Photoelectrochemical Water Oxidation Using Cobalt Phosphate‐Modified Nitrogen‐Doped Titania Nanotube Arrays

The synthesis of cobalt phosphate (CoPi)‐modified, nitrogen‐doped TiO2 nanotube arrays (N‐TNAs) for photoelectrochemical (PEC) oxygen evolution under visible light is reported. Because of the nitrogen doping, the N‐TNAs exhibit enhanced visible‐light activity toward the PEC water oxidation reaction. However, the performance is diminished as a result of self‐oxidation by photogenerated holes in the valence band of the N‐TNAs. Compared with the N‐TNAs, the CoPi/N‐TNAs prepared under optimal conditions show a twofold improvement in photocurrent generation and are also sensitive to light with wavelengths as long as 580 nm. Stable oxygen evolution with a Faradaic efficiency approaching unity is demonstrated using the CoPi/N‐TNA photoanodes under simulated sunlight for at least 2 h of operation.


Introduction
TiO 2 has attracted intensive interest over the past several decades because of its special optoelectronic and physiochemical properties. [1] One of the most studied applications of TiO 2 is solar energy conversion based on semiconductor photocatalysis. [2,3] TiO 2 is a typical polymorphic compound, and it exhibits different properties depending on its crystal structure. However, the bandgap of TiO 2 is too wide for it to efficiently absorb visible light (λ > 400 nm). Because the spectral irradiance of sunlight consists mainly of visible light with 4-5% UV light, the use of unmodified TiO 2 inevitably leads to low photocatalytic performance.
To take advantage of the remaining spectrum of sunlight, researchers have doped foreign elements into the lattice of TiO 2 and modified its surface with visiblelight-absorbing substances (e.g., metal complexes, organic dyes, and plasmonic metals). [2,3] The doping strategy is based on the idea that the dopants form donor or acceptor levels in the bandgap of TiO 2 , enabling the doped elements to function as visible-light-absorption centers. For the purpose of photoelectrochemical (PEC) water oxidation using an n-type semiconductor, making the flat-band potential of the semiconductor more negative than the water reduction potential (0 V versus reversible hydrogen electrode (RHE)) is desirable. If the flat-band potential of an n-type semiconductor is more positive than the water oxidation potential, an additional bias is theoretically required for overall water splitting. In this regard, doping of anions into TiO 2 is advantageous over doping of metal cations because the doped anion species can form new energy levels above the original valence band of TiO 2 formed by O 2p orbitals (Scheme 1a). [4] In addition to bandgap reduction, another important subject in solar-driven semiconductor photo(electro)catalysis is increasing the active surface area for redox reactions. [5] To this end, nanostructuring a light-absorbing semiconductor is a straightforward approach. TiO 2 nanotube arrays (TNAs) with a unique 1D porous nanoarchitecture are attractive, especially for use in PEC water oxidation. [6] TNAs can be readily synthesized via electrical anodization of Ti foil in a fluoride-containing aqueous solution at acidic pH. [7] The tube lengths and pore diameters can be tuned by changing the electrolyte or temperature, or by controlling the applied voltage/current, whereas the array size can be extended almost without limit. [8][9][10][11] Thus far, doped TiO 2 nanostructures such as nanotubes and nanowires with improved PEC performance under visible light have been reported. [12][13][14][15] These reports did not, however, clarify the origin of the anodic photocurrent because the O 2 gas, which is supposed to be formed via the visible-light photoelectrolysis of water, was not quantified. In photo(electro)catalysis using semiconductor materials (especially those made from organic precursors), great care should be taken by considering the possibility of side oxidation of the organic residuals instead of the desired reactions. [16,17] Depositing Co species onto a semiconductor anode can improve its activity and stability toward PEC water oxidation. [18][19][20][21] In particular, for semiconductors that have anions less electronegative than oxygen (e.g., mixed-anion compounds and non-oxides), photogenerated holes in the semiconductor can not only oxidize water into O 2 but also oxidize the material itself. [16,22,23] Therefore, improving the "selectivity" of photogenerated holes toward water oxidation is critical for these semiconductors, and a cocatalyst such as Co will be the key component.
Recently, modification of a wide-bandgap semiconductor powder such as TiO 2 with Co species has been reported to result in a new absorption band in the visible-light region, which originates from electron transitions from the loaded Co species to the semiconductor (Scheme 1b). [24][25][26] The new visible-light absorption enables the material to be used for photocatalytic water oxidation, where the loaded Co species function as "catalytic sensitizers". Using this concept, researchers achieved PEC water splitting by a Co-modified TiO 2 thin film under visible light (>440 nm). [27] Although this "catalytic Co photosensitizer" strategy might be another means of activating TiO 2 toward solar energy conversion, the visible-light-induced photocurrent was not stable and the Faradaic efficiency for H 2 O-to-O 2 conversion was low (12-14%). In this case, self-oxidation of the loaded Co species was suggested.
Given this background, we here report improved PEC oxidation of water to O 2 under visible light using N-doped TNAs further modified with a cobalt phosphate cocatalyst (CoPi/N-TNAs). The CoPi/N-TNA electrode generated a stable photocurrent under visible-light irradiation, with an O 2 -evolution Faradaic efficiency close to unity.

Preparation of Nitrogen-Doped TNAs and Modification with CoPi
The TNAs were synthesized by a method reported previously, with some modifications. Details are provided in the Experimental section. As shown in Figure S1, Supporting Information, the TNAs contained anatase as the main phase, with a small fraction of rutile, except for the Ti substrate. The N-doping treatment used to obtain the TNAs did not cause a change in the crystal phase. Figure 1 shows the top-view SEM images of TNA samples. The as-synthesized TNAs had a perpendicular, orderly tubular structure, with a tube diameter of %100 nm and a wall thickness of 10 nm. N-doping of the TNAs did not substantially alter their morphology. The incorporation of N species into the TNAs was confirmed by X-ray photoelectron spectroscopy (XPS) ( Figure S2, Supporting Information). The N 1 s XPS spectrum of N-TNA contained at least three peaks located at 396.4, 399.3, and 401.8 eV, respectively. These peaks could be assigned to the lattice nitrogen resulting from O 2À /N 3À substitution upon nitridation, the interstitial nitrogen in the TiO 2 lattice, and molecularly chemisorbed nitrogen species on the TiO 2 surface, respectively,  www.advancedsciencenews.com www.small-structures.com consistent with a report by Sun et al. on N-doped TNAs. [28] The incorporation of N species into the lattice of TNAs is also supported by a negative binding energy shift in the Ti 2p and O 1s XPS spectra, which results from the difference in electronegativity between nitrogen and oxygen. [29,30] Upon photo-assisted electrodeposition of CoPi, numerous particles with a diameter of %20 nm were observed on the TNAs. The uniform deposition of CoPi onto the N-TNAs was also confirmed by energy-dispersive X-ray spectroscopy (EDS) mapping images ( Figure S3, Supporting Information). XPS measurements revealed that the CoPi/N-TNA sample gave distinct Co 2p 1/2 and 2p 3/2 peaks at around 797.5 and 781.8 eV, with satellite peaks appearing at 803.4 and 785.7 eV, respectively ( Figure S2, Supporting Information). The Co 2p 1/2 and 2p 3/2 XPS peak positions are consistent with those reported in CoPi loaded on a Cu 2 O/BiVO 4 film by a photo-assisted electro-deposition method. [31] A clear photoelectron signal from phosphorous species was also observed, and can be deconvoluted into P 2p 1/2 and 2p 3/2 peaks at 134.4 and 133.5 eV, respectively. They are very close to the binding energies of phosphorous species in cobalt phosphate. [32] To examine the optical properties of the TNA samples, we measured the reflectance of the electrode samples in the ultraviolet-visible (UV-vis) region. As shown in Figure S4, Supporting Information, an abrupt increase in reflectance was observed at 350-400 nm for the TNAs, which is attributable to a band-to-band electron transition in TiO 2 . [33] This "bandgap absorption" for the N-TNAs was red-shifted relative to that for the TNAs, presumably because of the incorporation of N into the lattice of TiO 2 , which can reduce the bandgap of TiO 2 . [3,34,35] Modification of the N-TNAs with CoPi did not substantially alter the reflection spectrum of the N-TNAs. As mentioned in the Introduction, Co species loaded onto a semiconductor substrate can produce a new absorption (or reflection) as a result of electronic transitions from the loaded Co species to the semiconductor. [24][25][26] In the present case, the electronic interaction between the loaded CoPi and N-TNAs might not be sufficiently strong to substantially change the reflection spectra. Alternatively, the strong interference arising from the metallic Ti substrate in longer wavelength regions might conceal the change in the optical properties as a result of the hybridization of the N-TNAs with CoPi.

Electrochemical Properties
Electrochemical impedance spectroscopy (EIS) measurements for the TNA, N-TNA, and CoPi/N-TNA electrodes were conducted to examine the effect of N-doping and CoPi deposition on the charge-transfer resistance under dark conditions. The EIS results ( Figure 2) show that, among the three electrode materials, the TNAs gave the largest Nyquist plot arc (i.e., charge transfer resistance of R ct ¼ 2.3 kΩ cm À2 ). N-doping of the TNAs substantially reduced the charge-transfer resistance to R ct ¼ 87 Ω cm À2 . Further reduction of the resistance was achieved upon CoPi loading (R ct ¼ 50 Ω cm À2 ). The results show that the CoPi/N-TNAs possess a more favorable structure for prompt charge transport. Figure S5, Supporting Information, shows Mott-Schottky plots for the TNAs and N-TNAs measured in an aqueous K 3 PO 4 solution. The positive slopes of the plots indicate that these samples are n-type semiconductors. The flat-band potentials were determined from the x-axis intercept of the plots to be %À1.2 V versus Ag/AgCl (-0.23 V versus RHE). The identical values of the flat-band potentials for the TNAs and N-TNAs are reasonable given that the N-doping mainly affects the potential of the valence band (not the conduction band). [3,23] In addition, the slope of the Mott-Schottky plot for the N-TNAs is smaller than that for the TNAs, which suggests that the carrier density was higher in the N-TNAs than in the TNAs because the slope of the plot is inversely related to the carrier density. The higher carrier density in the N-TNAs can account for the lower charge-transfer resistance of the material, as revealed in the Nyquist plots ( Figure 2).

PEC Performance
As shown in Figure 3a, the unmodified TNAs showed a small photoresponse upon irradiation with visible light (λ ¼ 460 nm), presumably because of photoexcitation related to defect states in the material. [36] An improvement of the anodic photocurrent was observed for the N-TNAs, which outperformed the TNAs by a factor of 2. Deposition of CoPi onto the N-TNAs further enhanced the photoresponse. In particular, an anodic photocurrent was observed at the lower potential regions, indicating a clear catalytic effect of the CoPi. The performance of the CoPi/N-TNA electrode was found to depend on the deposition time for Co. As shown in Figure S6, Supporting Information, 5 min was the optimal deposition time. A longer deposition time (10 min) gave a similar result, but a dark current appeared at %þ0.5 V versus Ag/AgCl, suggesting excess deposition of CoPi onto the N-TNAs. Figure 3b shows the action spectra for the same electrodes. The incident-photon-to-current efficiency (IPCE) for the TNA electrode decreased with the increasing wavelength of incident   Figure 4a shows that the CoPi/N-TNA anode exhibited a very stable photocurrent for 2 h under visible-light irradiation. However, the anodic photocurrent from the N-TNA electrode degraded quickly, most likely because of the self-oxidation of the nitride component in the material by photogenerated holes. The optimized photoanode was stable in the operation of more than 7 h, as shown in Figure S7, Supporting Information.
From the viewpoint of constructing an energy conversion scheme in which the standard Gibbs energy change is positive, it is important to carefully check the oxidation product(s) obtained by PEC water oxidation using a semiconductor electrode. [16] In the present work, the oxidation product from the illuminated CoPi/N-TNAs was found to be O 2 gas, which was obtained with a Faradaic efficiency of unity (Figure 4b).
The amount of O 2 evolved at the initial stage of the photoelectrolysis (blue plots, Figure 4b) was slightly lower than one-fourth that of electrons that flowed (red dashed line). This is most likely due to the time lag of gas diffusion from the liquid phase to the gas phase that is connected to GC, as reported previously. [21] From these results, we concluded that the CoPi/N-TNAs are a stable photoanode material for oxidizing water to O 2 by responding to longer-wavelength visible light.

Conclusion
This study demonstrated that CoPi/N-TNAs are a suitable semiconductor photoanode material for absorbing visible light and driving the PEC oxidation of water to O 2 gas. The N-doped TiO 2 nanotubes can absorb visible light because of their reduced bandgap. The CoPi cocatalyst prevented self-oxidation of the nitride component of the N-TNAs and improved the stability of the TiO 2 photocatalyst. CoPi deposited onto the surface also reduced the charge-transfer resistance and improved the charge-transfer kinetics during the PEC oxidation of water.  The mixed effects led to an improvement in the PEC performance in terms of photocurrent, O 2 evolution, and IPCE.

Experimental Section
Materials Preparation: TNAs were prepared via a modified anodization method using Ti foils in a two-electrode system. [6] A Ti sheet (0.127 mm thick, 99.7% purity, Sigma-Aldrich) was cut into small pieces with equal dimensions (1.5 Â 3.0 cm 2 ), and the pieces were ultrasonically cleaned in ethanol for 10 min in a sonication bath and rinsed with deionized water. Several Ti foil pieces (0.127 mm thick, 99.7%, Sigma-Aldrich) and a stainless-steel plate were immersed in a mixed aqueous solution of 0.28 M NaF (98%, Junsei) and 1.0 M H 3 PO 4 (85%, Wako). A DC voltage of þ20 V was then applied to the Ti foil (anode) against the stainless steel (cathode) for 4 h with magnetic stirring. The Ti foil piece was then annealed at 773 K for 6 h in air to obtain crystallized TNAs. To obtain N-doped TNAs, the as-prepared TNAs were heated at 773 K for 3 h under an NH 3 gas flow (50 mL min À1 ).
For the deposition of Co phosphate, a three-electrode system was applied using N-TNAs as the working electrode material, a saturated calomel electrode (SCE) as the reference electrode, and Pt wire as the counter electrode. A bias of þ0.90 V versus SCE was applied in a 0.1 M sodium sulfate aqueous solution containing 100 μM cobalt nitrate and 100 μM potassium phosphate.
Characterization: X-ray diffraction (XRD, Rigaku MiniFlex600) and XPS (Shimadzu, ESCA-3400) were used to examine the crystalline patterns and binding states of sample elements, respectively. The binding energies of XPS spectra were corrected with respect to C 1 s (285.0 eV). Field-emission scanning electron microscopy (FE-SEM, Hitachi SU8230) was also used to analyze the morphologies of the samples. UV-vis diffuse-reflectance spectra were collected using a spectrophotometer (Shimadzu UV-2150) equipped with an integrating sphere attachment (ISR-2200).
PEC Tests: The PEC behavior of the samples was examined using a typical three-electrode configuration (Ag/AgCl (saturated KCl): reference electrode, Pt wire: counter electrode), and a potentiostat. The scan rate in linear sweep voltammetry experiments was 10 mV s À1 . The electrochemical potential versus Ag/AgCl can be converted to that versus RHE according to the following equation E RHE ¼ E AgCl þ 0.0591 pH þ E°A gCl ðE°A gCl ¼ 0.1976 V at 298KÞ All three electrodes were immersed in a single PEC cell containing aqueous solutions of potassium phosphate (K 3 PO 4 ) under an Ar atmosphere. A certain fraction of the sample surface (1 cm 2 ) was exposed to the electrolyte and irradiated. A 300 W xenon lamp (Asahi Spectra, USA) was used as a light source. For visible light, a bandpass filter (420 and 460 nm) was inserted between the light source and the electrochemical cell.
The IPCE was determined in aqueous K 3 PO 4 solutions using the same three-electrode system with a Xe lamp and was calculated using the following equation IPCE ð%Þ ¼ ð1240 Â J ph Þ Â 100=ðP light Â λÞ where J ph (mA cm À2 ), P light (mW cm À2 ), and λ (nm) represent the photocurrent density at þ0.10 V versus Ag/AgCl, the photon flux, and the wavelength, respectively.
For O 2 evolution tests, Ar was purged through the solutions and the sample electrodes were maintained at a constant potential of þ0.20 V versus Ag/AgCl in a fully sealed H-type reactor. After the irradiation process, the evolved O 2 in the headspace was quantified using a gas chromatograph (MGC3000A, Inficon) equipped with an MS-5A column.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.