Constructing highly efficient semiconductor photoelectrodes for desirable photoelectrochemical (PEC) conversion systems, such as PEC water-splitting cells and dye (or quantum dots) sensitized solar cells, has attracted tremendous attention owing to their great potential for using the abundance of solar energy ^{[1, 2, 3, 4]}. The mesoporous structure, with the advantages of a short charge carrier diffusion length and large surface area, has been actively pursued in the development of photoelectrodes ^{[5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19]}. A typical example is the mesoporous polycrystalline films of TiO₂ nanoparticles, which show the improved performance as photoanodes in both PEC water-splitting cells and sensitized solar cells ^{[14, 15, 16, 17, 18, 19, 20]}. However, the boundaries between TiO₂ nanoparticles in the films, where photo- generated charge carriers suffer from strong scattering and recombination processes induced by lattice distortions, may impair the efficiency of collecting charge carriers ^[21]. Mesoporous single crystals of anatase TiO₂ with a large percentage of (001) facets synthesized by a seeded-silica-template hydrothermal route have been demonstrated to exhibit an improved electric conductivity. The electrodes of these crystals used in nonporous-state dye-sensitized solar cells correspondingly showed an improvement in collecting charge carriers ^[22]. Inspired by this pioneering work, several studies have synthesized mesoporous TiO₂ single crystals with different morphologies as photocatalysts for water splitting or photoanodes for PEC water splitting ^{[23, 24, 25]}. Although charge carrier transport within each particle of these photoanodes is favored, the transport is still greatly limited by the boundaries between the mesoporous single-crystal particles in the photoanode film.

To completely remove the particle boundaries in an electrode film, all single-crystal particles need to directly grow on the substrate and penetrate the whole film. One-dimensional single-crystal rod arrays of rutile TiO₂ have emerged as a promising structure in improving the carrier collection efficiency of electrode films ^{[26, 27, 28]}. However, the number of surface-reactive sites of these rod array films is usually much less than that of mesoporous nanoparticle films owing to their relatively small specific surface area ^[26], which limits the performance of PEC cells. The integration of high carrier collection efficiency and abundant reactive sites in electrodes remains a desirable challenge. Some attempts to realize this purpose have been made to increase the specific surface area of rod arrays. One simple method is tuning the length and diameter of rods ^[29]. Further attempts have included growing TiO₂ nanodendrite arrays through two hydrothermal steps ^[30] and decorating the nanodendrite arrays with nanoparticles to further increase the specific surface area ^[31]. Besides these attempts, a promising alternative solution is to develop mesoporous single-crystal rod array films, which provide both a large specific surface area and high carrier collection efficiency. Here, we report a template-assisted hydrothermal method for the synthesis of mesoporous single-crystal rod array films of rutile TiO₂. To the best of our knowledge, this is the first report of mesoporous single-crystal rod array films on conductive substrates. This method could be generalized for the fabrication of other metal oxide films with the features of mesoporous structure and single crystals. The obtained film as the photoanode in a PEC water-splitting cell has an improved performance of 2.6 times that of the reference nonporous rod array film without mesopores.

H₂O (32 mL), ammonium hydroxide (18 mL), and tetraethyl orthosilicate (TEOS, Sigma Aldrich, 99 mL) were added to 750 mL ethanol to form the mixture. The mixture was stirred at 700 rpm for 24 h and then centrifuged at 5000 rpm for 5 h to collect SiO₂ spheres with a diameter of around 50 nm. The recovered spheres were heated at 500 °C in air for 30 min (ramping time 250 min) to remove surface-adsorbed organic reagents.

Titanium n-butoxide (1 mL) was added to 60 mL of an aqueous solution of hydrochloric acid (6 mol/L) in a 100-mL glass beaker to form a clear solution. The SiO₂ spheres powder (600 mg), prepared as above, was added to the clean solution. The mixture solution was sonicated for 1 h to form the solution with well-dispersed SiO₂ spheres. Such solution (40 mL) together with fluorine-doped tin oxide (FTO) glass substrates was transferred to a Teflon container with a volume of 80 mL, where the FTO substrates were placed at the bottom with their conductive surface facing upwards. The SiO₂ spheres were coated on the FTO substrates by centrifuging the Teflon container at 5000 rpm for 1 h. The Teflon container was then filled in a stainless steel autoclave and heated at 150 °C in an oven for 12 h. After heating, the autoclave was naturally cooled to room temperature. A white SiO₂ sphere/TiO₂ nanorod film was uniformly coated on the FTO glass substrate. The sample was thoroughly washed with deionized water and dried in air. Prior to removing SiO₂ spheres from the film by immersing the sample in 2 mol/L NaOH solution at 80 °C for 2 h, the sample was heated at 550 °C in air for 2 h to improve both the crystallinity of TiO₂ rods and their adherence to the substrate.

Titanium n-butoxide (1 mL) was added to 60 mL of an aqueous solution of hydrochloric acid (6 mol/L) in a 100-mL glass beaker to form a clear solution. The clear solution (40 mL) together with FTO glass substrates was transferred to a Teflon-lined stainless steel autoclave with a volume of 80 mL. The sealed autoclave was heated at 150 °C in an oven for 12 h, and then naturally cooled to room temperature. A white film of TiO₂ rods was uniformly coated on the FTO glass substrate. The sample was thoroughly washed with deionized water and dried in air. The sample was heated at 550 °C in air for 2 h to improve both the crystallinity of TiO₂ nanorods and their adherence to the substrate.

X-ray diffraction (XRD) patterns of the samples were recorded on a Rigaku diffractometer using Cu K_α irradiation. Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) images were obtained on a Nova NanoSEM 430 and JEOL 2100, respectively. Diffuse reflectance absorption spectra were recorded using a UV-visible spectrophotometer (JASCO-550). The chemical states were analyzed using X-ray photoelectron spectroscopy (XPS, Thermo Escalab 250, a monochromatic Al K_α X-ray source). Binding energy was referenced to the C 1s peak (284.6 eV) from adventitious carbon.

The photoelectrochemical water splitting was carried out in a three-electrode system, where the rod array film, Ag/AgCl electrode, and Pt foil act as working electrode, reference electrode, and counter electrode, respectively, in the electrolyte of aqueous solution (1 mol/L NaOH) under AM 1.5G illumination with light density of 100 mW/cm² (Newport). According to the Nernst equation (E_RHE = E_Ag/AgCl + 0.059pH + 0.196), the measured potentials versus the RHE scale can be obtained from the potentials versus Ag/AgCl.

The nonporous single-crystal TiO₂ array film grown on an FTO transparent conductive substrate was fabricated by directly treating the FTO substrate in a hydrochloric acid aqueous solution containing the titanium precursor under hydrothermal conditions according to the literature ^[27]. The FTO film itself can act as a seed layer to induce the growth of rutile TiO₂ rods owing to their identical crystal structures and the similar lattice parameters between rutile TiO₂ and SnO₂. The FTO film on the substrate consists of SnO₂ particles with a preferential aligning along the <002> direction. Each particle acts as a seed for the growth of single-crystal rutile TiO₂. Rutile TiO₂ crystals prefer to grow along the <001> direction, which leads to the formation of a rod-like morphology, because of the much higher surface energy of (001) facets than the thermodynamically stable (110) facets. This is the basis of the rutile TiO₂ film growing epitaxially on the FTO film usually showing a rod array morphology and growing along the <001> direction. To introduce mesopores into the single-crystal TiO₂ rods, a film of closely packed SiO₂ nanospheres as the template was deposited on the FTO substrate to regulate the shape of single-crystal TiO₂ rods grown up from the FTO layer during the subsequent hydrothermal process. Removing the template from the obtained SiO₂-TiO₂ composite film leads to the formation of the desired mesoporous single-crystal array film of rutile TiO₂. The key synthesis steps are summarized in Fig. 1.

Fig. 1. Schematic of the synthesis procedures of the films of nonporous single-crystal rod arrays (the top panel) and mesoporous single-crystal rod arrays (the bottom panel) of rutile TiO2 supported on fluorine-doped tin oxide (FTO) glass substrates.

The challenge for the fabrication of the mesoporous single-crystal array film is how to cast a robust template film of closely packed SiO₂ nanospheres that can firmly attach to the FTO film under the hydrothermal process. A high-speed centrifugation-assisted method was developed for the deposition of the SiO₂ nanosphere film. In brief, SiO₂ nanospheres with a diameter of approximately 50 nm were fabricated by a modified Stӧber method ^[22], and the SiO₂ nanospheres synthesized were then fully dispersed in the titanium precursor aqueous solution by ultrasonic treatment. Finally, these SiO₂ spheres were deposited on a FTO substrate located at the bottom of a Teflon-lined autoclave by centrifugation.

XRD patterns (Fig. 2) of the as-prepared mesoporous and nonporous rod array films on the FTO substrates suggest the formation of phase-pure rutile TiO₂ crystals (space group P4₂/mnm). The sharp diffraction peaks indicate the high crystallinity of the TiO₂ rods in both samples. The close full widths at half maximum of the two major diffraction peaks of rutile (110) and (101) suggest that the mesoporous rods have a similar crystal grain size to the nonporous rods. In contrast to the XRD pattern with the (110) peak being the most prominent for rutile TiO₂ powder, the XRD pattern of the rod array films exhibits the (101) peak as the most prominent. This feature is attributed to the oriented arrangement of the rutile rods in the films, as shown in the SEM images in Fig. 3(a) and (b). The top ends of nearly all rods are exposed to air. These rods have a width from 300 to 500 nm and length of approximately 2 μm. Both SEM and TEM images (Fig. 3(c) and (d)) show the presence of abundant mesopores in the rods synthesized with the silica template. The pore size in the rods is around 50 nm, which is consistent with the diameter size of the silica nanospheres used as the template. Compared with the uniform width of the nonporous rods, the top part of the mesoporous rods is narrower than the other parts. This is proposed to arise from the steric effect of the top silica nanospheres of the template layer in restraining the growth of the TiO₂ rods.

Fig. 2. XRD patterns of (1) solid and (2) mesoporous single-crystal nanorod array films of rutile TiO2 supported on FTO glass substrates. The solid squares indicate the diffraction peaks from the substrate.

Each single nonporous or mesoporous rod is a single crystal, as indicated by the selected area electron diffraction (SAED) patterns and high-resolution transmission electron microscopy (HRTEM) images (Fig. 3(e) and (f)). The set of periodic diffraction spots (insets of Fig. 3(e) and (f)) indicates the single-crystal nature for both nonporous and mesoporous rods. The different arrangements of the diffraction spots between nonporous and porous rods are explained as a result of the different zones of TiO₂ detected. For the nonporous rods, the rectangular diffraction patterns belong to the <1-10> zone axis of rutile TiO₂. The distances between the lattice fringes along two perpendicular directions are 0.325 and 0.296 nm, as shown in the HRTEM image (Fig. 3(e)), which are assigned to the (110) and (001) facets of rutile TiO₂, respectively. The parallelogram diffraction patterns recorded from the mesoporous rod belong to the <010> zone axis of rutile TiO₂. The lattice fringe spaces along two directions with an angle of 114.4° are 0.249 nm, as displayed in Fig. 3(f), which is assigned to the (101) facets of rutile TiO₂. The growth direction of the rods is in the <001> orientation according to the SAED patterns and HRTEM images (Fig. 3(e) and (f)).

Fig. 3. Top-view SEM images of (a) nonporous and (b) mesoporous rutile TiO2 rod array films supported on FTO substrates; low- magnification TEM images of a single (c) nonporous and (d) mesoporous rutile TiO2 rod scratched from the corresponding films; HRTEM images recorded from the (e) nonporous and (f) mesoporous rutile TiO2 rods; the insets are SAED patterns.

The chemical state of Ti and O in both samples was studied by XPS. The oxidation state of Ti and O in the high-resolution Ti 2p and O 1s XPS spectra of both samples (Fig. (4)) is clearly assigned to the full oxidation state of Ti⁴⁺ (binding energy: Ti2p_1/2 463.8 eV, Ti 2p_3/2 458.1 eV) and O^2- (binding energy: O1s 529.2 eV). Raman spectra (Fig. (5)) of the rods show the typical active modes of rutile TiO₂ and no obvious difference is observed between the samples. These results suggest that the mesoporous structure itself does not affect the chemical state and surface atomic structures of TiO₂ rods.

Fig. 4. High-resolution XPS spectra of Ti 2p and O 1s of the (1) solid and (2) mesoporous single-crystal nanorod array films of rutile TiO2 supported on FTO glass substrates.

Fig. 5. Raman spectra of the (1) solid and (2) mesoporous single-crystal nanorod array films of rutile TiO2 supported on FTO glass substrates.

UV-visible absorption spectra and XPS valence band spectra of the nonporous and mesoporous single-crystal rod array films were recorded to compare their optical and electronic structure properties. A gradually increased absorption from 500 to 900 nm observed in both mesoporous and nonporous samples results from the presence of oxygen vacancies in the samples. More oxygen vacancies are expected in the mesoporous sample owing to its larger specific surface area, which leads to a higher absorbance in this range for the mesoporous sample. The mesoporous rods show a higher absorbance in the UV region. This absorbance improvement could be attributed to light trapping and multiple reflecting effects arising from the mesopores in the rods. Moreover, the absorption edge of the mesoporous rods is blue-shifted by approximately 10 nm compared with that of the nonporous rods (Fig. 6). The bandgap of the mesoporous rods is larger than that of the nonporous rods by approximately 0.1 eV, as shown in their corresponding plots of transformed Kubelka-Munk function versus the energy of light (inset of Fig. 6). The blue-shift of the bandgap of mesoporous TiO₂ rods can be attributed to the incorporation of Si dopants from SiO₂ template into TiO₂ crystal lattices during the fabrication process. This is clearly confirmed by the high-resolution Si 2p XPS spectrum of the mesoporous rods (Fig. 7). The binding energy (101.6 eV) of Si 2p is much smaller than that of the reference SiO₂ (103.4 eV) as a result of the formation of Si-O-Ti bonds, which reduces the oxidation state of Si atoms. The blue-shift of absorption edge for Si-doped TiO₂ compared with pure TiO₂ was also observed in the porous TiO₂ nanotube array and compact TiO₂ films ^{[32, 33]}. The positions of valence band edges of the two samples at 1.8 eV are the same, as indicated by the valence band spectra (Fig. 8). The band alignments between the two samples are given in Fig. 9. The mesoporous rods have a higher conduction band edge by approximately 0.1 eV than the nonporous rods.

Fig. 6. UV-visible absorption spectra of the (1) solid and (2) mesoporous single-crystal nanorod array films of rutile TiO2 supported on FTO glass substrates. The inset is the plots of transformed Kubelka-Munk function versus the energy of light.

Fig. 7. High-resolution XPS spectrum of Si 2p of the mesoporous single-crystal nanorod array films of rutile TiO2 supported on FTO glass substrates.

Fig. 8. XPS valence band spectra of the (1) solid and (2) mesoporous single-crystal nanorod array films of rutile TiO2 supported on FTO glass substrates.

Fig. 9. The electronic energy band alignments of the solid and mesoporous single-crystal nanorods of rutile TiO2.

TiO₂-based materials are capable of playing versatile roles in photocatalytic H₂ generation ^{[34, 35, 36]}. PEC water splitting is considered a promising strategy in solar energy conversion and storage. The activity of the nonporous and mesoporous rod array films as photoanodes for PEC water oxidation was investigated and the results are shown in Fig. 10. The photocurrent density generated by the mesoporous rod photoanode is 0.74 mA/cm² at 1.23 V (versus a reversible hydrogen electrode) under AM 1.5 G simulator irradiation with a power density of 100 mW/cm², which is 2.6 times higher than that of the nonporous rod photoanode. The photoelectrochemical or photocatalytic activities of TiO₂ photocatalysts are known to be sensitive to the crystal phase, electronic structure, optical absorption, surface area, crystallinity, surface facets, and structural and chemical defects ^{[37, 38]}. The much-improved photocurrent density of the mesoporous rod photoanode in comparison with the nonporous rod photoanode can be attributed to the synergistic effect of four favorable factors as follows.

Fig. 10. Applied potential bias-dependent photocurrent density of the (1) nonporous and (2) mesoporous rutile TiO2 rod array photoanodes in the dark and under AM 1.5 G simulator irradiation with a power density of 100 mW/cm2. Measurement conditions: 1 mol/L aqueous NaOH electrolyte, scanning rate of 50 mV/s, Ag/AgCl reference electrode.

(1) The mesoporous structure endows a larger specific surface area of the mesoporous rods than that of the nonporous rods, which decorates the mesoporous rods with abundant surface-reactive sites for water oxidation. (2) The diffusion distance of the photo-generated holes from the bulk of the mesoporous rods to the interface between the rods and electrolyte is greatly shortened by the introduced mesopores. This can decrease the bulk recombination probability of the photo-generated charge carriers. (3) The increased absorbance favors the use of solar light to excite valence band electrons for the generation of more charge carriers for surface redox reactions. (4) Finally, but also importantly, the higher conduction band edge by around 100 meV facilitates the transport of photo-generated electrons from the mesoporous rods to the current collector (FTO layer) by providing a larger driving force. This is indicated by the negatively shifted photocurrent onset potential (approximately 100 mV) of the mesoporous rods relative to the nonporous rods.

The films of mesoporous single-crystal rod arrays of rutile TiO₂ directly growing on FTO conductive substrates were synthesized by a silica template-assisted hydrothermal method. The resultant films as photoelectrodes integrate several merits including abundant surface reactive sites, shortened diffusion distance of photoexcited charge carriers, and improved both light absorbance and driving force for the charge separation. As a demonstration, the mesoporous rod based photoanode for PEC water oxidation gives a much higher activity than the nonporous rod based photoanode. The developed films are anticipated to also act as an attractive electrode in other solar energy conversion systems including dye/quantum dot sensized solar cells.