VB) is dominated by O 2p and Cl 3p orbitals with additional contribution from Bi orbitals [5,15]. Indeed, one of the interesting properties of Bi oxyhalides, in comparison with other more ionic semiconductor such as TiO 2 , is the hybridised nature of the Bi 6s orbitals dispersed around the top of the VB and the bottom of the CB. This is expected to improve defect tolerance and mobility of photogenerated electron/hole pairs [6].
BiOCl has been primarily investigated in the context of photocatalysis for water remediation [16,17], with reports suggesting higher stability and activity than the benchmark TiO 2 Degussa P25 catalyst [6,12,18,19]. However, stability in some of the studies is probed by Xray diffraction, which may not be sensitive enough to changes in surface structure. Other studies have reported changes in the optical properties of the powder under UV illumination, which were linked to the formation of oxygen vacancies [20]. A number of studies have reported enhanced photocatalytic activity in the presence of these socalled oxygen vacancies, although the mechanism underlying these observations remains to be elucidated [21][22][23][24][25][26]. Ye et al. concluded that the visible absorption induced in BiOCl ("black" BiOCl) is linked to a disordered, reduced outer layer which generates a deep trap state close to the conduction band [22,26]. Several groups have carried out more in-depth studies of surface state defects (both oxygen vacancies and bismuth vacancies) in ultra-thin BiOCl nanosheets and their effect on band gap and recombination [18,27,28].
In this contribution, we report the photoelectrochemical properties of highly phase pure BiOCl microsheets obtained by an ionic liquid based method requiring low temperatures, ambient pressure and a short synthesis time. The as-grown material exhibits fascinating microstructure features such as Moire fringes in transmission electron microscopy arising from stacking of the platelet structure. Thin films obtained by spin-coating and low temperature sintering onto F:doped SnO 2 electrodes (FTO) were characterised by cathodic currents at potentials more negative than 0.05 V vs RHE, which is associated with electron injection into the conduction band of BiOCl. Electrochemical features emerging from electron injection into the conduction band provide evidence of cathodic surface decomposition, leading to the formation Bi sites. Photoelectrochemical responses were investigated in the presence of Na 2 SO 3 , a strong hole-acceptor which can minimise photoanodic decomposition of the material and surface recombination losses [40]. Unexpectedly, transient photocurrent under UV illumination were characterised by a sharp potential dependence featuring anodic and cathodic photoresponses. We rationalised the complex potential dependence of the photocurrent in terms of carrier transport and photoelectrochemical stability of the material.

Experimental
BiOCl microplatelets were synthesised via a low temperature method using 1-butyl, 3-methylimidazolium chloride as solvent [41][42][43][44][45]. Ionic liquid based synthesis is extremely versatile, enabling to dissolve a variety of metal precursors including highly acidic Bi salts in anhydrous conditions. 1-Butyl, 3-methylimidazolium chloride was prepared by mixing 1-methylimidazole (9.96 ml, 0.125 mol) (≥99%, Sigma Aldrich) with 1-chlorobutane (22.93 ml, 0.15 mol) (anhydrous, 99.5%, Sigma Aldrich) in acetonitrile (20 ml) and refluxed (70°C) for 24 h, with constant stirring. The solvent was then removed by rotary evaporation at 45°C. The remaining liquid was washed with a large excess of cold ethyl acetate, and centrifuged at 8000 rpm for 2 min. The ionic liquid is collected from the lower layer. To prepare BiOCl, 1 ml of 1-butyl, 3-methylimidazolium chloride was heated under stirring at 80°C to achieve dehydration. After 1 h, 105 mg Bi(NO 3 ) 3 ·5H 2 O (reagent grade, 98%, Sigma Aldrich) was added and stirred at 80°C until completely dissolved, yielding a clear solution without precipitants. Water was then added dropwise to the solution under stirring and white powder precipitated. The white BiOCl powder was collected by centrifugation and washed several times with ethanol.
X-ray diffraction (XRD) patterns were recorded using a Bruker AXS D8 Advance diffractometer with a θ-θ configuration and using CuKα radiation (λ = 0.154 nm). Transmission electron microscopy (TEM) studies were carried out using a JEOL JEM-1400Plus microscope and high resolution TEM (HRTEM) studies were performed with a JEOL JEM. Samples for TEM were prepared by placing 1 μl drops of the BiOCl particles dispersed in a 1:1 mix of ethanol and water on a 3 mm diameter carbon-coated copper grid. UV/Vis diffuse reflectance spectra were recorded using a Shimadzu UV-2600 UV-Vis Spectrophotometer.
Thin films in the range 2-3 μm were prepared by spin-coating 150 μl BiOCl suspension (150 mg suspended in 2 ml terpineol) onto a clean 1 cm 2 area of fluorine doped tin oxide (FTO). A Pt wire was used as a counter electrode and an Ag/AgCl placed in a lugging capillary as a reference electrode. Photocurrent measurements were conducted in Arsaturated Na 2 SO 3 aqueous solutions at pH 10, using an Ivium CompactStat. All potentials have been converted to RHE. A single compartment glass cell fitted with a quartz window was used in the photoelectrochemical studies. Illumination was provided using a LED with a narrow emission centred at 311 nm (Thorlabs) or a Newport 150 Watt Xenon Arc lamp in conjunction with a constant power supply (Bentham 605) and a monochromator (Bentham TMc300) and optical chopper (Scitec Instruments). The SEM image in Fig. 2A reveals the flake-like morphology of the as-prepared BiOCl powder. The TEM micrograph in Fig. 2B shows BiOCl sheets with diameter of 540 ± 160 nm. Fig. 2B shows most of the particles lying flat, with a small number of particles appearing as spikes when seen edge-on. The inset in Fig. 2B highlights an edge-on particle with a thickness of 20.8 ± 0.87 nm. Moiré fringes can clearly be seen  in Fig. 2C and Fig. 2D with a characteristic spacing of 7.9 ± 1.5 nm. These fringes are caused either by rotation between overlapping crystals or, in an open structure such as BiOCl, by rotation between the layers of the crystal. Fig. 3 shows a Tauc plot [46] constructed from diffuse reflectance measurements. The analysis shows a linear relationship of the Kubelka-Munk function with the excitation energy, consistent with an indirect transition at the band gap energy. The estimated band gap value was 3.41 eV (~364 nm) in agreement with the value calculated using hybrid functional including relativistic treatment [15]. Fig. 4a shows a cyclic voltammograms of BiOCl in a 0.1 M Na 2 SO 4 solution at pH 10 in the absence of light. The initial potential was set at 0.81 V vs RHE, scanning in the negative direction at a rate of 50 mV s − 1 . Upon scanning to potential more negative than −0.10 V, a sharp cathodic response can be observed. In the reverse scan (positive direction), the voltammogram features a well-defined anodic peak centred at 0.47 V. The anodic feature is only observed upon scanning the potential more negative than − 0.10 V. The red trace, obtained with a fresh sample, shows that if the potential is reversed at 0.10 V, then no anodic response is detected on the reverse scan. This behaviour clearly shows that changes in the composition at the semiconductor surface occur upon electron injection at potentials below 0.10 V. Cathodic corrosion of semiconductor electrodes has also been reported on a variety of materials such as II-VI quantum dots [47] further demonstrating that systematic analysis of voltammetric responses is essential to distinguish intrinsic electronic states from those generated by electrochemical reactions.

Results and discussion
Mott-Schottky plot constructed from analysis of impedance data in the frequency range of 67 to 3786 Hz is shown in Fig. 4b. The plot exhibit with positive slope consistent with a n-type semiconducting behaviour. The flat band potential is located at 79 mV vs RHE (4.52 eV vs Vacuum) which is close to the onset of the cathodic current in the cyclic voltammogram (Fig. 4a). This value is also consistent with the electron affinity of 4.5 eV calculated for BiOCl, which is also in the range typically observed for a variety of n-type semiconductors [15]. This behaviour confirms that the sharp cathodic current at − 0.10 V vs RHE can be associated with electron accumulation into the conduction band of BiOCl. The conduction band is dominated by Bi (III) orbitals, leading to the generation of metallic Bi at the surface, The standard potential for Reaction (1) is 0.16 V vs RHE. In the reverse reaction, Bi sites are oxidised to Bi(III) giving rise to the peak current centred at 0.47 V, This observation is in accordance with the Pourbaix diagram for Bi [48]. Considering an optical band gap of 3.34 eV and setting the conduction band edge at 79 mV vs RHE, the valance band edge is expected to be at 3.42 V. Recent reports have claimed p-type behaviour of BiOCl based on Mott-Schottky plots [49,50], with flat band potential values above 2 V vs RHE. Such unrealistic flat band potential values for a ptype semiconductor is a clear indication that the capacitance measurements are affected by Fermi level pinning most probably connected to high density of Bi sites. Fig. 5 displays a linear potential scan at 1 mV s − 1 in the negative direction in Ar-saturated 0.1 M Na 2 SO 3 solution at pH 10 under squarewave illumination. The BiOCl film was illuminated through the electrolyte solution with a LED featuring a narrow wavelength emission centred at 311 nm and a photon flux of 9.56 × 10 14 cm − 2 s − 1 . Photocurrent responses are observed across the full potential range. Intriguingly, the photocurrent switches from anodic to cathodic values at potentials close to 0.7 V vs RHE. This behaviour has not been reported previously, which is even more surprising given the high concentration of SO 3 2 − . As mentioned above, SO 3 2 − is a swift hole acceptor that promotes stabilisation of n-type semiconductors under illumination [40]. In the case of SO 4 2 − (i.e. water oxidation), the anodic photocurrents are somewhat dampened although further experiments are required to establish a more quantitative analysis of the difference in hole-extraction rate.
In order to ensure that the origin of the photocurrent reversal was not linked to rather high bias, photocurrent transients were recorded with a fresh film in the potential range close to 0.8 V. As shown in Fig. 6, the behaviour is clearly reproduced with a sharp transition between 0.7 and 0.8 V. The same photocurrent reversal is observed under oxygen saturated solutions, indicating that the cathodic photocurrents are associated with the BiOCl reduction (triggering Reaction (2)), rather than oxygen reduction. It is also observed that applying a more negative potential under illumination causes a change in the BiOCl film, from a pearlescent white appearance to silver/grey, confirming that BiOCl is being reduced.
The photocurrent spectra of BiOCl in a 0.1 M Na 2 SO 3 solution at 0.4, 0.6, 0.8 V (vs RHE) are contrasted in Fig. 7. It can be seen that the photocurrent onset is located around 370 nm, which is consistent with the optical bandgap (3.34 eV). Consequently, both photoanodic and photocathodic responses originate from the generation of charge carriers under band-gap illumination. It is also observed that the photocathodic responses have a larger magnitude and sharper potential dependence as shown in Figs. 5 and 6. The cathodic photocurrent is thus attributed to a photocathodic corrosion of the electrode, where some of the Bi 3 + is being reduced to Bi metal (Reaction (2)), whereas the anodic photocurrent is attributed to SO 3 2 − oxidation.
As discussed previously, reduced Bi sites can be generated upon injecting electrons into the conduction band of BiOCl. At potentials above 1 V, the surface states associated Bi 2 O 3 states are effectively empty, opening a pathway to electron from SO 3  Finally, the maximum quantum yield for the photoanodic and photocathodic responses were 0.3% and 1.8%, at wavelength below 310 nm. These values suggest significant recombination losses, most probably in the bulk of the film. Integrating the EQE spectra recorded at 0.4 V, we could estimate photocurrent responses in the range of 10.5 μA cm − 2 under AM 1.5G illumination. It is rather complex to benchmark this value with respect to reports in the literature given the significant differences in the experimental conditions and morphology of the electrodes. A recent study by Bachu et al. also mentioned the switch of photocathodic to photoanodic responses, reporting a value of 100 μA cm − 2 at approximately 1.8 V vs RHE under 100 mW cm − 2 illumination [49]. The data reported by Fan and co-workers are difficult to rationalize given that the BiOCl film was connected to the counterelectrode in a three-electrode potentiostatic assembly [50]. Liu et al. obtained photocurrent values below 0.5 μA cm − 2 for BiOCl heterostructure featuring two different phases, however the potential and illumination levels are not specified in the report [51].
Based on these observations, we conclude that pristine BiOCl particles are unsuitable as photoelectrodes for the water-splitting reaction, mainly due to the low-lying conduction band edge and the pore charge transport in highly textured films. However, metallic Bi sites are spontaneously generated under illumination in aqueous solutions which can potentially catalyse the hydrogen evolution or oxygen reduction reaction. It could be envisaged that under appropriate illumination levels and electrolyte composition, stable photocatalytic performance can be achieved by these materials. An alternative strategy to improve performance is to facilitate transport of majority carriers by generating composite materials, e.g. high surface area carbon [52,53].

Conclusions
This study reports on the photoelectrochemical responses of BiOCl thin films prepared by an ionic liquid synthesis method. This room temperature method generates crystalline, tetragonal phase BiOCl micro-platelets with the (110) plane being the most prominent. Thinfilms were prepared by spin-coating onto F:SnO 2 electrodes, showing a cathodic response at −0.1 V vs RHE assigned to electron injection into the conduction band edge. Potential dependence of the capacitance also suggests that the flat band potential is located at 0.079 V. Interestingly, electron injection into the conduction band leads to the appearance of a broad anodic peak at 0.47 V, which is linked to the oxidation of surface Bi sites. This peak suggest that electron accumulation leads to the cathodic decomposition of BiOCl, yielding metallic Bi sites. Photoelectrochemical responses in the presence of Na 2 SO 3 at pH 10 show a complex potential dependence across the range between 0.30 V and 1.10 V. At potentials above 0.7 V, the photocurrent is anodic and linked to the oxidation of SO 3 2 − . As the potential is shifted more negative, the photocurrent sign sharply switches to negative values. Comparing the potential dependence of photocurrent with the cyclic voltammogram in the dark, the reversal of the photocurrent sign occurs at the onset of reduction of Bi oxide surface sites. This interesting observation suggest that hole-injection to SO 3 2 − is mediated by surface Bi-oxide sites. Once electrons are populated in these sites, i.e. at potentials more negative than 0.7 V, the generation of metallic Bi promotes negative photocurrent responses. This photocathodic signals are most probably connected to further decomposition of BiOCl to form metallic Bi.