Enhanced Hydrogen Production from Ethanol Photoreforming by Site-Specific Deposition of Au on Cu2O/TiO2 p-n Junction

Hydrogen production by photoreforming of biomass-derived ethanol is a renewable way of obtaining clean fuel. We developed a site-specific deposition strategy to construct supported Au catalysts by rationally constructing Ti3+ defects inTiO2 nanorods and Cu2O-TiO2 p-n junction across the interface of two components. The Au nanoparticles (~2.5 nm) were selectively anchored onto either TiO2 nanorods (Au@TiO2/Cu2O) or Cu2O nanocubes (Au@Cu2O/TiO2) or both TiO2 and Cu2O (Au@TiO2/Cu2O@Au) with the same Au loading. The electronic structure of supported Au species was changed by forming Au@TiO2 interface due to the adjacent Ti3+ defects and the associated oxygen vacancies while unchanged in Au@Cu2O/TiO2 catalyst. The p-n junction of TiO2/Cu2O promoted charge separation and transfer across the junction. During ethanol photoreforming, Au@TiO2/Cu2O catalyst possessing both the Au@TiO2 interface and the p-n junction showed the highest H2 production rate of 8548 μmol gcat−1 h−1 under simulated solar light, apparently superior to both Au@TiO2 and Au@Cu2O/TiO2 catalyst. The acetaldehyde was produced in liquid phase at an almost stoichiometric rate, and C−C cleavage of ethanol molecules to form CH4 or CO2 was greatly inhibited. Extensive spectroscopic results support the claim that Au adjacent to surface Ti3+ defects could be active sites for H2 production and p-n junction of TiO2/Cu2O facilitates photo-generated charge transfer and further dehydrogenation of ethanol to acetaldehyde during the photoreforming.


Introduction
Hydrogen is extensively used in various industrial processes, e.g., in the petrochemical industry, metallurgy, fine chemical engineering, etc. [1]. Hydrogen, as a clean and renewable fuel, has aroused tremendous attention from both academic and industrial perspectives in the past decades because its energy-extraction process produces only water as a byproduct and emits no greenhouse gases, e.g., CO 2 or any pollutants [2,3]. At present, industrial production of hydrogen depends predominantly on steam reforming of CH 4 , an energy-consuming process accompanied by CO 2 emissions. Therefore, it is highly desirable to develop efficient, economical and energy-neutral processes for sustainable H 2 production.
Ethanol could be produced in a sustainable way from huge-amount, low-grade biomass, e.g., lignocelluloses and agriculture waste besides from the conventional petrochemical route [4]. X-ray diffraction patterns of the samples indicate that TiO2 exist in the form of anatase phase (JCPDS No. . The second phase in the samples with p-n junction can be indexed to cubic Cu2O ( Figure S1). The diffraction related to Au phase is not observed in the Au-supported samples because of the low loading of Au (~0.85 wt%) and/or high dispersion of small-sized Au particles (~2.5 nm) on the surface of the support. X-ray photoelectron spectra (XPS) were used to analyze the chemical valence and electronic structure of each element. Two peaks were fitted at binding energy (B.E.) of 457.9 and 458.3 eV in the Ti 2p3/2 region of Cu2O/TiO2 and Au-supported catalysts, corresponding to Ti 3+ and Ti 4+ species, respectively (Figure 2A) [22][23][24]. The B.E. values hardly shift among these samples. However, the peak area ratio of Ti 3+ to Ti 4+ is lower when Au is deposited on TiO2 i.e., Au@TiO2/Cu2O and Au@TiO2/Cu2O@Au (0.96 and 1.03), compared to that of Au-free X-ray diffraction patterns of the samples indicate that TiO 2 exist in the form of anatase phase (JCPDS No. . The second phase in the samples with p-n junction can be indexed to cubic Cu 2 O ( Figure S1). The diffraction related to Au phase is not observed in the Au-supported samples because of the low loading of Au (~0.85 wt%) and/or high dispersion of small-sized Au particles (~2.5 nm) on the surface of the support. X-ray photoelectron spectra (XPS) were used to analyze the chemical Catalysts 2020, 10, 539 4 of 18 valence and electronic structure of each element. Two peaks were fitted at binding energy (B.E.) of 457.9 and 458.3 eV in the Ti 2p 3/2 region of Cu 2 O/TiO 2 and Au-supported catalysts, corresponding to Ti 3+ and Ti 4+ species, respectively ( Figure 2A) [22][23][24]. The B.E. values hardly shift among these samples. However, the peak area ratio of Ti 3+ to Ti 4+ is lower when Au is deposited on TiO 2 i.e., Au@TiO 2 /Cu 2 O and Au@TiO 2 /Cu 2 O@Au (0.96 and 1.03), compared to that of Au-free TiO 2 /Cu 2 O and Au deposition on Cu 2 O i.e., Au@Cu 2 O/TiO 2 (1.17 and 1.11) ( Table 1). This indicates that the site-specific deposition of Au differs at defect-rich TiO 2 and Cu 2 O, leading to a variation to Ti 3+ content when Au is deposited to TiO 2 . In our previous studies, the ratio of Ti 3+ :Ti 4+ increased as the amount of NaBH 4 increased. When the ratio reaches around 1, even when increasing the amount of NaBH 4 , the ratio will not increase any more, which suggests that it is approaching saturation of Ti 3+ :Ti 4+ ratio [25].
X-ray diffraction patterns of the samples indicate that TiO2 exist in the form of anatase phase (JCPDS No. . The second phase in the samples with p-n junction can be indexed to cubic Cu2O ( Figure S1). The diffraction related to Au phase is not observed in the Au-supported samples because of the low loading of Au (~0.85 wt%) and/or high dispersion of small-sized Au particles (~2.5 nm) on the surface of the support. X-ray photoelectron spectra (XPS) were used to analyze the chemical valence and electronic structure of each element. Two peaks were fitted at binding energy (B.E.) of 457.9 and 458.3 eV in the Ti 2p3/2 region of Cu2O/TiO2 and Au-supported catalysts, corresponding to Ti 3+ and Ti 4+ species, respectively ( Figure 2A) [22][23][24]. The B.E. values hardly shift among these samples. However, the peak area ratio of Ti 3+ to Ti 4+ is lower when Au is deposited on TiO2 i.e., Au@TiO2/Cu2O and Au@TiO2/Cu2O@Au (0.96 and 1.03), compared to that of Au-free TiO2/Cu2O and Au deposition on Cu2O i.e., Au@Cu2O/TiO2 (1.17 and 1.11) ( Table 1). This indicates that the site-specific deposition of Au differs at defect-rich TiO2 and Cu2O, leading to a variation to Ti 3+ content when Au is deposited to TiO2. In our previous studies, the ratio of Ti 3+ :Ti 4+ increased as the amount of NaBH4 increased. When the ratio reaches around 1, even when increasing the amount of NaBH4, the ratio will not increase any more, which suggests that it is approaching saturation of Ti 3+ :Ti 4+ ratio [25].  The O 1s spectra were deconvoluted to three peaks at 530.0 eV, 532.0 eV and 533.5 eV (Table 2), respectively, which are assigned to lattice oxygen (O L ), adsorbed oxygen adjacent to surface oxygen vacancy (O V ) and surface chemisorbed or dissociated oxygen species (O C ) ( Figure 2B) [25]  Au 4f spectra of each Au-supported catalyst were shown and compared in Figure 2C. The main Au 4f peaks are located at 87.8 eV (4f 5/2 ) and 84.0 eV (4f 7/2 ). The fitted peak appears at B.E. of 84.0 eV in the Au 4f 7/2 region of Au@Cu 2 O/TiO 2 , which is assigned to Au 0 species [27]. In contrast, when Au is deposited on both TiO 2 and Cu 2 O, the B.E. shifts to a lower value of 83.7 eV, with a negative shift of 0.3 eV ( Table 3). The negative shift of B.E. hints at the increased electron density of Au 0 species on Au@TiO 2 /Cu 2 O@Au compared to that on Au@Cu 2 O/TiO 2 . The B.E. continually shifts to a lower value of 83.4 eV when Au is only deposited on TiO 2 . The more negative shift (−0.6 eV) indicates higher electron density of Au because of strong metal-support interaction (SMSI) between Au nanoparticles and defect-rich TiO 2 in Au@TiO 2 /Cu 2 O [28,29]. The negatively charged Au adjacent to Ti 3+ sites associated with oxygen vacancies could be active sites for ethanol reforming [25]. The higher electron density of Au on Au@TiO 2 /Cu 2 O facilitates proton reduction to hydrogen during ethanol photoreforming. XPS spectra were carried out to determine the chemical states of Cu species in the catalysts. Figure 3A shows the Cu 2p core level spectra. The binding energies at 932.1 eV and 951.9 eV, respectively, in Cu 2p 3/2 and Cu 2p 1/2 region were assigned to Cu + species [30]. The binding energies hardly shift among these catalysts. There is no shake-up satellite peak in the Cu 2p spectra. The absence of the satellite peak excludes the existence of Cu 2+ species [31,32]. However, the Cu + and Cu 0 species cannot be distinguished in Cu XPS spectra because the binding energies assigned to Cu + and Cu 0 are very close to each other. The difference in binding energies is merely around 0.1-0.2 eV [33,34]. Consequently, Cu LMM Auger spectra were analyzed to distinguish between Cu 0 and Cu + species ( Figure 3B). The spectra were fitted to two peaks at 917.0 eV and 919.0 eV, respectively, which is assigned to Cu + and Cu 0 species [35]. The Cu + is the dominating species on the basis of Auger spectra. The coexistence of Cu + and Cu 0 species in Cu 2 O nanocubes or nanoparticles is also observed in the previous literatures [36]. There is no shift in the kinetic energies regardless of Au deposition on either Cu 2 O or TiO 2 . This is an indication that the chemical states of Cu 2 O changes little in these catalysts.
Catalysts 2020, 10, 539 6 of 18 [33,34]. Consequently, Cu LMM Auger spectra were analyzed to distinguish between Cu 0 and Cu + species ( Figure 3B). The spectra were fitted to two peaks at 917.0 eV and 919.0 eV, respectively, which is assigned to Cu + and Cu 0 species [35]. The Cu + is the dominating species on the basis of Auger spectra. The coexistence of Cu + and Cu 0 species in Cu2O nanocubes or nanoparticles is also observed in the previous literatures [36]. There is no shift in the kinetic energies regardless of Au deposition on either Cu2O or TiO2. This is an indication that the chemical states of Cu2O changes little in these catalysts. The electron paramagnetic resonance (EPR) spectra are shown in Figure 4A. An intense EPR signal appears at a g value of 2.003 for Cu2O/TiO2. This signal is usually assigned to surface O-species, formed via the interaction of dioxygen and Ti 3+ defects in TiO2 [24,[37][38][39]. The EPR signal appears at the same g value for Au@Cu2O/TiO2, Au@TiO2/Cu2O@Au and Au@TiO2/Cu2O. The intensity of the signal in Au-supported catalysts is lower than that in un-supported Cu2O/TiO2. The intensity of Orelated EPR signal (g = 2.003) could be associated with the Ti 3+ /Ti 4+ ratio in TiO2. The lower signal intensity in Au-supported catalysts indicates the smaller Ti 3+ /Ti 4+ ratio. The Au@TiO2/Cu2O has the smallest Ti 3+ /Ti 4+ ratio of 0.96 among these samples, showing the lowest signal intensity of g value. This is consistent with the results presented in XPS Ti 2p analyses ( Table 1).
The digital photos of the catalysts display that all are in black color ( Figure 4B). The UV-Vis absorption spectra were compared in Figure 4C. Compared to the spectrum of TiO2, the Cu2O/TiO2 junction shows stronger absorption in the visible light region besides in the UV regime. The Ausupported catalysts possess more intensive absorption throughout the visible light wavelength. The absorption band (~560 nm) associated with Au surface plasmon resonance (SPR) is not clearly observed within the UV-vis absorption spectra of Au-supported catalysts because of the intense visible absorption of Cu2O/TiO2 itself [40]. The fluorescence emission spectra of TiO2 and Cu2O/TiO2 showed a broad band peaked at 423 nm under the excitation of 350 nm ( Figure 4D). The latter has lower emission intensity, indicating that the carrier recombination is inhibited due to the existence of Cu2O/TiO2 junction. After Au is deposited on either TiO2 or Cu2O, the emission intensity of supported The electron paramagnetic resonance (EPR) spectra are shown in Figure 4A. An intense EPR signal appears at a g value of 2.003 for Cu 2 O/TiO 2 . This signal is usually assigned to surface O-species, formed via the interaction of dioxygen and Ti 3+ defects in TiO 2 [24,[37][38][39]. The EPR signal appears at the same g value for Au@Cu 2 O/TiO 2 , Au@TiO 2 /Cu 2 O@Au and Au@TiO 2 /Cu 2 O. The intensity of the signal in Au-supported catalysts is lower than that in un-supported Cu 2 O/TiO 2 . The intensity of O-related EPR signal (g = 2.003) could be associated with the Ti 3+ /Ti 4+ ratio in TiO 2 . The lower signal intensity in Au-supported catalysts indicates the smaller Ti 3+ /Ti 4+ ratio. The Au@TiO 2 /Cu 2 O has the smallest Ti 3+ /Ti 4+ ratio of 0.96 among these samples, showing the lowest signal intensity of g value. This is consistent with the results presented in XPS Ti 2p analyses ( Table 1).
The digital photos of the catalysts display that all are in black color ( Figure 4B). The UV-Vis absorption spectra were compared in Figure 4C. Compared to the spectrum of TiO 2 , the Cu 2 O/ TiO 2 junction shows stronger absorption in the visible light region besides in the UV regime. The Au-supported catalysts possess more intensive absorption throughout the visible light wavelength. The absorption band (~560 nm) associated with Au surface plasmon resonance (SPR) is not clearly observed within the UV-vis absorption spectra of Au-supported catalysts because of the intense visible absorption of Cu 2 O/TiO 2 itself [40]. The fluorescence emission spectra of TiO 2 and Cu 2 O/TiO 2 showed a broad band peaked at 423 nm under the excitation of 350 nm ( Figure 4D). The latter has lower emission intensity, indicating that the carrier recombination is inhibited due to the existence of Cu 2 O/TiO 2 junction. After Au is deposited on either TiO 2 or Cu 2 O, the emission intensity of supported catalysts further decreases. It indicates that forming the interface of Au and semiconducting oxides facilitates the separation of photogenerated electrons and photogenerated holes by charge transfer across the metal-oxide interface [22,41].
Catalysts 2020, 10, x FOR PEER REVIEW 7 of 17 catalysts further decreases. It indicates that forming the interface of Au and semiconducting oxides facilitates the separation of photogenerated electrons and photogenerated holes by charge transfer across the metal-oxide interface [22,41]. The photocurrent tests were carried out to evaluate the charge separation within the catalysts. The photocurrent−time traces of photoelectrodes were obtained in the photoelectrochemical cell at a bias of 1.23 V vs. RHE under chopped AM 1.5G light illumination ( Figure 5). The Au loading on Cu2O/TiO2 enhances the photocurrent compared to Cu2O/TiO2 itself, which indicates the improved charge separation across the metal/oxide interface. The Au@TiO2 electrode shows a higher photocurrent than Au@Cu2O/TiO2. The interface of Au and TiO2 is more favorable to charge separation than that of Au and Cu2O due to the lower conduction band potential of TiO2 than that of Cu2O [42]. In addition, the existence of Ti 3+ defects associated with adjacent oxygen vacancies intensifies the interaction of Au and TiO2 [25]. The Au@TiO2/Cu2O electrode exhibits the highest photocurrent among all electrodes, 2.5 times higher than Au@TiO2. This hints at the contribution of TiO2/Cu2O p-n junction combined with the interface of Au and TiO2 to charge separation and transfer. The photocurrent of Au@TiO2/Cu2O is also higher than Au@TiO2/Cu2O/Au. This is an indication that the site-specific deposition of Au on TiO2 leads to charge separation more efficiently under light illumination than the random deposition of Au on both TiO2 and Cu2O. The photocurrent tests were carried out to evaluate the charge separation within the catalysts. The photocurrent−time traces of photoelectrodes were obtained in the photoelectrochemical cell at a bias of 1.23 V vs. RHE under chopped AM 1.5G light illumination ( Figure 5). The Au loading on Cu 2 O/TiO 2 enhances the photocurrent compared to Cu 2 O/TiO 2 itself, which indicates the improved charge separation across the metal/oxide interface. The Au@TiO 2 electrode shows a higher photocurrent than Au@Cu 2 O/TiO 2 . The interface of Au and TiO 2 is more favorable to charge separation than that of Au and Cu 2 O due to the lower conduction band potential of TiO 2 than that of Cu 2 O [42]. In addition, the existence of Ti 3+ defects associated with adjacent oxygen vacancies intensifies the interaction of Au and TiO 2 [25]. The Au@TiO 2 /Cu 2 O electrode exhibits the highest photocurrent among all electrodes, 2.5 times higher than Au@TiO 2 . This hints at the contribution of TiO 2 /Cu 2 O p-n junction combined with the interface of Au and TiO 2 to charge separation and transfer. The photocurrent of Au@TiO 2 /Cu 2 O is also higher than Au@TiO 2 /Cu 2 O/Au. This is an indication that the site-specific deposition of Au on TiO 2 leads to charge separation more efficiently under light illumination than the random deposition of Au on both TiO 2 and Cu 2 O. The photoreforming of ethanol was achieved using pure ethanol in an Ar atmosphere (1.4 bar) under simulated solar irradiation. The H2 production rate is contrasted on each catalyst ( Figure 6A). The rate of H2 production reaches 6932 μmol gcat −1 h −1 on Au@Cu2O/TiO2, 8.35 times higher than that on Cu2O/TiO2 (830 μmol gcat −1 h −1 , Table 4). It is apparent that proton reduction is greatly accelerated on the Au-supported catalyst for H2 production. Additionaly, the charge separation is improved by loading Au on the semiconductor, evidenced by photocurrent measurements ( Figure 5). The Au@TiO2 catalyst exhibits higher activity (7143 μmol gcat −1 h −1 ) towards H2 production than Au@Cu2O/TiO2. This suggests that constructing the interface of Au and TiO2 is more favorable than that of Au and Cu2O, which could direct the structural design of catalysts for ethanol photoreforming. The rate of H2 production on Au@TiO2/Cu2O (8548 μmol gcat −1 h −1 ) is apparently higher than that on either Au@TiO2 (7143 μmol gcat −1 h −1 ) or Au@TiO2/Cu2O@Au (7348 μmol gcat −1 h −1 ). Table S4 compares the H2 yield rates reported for TiO2-supported noble-metal catalysts with the result in this work. The p-n junction between TiO2 and Cu2O contributes the enhanced H2 production. However, the selective deposition of Au on defect-rich TiO2 is essential for higher activity because the Au-Ov-Ti 3+ sites are more active in ethanol photoreforming [25]. The quantity of Au-Ov-Ti 3+ sites is relatively less in randomly deposited Au@TiO2/Cu2O@Au than in site-specifically deposited Au@TiO2/Cu2O. As a consequence, the latter shows higher activity than the former. In addition, the lower H2 production rate of Au@Cu2O/TiO2may be due to the alloying of Au-Cu, leading to a smaller number of active sites on gold according to the previous report [43].   Table 4). It is apparent that proton reduction is greatly accelerated on the Au-supported catalyst for H 2 production. Additionaly, the charge separation is improved by loading Au on the semiconductor, evidenced by photocurrent measurements ( Figure 5). The Au@TiO 2 catalyst exhibits higher activity (7143 µmol g cat −1 h −1 ) towards H 2 production than Au@Cu 2 O/TiO 2 .
This suggests that constructing the interface of Au and TiO 2 is more favorable than that of Au and Cu 2 O, which could direct the structural design of catalysts for ethanol photoreforming. The rate of H 2 production on Au@TiO 2 /Cu 2 O (8548 µmol g cat −1 h −1 ) is apparently higher than that on either Au@TiO 2 (7143 µmol g cat −1 h −1 ) or Au@TiO 2 /Cu 2 O@Au (7348 µmol g cat −1 h −1 ). Table S4 compares the H 2 yield rates reported for TiO2-supported noble-metal catalysts with the result in this work. The p-n junction between TiO 2 and Cu 2 O contributes the enhanced H 2 production. However, the selective deposition of Au on defect-rich TiO 2 is essential for higher activity because the Au-O v -Ti 3+ sites are more active in ethanol photoreforming [25]. The quantity of Au-O v -Ti 3+ sites is relatively less in randomly deposited Au@TiO 2 /Cu 2 O@Au than in site-specifically deposited Au@TiO 2 /Cu 2 O. As a consequence, the latter shows higher activity than the former. In addition, the lower H 2 production rate of Au@Cu 2 O/TiO 2 may be due to the alloying of Au-Cu, leading to a smaller number of active sites on gold according to the previous report [43]. deposition of Au on defect-rich TiO2 is essential for higher activity because the Au-Ov-Ti 3+ sites are more active in ethanol photoreforming [25]. The quantity of Au-Ov-Ti 3+ sites is relatively less in randomly deposited Au@TiO2/Cu2O@Au than in site-specifically deposited Au@TiO2/Cu2O. As a consequence, the latter shows higher activity than the former. In addition, the lower H2 production rate of Au@Cu2O/TiO2may be due to the alloying of Au-Cu, leading to a smaller number of active sites on gold according to the previous report [43].  The products except H 2 in the gas phase, e.g., CH 4 , CO, CO 2 , are below 0.3% on all catalysts (Table 4), which clearly indicates the efficient inhibition of C-C cleavage during the photoreforming. Also, high-purity H 2 can be produced on these catalysts. For instance, the purity of H 2 is beyond 99.9% on Au@TiO 2 /Cu 2 O catalyst. Acetaldehyde is the only detectable product in the liquid phase, evidenced by GC-MS analysis ( Figure S2). The high selectivity of acetaldehyde is due to the absence of O 2 activation to peroxide intermediates and hydroxyl species on the surface of Au NPs under Ar ambient [43,44]. In addition, acetaldehyde is produced in almost stoichiometric yield with H 2 , which is consistent with the reforming pathway of ethanol via a redox process [11]. The photo-generated holes in the valence band of semiconductor oxidize ethanol to acetaldehyde and the electrons on the active sites adjacent to Au reduce protons to H 2 during the process.
The optimal Au@TiO 2 /Cu 2 O catalyst was recycled five times and the H 2 yield was hardly decreased ( Figure 6B). The structural stability of catalysts was further evidenced by HRTEM and Ti 2p, O 1s, Au 4f XPS and Cu LMM Auger analyses. The size of Au particles remains 2.5 nm, the same as that before the reactions ( Figure S3). In addition, the Au loading after reaction is not changed by ICP measurement (0.84 wt%). The Ti 3+ /Ti 4+ ratio and O V /O L ratio in the catalyst have no change before and after reactions ( Figure S4, Tables S1 and S2). The peak in the Au 4f region appears at 83.5 eV and has little shift compared to that before reactions ( Figure S4, Table S3). Additionally, the peaks in Cu 2p XPS and Cu LMM Auger spectra keep unchanged after photoreforming ( Figure S5). These findings verify that the catalyst is structurally stable and keeps activity after repeated use.
The in situ FTIR spectra of ethanol adsorption on the catalysts were first collected in the dark after adsorption of ethanol for 40 min, and then recorded every hour during photoreforming. The spectra of the three supported catalysts showed intensive absorption bands at 1000~1200 cm −1 , 1200~1500 cm −1 , 2700~3100 cm −1 , and 3600~3800 cm −1 , which are respectively assigned to the vibrations of ν(C-O), δ(CH  The band at 1755 cm −1 gradually increases with the extension of irradiation time, which corresponds to the vibration related with acetaldehyde molecule [46]. This is consistent with the increased acetaldehyde production in liquid phase with the reaction. The band shifts little among the three catalysts. The band at around 2350 cm −1 , assigned to the vibration of CO2 [47], does not appear in the FTIR spectra. This verifies that the C-C cleavage of ethanol is greatly inhibited on the three Au- The band at 1755 cm −1 gradually increases with the extension of irradiation time, which corresponds to the vibration related with acetaldehyde molecule [46]. This is consistent with the increased acetaldehyde production in liquid phase with the reaction. The band shifts little among the three catalysts. The band at around 2350 cm −1 , assigned to the vibration of CO 2 [47], does not appear in the FTIR spectra. This verifies that the C-C cleavage of ethanol is greatly inhibited on the three Au-supported catalysts during the photoreforming. Consequently, the dehydrogenation of ethanol to acetaldehyde is the dominating step, accompanied with high-purity H 2 production on these catalysts. The energy levels of semiconductors play a critical role in the activity and selectivity of ethanol photoreforming. In a catalyst with a p-n junction, the charge separation and transfer are strongly dependent on the relationships of energy levels [48]. The flat band potentials (E fb ) of the TiO 2 nanorods and Cu 2 O nanocubes were determined by electrochemical Mott-Schottky measurements at varied frequencies and room temperature. The Mott-Schottky plots of TiO 2 exhibit positive slopes, indicative of their characteristic of n-type semiconductor ( Figure 8A). The E fb values are estimated by extrapolating the linear portion of the plots measured at varied frequencies to the intercept of x-axis. It is generally considered that the bottom of the conduction band (CB) in n-type semiconductors is approximately equal to its flat band potential. [49,50] Consequently, the conduction band of TiO 2 nanorod is −0.63 V vs. Ag/AgCl. This potential was converted to −0.02 V vs. RHE according to the following formula [51].
The bandgap of TiO 2 nanorod was estimated to be 2.76 eV using the Kubelka−Munk equation, i.e., F(R) = (1 − R) 2 /2R, where R is the reflectance ( Figure S6). This value is smaller than that reported for bulk anatase TiO 2 of~3.2 eV [52]. It is documented that theTi 3+ defect-rich anatase TiO 2 exhibits narrower bandgap owing to the involvement of defect energy levels [53]. The valence band TiO 2 nanorod is calculated to be 2.74 eV from the bandgap and the conduction band level.
The Mott-Schottky plots of Cu 2 O show negative slope, an indication of p-type semiconductor ( Figure 8B). The flat band potential of 0.65 V vs. Ag/AgCl was converted to 1.26 V vs. RHE according to Formula (1). It is usually considered that the bottom of valence band (VB) in p-type semiconductors was 0.30 V more positive than the flat band potential [54,55]. The valence band level of Cu 2 O is 1.56 V vs. RHE. Based on the bandgap of Cu 2 O (2.2 eV) [56,57], the conduction band (CB) level of Cu 2 O is estimated to be −0.64 V.
Catalysts 2020, 10, x FOR PEER REVIEW 11 of 17 supported catalysts during the photoreforming. Consequently, the dehydrogenation of ethanol to acetaldehyde is the dominating step, accompanied with high-purity H2 production on these catalysts. The energy levels of semiconductors play a critical role in the activity and selectivity of ethanol photoreforming. In a catalyst with a p-n junction, the charge separation and transfer are strongly dependent on the relationships of energy levels [48]. The flat band potentials (Efb) of the TiO2 nanorods and Cu2O nanocubes were determined by electrochemical Mott-Schottky measurements at varied frequencies and room temperature. The Mott-Schottky plots of TiO2 exhibit positive slopes, indicative of their characteristic of n-type semiconductor ( Figure 8A). The Efb values are estimated by extrapolating the linear portion of the plots measured at varied frequencies to the intercept of x-axis. It is generally considered that the bottom of the conduction band (CB) in n-type semiconductors is approximately equal to its flat band potential. [49,50] Consequently, the conduction band of TiO2 nanorod is −0.63 V vs. Ag/AgCl. This potential was converted to −0.02 V vs. RHE according to the following formula [51].
The bandgap of TiO2 nanorod was estimated to be 2.76 eV using the Kubelka−Munk equation, i.e., F(R) = (1−R) 2 /2R, where R is the reflectance ( Figure S6). This value is smaller than that reported for bulk anatase TiO2 of ~3.2 eV [52]. It is documented that theTi 3+ defect-rich anatase TiO2 exhibits narrower bandgap owing to the involvement of defect energy levels [53]. The valence band TiO2 nanorod is calculated to be 2.74 eV from the bandgap and the conduction band level.
The Mott-Schottky plots of Cu2O show negative slope, an indication of p-type semiconductor ( Figure 8B). The flat band potential of 0.65 V vs. Ag/AgCl was converted to 1.26 V vs. RHE according to Formula (1). It is usually considered that the bottom of valence band (VB) in p-type semiconductors was 0.30 V more positive than the flat band potential [54,55]. The valence band level of Cu2O is 1.56 V vs. RHE. Based on the bandgap of Cu2O (2.2 eV) [56,57], the conduction band (CB) level of Cu2O is estimated to be −0.64 V. Based on the findings above, we proposed possible charge transfer pathways on the Au@TiO2/Cu2O catalyst during ethanol photoreforming. The p-n junction is formed when p-type Cu2O and n-typeTiO2 come into contact with each other (Scheme 2). The photo-generated holes are transferred from the valence band of TiO2 to that of Cu2O owing to the more positive valence band level of TiO2. The electrons generated on Cu2O under visible light irradiation are transferred to the conduction band of TiO2. In addition, the hot electrons of Au under visible light irradiation are transferred to the conduction band of TiO2 due to the SPR effect of Au. The strong interaction between Au and defect-rich TiO2 enhances the hot electron transfer from Au to the interface of Au/TiO2. As a consequence, the interface of Au and TiO2 is rich in the electrons, leading to high electron density on Based on the findings above, we proposed possible charge transfer pathways on the Au@TiO 2 /Cu 2 O catalyst during ethanol photoreforming. The p-n junction is formed when p-type Cu 2 O and n-typeTiO 2 come into contact with each other (Scheme 2). The photo-generated holes are transferred from the valence band of TiO 2 to that of Cu 2 O owing to the more positive valence band level of TiO 2 .
The electrons generated on Cu 2 O under visible light irradiation are transferred to the conduction band of TiO 2 . In addition, the hot electrons of Au under visible light irradiation are transferred to the conduction band of TiO 2 due to the SPR effect of Au. The strong interaction between Au and defect-rich TiO 2 enhances the hot electron transfer from Au to the interface of Au/TiO 2 . As a consequence, the interface of Au and TiO 2 is rich in the electrons, leading to high electron density on the Au-Ov-Ti 3+ interfacial sites. The proton reduction is promoted on the electron-rich interfacial sites, resulting in higher H 2 production rate than the Au@Cu 2 O/TiO 2 catalyst without the interfacial sites. Consequently, the dual effects of Au-O v -Ti 3+ active sites and p-n junction endow Au@TiO 2 /Cu 2 O catalyst with higher activity towards ethanol photoreforming by the site-specific deposition of Au on defect-rich TiO 2 nanorods.
Catalysts 2020, 10, x FOR PEER REVIEW  12 of 17 the Au-Ov-Ti 3+ interfacial sites. The proton reduction is promoted on the electron-rich interfacial sites, resulting in higher H2 production rate than the Au@Cu2O/TiO2 catalyst without the interfacial sites. Consequently, the dual effects of Au-Ov-Ti 3+ active sites and p-n junction endow Au@TiO2/Cu2O catalyst with higher activity towards ethanol photoreforming by the site-specific deposition of Au on defect-rich TiO2 nanorods.

Synthesis of Defect-Rich TiO2 Nanorods
The defect-rich TiO2 nanorods were synthesized by a NaBH4 reduction method according to a previous literature [22]. In a typical procedure, the mixture (TiO2/NaBH4 = 2:1 molar ratio) was continuously stirred for 30 min and treated at 350 °C for 1 h in a flow of N2. The sample was washed with diluted HCl solution and water thoroughly. The sample was named as TiO2.

Synthesis of Cu2O Nanocubes
The Cu2O nanocubes were prepared by a chemical precipitation method, modified from a literature's one [58]. In a typical run, 155 mg of CuSO4 and 400 mg of PVP were dissolved in 160 mL of deionized water to obtain a light blue solution. The pH was adjusted to 10 by dropwise addition of NaOH (1 M). An amount of 176 mg of ascorbic acid (AA) was dispersed in the solution to reduce Cu 2+ to Cu2O. The suspension was stirred for 20 min. Subsequently, the suspension was centrifuged and washed thoroughly with deionized water and ethanol to remove PVP. The product was collected after the color of solution changed from light blue to brownish yellow. The solid product was finally dried under vacuum at 60 °C.

Synthesis of Defect-Rich TiO 2 Nanorods
The defect-rich TiO 2 nanorods were synthesized by a NaBH 4 reduction method according to a previous literature [22]. In a typical procedure, the mixture (TiO 2 /NaBH 4 = 2:1 molar ratio) was continuously stirred for 30 min and treated at 350 • C for 1 h in a flow of N 2 . The sample was washed with diluted HCl solution and water thoroughly. The sample was named as TiO 2 .

Synthesis of Cu 2 O Nanocubes
The Cu 2 O nanocubes were prepared by a chemical precipitation method, modified from a literature's one [58]. In a typical run, 155 mg of CuSO 4 and 400 mg of PVP were dissolved in 160 mL of deionized water to obtain a light blue solution. The pH was adjusted to 10 by dropwise addition of NaOH (1 M). An amount of 176 mg of ascorbic acid (AA) was dispersed in the solution to reduce Cu 2+ to Cu 2 O. The suspension was stirred for 20 min. Subsequently, the suspension was centrifuged and washed thoroughly with deionized water and ethanol to remove PVP. The product was collected after the color of solution changed from light blue to brownish yellow. The solid product was finally dried under vacuum at 60 • C.

Synthesis of Cu 2 O/TiO 2
2 g of TiO 2 nanorod sample was added into the CuSO 4 and PVP solution, and the pH of the solution was adjusted to 10 with an aqueous solution of NaOH (1 M). Subsequently, AA was added to the solution and stirred for 20 min. The solid product was collected by centrifugation and washed with ethanol and water. The product was finally dried under vacuum at 60 • C.
Transmission electron microscopy (TEM) measurements were carried out on a JEOL JEM-3010 high-resolution transmission electron microscope.
Elemental analysis was performed on a Shimadzu ICPS-7500 inductively coupled plasma atomic emission spectrometer (ICP-AES).
The XPS spectra were recorded on a Thermo VG ESCALAB MK II X-ray photoelectron spectrometer at a pressure of 2 × 10 −9 Pa using Al Kα X-ray as the excitation source (1486.6 eV). The positions of all binding energies were calibrated using theC1s line at 284.6 eV.
UV-visible diffuse reflectance spectra were performed on a Shimadzu UV-3000 spectrometer equipped with an integrating sphere attachment with BaSO 4 (10 mg) as reference.
Photoluminescence (PL) emission spectra were recorded by a Hitachi F-7000 spectrofluorometer using laser excitation at a wavelength of 350 nm.
Electron paramagnetic resonance (EPR) measurements were performed on a Bruker E500 spectrometer with a 9.53 GHz X-band. The sample mass was 50 mg. The spectra were recorded in the magnet field range of 318-328 mT.
The liquid phase products were determined by GC-MS measurements using a Thermo Fisher ISQ Trace1300 and the sample volume was 1 µL.
Electrochemical impedance spectroscopy (EIS) was performed on an electrochemical workstation at an open-circuit voltage of 0.3 V vs. RHE under illumination with 10 mV amplitude of perturbation and a frequency between 1.0 kHz, 1.5 kHz and 2.0 kHz. Mott-Schottky plots were measured at room temperature in the dark.
In situ Fourier Transform Infrared Spectroscopy (FTIR) was conducted in an in situ reaction cell on a Bruker Tensor II spectrometer installed with MCT narrow-band detector. The sample was pretreated in a flow of high-purity N 2 at 100 • C for 1 h. After an initial scan as the background spectrum, ethanol was induced into the cell through a flow of N 2 for 30 min. After flowing N 2 to remove the residual ethanol vapor, the FTIR spectra were collected in the range of 4000~950 cm −1 at room temperature.

Photoreforming of Ethanol
In a typical run, 25 mg of catalyst was added to 25 mL of ethanol and ultrasonically dispersed for 15 min. The suspension was transferred into a high-pressure stainless steel reactor (volume: 50 mL) equipped with a sapphire crystal window. A flow of Ar was purged into the reactor for 30 min and the reactor was evacuated. The Ar was re-purged and the pressure in the reactor was maintained at 1.4 bar. The suspension was irradiated by a 300 W Xenon lamp equipped with an AM 1.5G filter (100 mW·cm −2 ) under magnetic stirring for 6 h. After the reaction, the catalyst was removed from the solution by filtration. The gaseous products were detected by an online gas chromatograph (GC-2014C, Shimadzu, Japan) equipped with a high-sensitivity thermal conductivity detector (TCD) and Ar was used as the carrier gas. The liquid products were analyzed by a gas chromatograph (GC-2014C, Shimadzu) equipped with a flame ionization detector (FID). The reaction rate of H 2 followed the formula below: Reaction rate of production = n production (µmol)

Recycled Use
The catalyst was separated from the solution after the reaction, washed with deionized water, and finally dried under vacuum at 60 • C for 6 h. The dried catalyst was reused in a next catalysis run under the same reaction conditions.

Conclusions
The TiO 2 /Cu 2 O-supported Au catalysts were delicately constructed by the site-specific deposition of Au on either defect-rich TiO 2 nanorods or Cu 2 O nanocubes. The selective anchoring of Au nanoparticles on TiO 2 nanorods combined with the p-n junction of TiO 2 /Cu 2 O leads to the highest activity towards ethanol photoreforming. The H 2 production rate reaches a record level of 8548 µmol g cat −1 h −1 under simulated solar light. The acetaldehyde in liquid phase was generated at an almost stoichiometric rate, which indicates the effective inhibition of C−C cleavage of ethanol to CH 4 or CO 2 . Extensive spectroscopic studies verified that Au species adjacent to Ti 3+ defects and the associated oxygen vacancies on TiO 2 nanorods activate the proton reduction to H 2 . The p-n junction between TiO 2 and Cu 2 O facilitates charge separation and transfer owing to the matching of energy levels, which accelerates photo-generated hole transfer and the dehydrogenation of ethanol to acetaldehyde on the junction. This site-specific deposition strategy could be applicable in the enhancement in other biomass hydrogen production by rationally designing the delicate structure of catalysts and maximizing the catalytic capability.

Conflicts of Interest:
The authors declare no competing financial interest.