Water splitting using photoelectrochemical (PEC) cells with metal oxides attracts a significant research interest, owing to the potentials to overcome energy and environmental issues [1]. Over the past four decades, various oxides have been investigated, including TiO₂ [2], ZnO [3, 4], Fe₂O₃ [5, 6], BiVO₄ [7, 8], WO₃ [9, 10], etc. [11-13], whose band structures are suitable with respect to the water redox potentials. However, the photoefficiency is still very low for photoanodes based on a bare material, mainly owing to the narrow absorption range in the visible spectrum and high probability of electron-charge recombination [14]. Therefore, the development of efficient methods to overcome the above challenges is required for the improvement of the photoefficiency for water splitting.

Several strategies were proposed to enhance the overall photoefficiency [13], including an introduction of foreign elements and formation of heterojunction structure [15]. As a typical example, a sub-stoichiometric TiO₂ was extensively studied for an enhanced performance in photocatalysis. This structure can be formed by annealing pristine TiO₂ under an oxygen-deficient/-reducing environment; it exhibits black-color characteristics [16]. In addition, it exhibits significantly improved optical properties, with a narrower band gap for an enhanced visible-light absorption. Furthermore, oxygen vacancies can form, which increase the charge carrier density. The abundant charge carriers could increase the efficiency of charge separation by enhancing the band bending at the interface between TiO₂ and electrolyte [5]. However, this method is time-consuming and complex [17]. Recently, a rapid electrochemical method was developed to facilitate the formation of a sub-stoichiometric metal oxide photoelectrode [18]. This method was widely employed, as it is simple and cost-effective [19-22]. It was also employed in other metal oxides; for example, reduced WO₃, ZnO, and BiVO₄ were successfully obtained in different environments to enhance the PEC water splitting efficiency [20]. In the electrochemical approach, the deposition of elements is driven by the formed electrostatic field between the anode and cathode. Therefore, for a determined electrostatic field strength, the deposition of elements can be easily controlled by the duration of the applied potential. In the transient state, before the completion of the formation of black TiO₂, Ti(Ⅲ) oxide is induced at the bottom, near the substrate, while the Ti(Ⅳ) oxide is preserved at the top, far away from the substrate, which exhibit distinct electronic and optical properties, and form a heterostructure.

In general, three types of heterojunction structure are employed in the design of a photoanode, as illustrated in Fig. 1. For an efficient use of the electrons and charges in the redox reaction, the type Ⅱ band alignment is the most suitable alignment, as it provides an efficient extraction of the photo-generated electrons and charges for the interfacial reactions of hydrogen and oxygen evolutions. It reduces the possibility for recombinations, and improves the overall water splitting efficiency. TiO₂/BiVO₄ [23], ZnO/BiVO₄ [24], and TiO₂/Fe₂O₃ [25], exhibited significantly improved PEC water splitting efficiencies. However, in these systems, at least two materials are needed for the formation of the heterojunction; therefore, the synthesis normally required sophisticated procedures with limited candidates of suitable materials, and the performances were not easily reproducible.

In this study, a type-Ⅱ-heterojunction TiO₂ nanorod arrays (NAs) are obtained by a combination of reduced and pristine TiO₂, through a conventional electrochemical reduction. The obtained TiO₂ heterojunction structure is characterized by electrochemical and spectroscopic measurements. It exhibited an increased photoefficiency by more than three times and by approximately 20%, with respect to those of the pristine TiO₂ NAs and completely reduced black TiO₂, respectively. Two factors contribute to the enhanced water splitting efficiency. For the partially reduced TiO₂, the light absorption is significantly increased, from the ultraviolet (UV) to the visible spectrum. Moreover, the type Ⅱ heterostructure provides efficient charge separation and transport, which leads to an improved PEC water splitting efficiency [26]. This highly controllable approach can be easily employed for various semiconductors for the control of the band structure [27], with applications in energy harvesting, conversion, and storage [28-30], as well as sensor and biology applications [31, 32].

TiO₂ NAs were synthesized through hydrothermal methods. First, 0.1 mol/L TiCl₃ was dissolved in a solution that consists of a mixture of 13 ml hydrochloric acid and 15 ml water. The as-prepared solution was transferred into a 50-ml teflon-line stainless-steel autoclave, and several clean fluorine-doped tin oxide (FTO) glasses were submerged into the solution. The sealed autoclave was heated in an oven at 150 ℃ for 5 h, and cooled down to room temperature. TiO₂ NAs were homogeneously deposited on the FTO glass, and the samples were thoroughly washed with deionized (DI) water. The as-prepared TiO₂ NAs were further annealed at 550 ℃ for 1 h to obtain a ratio between anatase and rutile of approximately 3:1. The electrochemical reduction was performed using a VSP-300 electrochemical station (Bio-Logic), with a typical three-electrode system. TiO₂ NAs, Pt foil, and Ag/AgCl were used as the anode, cathode, and reference electrode, respectively. The TiO₂ NAs underwent an electrochemical reduction under a negative potential (–0.5 V vs reversible hydrogen electrode (RHE)) in 0.05 M H₂SO₄ solution for different time periods. The reduced electrode was dried at ambient conditions for 1 h.

The morphology and elemental composition of the photoanode were studied using field-emission scanning electron microscopy (FE-SEM, Jsm-6700F). Ultraviolet-visible diffuse reflectance spectroscopy (UV-VIS DRS, Varian Cary 500 Scan UV-VIS system) was employed to study the optical properties of the photoanode. The crystallinity of the samples was investigated using X-ray powder diffraction (XRD, Bruker D8 Advance with Cu K_α1 radiation, λ = 1.5406 Å). In addition, electrochemical impedance spectroscopy (EIS) was employed, and Mott-Schottky and Nyquist plots were obtained.

The as-prepared TiO₂ NAs photoanode was used in the same electrochemical station with a typical three-electrode system. The edge of the modified TiO₂ electrode was sealed using insulating epoxy resin, except for an active surface area of 1×1 cm²; it was functionalized as a photoanode, while a Pt foil and Ag/AgCl were used as the cathode and reference electrode, respectively. Liner sweeps and electrochemical impedance spectra were obtained in 1 mol/L NaOH (pH = 13.6) using an AM 1.5 solar simulator (Newport 94023A). The incident-photon conversion efficiency (IPCE) was measured using the same electrochemical setup with an IPCE station (Newport, QEPV-B), coupled with an infrared filter and aligned monochromator. Cyclic voltammetry (CV) scans of the TiO₂ electrodes were performed in a solution of tetrabutylammonium hexafluorophosphate (TBAPF6) in anhydrous acetonitrile, using TiO₂, Pt, and Ag/AgNO₃ as the anode, cathode, and reference electrode, respectively.

The carrier density is estimated using the Mott–Schottky relation [33]:

(1)

where C_sc² is the capacitance of the space-charge region, e is the elementary charge of the electron (1.602×10^–19 C), ε₀ is the permittivity of vacuum (ε₀ = 8.854×10^–12 F/m), ε is the dielectric constant of TiO₂ (ε ~ 55) [34], A is the electrochemically active surface area (1 cm² in the employed system), V is the applied voltage, V_fb is the flat-band potential, and N_d is the donor density. The IPCE is obtained using the quantum-based equation:

(2)

where λ and P_input are the wavelength and power (mW/cm²) of the incident light (Xenon lamp), respectively, I is the pho-tocurrent density; the IPCE is a relative value expressed in percentages. The band gaps of the samples are obtained through Tauc plots, using the relation:

(3)

where A₀ is a constant that depends on the electron and hole mobilities of the material, h denotes the Planckʼs constant, v is the frequency of the incident photon, and hv is the light inten-sity; n can be either ½ or 2, for an indirect-or direct-band-gap semiconductor, respectively; TiO₂ was considered to exhibit a direct band gap.

The electrochemical treatment is an efficient approach for a precise control of the deposition of elements onto the electrode [35]. As illustrated in Fig. 2(a), when an electrostatic field is applied between the TiO₂ NAs anode and Pt cathode with a proton-abundant environment, the protons are driven towards the TiO₂ electrode owing to the negative charge. As the electrostatic force at the bottom side of the NA is larger than that at the top side, the reduction gradually occurs from the bottom to the top of the NAs, as illustrated in Fig. 2(b). The pristine, heterojunction, and black TiO₂ are denoted as N0, N1, and N2, respectively. N3 is a state that corresponds to an excessive reduction, when the TiO₂ NAs are peeled off. The dependence of the current as a function of the time (I-t plot) is shown in Fig. 2(e). The negative current gradually decreases, reaching a steady state. The corresponding images of the samples are shown as insets in Fig. 2(e). The behavior of the curve is highly dependent on the morphology and thickness of the film. In this study, the thickness of the TiO₂ film is approximately 2 μm, while the average diameter of the rods is approximately 150 nm (Fig. 2(c) and (d)). In the reduction curve, the decrease indicates the reduced resistance of the anode owing to the insertion of protons, which implies that the black TiO₂ exhibits a larger number of charge/donor carriers than the white TiO₂, owing to the formation of TiO_3–x(OH)_x [35]. Therefore, the conductivity of the black TiO₂ is significantly improved, compared with that of the pristine white TiO₂. The electron transport and migration at the bottom part of the TiO₂ NAs are increased, with respect to those of the pristine TiO₂. The reduction cycle is complete for approximately 52 min, longer than in previous studies, as the values of the proton concentration (0.05 mol/L H₂SO₄) and applied bias (–0.5 V) are relatively low [20, 36], in order to achieve a controlled electrochemical process. In contrast, the pristine FTO does not exhibit reduction under the employed conditions, as shown in Fig. S1. Fig. 2(c) and (d) shows a top and cross-sectional views of N2, respectively. There are no considerable differences in morphology compared with N0 (Fig. S2). A bottom view of the peeled film at the state N3 is also obtained (Fig. S3), and it shows that the rectangular shape of the rods is well preserved. The aggregates observed around them could be attributed to the irregular initial nucleation owing to the hydrothermal approach. This phenomenon indicates that the peeling of the film is mainly caused by the proton attack at the fragile section between the rods and FTO substrate.

The optical properties of the samples are studied using UV DRS, as shown in Fig. 3(a). The corresponding band gaps are calculated using the Tauc plots (Fig. S4). The pristine TiO₂ NAs exhibit absorption up to 394 nm, which confirms that the band gap is approximately 3.15 eV. The absorption of N1 is extended up to 431 nm; therefore, its band gap is narrower, with a value of approximately 2.88 eV. At complete reduction, the band gap of N2 is approximately 2.85 eV, corresponding to a wavelength of approximately 435 nm, very similar to those of N1. The bandgap is mainly determined by the dominant absorption peak. The black color is mainly attributed to the band bending owing to the reduction [37]. Both valence and conduction bands are tailed from the original positions, which decreases the band gap [16]. The absorption wavelength extends up to 700 nm, as shown in Fig. 3(a). However, the absorption of N2 in the UV spectrum is slightly lower than those of N0 and N1.

The crystalline structures of the samples are studied using powder XRD, as shown in Fig. 3(b). The as-prepared NAs are annealed at 550 ℃ for 1 h to form a combined TiO₂ phase of anatase (JCPDS No. 21-1272) and rutile (JCPDS No. 21-01276). The relatively high peak at approximately 37° originates from both anatase and FTO substrate (SnO₂: JCPDS No. 46-1088). Two small peaks at 20.78° and 22.87° can be observed for N2, which can be attributed to the formation of reduced TiO₂, such as Ti₄O₇ and Ti₆O₁₁ [26]. At these positions, in the XRD spectrum of N1, only a relatively high noise level can be observed.

The band edges of the samples are obtained by electrochemical measurements using CV scans, as shown in Fig. 4; the experiments are conducted with TBAPF6 in anhydrous acetonitrile. As the reduced TiO₂ exhibits an activity for a reaction with TBAPF6 in the organic environment, the scan was performed in the range of –1.5 to 1.5 eV, in order to avoid a redox reaction with Ti(Ⅲ) in the organic environment. The pristine TiO₂ is scanned from –2 to 2 eV, and exhibits a high chemical stability. Fig. 4(a) reveals that the conduction band edges of the pristine and completely reduced TiO₂ are located at –3.74 and –4.14 eV, respectively, with respect to the vacuum level. These values correspond to approximately 0.1 and –0.3 eV vs. RHE, at pH = 7, which is consistent with the values reported in the literature [35, 38]. The CV scan of N1 is performed (Fig. 4(b)) in a smaller scan range of 0 to –1.5 eV, in order to minimize the effects of trivial redox reactions. Peaks are observed at –3.74 and –4.14 eV, which indicate the formation of the TiO₂ heterojunction. Three addition peaks can be observed, which emerge most likely owing to the high reactivity between the semi-reduced TiO₂ (interfacial region between the reduced and pristine TiO₂) and TBAPF6. Both phases of TiO₂ and reduced black phase are very stable, as shown in Fig. 4(a). However, in the crystal structure of the dual-phase TiO₂, the interfacial layer could be a transient state from TiO₂ to black TiO₂. In addition, this layer is likely to be highly unstable, and cannot be maintained owing to the highly reactive TBAPF6; this could be attributed to the appearance of the three additional peaks in Fig. 4(b).

The PEC performances of the samples are investigated, and the results are shown in Fig. 5. The light-off current density is negligible for N1, which indicates that the TiO₂ NAs heterojunction is inactive under ambient conditions, as the other samples. The pristine TiO₂ exhibits a photocurrent density of approximately 0.8 mA/cm² at 1.23 V (vs. RHE) under the irradiation of AM 1.5, using 1 mol/L NaOH solution as an electrolyte. The photocurrent density of N1 (2.5 mA/cm²) is higher than that of N2 (2.1 mA/cm²). The stability of the photoanode is similar to the reported performance with a moderate value [16]. N1 preserves approximately 70% of its efficiency after 35 min, as shown in Fig. S5.

Furthermore, EIS is performed to characterize the interfacial reaction between the electrolyte and semiconductor surface. Nyquist plots of the three samples at 1.23 V (vs. RHE) with and without light irradiation are obtained, as shown in Fig. 5(c) and (d), respectively. The curve with a semicircle shape reveals the charge transfer between the solid semiconductor and liquid electrolyte, as the diameter of the semicircle is equal to the charge transfer resistance of the sample [39]. Among the samples, N1 exhibits the largest diameter of the semicircle under both light-on and -off conditions. It is worth noting that without the light irradiation, the semicircle's diameter of N1 is slightly larger than that of N2, which implies that the small portion of the unreduced TiO₂ acts as a resistance for the interfacial charge transfer. In contrast, when the photoanodes are irradiated under the AM 1.5 irradiation, the semicircle's diameter of N1 is slightly smaller than that of N2, owing to the enhanced charge transport and separation by the formed heterojunction structure. The IPCE is also investigated to provide insights into the relation between the wavelength of the incident photons and capability for a catalytic redox reaction; the details of the calculations are presented in the experimental section. The obtained performances, shown in Fig. 5(b), are consistent with those reported in a previous study [40]. N0 exhibited a low absorption in the UV range and relatively low conversion efficiency, compared with those of the other two samples. N1 and N2 exhibited very similar absorptions in the visible range. In the UV spectrum, at wavelengths in the range of approximately 330–360 nm, N1 exhibited a higher conversion efficiency than that of N2, which confirms that the white TiO₂ portion at the top of the rods enhances the absorption of UV light, leading to an enhanced PEC water splitting performance.

The band structure of N1 is illustrated in Fig. 6(a). The conduction band edge of the reduced TiO₂ is located at 0.4 eV (vs. RHE, pH = 0), while that of the pristine TiO₂ is located at 0.1 eV. The flat-band potentials of the samples are obtained from the EIS measurements through the Mott-Schottky plots at a frequency of 1 Hz (Fig. S6). The flat-band potentials of N0 and N2 are approximately 0.13 and 0.75 eV (vs. RHE). N1 exhibits almost the same curve as that of N2, which indicates that the TiO₂ heterojunction exhibits electrochemical properties as those of the reduced TiO₂ without irradiation. This demonstrates the formation of the type Ⅱ heterojunction structure, which increases the PEC performance for water splitting. In addition, the donor densities of the pristine and reduced TiO₂ can be estimated from the results; it significantly increased from ~10¹⁵ to ~10¹⁷ [34] upon the reduction of TiO₂. Therefore, in this case, as shown in Fig. 6(b), high-energy charges and electrons emerge in the pristine TiO₂ upon the absorption of the incident UV light. The reduced TiO₂ could effectively harvest visible light. In addition to the formation of the type Ⅱ heterojunction structure, the highly concentrated donor carriers facilitate the migration of the electrons, and reduce the probability for recombination. The high-energy electrons and charges injected from the pristine portion further facilitate the photocatalytic reaction towards the desired direction, promoting the oxygen and hydrogen evolutions.

Heterojunction-structured TiO₂ NAs were synthesized through a simple electrochemical approach. The obtained sample exhibited an increased PEC performance, with respect to those of pristine and reduced TiO₂, owing to a synergistic effect of light absorption, from the UV to the visible spectrum, and formation of the type Ⅱ heterojunction structure for an effective electron/charge separation and transport. This simple and highly controllable approach can be employed for various semiconductors, and exhibits large potentials for energy, environmental, sensor, and biology applications [41-47].