A simple , low-cost CVD route to thin fi lms of BiFeO 3 for e ffi cient water photo-oxidation †

Materials Chemistry Centre, Department of Gordon Street, London, WC1H 0AJ, UK. E-m Department of Chemical Engineering, Univ London, WC1E 7JE, UK UCL Healthcare Biomagnetics Laborator 4BS, UK National Physical Laboratory, Hampton 0LW, UK † Electronic supplementary information details, DSC–TGA pattern of [{Cp(CO)2F BiFeO3 lm recorded at 195 C, XPS sp thin lm aer etching, top-down SEM im glass at 300 C, top-down SEM images o 650 C, P–E hysteresis loop for a pure BiF and ZFC–FC curves for BiFeO3 lm, UV corresponding Tauc plot for a BiFeO3 lm Cite this: J. Mater. Chem. A, 2014, 2, 2922


Introduction
Water photolysis for H 2 fuel generation has the potential to meet increasing energy demands whilst reducing the emission of harmful greenhouse gases into the atmosphere. Water photolysis can be considered to be composed of two half reactions, water oxidation (to O 2 ) and water reduction (to H 2 ). However, water oxidation is widely considered to be more challenging given the fact that generation of one molecule of O 2 requires four holes, generated on a timescale ve orders of magnitude slower than the two electron proton reduction to H 2 . 1,2 Therefore the search for a stable, efficient water oxidation photocatalyst is widely regarded to be signicant for large-scale water photolysis. The most commonly used materials for photocatalytic water splitting are binary transition metal oxides but the band-gaps of these materials (over 3.0 eV) are too high to serve as efficient photocatalysts under visible light irradiation. 3,4 Perovskite bismuth ferrite (BiFeO 3 , "BFO") exhibits a direct band-gap of approximately 2.2 eV and is an active photocatalyst; [5][6][7] BiFeO 3 nanowires have been demonstrated to be promising oxygen evolution catalysts exhibiting high efficiencies under UV-light irradiation, and very recently Au-BiFeO 3 nanowires have been reported to be highly active for oxygen evolution under visible light (l > 380 nm) irradiation. 8,9 Chemical Vapour Deposition (CVD) has many potential advantages for deposition of BiFeO 3 thin lms including excellent substrate coverage, low-cost, ease of scale-up, control over thickness and morphology and high throughput capabilities, however the growth of phase-pure BiFeO 3 lms using chemical deposition techniques is challenging. [10][11][12] It has been suggested that single-source heterometallic precursors could be exploited in order to improve stoichiometry control in multicomponent materials. 12,13 However no examples for BiFeO 3 thin lms appear in the literature despite the availability of a number of bimetallic bismuth-iron containing complexes, i.e.
[Bi 2 (Hsal) 6 15 which could serve as potential precursors. Here, we describe the growth of BiFeO 3 lms onto a variety of substrates via a simple, low-cost solution based aerosol assisted (AA)CVD process utilizing the single-source precursor [CpFe(CO) 2 BiCl 2 ] (Cp ¼ cyclopendienyl, C 5 H 5 ). AACVD is advantageous as it does not rely upon the use of highly volatile precursors, essential for typically high molecular weight heterometallic cluster compounds. 16 Post-deposition heat treatment of as-deposited lms resulted in pure BiFeO 3 at 700 C, conrmed via XRD, low temperature Raman and XPS spectroscopy. BiFeO 3 lms displayed the expected ferroelectric and ferromagnetic behavior and possessed direct band-gaps of $2.1 eV and we have demonstrated the lms to be highly active photocatalysts for water oxidation under simulated solar irradiation. This new synthetic methodology enables large area thin lm deposition, and hence is relevant for high volume applications such as solar driven water oxidation or organic pollutant degradation for water treatment.

Experimental
Detailed experimental information on lm analysis and photochemical measurements are provided in ESI. †

Precursor synthesis
The bimetallic molecular precursor [{Cp(CO) 2 Fe}BiCl 2 ] was synthesised according to the literature from a simple equimolar reaction of [CpFe(CO) 2 ] 2 and BiCl 3 in dichloromethane, and possessed identical NMR and IR spectra to those previously reported. 17 CVD AACVD reactions were carried out using an in-house built coldwall CVD described elsewhere. 18,19 Nitrogen (99.96%) was obtained from BOC and used as supplied. AACVD experiments were initially conducted on SiCO cated oat glass substrates (150 mm Â 45 mm Â 3 mm) supplied by Pilkington Glass Ltd (NSG group). The glass substrates were cleaned thoroughly in commercial washing up detergent, dried, and then cleaned with isopropanol then dried using a heat gun. In order to anneal lms to temperatures greater than 600 C, depositions were carried out on 20 mm Â 20 mm Â 2 mm Corning 1737 AMLCD alkaline-earth boro-aluminosilicate transparent glass substrates. For ferroelectric measurements, lms were deposited onto silicon wafers which were sputtered for 180 seconds with a thin layer of platinum (argon pressure 0.1 torr, current 25 mA) prior to use, with lm deposition onto the platinum. 150 mg (0.33 mmol) of [CpFe(CO) 2 BiCl 2 ] precursor dissolved in ca. 40 cm 3 dry THF was used for each deposition. The nitrogen gas ow through the precursor was maintained at 0.8 l min À1 and regulated using a calibrated ow meter. Annealing was carried out in air for two hours at a heating ramp rate of 10 C min À1 . For each annealing experiment a fresh sample deposited via AACVD at a substrate temperature of 300 C was used.

Photoelectrochemical measurements
Chronoamperometry measurements were conducted using a potentiostat, a Pyrex cell with a glass window and a mechanical light chopper. A Pt wire was used as the counter electrode and an Ag/AgCl electrode was used as the reference electrode. An aqueous 0.2 M sodium sulphate (Na 2 SO 4 ) solution was purged for 15 minutes with argon and was used as the electrolyte (pH 6.5). The light source was a 150 W Xe lamp equipped with an AM 1.5G lter (100 mW cm À2 , 1 Sun, Newport, USA). The cell was sealed with a rubber septum. The scan rate was 10 mV s À1 . Mott-Schottky (impedance) measurements were measured in 0.2 M Na 2 SO 4 in the dark at a frequency of 1 kHz and scan rate of 10 mV s À1 . The potential was measured against an Ag/AgCl reference electrode and converted to RHE potentials using

Photocatalytic oxygen evolution
Selected lms were used to photo-oxidise water using sacricial reagents (alkaline sodium persulphate) under UVA (365 nm) and simulated solar irradiation (150 W Xe lamp). 20 In a typical experiment, the lm was immersed in 30 cm 3 aqueous solution under strong stirring conditions (55 rpm) in a quartz vessel with water-cooled walls (T ¼ 298 K). The photo-oxidation of water is biased through immersion in a solution containing a sacricial electron-acceptor (scavenger) composed of 0.01 M Na 2 S 2 O 8 in 0.1 M NaOH. 21 The MPD is comprised of a circular shaped silver electrode (counter and reference) and a platinum electrode disc (cathode) connected via a salt bridge (3 M KCl). 21 The Pt electrode is protected from the test solution by a gas-permeable PTFE membrane.

Results and discussion
AACVD of [CpFe(CO) 2 BiCl 2 ] (the TGA trace of this compound is shown in the ESI, Fig. S1 †) in THF solvent at a substrate temperature of 300 C resulted in the formation of adherent dark orange lms, passing the Scotch tape test, with complete substrate coverage. Compositional analysis via WDX revealed these lms contained 74 at% bismuth, 23 at% iron and 3 at% chlorine. X-ray diffraction showed only the presence of Bi 24 Fe 2 O 39 ( Fig. 1 421c, PDF no. 042-0201), indicating other non-crystalline iron or bismuth rich species must also be present. No evidence of BiFeO 3 formation was observed at this temperature (300 C) via X-ray diffraction. XPS analysis revealed the presence of Fe 2p 3/2 at 710.9 eV and Bi 4f 7/2 at 158.7 eV, characteristic of the presence of Fe 3+ and Bi 3+ . 22,23 In order to obtain BiFeO 3 the as-deposited lms were annealed at a variety of temperatures up to 700 C (between 450 and 550 C Pilkington SiCO oat glass was used and for 700 C Corning glass was used) with the average results, aer repeating the process several times to ensure reproducibility, summarised in Table 1.
Post-deposition annealing at increasing temperature led to a decrease in lm thickness and relative loss of bismuth (to iron), most likely as bismuth or bismuth oxide, 24 and progressive conversion of Bi 24 Fe 2 O 39 to Bi 2 Fe 4 O 9 and subsequently to BiFeO 3 in agreement with the phase diagram constructed by Scott 25 and Lu. 26 Raising the annealing temperature to 700 C caused the Bi : Fe ratio to become near unity (Bi 51 at%, Fe 49 at%) and the appearance of the lms changed from dark orange (as-deposited) to bright orange. At 700 C BiFeO 3 was the only phase present via XRD [ Fig. 1 27 Raman analysis further conrmed the assignment of phase pure BiFeO 3 (ESI, Fig. S2 †). Chlorine contamination from the precursor was initially high for the as-deposited lm and for those annealed at 400 C (2.5 and 1.6 at% respectively), however this contamination decreased dramatically as a function of annealing temperature and by 600 C was below the detection limit of WDX.
XPS of a lm annealed at 700 C showed the presence of iron in the +3 oxidation state [Fe 2p 3/2 , 711.3 eV (Fig. S3, ESI †)] with the expected Fe 3+ satellite peak observed at 718.9 eV. 12 Oxygen under stoichiometry, leading to a co-existence of Fe 2+ and Fe 3+ species, leads to broadening of the Fe 2p 3/2 peak to lower energies but this was not observed in our samples. A single Bi 4f 7/2 ionisation at 159.4 eV was observed, characteristic of bismuth in the +3 oxidation state, as expected for BiFeO 3 . 23 No chlorine contamination was detected via XPS; carbon was observed on the surface but decreased upon etching indicating it to be surface contamination. Analysis of the Bi 4f, Fe 2p and O 1s (B.E. of O 1s ¼ 530.2 eV) peak areas aer etching indicated the presence of the three components in an approximate 1 : 1 : 3 ratio, commensurate with BiFeO 3 .
As-deposited lms (300 C) possessed a globular morphology, with average particle diameters of 100 nm (ESI, Fig. S4 †). Heat treatment at higher temperatures led to coalescence of particles (ESI, Fig. S5 †), whilst at 700 C ( Fig. 2(a)) the sintering of the particles led to lms becoming rougher and less uniform; AFM (Fig. 2(c and d)) showed the lms to be comprised of larger aggregates with a rough texture (root mean squared roughness (rms) ¼ 62 nm) in agreement with SEM. Hence lms synthesised by this route are likely to possess high surface areas for enhanced catalytic activity.
Films of BiFeO 3 , grown by deposition directly onto 1 cm 2 Pt/ SiO 2 /Si wafers via AACVD followed by annealing at 700 C, displayed a maximum polarisation of 8.7 mC cm À2 , providing an effective relative permittivity of almost 800 (ESI, Fig. S6 †). The maximum polarisation is smaller than that for single crystal or epitaxial lm bismuth ferrite (around 50 mC cm À2 ), 28 but higher than those obtained for BiFeO 3 lms grown via sol-gel processing (P r ¼ 1.8 mC cm À2 ) 29 or PLD (P r ¼ 0.83 mC cm À2 ). 30 Measurement of the magnetic properties of a 320 nm thick lm (ESI, Fig. S7 †) revealed that the M-H hysteresis loops recorded at 5 K and 300 K were similar to those reported previously 28 (aer subtracting the diamagnetic contribution from the substrate), i.e. these BiFeO 3 lms show weak ferromagnetic behaviour with well-saturated hysteresis loops at both temperatures. Upon raising the temperature to 300 K the coercivity was measured as 135 Oe and À115 Oe with an expected decrease in saturation magnetisation to 8.9 emu cm À3 . The coercivity is lower than that observed for a 70 nm thick BiFeO 3 lm grown via PLD (200 Oe) 31 although a decrease in magnetisation as a  function of increasing lm thickness was observed, from 150 emu cm À3 for a 70 nm thick lm to 5 emu cm À3 for a 400 nm thick lm, consistent with our measurement of lm thickness ($320 nm). Spin-glass behaviour was also observed from M-T measurements (ESI †). A typical transmission spectrum of a lm containing BiFeO 3 as the only crystalline phase (ESI, Fig. S8 †) shows over 70% transmittance in the 800-2500 nm range with the cut-off from the glass substrate coming into effect below 380 nm. The bandgaps of the lms were calculated using Tauc plots (see Fig. S9, ESI †). 32 Extrapolating the linear part of the plot to the x-axis for a phase-pure sample resulted in an intercept of approximately 2.1 eV, in good agreement with experimentally derived bandgap values for BiFeO 3 ; lms with smaller band-gaps (2.0 eV) contained impurity Bi 2 Fe 4 O 9 and are in line with the expected decrease in band-gap of Bi 2 Fe 4 O 9 compared to BiFeO 3 . 33,34 For photoelectrochemical (PEC) measurements BiFeO 3 was grown on FTO coated glass substrates. In order to estimate the relative levels of the conduction and valence bands of BiFeO 3 electrical impedance measurements were carried out from which the at-band potential (E  ) was measured. Fig. 3(a) displays the Mott-Schottky plot for a BiFeO 3 lm deposited onto FTO coated glass. For n-type semiconductors the at-band potential (E  ) is considered to be located just under the conduction band, hence E  of BiFeO 3 was estimated to be À0.31 V (vs. Ag/AgCl) or + 0.18 V (vs. RHE), 35,36 which is similar to values previously estimated from atomic electronegativities. 37 Based upon this value and that determined for the band-gap of BiFeO 3 (2.1 eV), a band diagram (Fig. 3(b)) was constructed, indicating BiFeO 3 has signicant overpotential for photocatalytic water oxidation.
The photoanodic activity of BiFeO 3 was investigated using PEC measurements and also via oxygen evolution using a sacricial electron acceptor solution. Fig. 4 shows the chronoamperometry measurement obtained from a BiFeO 3 photoelectrode under 1 Sun (100 mW cm À2 ) AM 1.5G illumination. There is a steady increase in photocurrent with increasing potential and the photocurrent at 1.0 V vs. Ag/AgCl was appreciable (ca. 0.1 mA cm À2 (103 mA cm À2 )). This value is similar to that obtained by Yu et al. 38 for BiFeO 3 lms deposited via PLD onto platinised silicon substrates ($90 mA cm À2 at 1.0 V vs. Ag/ AgCl, 400 W Xe lamp) but substantially higher than the current density reported for hydrothermally synthesised BiFeO 3 nanocube electrodes in pure water (5.2 mA cm À2 at 1.0 V vs. SCE, 500 W Hg lamp) 39 and higher than BFO/SRO/STO lms grown via sputtering (10 mA at 0.64 V vs. Ag/AgCl). 40 In order to verify the water oxidation activity, tests for oxygen evolution from a Na 2 S 2 O 8 /NaOH sacricial solution were carried out using an MPD cell (see ESI † for more details) and compared to a commercial standard photocatalyst (Pilkington Activ® glass). 41 The overall photocatalytic reaction is given by eqn (1). Fig. 5 shows the typical output of the MPD cell during photogeneration of oxygen on BiFeO 3 lms under full-arc Xe-lamp irradiation (150 W) with the corresponding amounts of oxygen produced during the rst 10 h irradiation shown in Fig. 5(b). The output reading shows an increase in voltage during an initial illumination period followed by a plateau around 0.05 V, indicating steady oxygen production during prolonged irradiation (above 20 h). A small drop in the signal when turning on the Xe lamp, attributed to an interruption of the electrical supply, was observed however this did not affect the observed rate. Signal dri was also observed aer prolonged use of the MPD cell, as illustrated by the blank test carried out on uncoated glass, however signal drop was noted aer switching off the light source demonstrating the oxygen evolution was a photocatalytic effect. In addition no change in voltage, i.e. no oxygen evolution, was observed in the absence of illumination. The oxygen rates and % O 2 yields obtained during UVA irradiation of BiFeO 3 and Activ® glass lms are given in Table 2.
The % O 2 yield of the process has been determined according to eqn (2).  An average O 2 yield of 24.4% (365 nm) was obtained for BiFeO 3 (in the absence of co-catalyst) which represents a near 10-fold increase over optimised B-doped TiO 2 lms investigated under identical conditions. 42 The O 2 yield of the Activ® reference was $5% at 365 nm. It should be noted that the BiFeO 3 lms tested were signicantly thicker ($300 nm) than that reported for the TiO 2 layer in Activ® glass ($15 nm). XPS and Xray diffraction patterns of the lms conducted aer photocatalytic testing revealed no change in composition or phase indicating that these BiFeO 3 lms are robust under the basic conditions of the test and do not undergo photocorrosion, which was supported by the continued measurement of oxygen evolution on prolonged irradiation and subsequent tests which proved the effect was reproducible. Visible-light tests were also carried out using UV cut-off lters under similar experimental conditions however no increase in voltage could be distinguished above the signal dri of the MPD cell. This is perhaps not surprising given the photocatalytic activity of BiFeO 3 powders for oxygen evolution (from a FeCl 3 solution) using visible light irradiation (500 W Hg lamp, l > 420 nm) is less than 0.2 mmol h À1 , 39 and hence we are currently investigating more suitable methods for evaluating the visible light activity of BiFeO 3 thin lms. The high photocatalytic activity for water oxidation displayed by our BiFeO 3 lms may not be solely due to the high surface area of thin lms; it has recently been reported that ferroelectric domains in BiFeO 3 enhance charge-carrier separation through photocatalytic decomposition of AgNO 3 , resulting in selective Ag reduction on the surface and simultaneous oxygen evolution under visible light irradiation. 43,44 Conclusions For the rst time, BiFeO 3 lms were grown via a simple solution-based AACVD procedure at an unprecedented low temperature using the single-source precursor [{Cp(CO) 2 Fe} BiCl 2 ], followed by post-deposition annealing at elevated temperature in air. As-deposited lms were characterised as containing Bi 24 Fe 2 O 39 via XRD; annealed lms were identied as BiFeO 3 by XRD and Raman spectroscopy with compositional analysis revealing bismuth to iron ratios of 1 : 1. Magnetic hysteresis and ferroelectric polarisation measurements conrmed ferromagnetic and ferroelectric ordering at room temperature. Direct band-gaps between 2.0 and 2.2 eV were measured for all lms. Photocatalytic testing under both UV and solar irradiation conrmed appreciable activity of BiFeO 3 for the kinetically slow four hole process of water oxidation. 1 Despite having a less deep valence band potential than TiO 2 , facile water oxidation is found, with the resultant apparent quantum yield of our samples exhibiting a near six-fold increase over a commercial standard photocatalyst (TiO 2 Activ® glass) and a ten-fold increase over B-doped TiO 2 lms recently reported. 42 Furthermore, we have demonstrated the possibility of using relatively non-volatile molecular compounds for the growth of complex heterometallic oxides via AACVD. To the best of our knowledge, no precursors of this class have previously been utilised in CVD processes. Fig. 5 (a) Typical output of the MPD cell during photo-generation of oxygen on BiFeO 3 (black line) and Activ glass (blue line) films under solar (Xe) lamp irradiation conditions. The shaded areas represent running time in the dark. A test using a plain glass substrate is also included for reference (green line). The inset shows the typical output signal for air-and nitrogen-saturation conditions (1 and 0 V, respectively); (b) continuous oxygen evolution of both BiFeO 3 and TiO 2 (Activ) films under solar irradiation over a period of 10 hours.