Facilitating hole transfer on electrochemically synthesized p-type CuAlO2 films for efficient solar hydrogen production from water

Delafossite CuAlO2 photoelectrodes are synthesized via the electrodeposition of Cu(II) and Al(III) onto fluorine-doped tin oxide (FTO) substrates in water and dimethylsulfoxide (DMSO) solvents, followed by annealing in air and Ar. The surface properties, crystalline structure, and photoelectrochemical (PEC) performance of the as-synthesized samples are significantly affected by the synthetic conditions. Optimized CuAlO2 electrodes (synthesized in DMSO and annealed in air) possess suitable energetics for H2 production under sunlight (an optical bandgap of ∼1.4 eV and a conduction band level of −0.24 VRHE). They exhibit a photocurrent onset potential of ∼+0.9 VRHE along with a faradaic efficiency of ∼70% at +0.3 VRHE in an alkaline solution (1 M KOH) under simulated sunlight (AM 1.5; 100 mW cm−2). The addition of sacrificial hole scavengers (sulfide and sulfite) significantly improves the PEC performance of CuAlO2 by a factor of eight, along with providing a faradaic efficiency of ∼100%. This indicates that the hole transfer limits the overall PEC performance. This issue is addressed by employing a ∼150 nm-thick Au film-coated FTO substrate for the CuAlO2 deposition. In the absence of hole scavengers, the H2 production with the Au-underlain CuAlO2 photoelectrode (Au/CuAlO2) is three-fold higher than that with bare CuAlO2, while the faradaic efficiencies at +0.3 and +0.55 VRHE are ∼100%. The time-resolved photoluminescence emission decay spectra of the CuAlO2 and Au/CuAlO2 confirm the facilitated charge transfer in the latter.


Introduction
The semiconductor-based solar production of carbon-neutral chemicals (e.g., H 2 and formate from water and CO 2 , respectively) has been studied over the past four decades, 1-4 but has recently received greater attention because of the increase in the concentration of atmospheric CO 2 beyond 400 ppm. There are numerous semiconductors available, including oxides, 5-9 chalcogenides, 10-13 silicon, [14][15][16] and III-V composites. 3,17 However, they oen suffer from low efficiency, a complicated synthetic process, the use of expensive components, non-scalability, and low durability. Cu(I)-based delafossite materials are unique in terms of their structure (Cu I M III O 2 type, where M ¼ Fe, [18][19][20][21] Rh, 22 Al, 23,24 Ga, 25 etc.); various bandgap (E g ) energies (1.2-3.0 eV); high conduction band (E cb ) level, which is sufficient for H 2 production and CO 2 reduction; and relative stability in aqueous solutions compared to other p-type III-V and II-VI materials. 26,27 For example, electrodeposited p-CuFeO 2 (E g $ 1.36 eV) was shown to be capable of producing H 2 in an aqueous alkaline solution, 28 whereas intercalating Mg 2+ or oxygen atoms into CuFeO 2 enhanced the photoelectrochemical (PEC) performance. 18 Furthermore, CuFeO 2 coupled with CuO (E g $ 1.4 eV) could produce formate from CO 2 and water at a circumneutral pH with a $1% energy efficiency in the absence of any potential bias. [29][30][31] In comparison to CuFeO 2 , CuAlO 2 has been given less attention despite their similar physicochemical properties. The typical synthetic route of CuAlO 2 is annealing a Cu(I) and Al(III) salt mixture at high temperature (solid-solution process), 23,26 which results in irregular, coarse particles of several micrometers. 18,25 Although this method has some advantages (e.g., high yield), the as-synthesized particles are difficult to fabricate into durable lms on transparent conducting oxide (TCO) substrates because of the absence of particle-to-particle interaction. Even if they are formed, the lms have less intimate and looser interparticle connections undergoing a signicant charge recombination at the solid/solid interface. 18,23 This difficulty in synthesizing CuAlO 2 lms has caused this material to be less studied despite its potential as a promising photocathode.
With this in mind, we have, for the rst time, attempted to synthesize CuAlO 2 lms on TCO substrates via electrochemical deposition (ED) under various experimental conditions (e.g., ED potentials, times, solution media, and annealing atmospheres). The as-synthesized CuAlO 2 samples were characterized using various surface analysis tools (SEM, EDX, XRD, XPS, UV-vis, impedance, and time-resolved uorescence spectrometry). They were then further evaluated in terms of their PEC hydrogen production in aqueous alkaline solutions under simulated sunlight (AM 1.5; 1 sun). The results of this evaluation showed that the hole transfer limited the overall PEC performance, and the use of sacricial hole scavengers (electron donors) signicantly improved the H 2 production. To address this issue, thin Au layers ($150 nm), as a hole conductor, were pre-deposited onto TCO substrates via an electron-beam evaporation system to facilitate the hole transfer of CuAlO 2 . In the absence of the hole scavengers, the H 2 production with the Auunderlain CuAlO 2 photoelectrode (Au/CuAlO 2 ) was three-fold higher than that with bare CuAlO 2 , while the faradaic efficiencies were $100%.

Synthesis of samples
Pieces of uorine-doped SnO 2 (F:SnO 2 , FTO)-coated glass (Pilkington Co, 1 cm Â 3 cm) were ultrasonically cleaned in ethanol for 10 min, rinsed with deionized water (>18 MU cm, Barnstead), and dried in an N 2 stream. For the electrochemical plating of Cu and Al, the as-prepared FTO (working electrode), saturated calomel electrode (SCE, reference electrode), and Pt wire (counter electrode) were immersed in a deionized water or dimethylsulfoxide (DMSO, >99%, Wako) solution containing Cu(NO 3 ) 2 $3H 2 O (4 mM, >99%, Sigma Aldrich), Al(NO 3 ) 3 $9H 2 O (20 mM, >98%, Sigma Aldrich), and KClO 4 (50 mM, >99%, Sigma Aldrich). Then, the FTO substrates were held at constant potentials (À0.31 V SCE in water and À1.91 V SCE in DMSO) for 2 h using a potentiostat/ galvanostat (CompactStat, Ivium) ( Fig. S1 †). Aer the deposition, the samples were dried, washed with deionized water, and placed in a tube furnace (Ajeon Heating Industrial Co., LTD) at room temperature in the presence of atmospheric air or Ar. The furnace temperature was increased to 700 C at a rate of 2 C min À1 and held at 700 C for 1 h. If necessary, Au-layered FTO (FTO/Au) was used for the deposition of CuAlO 2 . For this, FTO substrates were coated with a 150 nm-thick Au layer using an electron-beam evaporation system (Dada Korea) with a metallic Au evaporation slug (99.999%) in a reactor chamber at a pressure of 2 Â 10 À5 Torr. The growth rate of the Au layer was estimated to be $0.05 nm s À1 . 32

Photoelectrochemical measurements and product analysis
The as-synthesized samples (working electrodes) were immersed in aqueous potassium hydroxide (1 M KOH at pH $13.5, Sigma Aldrich) pre-purged with N 2 for over 1 h in an airtight single (undivided) glass cell with an SCE (reference electrode) and a platinum wire (counter electrode). Light-chopped linear sweep voltammograms were obtained via a potential sweep from +0.3 to À0.9 V SCE at a scan rate of 5 mV s À1 under simulated solar light (100 mW cm À2 ) from a 150 W xenon arc lamp (ABET Technology) equipped with an air mass (AM) 1.5G lter. The light intensity was weekly re-calibrated to be 1 sun (100 mW cm À2 ) using a standard mono-Si solar cell (K801S-K009, McScience Inc.), as described elsewhere. 33 For the PEC H 2 production, constant potentials (À0.75 V SCE and À0.5 V SCE ) were applied to the samples under irradiation. The potentials of the reference electrode (SCE) were converted to those of a reversible hydrogen electrode (RHE) using the following relationship: Unless otherwise specied, the RHE was omitted for simplicity. The incident photon-to-current efficiency (IPCE) was estimated using a CS130 monochromator (Mmac-200, Spectro) with a 300 W Xe arc lamp using the following equation: where I ph , P light , and l refer to the photocurrent densities at 0.3 and 0.55 V, photon ux, and wavelength, respectively. For quantication of molecular hydrogen (H 2 ) evolved, varying volumes (10-250 mL) of a standard H 2 gas (99.999%) with Ar carrier gas were owed through a 5Å molecular sieve column equipped in a gas chromatograph (GC, YoungLin, ACME-6100) with a thermal conductivity detector (TCD) (detection limit of H 2 $ 0.01%), and a standard curve t between the standard gas concentration and the corresponding spectral area was obtained. The faradaic efficiencies for H 2 under constant potentials were estimated using the following equation: where I ph , A, and t are the photocurrent density (A cm À2 ), area (0.25 cm 2 ), and time (s), respectively. All of the photoelectrochemical experiments were repeated at least twice to obtain reliable results.

Surface characterization
The surface morphologies and side views of the samples were analyzed using a eld-emission SEM (FE-SEM, Hitachi S4800) equipped with an energy-dispersive X-ray (EDX) detector. The UV-vis diffuse reectance absorption spectra of the powders collected from the sample lms were obtained using a UV-vis spectrophotometer (UV-2450, Shimadzu), with BaSO 4 as a ref. 29. The obtained reectance (R) was then converted into absorbance via the Kubelka-Munk function ((1 À R) 2 /2R). 34 Xray diffraction (XRD) measurements were performed to examine the crystalline structures of the samples with a Philips X-pert powder diffractometer (PW3040/00) in a Bragg-Brentano geometry under Cu Ka radiation. X-ray photoelectron spectroscopy (XPS) and Auger spectroscopy (Cu LMM; XPS, ULVAC-PHI) analyses were performed on a PHI 5500 model spectrometer equipped with an Al Ka monochromator X-ray source at 20 kV, a hemispherical electron energy analyzer, and a multichannel detector.
On the other hand, the sample deposited in water and annealed in an air atmosphere (water/air) exhibited a more anodic E on ($1.0 V) and yet large dark currents. The sample deposited in DMSO and annealed under Ar (DMSO/Ar) was poor at generating a photocurrent and suffered from a signicant dark current. In contrast, the sample annealed in an air atmosphere (DMSO/air) displayed an E on of $0.9 V and a high photocurrent. Furthermore, a dark current started from E $0.42 V, indicating that the sample is more durable than the other samples. The DMSO/Ar sample was yellowish, whereas the other samples were blackish. This suggests that the latter three samples possess similar compositions and/or crystalline structures. The XRD patterns of the as-synthesized samples were further examined to gain knowledge about the crystalline structures (Fig. 2a). Most of the XRD peaks in the water/Ar sample were indexed to CuAlO 2 (e.g., 006, 101, 012, 009, 018, and 112 planes at 2q ¼ 32.5 , 35.7 , 38.8 , 48.9 , 57.1 , and 68.1 , respectively; JCPDS no. 73-9485), whereas there was a peak indexed to CuO (220) at 2q ¼ 58.4 . The water/air sample exhibited the same XRD pattern. In the DMSO/air sample, CuAlO 2 peaks were predominant, whereas the intensity of CuO (220) was signicantly reduced. It should be noted that the Cu(I) in the Cu-based delafossites (Cu I M III O 2 ) is readily oxidized to Cu(II) in the synthesis process and/or a partial oxidation of CuAlO 2 in air, leading to oxygen-enriched oxides. 23,29,31 The presence of the mixed valence state of Cu(I) and Cu(II) usually increases the charge transfer. 19,31,36,37 In contrast to these samples, the DMSO/Ar sample did not exhibit CuAlO 2 peaks, whereas only Al 2 O 3 -associated peaks (e.g., 111, 110, and 113 planes at 2q ¼ 31.1 , 36.7 , and 43.6 , respectively; JCPDS no. 75-0277) were found. This remarkable difference from the other samples was further conrmed by the yellowish color of the DMSO/Ar sample (Fig. 1).
Based on this knowledge, CuAlO 2 (i.e., the DMSO/air sample) was further analyzed using the XPS (Fig. 2b). The sample showed mixed bands at binding energies of 72 and 80 eV, the deconvolution of which indicated the co-presence of Al2p (73.6 eV) and Cu3p (74.9 and $76.8 eV) at an atomic ratio of 1 : 1. The EDX analysis of the sample conrmed the similar composition ratio (Fig. S2 †). In addition, the XPS O1s band could be resolved into a single oxygen atom coordinated to Cu(I) at 529.6 eV and two oxygen atoms coordinated to Al(III) (one from the AlO 6 edgesharing octahedral layer at 530.8 eV and the other from a surface hydroxide or hydrated species at 531.9 eV) (Fig. S3 †). Two satellite peaks in the Cu2p spectrum further suggested the existence of Cu(II) (Fig. 2b), 33 which was consistent with the results of the XRD analysis. The presence of Cu(II) was further conrmed by the XPS spectra of the Cu LMM Auger transition (Fig. 2b inset). The as-synthesized CuAlO 2 sample exhibited a kinetic energy peak at $899.7 eV, which was located between those of Cu 2 O ($898.8 eV) and CuO ($900.3 eV). The assynthesized CuAlO 2 samples (i.e., water/Ar, water/air, and DMSO/air) exhibited $7 mm-thick porous particulate lms composed of ower-like three-dimensional aggregates (Fig. 2ce). This morphology is quite similar to those of electrochemically synthesized Cu-based oxide lms (CuFeO 2 , CuO, Cu 2 O, etc.), 29-31 throughout which the component elements are uniformly distributed. 29,31 The elemental mapping of the CuAlO 2 conrmed the uniform distribution of Cu and Al horizontally and vertically in the sample (Fig. S4 †).

Use of CuAlO 2 for photoelectrochemical H 2 production
The PEC hydrogen evolution with CuAlO 2 (i.e., DMSO/air sample) was systematically examined in an alkaline electrolyte (1 M KOH) with sulde and/or sulte (0.1 M Na 2 S, 0.1 M Na 2 SO 3 , and their mixed solution) as the sacricial hole scavenger (Fig. 3a). It should be noted that sulde and sulte are typical products in the ue gas desulfurization processes of smelters and coal-red power plants. 38 The overall shapes of the LSVs of the sulde and sulde/sulte solutions were similar, with an E on of $0.85 V and a cathodic peak at $0.45 V. When sulte alone was present, a large dark current owed at E < $0.7 V, whereas no cathodic peak was found. This suggests that the cathodic peak of CuAlO 2 could be attributed to sulde, which has a higher reducing power (E (S/S 2À ) ¼ À0.476 V) than sulte (E (S 2 O 6 2À /H 2 SO 3 ) ¼ 0.564 V). Despite the cathodic peak of the mixed solution, the dark current was inhibited down to an E value of $0.37 V, which was 0.1 V more negative than that in water (Fig. 1d vs. Fig. 3a). In the absence of the hole scavengers, the application of a potential at +0.55 V did not produce H 2 for 3 h despite a photocurrent ow. We could observe measurable amounts of H 2 ($2.5 mmol for 3 h) only at E # +0.3 V (Fig. 3b). A photocurrent of $0.1 mA cm À2 was generated during the same period (Fig. S5 †), leading to a Faraday efficiency of $70% for H 2 production (Fig. 3c). In the presence of sulte alone, the H 2 production and Faraday efficiency were enhanced yet insignificantly ($5 mmol for 3 h and $80%, respectively). However, the addition of sulde to the water and sulte solutions markedly enhanced the H 2 production by $7 and $4 times, with Faraday efficiencies of $90 and $100%, respectively. Compared to sulde alone, the higher efficiency of the sulte/sulde mixture was attributed to the regeneration of hole-oxidized sulde (e.g., S 2 2À ) by sulte. [10][11][12]39 The Faraday efficiency of $100% with bare CuAlO 2 in the mixed solution reveals that electron injection at the p-type material/water interface was highly efficient, whereas the internal hole transfer limited the overall charge transfer (see below). For comparison, the PEC activities of the other samples (DMSO/Ar, water/Ar, and water/air) were examined in the mixed solution (Fig. S6 †). The DMSO/Ar sample produced neither photocurrents nor H 2 at +0.3 V because of the predominant structure of Al 2 O 3 (Fig. 2a). The other two samples with the crystalline structure of CuAlO 2 exhibited the PEC activity; however, their H 2 production values were less than 5% of the production with the DMSO/air sample, and their faradaic efficiencies were less than 20%.
The energetics of the as-synthesized CuAlO 2 (i.e., DMSO/air samples) were examined in detail. The UV-vis diffuse reectance absorption spectrum of the CuAlO 2 particles (collected from the lms) showed a broadband light absorption in the wavelength range of 400-900 nm, and the bandgap (E g ) was estimated to be $1.4 eV (corresponding to a l value of $885 nm; see Fig. 4a). This E g value was attributed to the indirectly allowed transition of CuAlO 2 (1.2-1.7 eV), 23,40,41 whereas the directly allowed transition usually leads to large E g values of 3-3.5 eV. 24,27 The indirect transition-induced photogeneration of charge carriers was conrmed by the IPCE proles (Fig. 4a). The IPCE value at l ¼ 400 nm was estimated to be $15% (E ¼ +0.3 V; see Fig. 1 for the photocurrent prole), which was not only far greater than the value reported in the literature ($0% at l > 400 nm) for CuAlO 2 synthesized via a sol-gel process 24 but also >2fold greater than those of other copper-based delafossites (e.g., CuFeO 2 at E ¼ +0.15 V) synthesized via the electrodeposition method. 29 In addition, the IPCE values decreased with increasing wavelength, with a wavelength onset of $610 nm. A Mott-Schottky analysis of the as-synthesized CuAlO 2 lm was performed at two frequencies (7 and 10 kHz) to estimate the at band potential (E  ) according to the following equation: 42 where C, 3 0 , 3 r , N D , E, k, and T refer to the space charge capacitance, permittivity of a vacuum, relative dielectric constant, donor density, applied potential, Boltzmann constant, and temperature, respectively. It should be noted that a decrease in 1/C 2 with increasing potential bias is a characteristic of p-type semiconductors (Fig. 4b).
The extrapolated x axis intercepts of the plots at the two frequencies coincided at E $1.25 V, which corresponded to E  . Assuming that the valence band (VB) level was lower than E  by $100 mV, 43 the conduction band (CB) of CuAlO 2 was determined to be À0.24 V, which is high enough to produce hydrogen from water (E ¼ 0 V) (Scheme 1a). However, even though the energetics appeared to be suitable for PEC H 2 production, the as-synthesized CuAlO 2 generated  a relatively small photocurrent density of $0.1 mA cm À2 at 0.3 V in the absence of a hole scavenger (Fig. S5 †). The addition of the hole scavenger enhanced the photocurrent density to $0.6 mA cm À2 and the faradaic efficiency to $100% (Fig. 3c), indicating that the hole transfer was limited. The hole transfer limit was straightforwardly examined by comparing the LSVs of the CuAlO 2 irradiated through the FTO substrate and electrolyte (Fig. S7 †). Compared to the substrate-side irradiation, the electrolyte-side irradiation led to signicantly reduced photocurrents, even though the light intensity arriving at CuAlO 2 in the substrate side-irradiation was $80% of the reference light (AM 1.5G; 100 mW cm À2 ) owing to the semi-transparent FTO (Fig. S8 †). 33 Assuming the light penetration depth is the same between the two irradiation directions, the photogenerated holes and electrons under the electrolyte and substrate-side irradiations, respectively, must travel further than their counter charge carriers. Therefore, the reduced photocurrent in the electrolyte-side irradiation reveals that the hole transfer is limited compared to the electron transfer.

Facilitating hole transfer
To facilitate the hole transfer in water without hole scavengers, a 150 nm-thick Au lm was overlaid onto FTO substrates via an electron-beam process, onto which CuAlO 2 was electrodeposited. The work function (W f ) of Au is 5.31-5.47 eV, 44 depending on the surface orientation, which can be estimated to be $0.3 V more negative than the CuAlO 2 VB level (Scheme 1b). Accordingly, the Fermi level (E F ) equilibration between the W f of Au and the E F of CuAlO 2 causes an enhanced upward band-bending, leading to efficient hole transfer while inhibiting the electron-hole charge recombination. A comparison between the LSVs of the CuAlO 2 samples deposited on bare FTO and FTO/Au (denoted Au/ CuAlO 2 ) showed that the E on of the latter was $0.1 V more negative than that of the former (Fig. 1) because of the upward shi in the E F of the latter (Scheme 1b). In addition, the presence of the Au underlayer signicantly inhibited the dark currents to E $0.4 V (Fig. 1), while enhancing the photocurrents (Fig. S9 †). The latter was further conrmed by the IPCE prole of Au/CuAlO 2 , particularly in the range of l > $450 nm (Fig. 4a). In contrast to the pristine sample, Au/CuAlO 2 exhibited an IPCE value of $10% at l ¼ 500 nm and IPCE onset at l ¼ $800 nm. These enhanced PEC properties could be attributed to the altered energetics produced by the Au underlayer. The Mott-Schottky plot of Au/ CuAlO 2 showed an E  of $1.1 V, which was approximately 0.15 V more negative than the E  of CuAlO 2 (Fig. 4b). Although slightly smaller than the estimated value of $0.3 V, this shi qualitatively explains the enhanced upward band-bending. Fig. 5a shows the PEC H 2 production values using Au/CuAlO 2 lms at 0.3 and 0.55 V in an aqueous KOH solution (1 M) without the sulde/sulte hole scavengers. The H 2 production rate at 0.3 V was $2.5 mmol h À1 , three-fold higher than the case of CuAlO 2 (Fig. 3b). The similar amounts of H 2 on the pristine CuAlO 2 at 0.3 V and Au/CuAlO 2 at 0.55 V (Fig. 3b vs. 5a) indicate that the deposition of the Au underlayer can save 0.25 V. Furthermore, the Au layer enhanced the photocurrent (Fig. S9 †), while the faradaic efficiencies were similar ($100%) in these cases (CuAlO 2 @0.3 V vs. Au/CuAlO 2 @0.55 V) (Fig. 5b).   Fig. S10-S12 † for more information on the emission spectrum.