Pillaring and NiO_x co-catalyst loading as alternatives for the photoactivity enhancement of K₂Ti₄O₉ towards water splitting

Abstract

Structural improvements of efficient and abundant photocatalysts and the employment of less expensive co-catalysts are appealing routes to promote the development of economically viable solar fuel production. We herein report a significant enhancement of the H₂ production from photocatalytic water splitting over K₂Ti₄O₉ after pillaring with TiO₂ pillars and loading of NiO_x as a co-catalyst. The pillaring favored the charge carrier separation within the photocatalyst framework under illumination, whereas the NiO_x co-catalyst, composed of nickel metal nanoparticles and nanostructured NiO, intensified the electron–hole separation at the surface. These are unprecedented results indicated by surface photovoltage spectroscopy, which was applied for the first time in this kind of photocatalyst composite. All modifications enhanced the H₂ evolution from 14.9 μmol, for the pristine K₂Ti₄O₉, to 982.0 μmol, for the final product, after 7 h of irradiation tests using methanol as a hole scavenger. This boost was also favored by the mesoporosity and high surface area created by the pillaring process. Moreover, NiO_x as the co-catalyst proved to be as effective as platinum and more effective than Ag and Au in this particular system.

---

Introduction

Finding efficient and inexpensive photocatalysts that can drive the photocatalytic production of hydrogen from water splitting is still a challenge to the scientific community. The growing interest in this subject comes from the fact that H₂ obtained from sunlight and water is a potential substitute for the vast use of non-renewable and greenhouse-gas-emitting fossil fuels.^1–5 However, the viability of such a trade is hindered by the limited efficiency of the known photocatalysts and the high costs of their constituents.

In the pursuit of high efficiencies, the literature describes well-known water splitting photocatalysts, such as TiO₂, SrTiO₃, K₂Ti₄O₉, K₄Nb₆O₁₇, and ZnO, modified by the noble metals rhodium, platinum, silver, and gold, among others, to increase the light absorption and electrical conductivity, and/or intensify the separation of the photogenerated charge carriers.^6–12 These procedures, usually substitutional doping or nanoparticle loadings, lead to remarkable improvements in the photoactivity of the pristine semiconductors.^6,13,14 However, noble metals are very expensive and most of the time these modifications are accompanied by costly procedures, further increasing the overall cost of the final product. Therefore, alternative non-expensive classes of co-catalysts with efficiencies comparable to those of noble metals are required to make the photocatalytic H₂ production economically attractive.

Nickel-based co-catalysts have been studied by different research groups, showing satisfactory results in the increment of the activity of the respective photocatalysts.^15–22 A thorough study regarding the function of the NiO_x (0 < x < 1) co-catalyst on SrTiO₃ was reported by Osterloh and co-workers in 2012.²³ The authors pointed out a striking feature of this component, which is rather a dual co-catalyst with nickel metal nanoparticles and nickel oxide nanoparticles having complementary roles to promote the proper electron–hole separation and consequent enhancement in the yield of the photocatalytic water splitting.

In light of this evidence, we consider that the employment of NiO_x is a good approach to the replacement of expensive noble metal co-catalysts in photocatalytic systems because nickel is cheaper and more abundant.

The photoactivity of a semiconductor also depends on its intrinsic structural properties, such as surface area, porosity, crystallinity, and structural defects. In general, periodic porous structures with high surface area tend to act as effective matrices for catalysis because the open-access framework exposes the active sites more efficiently and enables the diffusion of the reactants.^24–28

Although some layered solids like potassium tetratitanate (K₂Ti₄O₉) and potassium niobate (K₄Nb₆O₁₇) are stable and photoactive towards water splitting, their efficiencies are hindered by fast electron–hole recombination and low charge carrier diffusion lengths.^10,29 Nonetheless, most layered solids are structurally versatile due to their expansible interlayer region, which permits intercalation of guest species, leading to structural and compositional modifications that might completely alter their properties and applicability.^30–33

Intercalation is the first stage of the pillaring process, which is an effective and well-known strategy to obtain high-surface-area porous materials from layered structures.^31,34 This method has already been proven to be efficient for the enhancement of the photocatalytic activities of layered niobates, titanates,^24,34–37 and even graphene oxide and reduced graphene oxide, pillared with carbon nanotubes.³⁸

In 2002, Choy and colleagues created means for the consistent pillaring of layered titanates in spite of the high charge that maintains the interlayer attraction.³⁴ They obtained a periodical mesoporous structure by intercalating tetrabutylammonium hydroxide (TBAOH) within the layers of the acidic form of the titanate, followed by exfoliation under ultrasonication. The layers were then restacked after substituting TBAOH with TiO₂ nanosol coordinated by acetylacetone.

Integrating structural modifications of an active well-known photocatalyst with the incorporation of effective and cheap co-catalysts is an appealing route towards the development of economically viable H₂ production systems. This is a goal that will only be achieved if the solar-to-hydrogen efficiency of the processes increases in the same degree as the costs drop.^1,4

This work takes a step forward and presents for the first time the study of hydrogen evolution over the semiconductor K₂Ti₄O₉ pillared with TiO₂ (through an effective and non-expensive procedure) loaded with NiO_x co-catalysts (which are cheap and abundant), through photocatalytic water splitting.

We also carried out a detailed investigation of the surface photovoltage created under illumination in this class of composites by surface photovoltage spectroscopy. The results unveil the implications that the modifications in K₂Ti₄O₉ bring to the charge carrier separation and recombination and how these modifications affect the hydrogen evolution yields.

Results and discussion

A new photocatalyst was obtained through the pillaring of K₂Ti₄O₉ with TiO₂ nanopillars and the loading of Ni⁰ and NiO nanoparticles on the surface (NiO_x/TiO₂/Ti₄O₉). As one of the goals of this work is the investigation of the effects caused by these modifications over H₂ production, results of the intermediate material loaded solely with NiO (NiO/TiO₂/Ti₄O₉) are presented, as well. Other photocatalysts were prepared by substituting NiO_x with Ag, Au and Pt nanoparticles to compare the activity of this nickel-based co-catalyst with those of noble metals.

The X-ray diffraction (XRD) patterns of the materials, K₂Ti₄O₉, the pillared version (TiO₂/Ti₄O₉) and NiO_x/TiO₂/Ti₄O₉, displayed in Fig. 1, reflect the modification of the structure of K₂Ti₄O₉, into TiO₂/Ti₄O₉ after the pillaring with TiO₂.²⁴

Fig. 1 X-ray diffraction patterns of (a) pristine K₂Ti₄O₉, (b) TiO₂/Ti₄O₉ and (c) NiO_x/TiO₂/Ti₄O₉. The peaks marked with (●) and (◆) correspond to the Ti₄O₉²⁻ planes and the planes of TiO₂ anatase, respectively.

The XRD pattern indicates that the layered K₂Ti₄O₉ is a well-ordered monoclinic structure, pertaining to the space group C2/m (JCPDS 32-0861), with the distance between the adjacent Ti₄O₉²⁻ layers corresponding to 0.87 nm, based on the (200) reflection at 2θ = 10.1°.²⁹

The (200) reflection is attributed to the basal distance of pure K₂Ti₄O₉ and the absence of this reflection at 2θ values close to 10.0° in the pattern of TiO₂/Ti₄O₉ is a relevant indication of the structural alteration after the pillaring. This is a consequence of the insertion of TiO₂ nanopillars within the layers of K₂Ti₄O₉. The pillaring process causes a certain disturbance in the structure due to the impaired coherence between the lamellae caused by their random restacking during the reassembling for the formation of the porous structure.^31,32,34

The patterns of TiO₂/Ti₄O₉ and NO_x/TiO₂/Ti₄O₉ show the planes (201), (311) and (020) pertaining to the in-sheet structure of the Ti₄O₉²⁻ nanolayers, indicating that the original structure of the sheets was conserved. The XRD patterns also confirm the incorporation of TiO₂ nanoparticles by exhibiting broad diffraction peaks at 25.3, 38.1, 48.0, 54.6, 62.8 and 75.5° assigned to (101), (004), (200), (105), (204) and (215) reflections of the TiO₂ anatase structure of the space group I4₁/amd (JCPDS 83-2243).^24,34,37

The pattern of the mesoporous solid NiO_x/TiO₂/Ti₄O₉ (Fig. 1c) does not display any extra peaks related to either NiO or Ni because of the low amount of these co-catalysts (2 wt%) and high dispersion on the surface. Moreover, this pattern leads to favourable evidence that no structural modification took place in the porous material during the in situ synthesis of NiO under the thermal treatment at 400 °C.¹⁸ The same is true for the patterns of the mesoporous structures loaded with 2 wt% of Ag, Au and Pt, shown in Fig. S1.†

The N₂ adsorption–desorption isotherms and pore size distribution curves of the materials K₂Ti₄O₉, TiO₂/Ti₄O₉ and NiO_x/TiO₂/Ti₄O₉ (Fig. 2) depict the large enhancement of surface area of the photocatalyst after pillaring, with concomitant formation of porous structures with pores in the mesoporous range.

Fig. 2 Adsorption–desorption isotherms and corresponding pore size distribution curves of (a) K₂Ti₄O₉, (b) TiO₂/Ti₄O₉ and (c) NiO_x/TiO₂/Ti₄O₉.

Both the surface area and total pore volume of K₂Ti₄O₉ considerably increased from 25 to 145 m² g⁻¹ and from 0.049 to 0.25 cm³ g⁻¹, respectively, after pillaring with TiO₂. The high surface area of the pillared solid is predominantly due to the presence of well-distributed TiO₂ nanoparticles in the interlayer region and not to the formation of TiO₂ nanoparticles on the outer surface of the material.³⁴ This statement is supported by the BET specific surface area of TiO₂ anatase, formed through the same procedure without the addition of Ti₄O₉²⁻ nanosheets, which was 41 m² g⁻¹.

The isotherm of the pristine K₂Ti₄O₉ is a type III isotherm, attributed to non-porous solids. This isotherm shows a hysteresis in the desorption branch that may be related to the random aggregation of adjacent layers forming a “house-of-cards” configuration.^32,34

The porosity completely changed after the pillaring, as the isotherm of TiO₂/Ti₄O₉ is of type IV attributed to a mesoporous structure with a H1 hysteresis loop, characteristic of well-formed and periodically arranged mesopores.²⁷ This proves that the complexed TiO₂ nanosols could access the interlayer regions of the tetratitanate layers and form nanopillars homogeneously dispersed within the entire structure.

Similar characteristics are observed in the isotherm of NiO_x/TiO₂/Ti₄O₉ although the hysteresis slightly deviates from the H1 type due to the disturbance of the pore structure by the formation of NiO and Ni nanoparticles inside and in the entrance of the pores. This also led to a small reduction of the specific surface area to 134 m² g⁻¹. Interestingly, the variation in the specific surface area values was less pronounced for the materials loaded with Ag, Au and Pt (isotherms exhibited in Fig. S2†), which have specific surface areas of 153, 140 and 148 m² g⁻¹, respectively, probably due to the extra porosity that the incorporation of these nanoparticles generates on the surface, as indicated by the increase in the N₂ adsorption at very high P/P₀ values.^24,37

Nonetheless, despite the slight disturbance in the pore structure, none of the isotherms indicate loss of the mesoporosity feature after the loading of the co-catalysts, which still show high surface area values. In addition, the total pore volume remains practically the same (0.25 cm³ g⁻¹) in all cases.

Fig. 2 also displays an inset with the non-local density functional theory (NLDFT) pore size distribution curves of the materials K₂Ti₄O₉, TiO₂/Ti₄O₉ and NiO_x/TiO₂/Ti₄O₉. The curves show that no relevant porosity was identified for the layered precursor K₂Ti₄O₉, especially in the mesoporous range, whereas TiO₂/Ti₄O₉ and NiO_x/TiO₂/Ti₄O₉ have pore diameters between 2 and 6 nm, confirming that the built pores are in the mesoscale. The pore distribution is broad because of the combination of the pores created by the pillaring and the “house-of-card” disordered restacking of the exfoliated nanosheets.^24,34,37 The same characteristics are observed in the pore size distribution curves for the materials loaded with Ag, Au and Pt, exhibited in Fig. S2d.†

The pores in the mesoscale in a high-surface-area structure are beneficial to promote the efficient diffusion of reactants within the structure, easing the access to the photoactive sites.^25,27,28

Transmission electron microscopy (TEM) images (Fig. 3a and b) show that the structure of the layered precursor K₂Ti₄O₉ is composed of stacked rectangular sheets. The width of these layers varies from 50 to 150 nm and the length ranges from 300 nm to 1 μm.²⁹ Adjacent layers are parallel to each other with a distance of 0.9 nm, as evidenced by Fig. 3b, in good agreement with the XRD results.

Fig. 3 Transmission electron microscopy (TEM) images of (a, b) K₂Ti₄O₉, (c, d) TiO₂/Ti₄O₉ and (e) NiO_x/TiO₂/Ti₄O₉.

The pillaring incorporated a new component into the structure, as a closer view of TiO₂/Ti₄O₉ reveals the presence of some TiO₂ nanoparticles distributed onto the Ti₄O₉²⁻ layers (Fig. 3c). High-resolution TEM (HR-TEM) images of the samples K₂Ti₄O₉ and TiO₂/Ti₄O₉ (Fig. S3†) reveal the differences in the surfaces of both materials. The distance between the planes of the formed nanoparticles is 0.35 nm, in good correspondence with the (010) planes of TiO₂ anatase.

Fig. 3d shows that the nanosheets in TiO₂/Ti₄O₉ are connected by TiO₂ pillars located in the interlamellar region and the distance between the layers increased to 2.5 nm due to the insertion of these nanoparticles. This image proves the effective pillaring of K₂Ti₄O₉ through the adopted procedure.³⁴

The TEM image of NiO_x/TiO₂/Ti₄O₉ (Fig. 3e) exhibits the same characteristic as that of TiO₂/Ti₄O₉ with higher concentration of nanoparticles surrounding the Ti₄O₉²⁻ layers due to the addition of NiO and metallic nickel nanoparticles. The proper identification of the surrounding nanoparticles was achieved by EDS mapping, which provided a distinction between the particles of TiO₂ and those composed of nickel (Fig. 4).

Fig. 4 Energy-dispersive X-ray spectroscopy (EDS) mapping in (a) a selected region of NiO_x/TiO₂/Ti₄O₉ of the elements (b) titanium, (c) oxygen and (d) nickel; and (e) the EDS spectrum showing all the elements detected in the sample. The detected Cu, Co and Fe elements belong to the grid used for the measurements.

In general, the EDS mapping of NiO_x/TiO₂/Ti₄O₉ proves that the surface is composed of the elements titanium and oxygen, with nanoparticles of NiO_x with dimensions from 1.5 to 2.5 nm homogeneously distributed in random parts of the surface. The EDS spectrum (Fig. 4e) confirms the detection of the elements Ti, O and Ni, together with peaks of Cu, Fe and Co, which are the elements present in the grid holder composition (TEM image and EDS spectrum of the grid are presented in Fig. S4†). This proposition is further proven by the mapping of another region of the NiO_x/TiO₂/Ti₄O₉ surface, exhibited in Fig. S5.† This image contains three EDS spectra corresponding to different sections of the mapped region (nickel-rich and nickel-poor sections), confirming the presence of titanium and oxygen all over the surface and random distribution of Ni-containing nanoparticles.

The surface composition and electronic states of nickel, titanium and oxygen on the surface of the solids NiO/TiO₂/Ti₄O₉ and NiO_x/TiO₂/Ti₄O₉ were investigated by X-ray photoelectron spectroscopy (XPS). The XPS spectra at the Ti 2p, O 1s and Ni 2p regions for these materials are displayed in Fig. 5.

Fig. 5 X-ray photoelectron spectra at the (a) Ti 2p, (b) O 1s and (c) Ni 2p regions for the materials NiO/TiO₂/Ti₄O₉ and NiO_x/TiO₂/Ti₄O₉.

The spectra at the Ti 2p region display two signals at 458.3 and 464.0 eV related to the Ti 2p_3/2 and Ti 2p_1/2 binding energies, respectively, with the spin–orbit splitting of 5.7, which indicates that titanium is in the Ti⁴⁺ oxidation state.²⁴ In addition, the O 1s component attributed to the Ti–O bond type is observed at 529.5 eV in both cases.^16,24

The Ni 2p spectra of NiO/TiO₂/Ti₄O₉ and NiO_x/TiO₂/Ti₄O₉ (Fig. 5c) show two peaks attributed to the Ni 2p_1/2 (873.1 eV) and Ni 2p_3/2 (855.2 eV) spin–orbit levels, characteristic of Ni²⁺. Two shake-up satellite signals at 880.1 and 861.1 eV, which result from the charge transfer from O 2p to Ni 3d band in NiO, are also observable, confirming that nickel is in the +2 oxidation state. Moreover, the energy separation between Ni 2p_3/2 and Ni 2p_1/2 signals is 17.6 eV, which agrees with the value for NiO crystals.^15,16,39 A peak related to the Ni 2p_3/2 level of metallic nickel was expected to appear at a binding energy of 852.6 eV in the Ni 2p spectrum of NiO_x/TiO₂/Ti₄O₉. However, only a small shoulder at this region, which is not present in the spectrum of NiO/TiO₂/Ti₄O₉, could be observed due to the unavoidable oxidation of the Ni⁰ nanoparticles to Ni²⁺ upon exposure to air.^40,41

The incorporation of both TiO₂ nanopillars and NiO_x co-catalyst not only resulted in structural changes, but also optical alterations, as depicted by the ultraviolet-visible diffuse reflectance spectra (UV/Vis DRS) of the materials (Fig. 6).

Fig. 6 Ultraviolet-visible diffuse reflectance spectra of (a) K₂Ti₄O₉, (b) TiO₂/Ti₄O₉, (c) NiO_x/TiO₂/Ti₄O₉, (d) Ag/TiO₂/Ti₄O₉, (e) Au/TiO₂/Ti₄O₉, and (f) Pt/TiO₂/Ti₄O₉.

Pristine K₂Ti₄O₉ has intense absorbance in the ultraviolet region, with onset at 337 nm, credited to the intrinsic band gap energy of 3.40 eV, as previously reported.^29,34 The pillaring of this layered precursor led to a red shift in the onset to 375 nm and a colour change from white to pale yellow. This shift is a consequence of the combination of the layered protonic titanate with anatase TiO₂ nanosol. This is confirmed by the UV/Vis spectrum of the pure TiO₂ anatase synthesized in this work, which presents an onset exactly at 375 nm (Fig. S6†).³⁵

Despite the darkening of the composite after the incorporation of metallic nickel and NiO nanoparticles onto the surface of TiO₂/Ti₄O₉, the UV/Vis DRS spectrum of NiO_x/TiO₂/Ti₄O₉ (Fig. 6c) does not show any relevant increase in the light absorbance, which is mostly due to the high band gap of NiO (3.5 eV) and the low Ni-containing particle content in the sample (2 wt%).^23,42

An extra absorption of light in the visible region occurs for the materials loaded with the noble metals Ag, Au, and Pt because of the resonance of the incident photons with the localized surface plasmons of these nanoparticles (Fig. 6d–f). This effect is most visible in the case of Au/TiO₂/Ti₄O₉ because the surface plasmon resonance absorption band of gold nanoparticles is centred at 550 nm. The spectrum of Pt/TiO₂/Ti₄O₉, presents additional light absorption from 375 to 800 nm,^6,13,43,44 whereas the surface plasmon resonance band of silver nanoparticles is not observed because it lies in the region of 390 nm and overlaps with the absorption band of TiO₂/Ti₄O₉.⁴⁵

Surface photovoltage spectroscopy (SPS) is a reliable and precise technique to assess the photochemical charge carrier dynamics within the components of the nanocomposites and to obtain additional information about the photochemistry of the semiconductors. The information provided by SPS, such as the photovoltage produced by thin films of semiconductors and their effective band gap values, are of paramount importance to the understanding of the variation of H₂ production through water splitting by the different photocatalysts.^46–48 In this method, a contactless semi-transparent Kelvin probe is employed to measure the light-induced changes of the contact potential difference (CPD) of a sample as a function of the irradiation energy.⁴⁸

Compared to photoelectrochemistry, SPS is much more sensitive for the determination of photochemical parameters of a semiconductor, as even small potential difference values on the millivolt scale, produced by the transfer of small charge carrier densities, can be detected. Moreover, it is contactless and free of any interaction with an electrolyte that might affect the determination of the intrinsic photochemical parameters. These features allow the direct determination of the majority carrier type, mid-gap states, effective band gap and light-induced contact photovoltage difference (CPD) of any photocatalyst.^47,49–51

The SPS spectra of films of the materials K₂Ti₄O₉, TiO₂/Ti₄O₉, NiO/TiO₂/Ti₄O₉, and NiO_x/TiO₂/Ti₄O₉ on fluorine-doped tin oxide glass (FTO) are displayed in Fig. 7a.

Fig. 7 Surface photovoltage spectra of films of the photocatalysts (a) K₂Ti₄O₉, TiO₂/Ti₄O₉, NiO/TiO₂/Ti₄O₉, and NiO_x/TiO₂/Ti₄O₉; and (b) Ag/TiO₂/Ti₄O₉, Au/TiO₂/Ti₄O₉ and Pt/TiO₂/Ti₄O₉ on FTO substrates, recorded under vacuum (6.8 × 10⁻⁵ mbar).

Under illumination, the film of the pristine K₂Ti₄O₉ produced a maximum contact potential difference variation (ΔCPD) of −221 mV at a photon energy of 3.70 eV, above the optical bandgap of this semiconductor, with an onset at 3.37 eV, confirming that the photogenerated charge carriers were produced by the excitation of electrons from the valence band (VB) to the conduction band (CB) of K₂Ti₄O₉. The effective band gap of 3.37 eV given by the SPS spectrum of potassium tetratitanate agrees with the band gap indicated by UV/Vis diffuse reflectance spectroscopy (3.40 eV). The negative variation of the CPD signal indicates that the electrons of the film flow towards the conducting substrate and away from the Kelvin probe, whereas the holes accumulate in the film. This is a characteristic of n-type semiconductors, in which electrons are the majority charge carriers.⁵²

After pillaring, the intensity of the ΔCPD signal substantially increased from −221 mV to −709 mV at an energy value of 3.70 eV, representing the enhancement of the electron density that is transferred to FTO because of the proper coupling between TiO₂ nanoparticles and Ti₄O₉²⁻ nanosheets. TiO₂ anatase is also an n-type semiconductor with a band gap energy of 3.2 eV,^53,54 and its incorporation created a three-dimensional interface for charge separation and transport. The onset of the spectrum of TiO₂/Ti₄O₉ decreased to 2.88 eV (effective band gap), much lower than that of K₂Ti₄O₉ (3.37 eV), reflecting the formation of intermediate states by the pillaring of the layered structure.

The loading of NiO nanoparticles onto TiO₂/Ti₄O₉, enhanced the photovoltage from −709 to −768 mV at a photon energy of 3.70 eV, with a slight decrease in the onset value. These results imply that nickel oxide (a p-type semiconductor) acts as a hole-trapping site in the structure, which enhances the positive polarization of the sample, intensifying the charge carrier separation under illumination. The proper alignment of the energy levels of the VB of nickel oxide with those of TiO₂ and the layered precursor in the protonic form (H₂Ti₄O₉), displayed in Fig. 8, represents the action of NiO as a hole trap centre in this system and the favourable flow of excited electrons into the FTO substrate, as indicated by the SPS measurements.^23,42

Fig. 8 Energy diagram depicting the charge carrier flux in the composite NiO_x/TiO₂/Ti₄O₉ under illumination.

The incorporation of Ni nanoparticles onto the surface of NiO/TiO₂/Ti₄O₉, resulting in NiO_x/TiO₂/Ti₄O₉, considerably decreased the photovoltage to −195 mV at 3.70 eV (Fig. 7a), suggesting a retention of electrons in the film. This effect opposes that observed for the sole incorporation of NiO. The combined results lead to the conclusion that Ni⁰ co-catalysts act as electron traps on the TiO₂/Ti₄O₉ system, also represented in Fig. 8, whereas NiO acts as a hole trap, resulting in a joint action for the efficient separation of the charge carriers in the surface of the mesoporous absorber. Having separate reduction and oxidation sites on the surface of the photocatalyst prevents the recombination of electrons and holes and hence leads to the enhancement of the hydrogen production yields, as shown by the photocatalytic tests performed on our materials.

The results described above corroborate those reported for strontium titanate (SrTiO₃) modified with the Ni⁰/NiO combined co-catalyst, in which both species perform complementary roles in the improvement of the overall water splitting.²³ The combined loading of both co-catalysts favours the effective charge carrier separation. This conclusion is consistent with the general concept of photochemical diodes where metal particles on photocatalysts act as electron acceptors. It is also supported by the proper alignment of the valence band energy of NiO, which is thermodynamically suitable to accept holes from both K₂Ti₄O₉ and TiO₂ semiconductors.^22,42 This is the first time an investigation describes this co-catalyst combination with a TiO₂-pillared tetratitanate, showing the effects of the charge carrier separation on the activity of the resulting nanocomposite towards the photocatalytic H₂ production.

As a comparison, Fig. 7b exhibits the SPS spectra for the materials loaded with silver, gold and platinum, all showing certain decrease in photovoltage compared to TiO₂/Ti₄O₉. These noble metal photocatalysts act as electron trap sites, preventing their flow to the FTO substrate, like metallic nickel nanoparticles. The degree of this decrease in the order Pt > Au > Ag is intimately related to the work functions of the metals, which are 5.65, 5.10 and 4.26 eV, respectively.⁵⁵

The effect of the improved charge separation by loading NiO_x on TiO₂/Ti₄O₉ could be verified by the photocatalytic hydrogen evolution tests, over time, through water splitting employing the photocatalysts K₂Ti₄O₉, TiO₂/Ti₄O₉, NiO/TiO₂/Ti₄O₉ and NiO_x/TiO₂/Ti₄O₉ using methanol as a sacrificial agent. The average hydrogen evolution rates are depicted in Fig. 9 together with results of similar experiments with the mesoporous photocatalysts loaded with silver, gold and platinum nanoparticles.

Fig. 9 Hydrogen evolution over 50 mg of (a) K₂Ti₄O₉, TiO₂/Ti₄O₉, NiO/TiO₂/Ti₄O₉, and NiO_x/TiO₂/Ti₄O₉; and (b) Ag/TiO₂/Ti₄O₉, Au/TiO₂/Ti₄O₉, and Pt/TiO₂/Ti₄O₉ suspended in 50 mL of a 20% (v/v) aqueous methanol solution. The suspensions were irradiated with a 300 W Xe arc lamp with light power of 224 mW cm⁻².

An abrupt increase in the H₂ production was observed after the introduction of TiO₂ nanopillars within the layered tetratitanate interlamellar region. Apparently, the improved diffusion length of the electrons and holes within the structure has enhanced the rates of the redox reactions.^34–36 In terms of values, pristine K₂Ti₄O₉ produced 14.9 μmol of H₂ after 7 h of test, whereas TiO₂/Ti₄O₉ generated 766.8 μmol in the same period of time, which represents an increase of 51 times in hydrogen evolution.

The enhanced photoreactivity of the mesoporous material was also favoured by the pronounced increase of surface area, as photocatalytic reactions tend to be accelerated when more active sites are available for interaction with the reactants.^24,37,56 The existence of pores in the mesoscale is an extra factor that promotes the diffusion of the reactants within the structure, aiding their accessibility to the photoactive sites.^24,27,35,36

The loading of NiO_x aided the enhancement of the H₂ evolution. SPS results are in accordance with the difference of H₂ production before and after the co-catalyst incorporation. Even with the sole deposition of nickel oxide nanoparticles, the H₂ production was improved from 766.8 μmol, for TiO₂/Ti₄O₉, to 892.4 μmol, for NiO/TiO₂/Ti₄O₉, after 7 h of irradiation, due to the hole-trapping effect brought by the NiO co-catalyst.

A boost in the H₂ evolution yield to 982.0 μmol in 7 h was acquired after the incorporation of nickel metallic nanoparticles onto NiO/TiO₂/Ti₄O₉. This H₂ production is 66 times higher than that obtained for pristine K₂Ti₄O₉.^17,20,22,57 As already stated by Townsend et al.,²³ Ni⁰ nanoparticles act as a complement to the NiO co-catalyst function in the system by further favouring the charge carrier separation on the surface by trapping the electrons, and then preventing the charge recombination. Thus, the surface of the mesoporous material NiO_x/TiO₂/Ti₄O₉ contains separate sites for reduction of protons (Ni⁰) and oxidation of methanol (NiO), which are accessible through the pores in the mesoscale and the high surface area.

The effective charge carrier separation promoted by the Ni⁰/NiO co-catalyst system makes the photocatalytic activity of this material comparable to those containing the noble metal nanoparticles silver, gold and platinum. In fact, the materials Ag/TiO₂/Ti₄O₉, Au/TiO₂/Ti₄O₉ and Pt/TiO₂/Ti₄O₉ generated 771.9, 862.4 and 1151 μmol of hydrogen after 7 h of test, whereas NiO_x/TiO₂/Ti₄O₉ produced 982.0 μmol.^19,21,23

The use of noble metal co-catalysts has proven to be highly efficient for the enhancement of yields of photocatalytic processes due to various features, as discussed in previous reports.^14,58,59 However, the specificity of the Ni⁰/NiO system for the charge carrier separation made it more efficient than the mesoporous materials loaded with Ag and Au. Besides, a decrease in the H₂ evolution rate was detected for Ag/TiO₂/Ti₄O₉ after 5 h of test, probably due to the poisoning of the co-catalyst.

Photocatalytic hydrogen production tests were also carried out over pure and NiO_x-loaded (2 wt%) commercial TiO₂ P25, presenting yields of 51.3 and 260.9 μmol of H₂ after 7 h, respectively (Fig. S7†). These results confirm the advantage of using this mesoporous pillared structure as the main photocatalyst, compared to a recognized photoactive semiconductor.

The noble-metal-free characteristic of this photocatalyst is important because the high cost of the overall process is a constant hindrance to the large-scale application of photocatalysis in the production of fuels. Even though the porous material loaded with Pt presented higher yield compared to NiO_x/TiO₂/Ti₄O₉, the difference in the H₂ production (169.0 μmol after 7 h) does not compensate for the high cost attached to the use of such noble metals. The substitution with the much cheaper NiO_x co-catalyst is very advantageous to reduce the price of the whole process, maintaining the high efficiency.

Experimental section

Chemicals

The synthesis and pillaring of K₂Ti₄O₉ were carried out by using titanium(IV) oxide, anatase ≥99.99% (Aldrich); potassium carbonate, ≥99% (Aldrich); tetrabutylammonium hydroxide, 40 wt% (1.5 mol L⁻¹) solution in water (Acros Organics); titanium(IV) isopropoxide, 97% (Aldrich); and acetylacetone P. A. (Merck). Nickel(II) nitrate hexahydrate, >99% (Acros Organics) was employed for the loading of NiO and Ni⁰ on the surface of the porous material. The photodepositions of Ag, Au and Pt nanoparticles were performed by using silver nitrate, 99.9% (Alfa Aesar); hydrogen tetrachloroaurate(III) hydrate, 99.99% (Alfa Aesar); and chloroplatinic acid hydrate, ≥99.9% (Aldrich) as precursors, respectively. Titanium(IV) dioxide P25, ≥99.5% (Sigma-Aldrich), used for photoactivity comparison; ethanol, ≥99.8% (Synth); nitric acid, 68.0–70.0% (Synth); 2-propanol, ≥99.5% (Sigma-Aldrich) and methanol, ≥99.8% (Sigma-Aldrich) were high-grade reagents and were used as purchased. Deionized water used in the experiments was purified to 18.2 MΩ cm resistivity using an ultra-pure Milli-Q Millipore system.

Synthesis and pillaring of K₂Ti₄O₉

Potassium tetratitanate (K₂Ti₄O₉) was prepared by suspending K₂CO₃ and TiO₂ anatase (molar ratio of 1 : 3.5) in acetone (20 wt% of solid). The suspension was sonicated for 10 min, wet-ground in a mortar and, after evaporating the solvent at 60 °C, heated at 800 °C for 20 h under a ramp of 10 °C min⁻¹. The grinding and heating procedures were repeated. The resulting white solid was washed three times with fractions of 30 mL of deionized water, and dried at 100 °C, under vacuum, for 8 h.

The obtained K₂Ti₄O₉ sample was converted into its layered protonic form (H₂Ti₄O₉) by stirring it in 100 mL of a 1.0 mol L⁻¹ HCl aqueous solution at room temperature for 3 days (the acidic solution was replaced by a fresh one every day). The product was then centrifuged, washed with deionized water and dried at 100 °C under vacuum for 8 h.

In a next step, 5.0 g (14.8 mmol) of H₂Ti₄O₉ was exfoliated by dispersing it in 250 mL of an aqueous solution containing 11.5 g (44.4 mmol) of tetrabutylammonium hydroxide (TBAOH), through ultrasonication at room temperature, for 2 h. The molar ratio TBAOH : H₂Ti₄O₉ employed was 3 : 1. The pH of this exfoliated suspension was carefully adjusted to 9 with 0.5 mol L⁻¹ HNO₃ prior to the addition of the TiO₂ nanosol.

In a separate flask, 8.42 g (29.6 mmol) of titanium isopropoxide was slowly added to a mixture of 34.6 mL (337 mmol) of acetylacetone and 85.4 mL (1.12 mol) of 2-propanol. This solution was stirred for 30 min at room temperature. The monodisperse TiO₂ nanosol was then prepared by adding dropwise this titanium isopropoxide solution to 60 mL of a 0.1 mol L⁻¹ HNO₃ aqueous solution, under vigorous stirring. This mixture was maintained under vigorous stirring for 24 h at room temperature.

The TiO₂ nanosol was added dropwise to the exfoliated Ti₄O₉²⁻ suspension and stirred at 60 °C for 24 h. In such a material, the molar ratio between TiO₂ and Ti₄O₉²⁻ (TiO₂ : Ti₄O₉²⁻) is 2 : 1, according to the quantity of titanium isopropoxide used in the synthesis. The product was centrifuged, washed several times with a 1 : 1 v/v water : ethanol solution, to remove the excess of TiO₂ nanosol, and dried at 60 °C under vacuum for 8 h. The pillaring of the material was concluded by heating it at 300 °C for 2 h under a heating ramp of 1 °C min⁻¹. The as-prepared porous material was named TiO₂/Ti₄O₉. As the organic cation TBA is removed after heating, the negative charge is balanced by H⁺ ions from the HNO₃ aqueous solution.

For comparison purposes, nanoparticles of TiO₂ anatase were prepared through a similar route, excluding the addition of the exfoliated nanosheets in the synthesis.

NiO_x loading

For the incorporation of the NiO_x co-catalyst, 1.0 g of TiO₂/Ti₄O₉ was added to 20 mL of an aqueous solution containing 0.0390 g (134 mmol) of Ni(NO₃)₂·6H₂O (1 wt% loading of NiO), and the components were thoroughly mixed in a sonication bath for 10 min. The solvent was removed after centrifuging and the solid was annealed at 400 °C for 60 min, under an O₂ flow. This material, named NiO/TiO₂/Ti₄O₉, was washed with three 30 mL fractions of deionized water and dried at 70 °C for 8 h. The photodeposition of Ni⁰ nanoparticles was carried out by suspending 0.5 g of NiO/TiO₂/Ti₄O₉ in 20 mL of an aqueous solution containing 0.0248 g (0.0851 mmol) of Ni(NO₃)₂·6H₂O (equivalent to 1 wt% Ni⁰ loading) and 10 mL of pure methanol, in a round bottom quartz flask. This flask was connected to an air-tight irradiation setup linked to a Varian 3800 gas chromatograph (with a 60/80 Å molecular sieve column and thermal conductivity detector) and the mixture was irradiated using a 300 W Xe lamp (175 mW cm⁻²) under an argon atmosphere and stirred for 12 h. The resulting grey solid, named NiO_x/TiO₂/Ti₄O₉, was centrifuged, washed with three 3 mL fractions of deionized water, and dried in an oven at 70 °C for 8 h.

Noble metal loadings

Briefly, 0.3 g of TiO₂/Ti₄O₉ was suspended in 20 mL of deionized water, in a round bottom quartz flask, followed by the addition of 0.0094 g (55.6 mmol) of AgNO₃, which corresponds to 2 wt% Ag⁰ loading. The mixture was sonicated for 10 min and subsequently connected to an air-tight irradiation setup linked to a Varian 3800 gas chromatograph. The system was constantly stirred and irradiated by a 300 W Xe lamp (175 mW cm⁻²) for 6 h. The resulting solid was then centrifuged, washed with three 30 mL fractions of deionized water and dried at 70 °C for 8 h. This Ag-loaded mesoporous material was named Ag/TiO₂/Ti₄O₉.

Similar procedures were followed for the loading of Au and Pt nanoparticles onto TiO₂/Ti₄O₉ by substituting AgNO₃ with 0.0109 g (30.5 mmol) of HAuCl₄·H₂O and 0.0126 g (30.8 mmol) of H₂PtCl₆·H₂O, also corresponding to loadings of 2 wt% of Au and Pt on TiO₂/Ti₄O₉, respectively. These materials were named Au/TiO₂/Ti₄O₉ and Pt/TiO₂/Ti₄O₉.

Surface photovoltage spectroscopy (SPS) measurements

SPS measurements were conducted under vacuum (8.6 × 10⁻⁵ mbar) on films on fluorine-doped tin oxide (FTO) substrates. A gold Kelvin probe (Delta PHI Besocke) mounted inside a vacuum chamber served as the reference electrode. The samples were illuminated with monochromatic light from a 150 W Xe lamp filtered through an Oriel Cornerstone 130 monochromator (0.1–0.3 mW cm⁻²). Films of the samples with average thickness of approximately 6 μm, as shown by the profilometer scan in Fig. S8,† were prepared by drop-coating 4 mg mL⁻¹ aqueous dispersions of the photocatalysts on FTO slides. After drying in air, the films were annealed at 150 °C for 90 min.

Photocatalytic hydrogen production

Dispersions of 50 mg of each photocatalyst in 50 mL of 20 vol% aqueous methanol solutions were prepared in a round bottom quartz flask to measure the rate of hydrogen evolution. The flask was evacuated down to 50 torr and purged with argon several times until the chromatogram of the headspace indicated an air-free environment. The system was irradiated with a 300 W xenon arc lamp (power density of incident light = 224 mW cm⁻²), after filtering out the infrared portion with a water filter. The air-tight irradiation system was connected to a vacuum pump and to a Varian 3800 gas chromatograph (with a 60/80 Å molecular sieve column and thermal conductivity detector – TCD) to quantify the amount of evolved H₂, using area counts of the recorded peaks. The measurements were repeated at least two more times for each material.

Characterization

X-ray diffraction (XRD) patterns were collected on a Shimadzu model XRD-7000 diffractometer using Cu Kα radiation (0.154 nm) at 40 kV and 30 mA with a rate of 0.01° s⁻¹ for 2θ values from 1.5 to 80°. N₂ adsorption–desorption isotherms of samples previously outgassed at 150 °C for 12 h were collected on a Quantachrome Nova 4200 instrument at −195.8 °C. The specific surface areas and pore volumes were calculated through the Brunauer–Emmett–Teller (BET) equation, whereas the pore size distributions were assessed by employing the Kernel non-local density functional theory (NLDFT) on the desorption branch of the isotherms. Transmission electron microscopy (TEM) and high-resolution TEM (HRTEM) images were obtained using a HRTEM-JEM 3010 URP microscope at an accelerating voltage of 300 kV at the Electron Microscopy Laboratory (LME) located at CNPEM. The samples were prepared by dispersing the materials in water, followed by drop-casting onto a carbon-coated copper grid (400 mesh). A JEOL JEM 2100F microscope equipped with an EDS detector and a GIF Tridiem 863 filter was employed for the Energy Dispersive Spectroscopy (EDS) mappings. X-ray photoelectron spectroscopy (XPS) measurements were carried out on a Thermo VG Scientific Sigma Probe instrument with the monochromatic Al Kα X-ray source (1486.6 eV) focused on the samples. The high-resolution spectrum was collected with pass energy of 20 eV and energy step size of 0.1 eV with a flood gun on to eliminate surface charge. The ultraviolet-visible spectra were obtained through the diffuse reflectance spectroscopy (DRS) method using a Thermo Scientific Evolution 220 spectrometer. The thicknesses of the films analysed through SPS were measured with a Veeco Dektak profilometer.

Conclusions

Our results exemplify how the thoughtful alteration of some properties of well-known semiconductors like K₂Ti₄O₉ can turn them into more powerful systems for the photocatalytic production of hydrogen from water splitting.

Novel photocatalysts were synthesized by the modification of the photoactive layered semiconductor K₂Ti₄O₉ through the pillaring with TiO₂ nanoparticles and surface decoration with NiO and Ni⁰ (NiO_x) co-catalysts. The pillaring conferred a 51-fold increase in the H₂ production due to an improved charge separation within the structure with a concomitant increase in the surface area and formation of pores at the mesoscale range. These characteristics reflected in a large enhancement in the photovoltage from −221 mV to −709 mV at a photon energy of 3.70 eV.

The loading of nickel metal nanoparticles and NiO intensified the charge carrier separation at the surface and further boosted the H₂ evolution because they function as complementary components. Ni⁰ acts as an electron trap and a reduction site while NiO accumulates holes and it acts as an oxidation site, as indicated by surface photovoltage spectroscopy. This feature makes the efficacy of this co-catalyst system comparable to those of noble metal co-catalysts, as confirmed by the photocatalytic tests, which makes the system more commercially attractive.

Conflicts of interest

The authors declare no competing financial interest.

Acknowledgements

The authors acknowledge São Paulo Research Foundation (FAPESP – Grants 2013/05654-9 and 2014/22388-3) for financial support and fellowship, the National Centre for Energy and Materials Research (CNPEM) for the TEM images (LME), and Dr Frank E. Osterloh for access to the surface photovoltage spectrometer and H₂ testing equipment.

References

1. J. W. Ager, M. R. Shaner, K. A. Walczak, I. D. Sharp and S. Ardo, Energy Environ. Sci., 2015, 8, 2811–2824.
2. A. Kudo and Y. Miseki, Chem. Soc. Rev., 2009, 38, 253–278.
3. F. E. Osterloh, Chem. Soc. Rev., 2013, 42, 2294–2320.
4. D. M. Fabian, S. Hu, N. Singh, F. A. Houle, T. Hisatomi, K. Domen, F. E. Osterloh and S. Ardo, Energy Environ. Sci., 2015, 8, 2825–2850.
5. T. Shirakawa, M. Higashi, O. Tomita and R. Abe, Sustainable Energy Fuels, 2017, 1, 1065–1073.
6. J. Yu, L. Qi and M. Jaroniec, J. Phys. Chem. C, 2010, 114, 13118–13125.
7. T. Puangpetch, S. Chavadej and T. Sreethawong, Energy Convers. Manage., 2011, 52, 2256–2261.
8. A. Iwase, H. Kato and A. Kudo, Catal. Lett., 2006, 108, 7–10.
9. Y.-G. Lin, Y.-K. Hsu, Y.-C. Chen, S.-B. Wang, J. T. Miller, L.-C. Chen and K.-H. Chen, Energy Environ. Sci., 2012, 5, 8917–8922.
10. S. Uchida, Y. Yamamoto, Y. Fujishiro, A. Watanabe, O. Ito and T. Sato, J. Chem. Soc., Faraday Trans., 1997, 93, 3229–3234.
11. R. Konta, T. Ishii, H. Kato and A. Kudo, J. Phys. Chem. B, 2004, 108, 8992–8995.
12. O. Rosseler, M. V. Shankar, M. K.-L. Du, L. Schmidlin, N. Keller and V. Keller, J. Catal., 2010, 269, 179–190.
13. Y. Zhong, K. Ueno, Y. Mori, T. Oshikiri and H. Misawa, J. Phys. Chem. C, 2015, 119, 8889–8897.
14. Y. Wang, Y. Wang and R. Xu, J. Phys. Chem. C, 2013, 117, 783–790.
15. M. Wang, J. Han, Y. Hu, R. Guo and Y. Yin, ACS Appl. Mater. Interfaces, 2016, 8, 29511–29521.
16. W. Wang, W. Zhang, C. Hao, F. Wu, Y. Liang, H. Shi, J. Wang, T. Zhang and Y. Hua, Sol. Energy Mater. Sol. Cells, 2016, 152, 1–9.
17. X. Chen, W. Chen, H. Gao, Y. Yang and W. Shangguan, Appl. Catal., B, 2014, 152–153, 68–72.
18. C.-J. Chen, C.-H. Liao, K.-C. Hsu, Y.-T. Wu and J. C. S. Wu, Catal. Commun., 2011, 12, 1307–1310.
19. Q. Liu, L. Zhang and P. A. Crozier, Appl. Catal., B, 2015, 172, 58–64.
20. Z. Zou, J. Ye and H. Arakawa, J. Phys. Chem. B, 2002, 106, 13098–13101.
21. M. Zhong, T. Hisatomi, Y. Kuang, J. Zhao, M. Liu, A. Iwase, Q. Jia, H. Nishiyama, T. Minegishi, M. Nakabayashi, N. Shibata, R. Niishiro, C. Katayama, H. Shibano, M. Katayama, A. Kudo, T. Yamada and K. Domen, J. Am. Chem. Soc., 2015, 137, 5053–5060.
22. K. R. Yoon, J. W. Ko, D.-Y. Youn, C. B. Park and I.-D. Kim, Green Chem., 2016, 18, 944–950.
23. T. K. Townsend, N. D. Browning and F. E. Osterloh, Energy Environ. Sci., 2012, 5, 9543–9590.
24. M. Alves Melo Júnior, A. Morais and A. F. Nogueira, Microporous Mesoporous Mater., 2016, 234, 1–11.
25. A. Corma, Chem. Rev., 1997, 97, 2373–2419.
26. X. Chen and S. S. Mao, Chem. Rev., 2007, 107, 2891–2959.
27. F. J. V. E. Oliveira, M. A. Melo Jr and C. Airoldi, Mater. Res. Bull., 2013, 48, 1045–1056.
28. L. Mercier and T. J. Pinnavaia, Adv. Mater., 1997, 9, 500–503.
29. M. R. Allen, A. Thibert, E. M. Sabio, N. D. Browning, D. S. Larsen and F. E. Osterloh, Chem. Mater., 2010, 22, 1220–1228.
30. M. A. Melo Jr and C. Airoldi, Dalton Trans., 2010, 39, 10217–10227.
31. M. Machida, X. W. Ma, H. Taniguchi, J.-I. Yabunaka and T. Kijima, J. Mol. Catal. A: Chem., 2000, 155, 131–142.
32. W. Hou, Q. Yan, B. Peng and X. Fu, J. Mater. Chem., 1999, 5, 109–114.
33. M. A. Melo Jr, F. J. V. E. Oliveira and C. Airoldi, Appl. Clay Sci., 2008, 42, 130–136.
34. J.-H. Choy, H.-C. Lee, H. Jung, H. Kim and H. Boo, Chem. Mater., 2002, 14, 2486–2491.
35. B. Li, B.-Z. Lin, O. Zhang, L.-M. Fu, H. Liu, Y.-L. Chen and B.-F. Gao, J. Colloid Interface Sci., 2012, 386, 1–8.
36. B.-H. Xu, B.-Z. Lin, Q.-Q. Wang, X.-T. Pian, O. Zhang and L.-M. Fu, Microporous Mesoporous Mater., 2012, 147, 79–85.
37. Y. Zhang, H. Liu, G. Zhang, L. He, P. Liu and B. Lin, Mater. Res. Bull., 2014, 60, 510–515.
38. L. L. Zhang, Z. Xiong and X. S. Zhao, ACS Nano, 2010, 4, 7030–7036.
39. D. Escalera-López, Y. Niu, J. Yin, K. Cooke, N. V. Rees and R. E. Palmer, ACS Catal., 2016, 6, 6008–6017.
40. K. Kamonsuangkasem, S. Therdthianwong, A. Therdthianwong and N. Thammajak, Appl. Catal., B, 2017, 218, 650–663.
41. S. Song, S. Yao, J. Cao, L. Di, G. Wu, N. Guan and L. Li, Appl. Catal., B, 2017, 217, 115–124.
42. B. A. Nail, J. M. Fields, J. Zhao, J. Wang, M. J. Greaney, R. L. Brutchey and F. E. Osterloh, ACS Nano, 2015, 9, 5135–5142.
43. C. G. Silva, R. Juárez, T. Marino, R. Molinari and H. García, J. Am. Chem. Soc., 2011, 133, 595–602.
44. K. Ueno, T. Oshikiri and H. Misawa, ChemPhysChem, 2016, 17, 199–215.
45. M. A. Melo Jr, L. S. S. Santos, M. C. Gonçalves and A. F. Nogueira, Quim. Nova, 2012, 35, 1872–1878.
46. J. Wang, J. Zhao and F. E. Osterloh, Energy Environ. Sci., 2015, 8, 2970–2976.
47. J. Zhao and F. E. Osterloh, J. Phys. Chem. Lett., 2014, 5, 782–786.
48. L. Kronik and Y. Shapira, Surf. Sci. Rep., 1999, 37, 1–206.
49. F. E. Osterloh, M. A. Holmes, J. Zhao, L. Chang, S. Kawula, J. D. Roehling and A. J. Moulé, J. Phys. Chem. C, 2014, 118, 14723–14731.
50. J. Zhao, B. A. Nail, M. A. Holmes and F. E. Osterloh, J. Phys. Chem. Lett., 2016, 7, 3335–3340.
51. T. L. Shelton, N. Harvey, J. Wang and F. E. Osterloh, Appl. Catal., A, 2016, 521, 168–173.
52. J. Wang and F. E. Osterloh, J. Mater. Chem. A, 2014, 2, 9405–9411.
53. A. Fujishima and K. Honda, Nature, 1972, 238, 37–38.
54. C. Burda, Y. Lou, X. Chen, A. C. S. Samia, J. Stout and J. L. Gole, Nano Lett., 2003, 3, 1049–1051.
55. H. B. Michaelson, J. Appl. Phys., 1977, 48, 4729–4733.
56. F. E. Osterloh, ACS Energy Lett., 2017, 2, 445–453.
57. L. Guo, H. Hagiwara, S. Ida, T. Daio and T. Ishihara, ACS Appl. Mater. Interfaces, 2013, 5, 11080–11086.
58. S. Linic, P. Christopher and D. B. Ingram, Nat. Mater., 2011, 10, 911–921.
59. I. Thomann, B. A. Pinaud, Z. Chen, B. M. Clemens, T. F. Jaramillo and M. L. Brongersma, Nano Lett., 2011, 11, 3440–3446.

---

Footnote

† Electronic supplementary information (ESI) available: Some additional characterization results, including XRD patterns, N₂ adsorption–desorption isotherms, TEM images and EDS mappings. See DOI: 10.1039/c7se00589j

---