Photoelectrochemical (PEC) water splitting is a promising strategy for the generation of solar hydrogen [1]. Although many photoactive materials have been extensively studied for PEC water splitting in the past decades [2][3][4], TiO2 is still undoubtedly the most widely used photoanode material due to its high natural abundance, low toxicity, excellent photostability, suitable band edge potentials, and high absorption coefficient for ultraviolet (UV) radiation [4][5][6]. However, the light harvesting capability of TiO2 is severely limited by its large bandgap [7]. Various approaches such as heteroatom doping [5], self-doping with Ti 3+ [8], and the introduction of surface disorder [6] have been implemented to improve the visible-light activity of TiO2. Among these strategies, heteroatom doping is currently the most commonly used method [5,[9][10][11].
TiO2 materials doped with non-metal elements such as nitrogen [5,10] and sulfur [11] have demonstrated visible-light responses. However, their visible-light photoactivities are still limited. In addition, the photocorrosion induced by the oxidation of the non-metal ion during water oxidation represents a major hurdle for their applications [10,11]. Recent theoretical calculations demonstrated that metal dopants (e.g., Cr, Fe, and Co) can extend the range of absorption to the visible region and improve the photostability of TiO2 [9,12]. Among these metal dopants, Cr was found to be effective in enhancing the visible-light absorption of TiO2, due to its suitable ion radius and the position of its energy states within the forbidden band [13].
However, to date, the reported visible-light photoactivity of Cr doped TiO2 is far lower than that of TiO2 doped with non-metals [10,11]. This can be attributed to two main factors. First, it is difficult to incorporate metal ions in the TiO2 lattice without aggregation using conventional chemical methods [14]. Second, metal doping may negatively affect the charge separation/transfer efficiency and, thus, the performance of the photoelectrode [2,[15][16][17]. Therefore, an effective doping approach resulting in a metal-doped TiO2 with favorable electronic properties and good visible-light photoactivities is highly desirable.
Ion implantation is a conventional material modification strategy that has been extensively used to modulate the electronic properties of semiconductors in the past decades. Compared with conventional chemical doping methods, ion implantation can incorporate virtually any metal-ion dopants of interest into the bulk lattice with controlled dose and depth. Additionally, the implanted metal ions are highly dispersed and isolated in the matrix material [18]. Herein, we employed an ion implantation method to dope Cr into TiO2 nanowire arrays for PEC water oxidation. Although Cr doping effectively narrows the bandgap and significantly enhances the visible-light absorption of TiO2, its visible-light photoactivity is still relatively low. Surprisingly, we found that post vacuum annealing played a decisive role in activating the visible-light photoactivity of Cr-doped TiO2. A vacuum-treated Cr-doped TiO2 (denoted as Cr-TiO2-vac) with a Cr dose of 3×10 16 cm -2 and subsequent vacuum annealing treatment at 550 °C for 3 h exhibited a visible-light (> 420 nm) current density of 0.53 mA/cm 2 at 1.8 V vs. reversible hydrogen electrode (RHE) without any other co-catalysts. A combination of experimental characterizations and theoretical calculations showed that the increased photoactivity can be attributable to the synergic effect of the Cr-dopant and oxygen vacancy that was created during vacuum annealing, improving both the visible-light absorption and electrical conductivity of TiO2. Furthermore, Cr-TiO2-vac exhibited impressive photostability. There was almost no photocurrent decay after testing for 100 h.
Rutile TiO2 NW arrays on FTO glass were synthesized using a well-developed hydrothermal method [19]. Briefly, 0.6 mL of titanium (IV) butoxide was added to 36 mL of an aqueous HCl solution (18 mL of deionized (DI) water + 18 mL of concentrated HCl (38%) with magnetic stirring). After stirring for 5 min, the solution was poured into a Teflon-lined stainless-steel autoclave (50 mL capacity). A piece of the FTO substrate (2.5×4 cm 2 ), which was cleaned for 60 min using an ultrasonicator in a mixed solution of DI water, acetone, and isopropanol (IPA), was immersed in the solution. The autoclave was sealed and heated to 150 °C for 8 h in an electric oven.
The obtained samples were washed with DI water and finally annealed in air at 550 °C for 3 h. Chromium ions were implanted into the as-prepared TiO2 NW arrays at 80 kV, with different ion dosages (1×10 15 -5×10 16 cm -2 ), using a metal vapor vacuum arc (MEVVA) ion source implanter. The implanted samples were then annealed in vacuum or in air at 550 °C for 3 h.
Photoelectrochemical (PEC) measurements were carried out in a conventional three-electrode system connected to a computer-controlled electrochemical workstation (CHI 650E) under a solar simulator (Abet technologies, SunLite TM solar simulator, Model: 11002) at an irradiation intensity of 100 mW/cm 2 with and without a visible light cutoff filter (λ > 420 nm). The PEC reactor contained a photoanode, a Pt foil, and an Ag/AgCl/KCl (sat) electrode as the working, counter, and reference electrodes, respectively. An aqueous solution of 1.0 M NaOH (pH = 13.6) was used as the electrolyte. IPCE spectra were obtained by using a xenon lamp (Newport model 66902) coupled with a grating monochromator (Newport Model 74125) and a power meter (Newport model 2936-R) with a photodiode (Newport model 71675_71580) in the wavelength range of 300-600 nm with an applied potential of 1.8 V (vs. RHE).
The measured potentials vs. Ag/AgCl were converted to a reversible hydrogen electrode (RHE) scale according to the Nernst equation:
where E o Ag/AgCl = 0.1976 V at 25 °C, EAg/AgCl is the experimentally measured potential against the Ag/AgCl reference electrode, and ERHE is the converted potential vs. RHE.
Field-emission scanning electron microscopy (FE-SEMs) was conducted using a JEOL S-4800 microscope. Transmission electron microscopy (TEM) and high-resolution transmission electron microscopy (HRTEM) were carried out on a Titan G2 60-300 Probe Cs Corrector high-resolution scanning transmission electron microscope (HRSTEM). X-ray diffraction patterns were collected on a Bruker AXS, D8 Advance X-ray powder diffractometer with Cu-Kα radiation (λ = 0.15418 nm).
Raman measurement was performed on a laser confocal microRaman spectrometer (RenishawinVia, Renishaw) with laser excitation at 532 nm. UV-vis diffuse reflectance spectra (DRS) were recorded on a PerkinElmer Lambda 750 S spectrometer equipped with an integrating sphere. X-ray photoelectron spectroscopy (XPS) experiments were conducted on a Thermo Scientific ESCALAB 250Xi system with Al-Kα (1486.6 eV) as the radiation source. Peak positions were internally referenced to the C 1s peak at 284.8 eV. X-ray absorption near-edge structure (XANES) spectra were recorded at the BL10B beamline in the National Synchrotron Radiation Laboratory (NSRL) in Hefei, China. Electron paramagnetic resonance (EPR) spectra were collected at room temperature using a JES-FA200 spectrometer.
Samples of TiO2, Cr-TiO2-air, and Cr-TiO2-vac nanowire powder were scraped from the growth substrates using a blade. We used the same amount of TiO2, Cr-TiO2-air, and Cr-TiO2-vac samples for EPR and XANES characterizations.
All calculations were performed with the CASTEP code as implemented in the Materials Studio package of Accelrys Inc. A high-level, reliable HSE06
TiO2 nanowire (NW) arrays were grown on FTO glass using a well-developed hydrothermal approach [19]. Cr ions were implanted into TiO2 by an ion implanter and followed by a thermal treatment under vacuum, as shown in the material processing scheme in Fig. 1(a). SEM images (Fig. 1b) show that the FTO glass substrate was uniformly covered with vertically aligned Cr-TiO2-vac NWs. Each NW has a rectangular cross section and a smooth surface facet. SEM images collected from freshly prepared TiO2 NWs and Cr implanted TiO2 NWs with air annealing treatment (Cr-TiO2-air) further demonstrated that these modification methods do not change the morphology of the TiO2 NW arrays (Fig. S1, Supporting Information). The crystallinity of the NWs was similar before and after Cr doping (Fig. S2, Supporting Information). Element mapping images show that Cr dopants are uniformly distributed over the TiO2 NWs (Fig. 1c). The Cr/Ti mole ratio was estimated to be 2.29 at% via a quantitative analysis of the EDX spectrum (Fig. S3, Supporting Information).
X-ray diffraction (XRD) patterns indicate that all the samples exist in the rutile phase [19] (Fig. 1d). After Cr ion implantation, the intensity of the TiO2 diffraction signals considerably decreased, suggesting that high-speed ion bombardment indeed creates defects in the TiO2 host crystal structure. Significantly, the diffraction peak intensities of the rutile phase were recovered upon post-implantation annealing. The important role of heat treatment in repairing the lattice defects formed by ion bombardment was also confirmed by Raman spectroscopy. Four characteristic Raman-active modes of TiO2 with frequencies at 142, 234, 446, and 609 cm -1 were observed (Fig. S4, Supporting Information), corresponding to the B1g, multi-photon, Eg, and A1g modes of the rutile space group (P42/mnm), respectively [20]. Similarly, the intensity of these vibrational peaks decreased after ion implantation, while the signals were recovered by the post-implantation annealing treatment. It is also noteworthy that a Raman-forbidden mode at approximately 690 cm -1 (forbidden A2g transition) [21] emerged for Cr-doped TiO2, suggesting the occurrence of structural changes after Cr doping. High-resolution transmission electron microscopy (HRTEM) and fast Fourier transform (FFT) studies confirmed that the Cr-implanted TiO2 NWs maintained good crystallinity (Fig. 1e). nm, consistent with the observation for rutile TiO2 [7]. However, the Cr-TiO2, Cr-TiO2-air, and Cr-TiO2-vac samples exhibited a significant absorption from 400 to 600 nm, which could be ascribed to the localized Cr-dopant energy levels within the bandgap. Notably, the absorption edge of the Cr-TiO2-vac sample showed a ∼20 nm shift to the longer wavelength region as compared to the Cr-TiO2-air sample. These results clearly indicate that both Cr ion implantation and subsequent annealing treatment can change the optical and electronic structure of TiO2.
X-ray photoemission spectroscopy (XPS) measurements were performed to probe the surface chemical states of Cr-implanted TiO2. As shown in the XPS survey spectra (Fig. S5a Cr-TiO2-vac, respectively (Fig. 2b). The negative shift in binding energy also suggests that the Ti 3+ species was possibly present on the surface of Cr-TiO2-vac [10,15]. In addition to the surface nature of the Cr-TiO2-vac NWs, we also aimed to understand how the Cr ion implantation and post-implantation annealing affect the bulk structure of TiO2. EPR was employed to characterize the chemical states in bulk TiO2 [23]. The strong EPR signal obtained from Cr-TiO2-vac with a g-value of 2.001
indicates the existence of oxygen vacancies [23], while only a trace amount of oxygen vacancies was observed for pristine TiO2 and Cr-TiO2-air (Fig. 2c). We believe the abundant oxygen vacancies present in Cr-TiO2-vac results in the increase in electron density around the metal atoms, causing the negative shift in binding energy observed for Ti and Cr in the XPS measurements.
X-ray absorption near-edge structure (XANES) spectra in the soft X-ray ranges were also collected to study the electronic structures of the Cr-doped TiO2 samples (Fig. 2d-f). were assigned to the transfer to the t2g and eg orbitals, respectively [25,26]. Compared with pristine TiO2, the L3-edge eg orbital of the Cr-implanted TiO2 samples (Cr-TiO2-air and Cr-TiO2-vac) slightly shifted to a lower energy. The split energy between the t2g and eg orbitals reflects the symmetry in the first coordination shell of the Ti atoms, indicating that the implanted Cr ions caused a slight lattice distortion [26]. In the XANES O K-edge spectra (Fig. 2f), the two relatively sharp features (530-535 eV) corresponding to the electronic transitions from the O 1s orbital to the t2g and eg orbitals, were observed. The higher energy signal from 535 to 547 eV (peaks C, D and E) originates from the transitions from the O 1s orbital to the hybridized orbitals induced by the antibonding Ti 4sp and O 2p orbitals [23]. We also found that the t2g orbital of Cr-TiO2-vac exhibited an obvious decrease in the absorption intensity and slightly shifted to a higher energy, as compared to pristine TiO2 and Cr-TiO2-air, indicating the existence of oxygen vacancies on its surface [23,25], again in good agreement with the XPS results.
Density functional theory (DFT) calculations were performed to understand the separate effects of the Cr dopants and the oxygen vacancies (Ovac) as well as the possible interaction between the Cr dopant and Ovac on the electronic and optical properties of TiO2. Fig. 3(a-d) respectively shows the structures of pristine TiO2, Cr-doped TiO2 (Cr-TiO2), oxygen-deficient TiO2 (Ovac-TiO2), and Cr-doped and oxygen-deficient TiO2 (Cr-Ovac-TiO2) with the main charge values and the electron density difference. The charge distribution on Ti and O decreased after Cr doping (Cr-TiO2) as the Cr-O interaction is weaker than the Ti-O interaction (Fig. 3a,b).
Additionally, the state of oxygen vacancies in Ovac-TiO2 (Fig. 3c) mainly spreads to the adjacent 5-coordinated Ti, which is attributed to the electron distribution on the eg orbital of Ti pointing towards the oxygen vacancy site. It is noteworthy that the excess electrons generated by the oxygen vacancy only partially transfer to the adjacent atoms and still retain some charges at the oxygen vacancy sites (Fig. 3c). For Cr-Ovac-TiO2, the charge distribution on Cr significantly decreased from 1.54 to 1.18, suggesting that the excess electrons on the oxygen vacancies changed the oxidation state of Cr (Fig. 3d). These results support the experimental observations of the negative shift in binding energy and lower absorption threshold in the XPS and XANES studies of the Cr-TiO2-vac samples, respectively. The projected densities of the states of pristine TiO2, Cr-TiO2, Ovac-TiO2, and Cr-Ovac-TiO2 were also obtained and displayed in Fig. 3(e). The bandgap of pristine TiO2 was calculated to be 3.5 eV.
By introducing Cr dopants into TiO2, one impurity state below the conduction band minimum in the forbidden band was generated, which was mainly contributed by the Cr 3d orbital. This impurity state contributes to the observed visible-light absorption.
However, the Fermi level of Cr-TiO2 is still located at the valence band maximum (VBM), as compared to pristine TiO2. This indicates that only the Cr 3d impurity state is localized in TiO2, which does not act as an electron donor. For TiO2 doped with oxygen vacancies, the Fermi level shifted up to the produced impurity state by the two excess electrons donated by the oxygen vacancy, which is consistent with the electron density difference results. The transfer of these two excess electrons from the oxygen vacancy to the Cr 3d state (Fig. 3d) resulted in a downward shift of the Cr 3d impurities states in TiO2 and a further narrowing of the bandgap (Fig. 3e). The bandgap shift observed for Cr-doped TiO2 and Cr-TiO2-vac is consistent with the absorption properties shown in Fig. 1(f). Moreover, the excess electrons transferred from the vacancy resulted in an upward shift of the Fermi level to the Cr 3d state. As a result, the degree of band bending increased at the surface of TiO2, thus facilitating charge separation [10,15]. These results revealed the synergistic effect between the oxygen vacancies and the Cr dopants on the modification of the electronic and optical properties of TiO2.
PEC performances of the TiO2 photoanodes were investigated under simulated solar illumination of 100 mW/cm 2 , in a three-electrode configuration with the TiO2 sample as the working electrode, Pt foil as the counter electrode and Ag/AgCl/KCl (sat) as the reference electrode, in a 1.0 M NaOH solution (pH = 13.6). Fig. 4(a) shows the results of chopped linear-sweep voltammetry (LSV) for the Cr-TiO2-vac sample. Cr-TiO2-vac achieved a photocurrent density of 1.29 mA/cm 2 at 1.8 V vs.
RHE under simulated solar light illumination, which is three times higher than that of the pristine TiO2 NWs (0.41 mA/cm 2 ) (Fig. 4a, inset) obtained at the same potential.
Most importantly, the photocurrent density of Cr-TiO2-vac reached 0.53 mA/cm 2 under visible-light (> 420 nm) illumination, which accounts for 41.1% of the total photocurrent density generated under simulated solar illumination. For pristine TiO2, the photocurrent contributed from visible light is only about 0.5% at 1.8 V vs. RHE.
Fig. 4(b) shows the I-t curves of TiO2, Cr-TiO2-air, and Cr-TiO2-vac collected under chopped visible-light illumination at 1.8 V (vs. RHE). Among them, Cr-TiO2-vac has the highest photocurrent density, which is 2 orders of magnitude higher than that of the pristine TiO2 NWs. Cr-TiO2-air surprisingly only showed a slightly higher visible-light photoactivity relative to pristine TiO2, although it absorbs visible light, suggesting that the localized Cr 3d impurity states are inefficient in separating and transferring the visible-light induced carriers, in the absence of oxygen vacancies.
Furthermore, we also evaluated the PEC performance of a series of TiO2 samples at various Cr implantation dosages and vacuum annealing temperatures (Fig. 4c and Fig. S6, Supporting Information). The Cr-TiO2-vac photoanodes with an implantation ion fluence of 3×10 16 cm -2 and followed by a vacuum annealing temperature of 550 °C showed the maximum photocurrent density. Further increases in the Cr ion dose or vacuum annealing temperature caused a reduction in photoactivity. Numerous recombination centers created by excessive Cr doping can be introduced into TiO2.
Additionally, annealing temperatures beyond 550 °C can severely damage the electrical conductivity and morphology of the FTO substrate. Moreover, a pristine TiO2 NW sample annealed in vacuum under the same conditions (TiO2-vac) was prepared and tested as a control sample. The I-t curves of TiO2-vac are given in Fig. S7, Supporting Information. The sample exhibited a comparable photocurrent density as that of the pristine TiO2 sample. These results further demonstrated that the increased visible-light photoactivity of Cr-TiO2-vac is associated with the synergic effect of oxygen vacancies and Cr doping.
To understand the wavelength dependent photoactivity, we performed IPCE measurements (Fig. 4d). The IPCE values were calculated according to the equation IPCE = (1240 × I)/(λ × Jlight), in which I is the measured photocurrent density, λ represents the wavelength of the incident light, and Jlight is the measured irradiance density. The maximum IPCE of pristine TiO2 was 27.5% at 390 nm with the response wavelength at approximately 420 nm, corresponding with the reported results for rutile TiO2 [7]. Significantly, Cr-TiO2-vac exhibited a substantially enhanced UV-light IPCE and remarkable visible-light photoactivity, as compared to pristine TiO2.
Cr-TiO2-vac achieved an IPCE of ∼6.8% at 450 nm, representing one of the highest values among all reported metal-ion doped TiO2 based photoanodes without co-catalysts. The IPCE profile of the Cr-TiO2-vac sample also fits well with its diffuse reflectance spectrum, confirming that the active photoactivity is indeed extended to ∼540 nm. On the other hand, we observed a steep decrease in photoactivity in the entire UV-light region for the Cr-TiO2-air sample, suggesting that Cr doping alone can negatively affect the PEC performance of TiO2. The Cr-TiO2-air sample displayed negligible IPCE values in the visible-light region, again suggesting that oxygen vacancies are essential for the increased visible-light photoactivity of Cr-TiO2-vac.
While the PEC performance depends on light absorption, charge separation and transport play decisive roles in the photoelectrochemical process [10,27]. To understand how the oxygen vacancies and Cr doping affect the charge separation and transport efficiency, electrochemical impedance studies were performed. Positive slopes of the Mott-Schottky curves from TiO2, Cr-TiO2, and Cr-TiO2-vac indicate all these samples are n-type semiconductors, as expected (Fig. 5a). Importantly, a smaller slope observed for Cr-TiO2-vac, as compared to pristine TiO2 and Cr-TiO2-air, suggests a higher donor density in the Cr-TiO2-vac samples. The donor density can be obtained from the Mott-Schottky plots according to the equation Nd =
(2/e0εε0)[d(1/C 2 )/dV] -1 , in which e0, ε, ε0, Nd, C, and V respectively represent the electron charge, dielectric constant (ε=170) [15], permittivity of vacuum, donor density, space charge capacitance, and electrode potential. A carrier density of 1.95×10 19 cm -3 , more than 2 orders of magnitudes higher than pristine TiO2
(1.53×10 17 cm -3 ), was achieved in the Cr-TiO2-vac samples. Since the Mott-Schottky equation was derived from the flat electrode model, it is expected to produce errors in determining the absolute value of the carrier density for nanostructured electrodes.
Nevertheless, a qualitative comparison of the Mott-Schottky slopes (donor densities) of the different samples is reasonable because there was no obvious change in the NW morphology after both Cr ion implantation and the annealing treatment. The increased donor density is mainly attributed to the formation of oxygen vacancies, which serves as a shallow donor in TiO2, consistent with DFT calculations, as shown in Fig. 3(e) and previous reports [28]. Charge transfer properties of these samples were further tested by electrochemical impedance spectroscopy (EIS) measurements. Fig. 5(b) shows that the Nyquist curves of all the samples exhibited single capacitive arcs. By fitting the data to a typical Randles circuit (upper inset, Fig. 5b), Cr-TiO2-vac was found to exhibit the smallest value of charge transfer resistance (RCT) (Table 1). These results indicate that vacuum treatment significantly decreased the charge transfer resistance at the interface between the photoelectrode and electrolyte. By contrast, the Cr-TiO2-air samples exhibited strong visible-light absorption a higher carrier density than that of pristine TiO2, although they have almost equal charge transfer resistances (Fig. 5b). The results showed that the effect of Cr dopants alone on charge transfer is minimal, while the presence of oxygen vacancies on the TiO2 surface is critical. A recent study reported that surface defects such as oxygen vacancies can lead to the formation of surface hydroxyl species, which can further mediate the transfer of photoinduced holes from the metal oxide to water [29]. Photostability has been a major concern for element-doped TiO2 materials.
Therefore, we performed a long-term stability test for Cr-TiO2-vac under visible-light illumination with a fixed potential of 1.8 V vs. RHE (Fig. 5c). Fig. S8 (Supporting Information) shows a digital picture of the PEC cell for the test. Gas evolution was observed at both the photoanode and Pt cathode. Significantly, Cr-TiO2-vac did not show obvious photocurrent decays and morphology changes (Fig. S9, Supporting Information) under continuous light illumination for 100 h, performing much better than N-implanted TiO2 which was reported to lose 60% of its initial photocurrent after 1 week [10].
In summary, we demonstrated that Cr-ion implantation followed by a vacuum annealing process can significantly improve the visible-light photoactivity of TiO2 photoanodes. The Cr-TiO2-vac sample without the addition of any co-catalysts yielded a 0.53 mA/cm 2 visible-light photocurrent density and an IPCE of 6.8% at 450 nm at 1.8 V vs. RHE, which is one of the best values reported for metal ion-doped