Exploiting defects in TiO2 inverse opal for enhanced photoelectrochemical water splitting

In this work, we report on defects generation in TiO2 inverse opal (IO) nanostructures by electrochemical reduction in order to increase photocatalytic activity and improve photoelectrochemical (PEC) water splitting performance. Macroporous structures, such as inverse opals, have attracted a lot of attention for energy-related applications because of their large surface area, interconnected pores, and ability to enhance light-matter interaction. Photocurrent density of electrochemically reduced TiO2-IO increased by almost 4 times, compared to pristine TiO2-IO photoelectrodes. Raman spectroscopy and X-ray photoelectron spectroscopy (XPS) analyses confirm the presence of oxygen vacancies in electrochemically reduced TiO2-IO photoelectrodes. Oxygen vacancies extend the absorption of TiO2 from the UV to visible region. The incident photon-to-current efficiency (IPCE) increased by almost 3 times in the absorption (UV) region of TiO2 and slightly in the visible region. Impedance studies show improved electrical conductivity, longer photogenerated electron lifetime, and a negative shift of the flatband potential, which are attributed to oxygen vacancies acting as electron donors. The Fermi level shifts to be closer to the conduction band edge of TiO2-IO. © 2019 Optical Society of America under the terms of the OSA Open Access Publishing Agreement

ethanol for 1 hour before being dipped in a mixed suspension of Thermo Scientific carboxylate modified polystyrene particles of 530 nm in diameter and deionized water in the proportion of 0.15 wt%. FTO substrates dipped in the mixture were placed in an oven for approximately 22 hours, allowing lateral capillary forces to deposit PS particles on the substrate as water evaporated. The infiltration of sacrificial PS template with TiO 2 was carried by atomic layer deposition (ALD), using titanium tetrachloride and deionized water as sources for Ti and O respectively in a Picosun R200 ALD system at 85°C. Annealing the infiltrated opal templates at 450° C for 30 mins in air converted them from amorphous phase to crystalline phase and remove the PS sacrificial template, leaving behind anatase TiO 2 -IO. Samco ICP-RIE (RIE-4100iP) was used to etch the top layer of TiO 2 -IO to allow electrolyte to infiltrate deep into the porous structure.
Electrochemical reduction of TiO 2 inverse opals was carried by an anodic reduction process using a two-electrode configuration and a potentiostat (Autolab, PGSTAT302N) setup, with TiO 2 -IO as the working electrode and Pt wire as the counter electrode. With the two photoelectrodes placed approximately 2.5 cm apart, electrochemical reduction was carried at a voltage of − 3V from 300 to 600 seconds in an electrolyte bath of 0.5 M Na 2 SO 4 . After the anodic reduction, the samples were rinsed in deionized water and dried with nitrogen.
The surface morphology and structure were characterized by scanning electron microscopy (SEM, FEI Helios Nanolab 600) operating at a voltage of 3 kV. Phase identification of TiO 2 was carried out by grazing incidence X-ray diffraction (GIXRD, PANanalytical), using Cu-Kα radiation (λ = 1.54 angstrom) with a voltage of 45 kV from 20° to 80°. Raman spectroscopy (Renishaw) using a 532 nm laser with Raman shift from 100 to 800 cm −1 , was carried out to confirm the defects generated from the electrochemical reduction process. X-ray photoelectron spectroscopy (XPS) using monochromated Al K alpha x-ray source (energy 1486.68 eV) and ultra-violet photoelectron spectroscopy (UPS) using with a He(I) beam (21.1 eV) was performed with Thermo Scientific ESCALAD 250Xi. Diffuse reflectance and transmittance were measured using a Perkin Elmer UV-Vis spectrophotometer equipped with an integrating sphere.
PEC performance, electrochemical impedance spectroscopy (EIS) and Mott-Schottky analyses of the photoanodes were evaluated in a three-electrode PEC cell configuration using a potentiostat (Autolab, PGSTAT302N). All potentials were reported against the reversible hydrogen electrode (RHE) using Nernst's equation E (V RHE ) = E (V Ag/AgCl ) + 0.059 pH + 0.197. Ag/AgCl in saturated KCl and a Pt wire were used as reference and counter electrodes, respectively in 1 M NaOH (pH = 13) as the electrolyte. Photoresponse characteristics were measured using a solar simulator (Newport) equipped with a 150 W Xe lamp and an AM 1.5G filter, calibrated with a standard Si solar cell. Transient photocurrent measurements were performed under chopped light irradiation (on-off light cycles of 50 s) at an applied bias of 1.23 V RHE in 1 M NaOH. Photocurrent stability measurement was performed at an applied bias of 1.23 V RHE in 1 M NaOH for a duration of 3.5 hours. Open circuit voltages for all samples were measured before EIS was performed from 10 −2 to 10 5 Hz at 10 mV amplitude. Mott-Schottky analyses were performed at a frequency of 500 Hz and an applied voltage of 10 mV. ZView from Scribner Associates was used to analyse EIS data. The incident-photon-to-current conversion efficiency (IPCE) measurements were taken in a three-electrode configuration as a function of wavelength from 300 to 600 nm using a 1000 W Xe lamp coupled to a monochromator. The incident light intensity was calibrated using a Thorlabs PM100D power meter and a standard silicon photodiode power sensor of wavelength range 200 -1100 nm. IPCE was calculated according to the following equation: where 1240 V.nm is the multiplication of Planck's constant and the speed of light, j photo is the measured photocurrent density in mA/cm 2 , P mono is the calibrated and monochromated illumination power intensity in mW/cm 2 and λ (nm) is the wavelength at which this illumination power is measured. Figure 1(a) shows the highly ordered PS sacrificial template consisting of a face-centered cubic array structure with PS spheres stacked in an orderly manner. By using a stop-flow ALD method, we can effectively infiltrate the PS sacrificial template to achieve uniform and high filling fraction of the macroporous structure with an average pore size of approximately 500 ± 10 nm as seen in Figs. 1(b) and 1(c). The magnified cross-section of TiO 2 -IO in Fig.  1(d) shows a well-connected macroporous structure with unclogged and well-defined pore openings. This would allow infiltration of the electrolyte deep into the structure ensuring shorter pathways for photogenerated carriers. Photogenerated holes can easily reach reactive sites to perform the oxidation process, while photogenerated electrons can travel quickly to the FTO ohmic contact, through the cell and to the counter electrode to perform the reduction process. Recombination of electron-hole pairs is reduced since photogenerated carriers can access their respective sites quickly. After undergoing electrochemical reduction and stability test for 3.5 hours, the TiO 2 -IO maintains its structure and the pores remained unclogged, as shown in Figs. 1(e) and 1(f). The crystalline structure of both pristine and ER TiO 2 -IO was characterized by GIXRD as shown in Fig. 2 (215) respectively. There is no change in the phase after the electrochemical reduction process and after 3.5 hours of photoelectrochemical stability test. The Raman spectrum as seen in Fig. 2(b) for pristine TiO 2 -IO shows six active modes with three E g modes at 144.1 cm −1 , 196.8 cm −1 and 637.2 cm −1 ; two B 1g modes at 395 cm −1 and 516.5 cm −1 ; and one A 1g mode at 516.5 cm −1 . These modes are typical of anatase TiO 2 [22]. The active mode E g is caused by symmetric stretching vibration, B 1g by the symmetric bending and A 1g by the antisymmetric bending of O-Ti-O in TiO 2 [23]. Both the E g (1) and E g (2) modes at 144.1 cm −1 and 196.8 cm −1 from ER400s TiO 2 -IO are blue-shifted, while the rest of the Raman modes remain unchanged. These two peak shifts have been attributed to lattice disorder resulting from a phonon confinement effect or a stoichiometric change [24]. From the XRD pattern and using Scherrer's equation, the crystallite size of both pristine and reduced TiO 2 was estimated to be 22 nm. Phonon confinement effect can be ruled out since there was no change in grain size of the pristine and ER400s TiO 2 -IO; and only the E g modes were shifted. Removing an oxygen atom from the TiO 2 lattice forces the three Ti atoms to shorten their overlapping dangling bonds with each other and strengthen their bonds with the rest of the lattice, thus decreasing the Ti-O bond length [25]. This reduction in the bond length is reflected in the shift of E g modes to a higher wave number, since the E g modes are mainly caused by the symmetric stretching of the Ti-O bond and is sensitive to change in bond length. All the peaks of the ER400s broadened, as compared to the pristine sample, with full width half maximum (FWHM), shown in Table 1. Peak broadening can be attributed to the excess oxygen vacancies leading to non-stoichiometry of TiO 2 [26]. Thus, the peak shift, broadening of the peaks and decrease in intensity of surface structure of TiO 2 are due to the defects, in this case oxygen vacancies generated from the reduction process [27].  XPS analysis was carried out to determine the chemical environment and type of surface defects induced by the reduction process. For the XPS core-level spectra of Ti 2p in Fig. 3(a), the peaks at binding energy (BE) 459 eV for pristine and 458.5 eV for ER TiO 2 -IO have been assigned to Ti 4+ species. No peaks for Ti 3+ species have been observed hence suggesting there was no reduction of Ti 4+ to Ti 3+ [28,29]. The spectra of O 1s in Fig. 3(b) has one main peak at BE 530.2 eV (pristine) and 529.8 eV (ER TiO 2 -IO) characteristic of Ti-O. The two smaller shoulder peaks are assigned to the surface oxygen vacancies, with BE 531.6 eV (pristine) and 531.1 eV (ER) TiO 2 -IO; and the surface OH group with BE 532.6 eV (pristine) and 532.2 eV (ER) TiO 2 -IO as shown in Fig. 3(c). With the reduction process, there is a decrease in atomic percentages in Ti 4+ and Ti-O peaks while an increase in the peaks is attributed to oxygen vacancies and the OH group, shown in the inset of Fig. 3(c). A shift to lower binding energies of both O 1s and Ti 2p spectra of ER-TiO 2 is observed across all the peaks. A lack of Ti 3+ in ER TiO 2 -IO with a shift to lower BE in the spectra indicates that electrons bound to the oxygen and titanium ions migrated to the oxygen vacancies; with the latter serving electrons to th UPS meas work function did not chang indicated in F where hυ is th electron cut-o values for po intensity due t For an n semiconducto ohmic contac they are illum regardless of direction of i fact that phot as electron tra he neighbourin surements wer n of TiO 2 -IOs b ge the valence Fig. 3(d) photoelectrodes confirms that the charge recombination is significant in pristine TiO 2 -IO. This is addressed by introducing oxygen vacancies in TiO 2 -IO. Oxygen vacancies not only act as electron donors thus lowering the Fermi level towards the conduction band, they also serve as electron traps to prevent charge recombination. Since the depletion layer width is dependent on the donor density, oxygen vacancies present in the TiO 2 lattice can affect the depletion layer width [34]. With increased oxygen vacancies generated by the electrochemical reduction process, the depletion layer width is shortened, recombination is reduced, thus improving charge carrier migration to their respective sites. However, increasing the reduction time from 400 s to 500 s leads to a decrease in photocurrent density, which can be attributed to the excessive generation of oxygen vacancies. With excess oxygen vacancies, more electrons are trapped thus lowering density of electrons collected at the FTO ohmic contact.  The photocurrents for both pristine and ER400s-IO photoelectrodes reflect the measured values from LSV, and is maintained for over a period of 600 s. The transient features indicate whether recombination is limiting the charge transfer process and the type of recombination that occurs at the SEI [35]. An anodic transient would correspond to holes accumulated, while a cathodic transient corresponds to electrons being dissipated at the SEI. For the ER400s TiO 2 -IO photoelectrode, an anodic photocurrent spike is observed due to the sudden generation of charge carriers, which then decays to a steady-state photocurrent due to recombination processes or a decrease in band bending [36]. For pristine TiO 2 -IO photoelectrode, photogenerated holes travel to the surface at a slow but sufficient rate to react with the electrolyte, thereby avoiding accumulation. By creating oxygen vacancies, electrons are trapped in the mid-gap states as well as absorption sites where hole transfer to the absorbed ionic species can prevent further recombination [37]. Long-term stability of ER400s TiO 2 -IO photoelectrode was subjected to continuous AM1.5 simulated sunlight illumination for 3.5 hours. Under an applied potential of 1.23 V RHE , the initial photocurrent density of 0.35 mA/cm 2 decreased by 14% to 0.29 mA/cm 2 , as shown in Fig. 5(b). To quantify light absorption and photoconversion efficiencies of the TiO 2 -IO photoelectrodes, we measure the incident photon-to-current efficiency (IPCE) where the photocurrents are obtained as a function of wavelength. Figure 5(c) shows the IPCE increasing almost 3 times in the absorption region of TiO 2 after electrochemical reduction, as well as increase slightly in the visible region. The increase in photoconversion efficiency of TiO 2 -IO in the UV as well as in the visible regions can be attributed to the enhanced photocurrent density in ER400s TiO 2 -IO photoelectrodes. The UV-Vis spectra in Fig. 5(d) shows the absorption band edges for both TiO 2 -IO photoelectrodes at the wavelength of 400 nm, which is at the absorption bandgap of anatase TiO 2 . The ER400s TiO 2 -IO photoelectrode exhibits a strong sub-bandgap absorption from 400 nm to the infrared region due to an extended Urbach tail, which has been attributed to lattice disorder or vacancies [38]. EIS was used to evaluate charge transfer kinetics at the SEI by measuring the AC response of pristine and ER400s TiO 2 -IO photoelectrodes. Nyquist and Bode plots of pristine and ER400s TiO 2 -IO photoelectrodes obtained from EIS was carried out at open-circuit potential, under dark and illuminated conditions. The internal resistances corresponding to the overall charge transfer process under illumination were significantly lower than those in dark conditions due to the generation of electron-hole pairs. In Nyquist plots, the real and negative imaginary components of the complex impedance Z are plotted on the horizontal and vertical axis, respectively. The presence of one semicircle indicates that an electrochemical interface is present in the system. In Fig. 6(a), all TiO 2 -IO photoelectrodes under illumination have lower impedances than those performed in the dark indicating that photogenerated electron-hole pairs lower the resistance of the photoelectrodes. The ER400s TiO 2 -IO photoelectrode has a lower impedance compared to the pristine electrode which shows that oxygen vacancies improved charge transfer kinetics at the SEI. A Randles circuit as shown in the inset of Fig. 6(b), consists of a series resistance (R s ) element followed by a parallel RC, which is a combination of parallel resistance (R p ) and a constant phase element (CPE), and is used for modeling the response of the photoelectrodes. The series resistance represents the ohmic resistance between FTO and TiO 2 -IO, while the parallel resistance is associated with charge transfer from TiO 2 -IO to the electrolyte, and the CPE is the impedance originating from the charge in the space charge layer in TiO 2 -IO. The CPE is a fitting element with impedance of Z CPE = 1 / Q(jω) α , and when α equals 1, 0, or − 1 a CPE behaves as a capacitor, a resistor, or an inductor. By fitting the Nyquist plots using the equivalent Randles circuit, we can ascertain the appropriate behavior of system to AC response. Through the creation of oxygen vacancies, Ti-O bonds are broken, producing excess valence electrons that are bonded to oxygen atoms in the valence band, and in migrating from a lower to higher energy level, conductivity of the electrode is improved [39]. From the Bode magnitude plots shown in Fig. 6(c), the impedance measured corresponds to R s and R p at low frequencies, while at higher frequencies, we see a drop in the overall impedance indicating that the charge transfer resistance at the interface has dropped since the resistance between the ohmic contact and TiO 2 -IO did not change. From the frequency peaks in the Bode phase plots in Fig. 6(d), lifetime of photogenerated electrons can be calculated using, τ n = 1/ (2πƒ max ), where ƒ max is the maximum oscillation frequency of the impedance. Photogenerated electron lifetimes for pristine and ER400s TiO 2 -IO photoelectrodes were calculated to be 75 and 234 ms respectively. By generating oxygen vacancies, electron lifetimes have been extended leading to the increase of electron diffusion length and inhibition of electron-hole recombination.

Results and discussion
where C is the capacitance at the space charge layer, ε is the dielectric constant (anatase TiO 2 = 45), ε 0 is the permittivity in free space, A is the area, q is the elementary charge, N D is the donor carrier concentration, E app is the applied electrochemical potential, E fb is the flatband potential, k B is the Boltzmann constant, and T is the temperature (25°C). The x-intercept of the Mott-Schottky plot provides us the flatband potential and the slope can be used to determine the donor density. The flatband potential pristine and ER400s TiO 2 -IO photoelectrodes are determined to be 0 eV (vs RHE) and − 0.56 eV (vs RHE) respectively. The flatband potential, E fb , is the electrochemical potential applied to the semiconductor that reduces this band bending to zero, thus the cathodic shift of the flatband potential can lower the onset potential. The gentler slope of ER400s TiO 2 -IO indicates higher donor concentration compared to that of pristine TiO 2 -IO photoelectrode. The introduction of oxygen vacancies which act as electron donors, shifts the Fermi level towards the conduction band and improves conductivity, as shown in impedance studies.

Conclusions
Defects in TiO 2 can be generated easily and quickly by electrochemical reduction process. Electrochemically reduced TiO 2 macroporous nanostructures such as TiO 2 -IO photoelectrodes show an improvement in photon absorption and PEC water splitting performance. The defects generated in these TiO 2 -IO photoelectrodes are oxygen vacancies. The performance of the electrochemically reduced photoelectrodes is dependent on the reduction times, with 400 s being the optimal reduction time. Longer reduction time produces excessive oxygen vacancies that act as electron traps, thereby limiting the density of electrons collected at the FTO ohmic contact. There is an increase in the photon absorption of the ER400s TiO 2 -IO photoelectrode in the visible region and photoconversion efficiency in the UV region increasing by almost 3 folds. This improved performance ostensibly results from the creation of oxygen vacancies in the TiO 2 -IO photoelectrodes due to shift in Fermi level towards the conduction band, which in turn leads to lower electron-hole recombination rates and improved charge transport. However, in the experiment, the valence band did not change and the reduction of the bandgap of TiO 2 -IO was not achieved. This work has shown that by introducing oxygen vacancies, we can increase photogenerated carrier lifetime and thus improve the conductivity of TiO 2 to enhance PEC water splitting performance.

Funding
Australian Research Council (ARC).