Experimental results

Typical fabrication methods to form a 3D substrate include lithographic patterning, plasma etching, and vacuum deposition.^[Reference Pease and Chou^20] However, these methods are generally expensive and hence are less viable for high-throughput commercial manufacturing of high-aspect-ratio nanostructures. As an alternative approach, we have utilized a titanium mesh to obtain such a 3D multi-layered electrode configuration by folding the mechanically ductile and easily bendable mesh repeatedly. Figure S2 shows photographs of a single-layer Ti mesh, [Fig. S2(a)], and a folded multi-layered Ti mesh [Fig. S2(b)] [see Fig. S2 in Supplementary materials].

The multi-layered mesh described in Supplementary Figs. S1(c) and S2 actually leads to a substantially increased surface area as compared with a single-layer mesh, the extent of which depends on the number of folding events. The folding of the Ti mesh was performed in such a way as to produce parallel 10 mm × 10 mm size layers with relatively uniform spacing of ~0.5 mm. It is apparent from the micrographs that the 3D folded mesh structure allows light to be passed, with a schematic representation presented in Supplementary Fig. S1(c). Besides the passage of light, this hole spacing also enables a smooth water flow in a continuous operation water splitting. Compared with the batch photocatalytic reaction, the continuous-flow photocatalysis process using the mesh design electrodes would be uninterrupted. The movement of the flowing solution can help to remove generated gas products on the surface of the photocatalysts and rapidly provide a fresh solution to the photocatalyst surface, which is beneficial for enabling additional photocatalytic reactions. Also, the mesh architecture can easily be contacted with liquid with minimal air bubble trapping thus enabling intimate contact with the electrolyte. Moreover, the 0.5 mm level spacing between the adjacent Ti mesh layers is adequate enough to allow the reacting chemical solution to bathe the mesh layers surface during the hydrothermal growth process for easier and uniform formation of the TiO₂ nanotubes as a large-surface-area photocatalyst.

A hydrothermal reaction was conducted on the entire folded Ti mesh assembly. A high-resolution scanning electron microscopy (SEM) photograph [Figs. 1(b) and 1(c)] reveals that a well-aligned, conformal TiO₂ nanotube array was formed on the surface of the Ti substrate. The nanotubes of TiO₂ grew mostly in a conformal and perpendicular geometry onto the local Ti surface. Figure 1(c) shows the tilted view of the TiO₂ nanotube array, about ~1 µm tall. The profile of the substrate structure is retained during the hydrothermal process for TiO₂ nanotube formation. In our previous work, we have reported characterization of the hydrothermally grown TiO₂ nanotubes.^[Reference Brammer, Kim, Noh, Loya, Frandsen, Conelly, Chen and Jin^21] Since the nanotube film is formed by etching of the Ti metal itself, the adhesion of the 8 nm nanotubes on the Ti substrate appears to be good as indicated by the preliminary shear deformation experiments. A thin and compact titanium oxide layer is produced at the early growth stage of the hydrothermal reaction prior to the formation of the TiO₂ nanotubular structure. Such a compact barrier layer could effectively prevent direct contact between the metallic titanium mesh surface and the electrolyte, thus could help to minimize charge recombination at the titanium/electrolyte interface. The transmission electron microscopy (TEM) micrograph, [Fig. 1(d)], reveals the multi-walled nature of the hydrothermally formed TiO₂ nanotubes having ~8 nm average diameter. Thus, the hydrothermal processing technique is a simple and economic approach for creating nanotube architecture on multi-layered Ti substrate, which provides additional surface area.

Supplementary Fig. S3 shows the UV–visible absorption spectrum of hydrothermally grown TiO₂ nanotubes on sapphire (a transparent substrate utilized here for optical measurements). The spectrum of the TiO₂ nanotube sample shows the UV light absorption below 390 nm in wavelength. It is known that TiO₂ is mostly operative in the UV region since its bandgap of 3.2 eV requires high-energy UV light.^[Reference Hernandez-Alonso, Fresno, Suarez and Coronado^22]Figure 2 shows the current density–potential response of various layer 3D structured TiO₂ electrodes in a 1 M KOH solution with the potential measured relative to an Ag/AgCl standard electrode in the a.m. 1.5 simulated solar illumination. The photoelectrochemical measurement in aqueous solution is a well-accepted general method for evaluating the performance of water splitting. The photocurrent density at +0.5 V of the hydrothermal TiO₂ nanotubes on a single-layer Ti mesh electrode was 26 µA/cm². In comparison, the 3, 5, and 7-layered mesh electrodes show progressively higher photocurrent densities than the single-layer mesh electrode, exhibiting an enhancement in photoresponse with a photocurrent density of 50, 65, and 118 µA/cm², respectively, under the identical illumination conditions. The photocurrent density improvement by increasing the number of the folded mesh layers seems to saturate around 7 layers, with the 9-layered electrode showing essentially the same photocurrent density as that for the 7 layer electrode. The enhanced performance of the multi-layered mesh-type photoanode was attributed to three prominent advantageous characteristics. First, the multi-layered mesh fully covered with TiO₂ nanotubes provided a larger surface area that can be used to support more photocatalytic material for light harvesting with further reaction sites than that of TiO₂ nanotubes on an equivalent footprint size single-layer mesh. The surface area of the mesh screen (from a top view) is ~30% area fraction of the footprint area, for example, by calculating the relative surface area of the Ti mesh material versus the rhombus unit cell mesh area in Fig. 1(a). With the 7-layered folded mesh electrode, the overall surface area of the Ti metal is 30% × 7 = 210%, which is increased ~110% as compared with the flat footprint Ti sheet. Therefore, 3D architecture by folding mesh is indeed a simple yet effective structure to enhance the areal amount of nanotubes so as to improve the performance of the photocatalytic nanomaterial although the photocurrent density did not increase as much as the calculated area (210%) due to the gradually decreased lights on added mesh layers [Supplementary Fig. S4]. Moreover, the better catalytic performance can be ascribed to larger effective contact area between photocatalytic material and electrolyte solution. Of course, the TiO₂ nanotube array structure provides a much increased surface area with further catalytic reaction sites and more rapid reaction rates. The hydrothermally grown TiO₂ nanotubes are extremely fine (8 nm in diameter). The distance between the neighboring nanotubes is roughly about 50 nm. Correspondingly, the calculated total surface areas of the plain (flat) Ti mesh surface layer versus that of the TiO₂ nanotube containing mesh surface having a 1 µm tall TiO₂ nanotube layer (assuming a square array of 8 nm diameter and 50 nm spaced apart vertical nanotubes) are estimated to be 1 versus 14.2 µm², respectively, for the identical unit substrate area of 1 µm × 1 µm. Therefore, the hydrothermally grown TiO₂ nanotube array exhibits a significantly increased surface area than the plain substrate without the nanotube structure.

Figure 2.

Photocurrent versus bias voltage of TiO₂ nanotube layered mesh electrodes on various numbers of mesh layers.

Second, the metallic titanium underneath the nanotube array (and TiO₂ surface in the gap between the nanotubes) serves as the highly conducting substrate to transport the photogenerated electrons.^[Reference Li and Li^23–Reference Quan, Yang, Ruan and Zhao^25] After the absorption of solar energy, electrons generated from the photocatalytic materials should be transported through the TiO₂ layer to the metallic anode substrate and finally to the external circuit. Since the resistivity of the metallic Ti mesh (~40 µΩ-cm) is much lower than that of the semiconducting TiO₂, the mesh-type photoelectrode all electrically connected even for the multi-layered folded mesh electrode, acts as a low resistivity current path to drain away the electrons, which improves the charge transportation/collection in part by suppressing charge recombination.^[Reference Koide, Islam, Chiba and Han^26] Therefore, the TiO₂ nanotube mesh not only offers high surface area and enhanced light absorption in its 3D structure, but its underlying Ti metal also provides direct paths to easily move the photogenerated electrons toward the counter electrode.

Third, the multi-layered TiO₂ nanotube mesh can absorb more light compared with the conventional photoanodes based on planar electrodes. After the incident light passes through the apertures of the preceding mesh, the backside of the preceding mesh is exposed to the reflected light from the front mesh layer. Thus, in the multi-layered mesh electrode structure, there may be more opportunities to absorb light through recapturing of multiple reflected light that would otherwise be lost. Consequently, the geometric effect of the mesh-type multi-layers facing each other may have a beneficial effect on photocatalytic activity and therefore contribute to the photocurrent. In order to quantitatively assess the amount of light that passes through each mesh layer, the optical transmittance was measured versus the number of the layers in the multi-layered Ti electrodes as shown in Supplementary Fig. S4. While the transmittance is gradually decreased with the added mesh layer, the multi-reflections from various mesh layers further increase the overall photoinduced current. The reflection of the following Ti mesh can act as another incident light toward the backsides of the preceding meshes. For example, the back-reflections from the mesh layers #6 and #7 will influence and increase the photoinduced currents on all the preceding mesh layers #1 through #5, which can be much larger than the back-reflection influence by the mesh layers #4 and #5 on #1 through #3. Thus, maximized efficiency of the reflected lights from the front sides of the meshes was obtained in the 7-layered mesh which has six backsides. Additional studies on the backside light absorption aspects will be conducted in the future. Also, the 7-layered mesh TiO₂–nanotube-containing electrode shows highest photocurrent density among various configurations as described in Supplementary Fig. S5. With such a 3D photoanode, we have clearly demonstrated the advantages of the multi-layered mesh architecture as an advanced 3D photoanode.

It should be emphasized that such an architecture is applicable to many other types of water-splitting catalysts or other photoelectrochemical catalysts generally deposited on such a multi-layered mesh electrode, especially with a 8 nm TiO₂ nanotube array as the catalyst basis. There are many photoelectrochemical catalysts being investigated for water splitting by various groups, and it will be interesting and meaningful to utilize some of the more recently developed, advanced photocatalysts on the multi-layered Ti mesh electrode configuration, especially on the surface of the high-surface-area 8 nm diameter TiO₂ nanotube array. Since the inability of the Ti oxide catalyst to utilize visible and infrared light limits the efficiency of solar photocatalytic activity. Metal doping, or surface loading method has been widely adopted to reduce the limitations caused by the wide bandgap and the high percentage of recombination of excited electrons and holes which decreases photocatalytic activity.^[Reference Rupa, Divakarand and Sivakumar^27] To assess the applicability of the folded multi-layered Ti mesh-type electrode as an advanced 3D photoanode basis for water splitting, we performed similar photoelectrochemical measurements on plasmonic Au nanoparticles on TiO₂ mesh electrode under identical illumination with incident light of 100 mW/cm² and identical photoelectrochemical reaction conditions. Au nanoparticles with 10 nm average diameter were deposited on the TiO₂ nanotubes by solid-state dewetting of 3-nm-thick Au films deposited on TiO₂ nanotube surfaces by high-pressure sputtering to enhance non-line-of-sight deposition on the nanotubes.

Supplementary Fig. S6 shows the schematic illustration of the solid dewetting process to fabricate Au nanoparticles dispersed on TiO₂ nanotube surface and the SEM image showing the Au nanoparticles attached to the 8 nm diameter TiO₂ nanotube surface after annealing. The photoelectrochemical measurements, [Fig. 3(a)], clearly indicate that the Au/TiO₂ mesh sample (on 7-layered Ti mesh electrode surface coated with 8 nm TiO₂ nanotubes) exhibited a significant enhancement in photoresponse with a photocurrent density of 244 µA/cm², which is 2 times higher than that of the 7-layered TiO₂ mesh alone without Au nanoparticles. The dark currents of all of the samples including the 7-layered TiO₂ mesh with Au nanoparticles were also measured and presented in Supplementary Figs. S7–S9. The addition of Au nanoparticles (~10 nm average diameter) also enabled photoelectrochemical reaction and water splitting in the visible spectrum as presented in Fig. 3(b) for which only visible light was used for the measurement. Figure 3(b) data show the photocurrent density values of the 7-layered Au/TiO₂ nanotube mesh, irradiated with mostly visible light using a high-pass filter with a “cut on” wavelength of 400 nm (FSQ-GG400, Newport Corp.) to remove the UV component of sunlight. The 7-layered TiO₂ mesh sample with fine Au nanoparticles exhibited a significant enhancement in photoresponse with a photocurrent density of 46 µA/cm² under visible illumination, which is ~11 times higher than that of the TiO₂ mesh alone without the Au nanoparticles. Visible light absorption spectra of the gold nanoparticles on TiO₂ nanotubes indicate that the Au nanoparticles are photoexcited, and charge separation is accomplished by the transfer of photoexcited electrons from the gold particle to the TiO₂ conduction band, thereby enhancing the photoelectrochemical performance. An incident photon-to-current conversion efficiency (IPCE) measurement would be useful for advanced characterization and understanding of the mechanisms, which we will be conducted in future experiments.

Figure 3.

Photocurrent versus bias voltage of TiO₂ nanotubes with and without Au nanoparticles irradiated with (a) UV–visible light and (b) visible light only.

The photocurrent versus bias voltage linear sweep voltammetry curves of Fig. 2 and Supplementary Fig. S5 show an interesting behavior of the limiting photocurrent never being reached even at very positive applied potentials. Hierarchical 3D structures have different geometric areas, which have different incident angles to the directional light in photoelectrochemical (PEC) measurement. Unlike flat electrodes, these different local areas have somewhat different kinetics of charge separation and recombination of the electron–hole pairs. Such structural and reaction variations may cause a different limiting photocurrent behavior. Further details of such studies will be conducted in the future. Another possibility that may be speculated is that the TiO₂ used in this work may have relatively poor effective minority carrier diffusion lengths, requiring large bias potentials (band bending) to successfully collect carriers at the TiO₂/electrolyte interface. In contrast, the linear sweep voltammetry curve for the Au-coated photoelectrode under visible light [Fig. 3(b)] exhibits a limiting photocurrent behavior. This may be caused by Au nanoparticles plasmonic enhancement that allows the charge carriers to be generated immediately adjacent to the Au nanoparticles, requiring very little band bending for the minority holes to reach the TiO₂/electrolyte or Au/electrolyte interfaces. Further research is needed to clearly understand the limiting photocurrent behavior in these samples.

In Fig. 3, a positive shift in the photocurrent onset potential of ~150 mV was observed for the sample with the plasmonic Au nanoparticles compared with that without Au nanoparticles. Photocurrent onset potential which is known as open-circuit voltage is basically related to the bandgap of photocatalytic material. When the Au nanoparticles are added onto the TiO₂ nanotubes, this co-catalyst system acts like a photocatalyst having a lower bandgap, which can be activated with low-energy visible light resulting in the shift of onset potential. Also, there is a similar shift in the photocurrent onset seen in Supplementary Fig. S5 for the samples with TiO₂ nanotubes compared with control Ti foil and mesh samples without TiO₂ nantubes. The samples without TiO₂ nanotubes may in a sense act as a lower bandgap material than the samples with well-crystallized TiO₂ nanotubes. Further details of such studies will be reported in future publications.