Over the past decades,semiconductor photocatalysis has attracted considerable attention owing to its wide applications especially in solar energy conversion,environmental remediation,and organic synthesis ^{[1, 2, 3, 4]}. Among the many semiconductor materials examined,TiO₂ has been regarded as a promising candidate for application in solar energy conversion because of its high chemical stability,high abundance,non-toxicity,and low cost ^{[5, 6]}. However,the low quantum efficiency generated during solar energy conversion and limited utilization of visible light or sunlight,which respectively result from the high rate of electron-hole recombination under irradiation and the relatively wide band gap of TiO₂,greatly hinder the application of TiO₂ ^{[7, 8]}.

To overcome these drawbacks,some strategies have been explored to improve the photocatalytic efficiency of TiO₂ such as surface modification and element doping ^{[9, 10, 11, 12, 13, 14, 15, 16, 17]}. Regarding element doping into TiO₂,a suitable dopant and suitable level of doping are required because the defects induced by the dopant unavoidably introduce new charge carrier traps and recombination centers that could be detrimental to the photocatalytic performance of the resulting TiO₂ material. In contrast,surface modification of TiO₂ with metals,especially noble metals,possesses unique advantages such as its simplicity in operation and ability in significantly enhancing the efficiency of the photocatalytic process. Noble metals,such as Pt and Au,deposited on TiO₂ nanoparticles have been proved to feature high Schottky barriers. Furthermore,the noble metals can act as electron traps to enhance charge separation at the interface between the metal particles and TiO₂,thus retarding the recombination of photo-excited electron and holes,consequently prolonging the lifetime of the carriers for efficient photocatalysis ^{[18, 19, 20, 21]}. Pt,which features the largest work function and lowest overpotential for hydrogen production,has been demonstrated to be the best candidate for TiO₂ surface modification for photocatalytic hydrogen production ^[19].

In the absence of a co-catalyst,TiO₂ shows very low activity toward hydrogen production from water splitting. Thus,in addition to the light-harvesting ability of the semiconductor,the photocatalytic reaction highly depends on the nature of the co-catalyst ^{[20, 22]}. A comprehensive investigation of such catalytic system will be helpful for achieving highly efficient production of hydrogen from water or other derivatives,as well as contributing to the fundamental understanding of the photocatalytic process. However,to date,systematic studies on the photocatalytic performance of such noble metal-loaded TiO₂ nanocomposites are limited and several fundamental issues remain unresolved. A primary issue is how the type of noble metal affects the photocatalytic activity of TiO₂ under irradiation. Furthermore,TiO₂ exists in three crystal forms i.e.,anatase,rutile,and brookite. Among these crystal forms,rutile and anatase have attracted considerable attention,and thus many extensive studies relating to their photocatalytic activity feature in the literature. In contrast,very limited research studies on brookite are available ^{[23, 24]}.

Herein,a series of noble metal-loaded TiO₂,M-TiO₂ (M = Pt,Au,Pd; TiO₂ = Anatase,Rutile),samples with ultrafine metal clusters were prepared and examined as photocatalysts for water splitting with methanol (methanol photo-reforming). As the properties of the co-catalysts may be influenced by the preparation method,photo-deposition,a simple,but popular method,was employed in the present study. Rutile and brookite TiO₂ were employed as the semiconductors. The effects of noble metal surface deposition on the photocatalytic activity of TiO₂ toward hydrogen production were examined.

All chemical reagents (analytical grade) were purchased from Alfa Aesar Chemical Co. and used as received without further purification. Rutile and brookite TiO₂ nanorods were prepared according to previous literature reports ^{[25, 26]}. The obtained fresh TiO₂ powder was calcined at 400 °C for 4 h in flowing air before further use. To obtain the noble metal-loaded TiO₂ samples,a modified photo-assisted loading process was developed. The process was conducted under high vacuum conditions for minimizing the heterogeneous nucleation rate of the loaded metals. Typically,a known amount of metal precursor solution (H₂PtCl₆,PdCl₂,or HAuCl₄) was added to a TiO₂ methanolic aqueous solution (10 mL methanol and 90 mL H₂O; 0.5 g TiO₂) under stirring without pH adjustment. Then,the entire closed reaction system was evacuated for 10 min,then irradiated using a 200 W Xe lamp (wavelength: 320-780 nm) for 6 h. The resulting precipitates were filtered and washed thoroughly with methanol and deionized water,and then dried under ambient conditions.

X-ray diffraction (XRD) patterns of the samples were recorded on a Bruker D8 ADVANCE powder diffractometer using Cu K_α radiation (λ = 0.1542 nm) at a scanning rate of 4°/min inthe region of 2θ = 20°-60°.

The metal contents of the samples were determined by inductively coupled plasma-atomic emission spectroscopy (ICP-AES; IRIS Advantage). The specific Brunauer-Emmett- Teller (BET) surface areas (S_BET) of the samples were determined from N₂ adsorption-desorption isotherms collected on a Quantachrome iQ-MP gas adsorption analyzer. Prior to measurement,the samples were dehydrated at 300 °C for 2 h.

Transmission electron microscopy (TEM) images were taken on a Philips Tecnai G² 20 S-TWIN electron microscope,operating at an acceleration voltage of 200 kV. A few drops of alcohol suspension containing the sample were placed on a carbon-coated copper grid,followed by evaporation at ambient temperature.

Diffuse reflectance ultraviolet-visible (UV-Vis) absorption spectra of the samples were recorded on a Varian Cary 300 UV-Vis spectrometer using BaSO₄ as reference.

The photoluminescence (PL) spectra of the samples were recorded on a Spex FL201 fluorescence spectrophotometer. The samples were dry-pressed into self-supporting wafers and then illuminated using a 325-nm He-Cd laser as an excitation source at ambient temperature.

The photocatalytic reforming of methanol (also known as photocatalytic water splitting with methanol (as the sacrificial agent)) was performed in a top-irradiation-type Pyrex reaction cell connected to a closed gas circulation and evacuation system under irradiation of a Xe lamp fitted with a filter (wavelength: 320-400 nm). In a typical experiment,about 100 mg catalyst sample was added to 100 mL methanol aqueous solution (10%) in the reaction cell. After evacuation for 30 min,the reactor cell was irradiated by the Xe lamp at 200 W under stirring. The evolved gas product H₂ was analyzed using an online gas chromatograph (Varian CP-3800) equipped with a thermal conductivity detector.

The XRD patterns of the prepared TiO₂ materials are shown in Fig. 1. The diffraction pattern of the prepared rutile TiO₂ (Rutile) sample (Fig. 1(a)) agreed with that of rutile (JCPDF#21-1276),whereas the prepared brookite TiO₂ (Brookite) (Fig. 1(b)) displayed a pattern typical of brookite (JCPDF#29-1360). No other crystalline phases were detected in either sample,thereby indicating the phase (rutile or brookite) purity of the prepared materials. After metal loading,no diffraction peaks corresponding to Pt,Pd,or Au species could be observed in the M-Rutile and M-Brookite samples. This result was attributed to the low loading of the metal species (~1.0 wt%) and the formation of ultrafine clusters (1.0-2.0 nm),as discussed later.

Fig. 1. XRD patterns of the prepared rutile TiO2 and M-Rutile (a) and brookite TiO2 and M-Brookite (b) samples.

Diffuse reflectance UV-Vis spectroscopy was performed to study the optical properties of the prepared TiO₂ materials,and the results are shown in Fig. 2. All TiO₂ samples displayed a typical UV absorption band at ~250-400 nm,which can be primarily ascribed to electron promotion in TiO₂ from the valence band (VB) to the conduction band (CB) ^[20]. Compared with the bare rutile and brookite samples,a slight red shift in the absorption edge was observed for the M-TiO₂ samples,thereby indicating that noble metal deposition can slightly extend the absorption of light of TiO₂ to the visible region. Both sets of M-Rutile and M-Brookite samples displayed similar shifts. This can further be confirmed by calculating the band gap values of the samples according to the plot in Fig. 2(c,d),obtained via transformation based on the Kubelka-Munk function. The estimated band gap values of Pd-Rutile,Pt-Rutile,and Au-Rutile nanocomposites were 2.80,2.91,and 2.96 eV,respectively. These band gap values were smaller than that of pristine rutile (3.04 eV). Furthermore,the estimated band gap values of Pd-Brookite,Pt-Brookite,and Au-Brookite nanocomposites were 3.33,3.30,and 3.25 eV,respectively,which are likewise lower than that of pristine brookite (3.37 eV). Therefore,it can be concluded that deposition of the noble metals studied herein induced very mild effects on the optical properties of the M-TiO₂ nanocomposites. Furthermore,an additional wide absorption band at ca. 580 nm was observed for the Au-loaded rutile and brookite samples,revealing the formation of plasmonic Au nanoparticles on TiO₂ surface ^{[27, 28]}. This observation illustrates the presence of surface plasmon resonance effect in the synthesized Au-TiO₂ samples.

Fig. 2. UV-Vis spectra of the TiO2 samples. (a,c) Rutile; (b,d) Brookite.

The morphologies of the prepared M-TiO₂ samples were studied by TEM to analyze the distribution and size of Au,Pd,and Pt. Fig. 3 shows that the ultra small uniform Au,Pd,and Pt nanoparticles were evenly distributed on the surface of rutile and brookite. Specifically,the M-Rutile samples featured uniform Au,Pd and Pt nanoparticles with average sizes of 1.3-1.5 nm. In contrast,the M-Brookite samples featured slightly larger metal particles with average sizes of 1.9-2.1 nm. It should be noted that no agglomerated particles were observed,indicating that the photo-deposition method,which does not require a calcination step,employed in the present study can effectively avoid aggregation of noble metal nanoparticles. Accordingly,the formation of uniform ultrafine nanoparticles with a narrow size distribution could be achieved on the surface of TiO₂. For direct comparison,some relevant physicochemical properties of the prepared M-TiO₂ samples are summarized in Table 1.

Fig. 3. TEM images of the M-TiO2 samples and associated size distribution histograms of the metal nanoparticles loaded on rutile or brookite TiO2.

Table 1 Physicochemical properties of the prepared M-TiO2 samples. Sample Size of M particles a (nm) Loading of M particles b (wt%) SBET (m2/g) Rutile — — 58.2 Au-Rutile 1.4 0.96 56.4 Pd-Rutile 1.5 0.94 57.3 Pt-Rutile 1.5 0.93 56.8 Brookite — — 53.4 Au-Brookite 2.1 0.95 52.7 Pd-Brookite 2.0 0.96 51.8 Pt-Brookite 1.9 0.92 52.3 a Estimated by TEM observations; b Determined by ICP-AES.

Table 1 Physicochemical properties of the prepared M-TiO2 samples.

The photocatalytic activities of TiO₂ and M-TiO₂ samples were investigated through photocatalytic reforming of methanol,and the results are shown in Fig. 4. As observed,both bare rutile and brookite TiO₂ samples displayed some level of activity toward hydrogen production from photocatalytic reforming of methanol; however,the yield of hydrogen was relatively low within the time-on-stream (TOS) of 5 h studied. After co-catalyst loading,the photocatalytic activity of TiO₂ greatly improved. Pt-TiO₂ displayed the highest photocatalytic activity among all samples studied,followed by Pd-TiO₂ and then Au-TiO₂. This trend was consistent with results of previous reports ^{[20, 29]}. The enhanced photocatalytic activity was attributed to the presence of the noble metal particles that can act as traps for the photogenerated electrons and reduce the extent of recombination between the electrons and holes. It is well known that Pt is an excellent co-catalyst for water splitting,exhibiting a lower overpotential and better activity than Pd and Au toward hydrogen evolution. Furthermore,for the same type of noble metal deposition,similar photocatalytic activity trends were observed over the rutile and brookite supports.

Fig. 4. Photocatalytic activity of hydrogen production over the different TiO2 samples. (a) Rutile and M-Rutile; (b) Brookite and M-Brookite.

The stability of a catalyst is a very important issue for practical application in photocatalytic hydrogen production from water splitting. Thus,the recyclability of the most active samples,i.e.,Pt-Rutile and Pt-Brookite,was investigated; three successive runs (5 h per run) were conducted. The results are shown in Fig. 5. After each run,the produced hydrogen was removed by evacuation. As observed,Pt-TiO₂ displayed good recyclability,and no significant decrease in the photocatalytic activity after three cycles was observed. Furthermore,the TEM observations (not shown here) confirmed that the morphology of the Pt clusters on TiO₂ remained unchanged following the successive catalytic runs. Thus,the experimental results demonstrate that the prepared Pt-TiO₂ samples exhibit a stable activity toward photocatalytic hydrogen production.

Fig. 5. Recycling of Pt-TiO2 catalysts in the hydrogen production process. (a) Pt-Rutile; (b) Pt-Brookite.

Hydrogen production from photocatalytic reforming of methanol not only depends on the light-harvesting properties of photocatalysts,but also relies highly on the properties of the co-catalyst. In the present study,rutile and brookite were used as the intrinsic photocatalysts,and Pt,Pd,or Au was used as co-catalysts. For the same intrinsic photocatalyst,the morphologies and nature of the co-catalyst would directly influence hydrogen production. Considering that the average particle sizes,particle distribution,and surface areas were comparable across the samples studied (for a given semiconductor support) (Table 1),the properties of the co-catalyst,e.g.,work function and hydrogen overvoltage,are thought to be the key factors controlling the photocatalytic activity.

Generally,photo-induced holes (h⁺) and electrons (e⁻) would rapidly recombine in excited bare TiO₂,consequently resulting in a low quantum yield and poor efficiency of the photocatalytic reactions ^[22]. As the work functions of the co-catalysts (Pt 5.7 eV,Pd 5.1 eV,Au 5.1 eV) are larger than that of TiO₂ (4.2 eV) ^{[19, 30, 31]},the photo-induced e⁻ will transit from the TiO₂ matrix to the co-catalyst until their Fermi levels are aligned. This phenomenon can effectively suppress the recombination of e⁻ and h⁺ owing to the formation of Schottky barriers at the M/TiO₂ interface,and promote the evolution of H₂. Among the co-catalysts examined in the present study,Pt exerted the greatest effect on hydrogen evolution,followed by Pd and Au. The obtained photocatalytic activity can be further explained by determining the hydrogen overvoltage (HOV) values of the electrodes composed of the different metals studied herein,as shown in Table 2. An inverse correlation between the hydrogen formation rates and HOV values is expected. In other words,the higher the HOV,the more difficult is the reduction of protons (H⁺) by electrons. Therefore,Pt which features the lowest HOV value,can effectively reduce H⁺ by electrons,and thus exhibits the best photocatalytic activity,which is in agreement with previous reports. These results indicate that the co-catalyst loaded on TiO₂ acted as sites for H⁺ reduction (hydrogen formation). Furthermore,this phenomenon was not influenced by the nature of the semiconductor photocatalyst,i.e. similar photocatalytic activity trends were achieved over rutile and brookite.

Table 2 List of LEDs used for the removal of dyes. Metal Work function (eV) [30] Hydrogen overvoltage (eV) [31] Pt 5.7 0.01-0.09 Pd 5.1 0.04 Au 5.1 0.18

Table 2 List of LEDs used for the removal of dyes.

The separation and transfer of electron-hole pairs under irradiation are illustrated in Scheme 1. Under UV light irradiation,the TiO₂ matrix is excited,and the excited electrons transit from the VB of TiO₂ to the CB,of TiO₂ forming e⁻-h⁺ pairs. Generally,the e⁻ and h⁺ will recombine quickly,and only a small fraction of e⁻ and h⁺ participate in the photocatalytic reaction,resulting in low photocatalytic efficiencies. In the presence of co-catalysts,the photogenerated electrons can transit to the metal nanoparticles to participate in the reduction reaction. According to a previous study,extremely small noble metal particles with sizes of less than 1.0 nm have higher energy band separation owing to quantum confinement,thereby preventing electron transfer from the CB of TiO₂ to metal ^[32]. As the particle sizes of the co-catalysts prepared in the present study are in the range of 1.3-2.1 nm,and the new Fermi energy levels of the M-TiO₂ samples are lower than that of the energy level of the bottom of the CB of TiO₂,the photo-excited electrons could effectively transit from the TiO₂ matrix to the metal nanoparticles driven by the above potential energy. At the metal-semiconductor interface,Schottky barriers can form and serve as efficient electron traps,thereby retarding photogenerated electron-hole recombination and increasing the photocatalytic activity of the M-TiO₂ samples.

Scheme 1. Schematic diagram of H2 production mechanism over M-TiO2 samples.

To investigate the efficiency of charge carrier trapping,migration,and transfer in the TiO₂ samples,room temperature emission PL spectra of the samples were recorded,and the results are shown in Fig. 6. As observed,after co-catalysts loading,the PL intensities decreased distinctly. And the changes in the intensity of the PL signals correlated well with the observed photocatalytic activity (i.e.,a lower PL signal intensity corresponds to a higher photocatalytic activity). These results indicate that the loaded co-catalyst can act as a trapping site,capturing photogenerated electrons from the CB and effectively inhibiting the recombination of the photogenerated electron-hole pairs ^[25]. Furthermore,the peak positions of the M-TiO₂ samples and pure TiO₂ were comparable,thus indicating that co-catalysts loading did not lead to new light-emitting phenomena,however,only affected the intensity of the PL signals.

Fig. 6. PL spectra of Rutile (a) and M-Rutile and Brookite and M-Brookite (b) TiO2 samples recorded at room temperature.

Ultrafine noble metal (Pt,Pd,Au) nanoparticles with sizes of ~1.0-2.0 nm were successfully loaded on rutile and brookite TiO₂ surface through a simple photo-deposition strategy under high vacuum conditions. Compared with bare rutile and brookite TiO₂,the prepared M-TiO₂ photocatalysts exhibited remarkable photocatalytic activity toward hydrogen evolution. This enhancment greatly related to the work function and HOV of the co-catalysts. Higher co-catalyst work functions enabled effective transfer of the photo-induced e⁻ from the TiO₂ matrix to the co-catalyst and suppression of the recombination of the e⁻ and h⁺ owing to the formation of Schottky barriers at the M/TiO₂ interface. In contrast,the co-catalysts in the M-TiO₂ samples acted as sites for H⁺ reduction,and the hydrogen formation rates inversely correlated to the HOV values. Furthermore,for the same type of noble metal deposition,similar photocatalytic activity trends were observed over rutile and brookite.