Cu(ii) nanocluster-grafted, Nb-doped TiO2 as an efficient visible-light-sensitive photocatalyst based on energy-level matching between surface and bulk states

Department of Metallurgy and Ceramics S Engineering, Tokyo Institute of Technolog 152-8552, Japan. E-mail: mmiyauchi@ceram Research Center for Advanced Science and T Komaba, Meguro-ku, Tokyo 153-8904, Japan Graduate School of Engineering, The Univ Tokyo 113-8656, Japan Japan Science and Technology Agency (J Saitama 332-0012, Japan † Electronic supplementary information ( Nb in the samples derived from ICP-AE samples. Fig. S1, XRD patterns of N NbxTi1 xO2. Fig. S3, TEM images and t Fig. S4, full-scale XPS spectra of the prep spectra of NbxTi1 xO2. Fig. S6, ESR sp core-level spectra of NbxTi1 xO2, Nb2O5 Fig. S8, valence band XPS spectra of th samples. Fig. S10, light source for the vi of the quantum efficiency calculations. Cu(II)–FexTi1 xO2 and Cu(II)–WxTi1 xO2 w comparative evaluation of CO2 generatio doping densities. Fig. S14, comparativ Cu(II)–NbxTi1 xO2 with different amoun preparation of Cu(II)–TiO2@NbxTi1 xO2. Cu(II)–NbxTi1 xO2 and Cu(II)–TiO2@ 10.1039/c4ta02211d Cite this: J. Mater. Chem. A, 2014, 2, 13571


Introduction
Titanium dioxide (TiO 2 ) has attracted considerable recent attention as an efficient photocatalyst for applications such as water splitting, organic decomposition, and solar cells. 1 However, because TiO 2 is a wide band-gap semiconductor, with band gap values of 3.2 and 3.0 eV for the anatase and rutile forms, respectively, it can only be activated under ultraviolet (UV) light irradiation, thereby limiting its practical applications. 2 To increase the utilization of solar and indoor light sources, TiO 2 has been doped with various transition metal cations, such as Cr, Mn, Fe, Pb, and Cu, and anions, including N, C, and S, in an attempt to extend the light absorption capacity of TiO 2 into the visible light region. 3 Despite extensive research effort being made to modify the properties of TiO 2 , most doped TiO 2 systems remain unsuitable for practical use because their quantum efficiencies (QEs) under visible light are too low to support the efficient photocatalytic reactions. 3,4 In the case of cationic doping, the increase in visible light sensitivity is mainly caused by impurity levels in the forbidden band, which act as recombination centers for photogenerated charge carriers. 2a-c In contrast, anionic doping of TiO 2 generally introduces isolated levels above the valence band (VB) that contain a Department of Metallurgy and Ceramics Science, Graduate School of Science and Engineering, Tokyo Institute of Technology, 2-12-1 Ookayama, Meguro-ku, Tokyo 152-8552, Japan. E-mail: mmiyauchi@ceram.titech.ac.jp b Research Center for Advanced Science and Technology, The University of Tokyo, 4-6-1 Komaba,  c Graduate School of Engineering, The University of Tokyo, 7-3-1 Hongo, Bunkyo-ku, Tokyo 113-8656, Japan d Japan Science and Technology Agency (JST), ACT-C, 4-1-8 Honcho Kawaguchi, Saitama 332-0012, Japan † Electronic supplementary information (ESI) available: Table S1, the amount of Nb in the samples derived from ICP-AES. holes generated with lower oxidation power and mobility than those in the VB, 4 resulting in low photocatalytic performance.
Recently, the surface modication of TiO 2 with Cu(II) or Fe(III) nanoclusters was shown to increase its visible-light sensitivity without inducing impurity levels in the band gap. 5 Under visible-light irradiation, electrons in the VB of TiO 2 are excited to these surface nanoclusters through an interfacial charge transfer (IFCT) process. 6 Simultaneously, the excited electrons are consumed in the multi-electron reduction of oxygen mediated by the nanoclusters. 7 Therefore, Cu(II) or Fe(III) nanocluster-graed TiO 2 exhibits a high QE under visible light. However, the capacity of this photocatalytic system for visiblelight absorption is limited because IFCT only occurs at TiO 2 particle/nanocluster interfaces.
Surface nanoclusters have also been demonstrated to increase the visible-light activities of doped semiconductors. 5d,f,8 For example, Ti 3+ self-doped TiO 2 , which is inactive even under UV light irradiation, is converted into an efficient visible-lightsensitive photocatalyst by the surface graing of Cu(II) or Fe(III) nanoclusters. 5d Although this nding indicates that photogenerated electrons are transferred from the doped levels to the surface Cu(II) or Fe(III) nanoclusters, the QEs of metal-doped TiO 2 photocatalysts cannot exceed that of Cu(II)-or Fe(III)-grafted TiO 2 , because the charge transfer is limited from doped levels to surface nanoclusters. Thus, efficient charge transfer between the doped levels and surface nanoclusters is critical for the photocatalytic performance of these systems. To achieve this criterion, the concept of energy level matching between the doped ions and surface-graed nanoclusters has been recently examined using Fe(III) ion-doped and -graed TiO 2 as a model system. 5e The Fe(III)-doped and -graed TiO 2 system exhibits a QE of 47.3% and a reaction rate of 0.69 mmol h À1 , suggesting the feasibility of this approach.
Compared with Fe(III)-based compounds, cupreous compounds have the added advantage of intrinsic anti-pathogenic properties under both dark and light conditions, 9 as was demonstrated for a Cu-deposited thin lm of TiO 2 . 9a Despite this advantageous characteristic, the visible-light absorption and quantum efficiencies of Cu(II)-graed TiO 2 are less than those of Fe(III)-graed TiO 2 . 5a,b Thus, for the development of Cu(II)-based photocatalysts that are suitable for indoor applications aimed at reducing chemical irritation or spread of infectious disease, it is necessary to markedly increase the visible-light sensitivity and photocatalytic efficiency of these materials. 9 Based on the concept of energy level matching, we hypothesized that this could be achieved by modifying TiO 2 to have similar interband energy levels to the redox potential of Cu(II) nanoclusters.
Niobium (Nb)-doped TiO 2 has attracted recent attention due to its intriguing electronic properties and potential applications in transparent conductive oxide, photovoltaic cells and sensors. Theoretical and experimental studies have shown that the doping of TiO 2 with Nb ions generates an energy level at 0.1 to 0.2 eV below the conduction band (CB), 10 a value that closely matches the redox potential of Cu 2+ /Cu + (0.16 V vs. SHE, pH ¼ 0). 5a,b However, Nb-doped TiO 2 exhibits low efficiency for photocatalytic reactions because pentavalent Nb ions are substituted for tetravalent Ti leading to the bulk formation of Ti 3+ species that serve as recombination centers. 5d,10 Instead, Cu(II) nanocluster-graed (Cu(II)-TiO 2 ) can increase its visiblelight activity without introduction of impurity levels in the band gap. However, the visible-light absorption of Cu(II)-TiO 2 is limited because IFCT proceeds only at the bulk/nanocluster interface.
In the present study, we investigated the coupling of the bulk doping of Nb ions and surface graing of Cu(II) clusters on TiO 2 . We speculated that under visible light irradiation, excited electrons in the doped Nb levels would transfer to the Cu(II) nanoclusters owing to their similar energy levels ( Fig. 1), thereby increasing the visible-light sensitivity and QE of TiO 2 photocatalysts.

Synthesis of Nb x Ti 1Àx O 2
Nb-doped TiO 2 (Nb x Ti 1Àx O 2 ) nanocomposites were prepared using a simple impregnation method with commercial TiO 2 (rutile phase, 15 nm grain size, 90 m 2 g À1 specic surface area; MT-150A, Tayca Co.) as the starting material. 5e In a typical synthesis, 1.5 g TiO 2 powder was mixed with 10 mL ethanol to form a TiO 2 suspension, to which niobium(V) chloride (NbCl 5 , Wako, 95%) was added as the source of Nb at a weight fraction to TiO 2 of 0.1%. The resulting suspension was stirred for 0.5 h in a vial reactor and was then dried at room temperature. The obtained residue was calcined at 950 C for 3 h to form Nb x Ti 1Àx O 2 which was then further treated with a 6 M HCl aqueous solution at 90 C for 3 h under stirring. The products were ltered through a 0.025 mm membrane lter (Millipore) and then washed with sufficient amounts of distilled water. Nb x Ti 1Àx O 2 was obtained as a clear powder and was dried at 110 C for 24 h before being ground to a ne powder using an agate mortar and pestle for the preparation of Cu(II) nanocluster-graed Nb x Ti 1Àx O 2 (Cu(II)-Nb x Ti 1Àx O 2 ) nanocomposites. Pure TiO 2 was obtained using the same annealing and acid treatment process without adding NbCl 5 solution, and was used to prepare Cu(II)-TiO 2 nanocomposites.

Modication of Nb x Ti 1Àx O 2 with Cu(II) nanoclusters
The graing of Cu(II) nanoclusters onto Nb x Ti 1Àx O 2 was performed using an impregnation method. 5a,b Briey, 1 g Nb x Ti 1Àx O 2 powder was dispersed in 10 mL distilled water and CuCl 2 $2H 2 O (Wako, 99.9%) was then added to the TiO 2 suspension as a source of Cu. The weight fraction of Cu relative to TiO 2 was set to 0.1%. The resulting suspension was heated at 90 C under stirring for 1 h in a vial reactor. The products were ltered through a 0.025 mm membrane lter (Millipore) and then washed with sufficient amounts of distilled water. The obtained residue was dried at 110 C for 24 h and subsequently ground to a ne powder using an agate mortar and pestle. Cu(II)-TiO 2 was also prepared by the same impregnation method.

Sample characterization
The structural characteristics of the prepared nanocomposites were measured by powder X-ray diffraction (XRD) at room temperature on a Rigaku D/MAX25000 diffractometer with a copper target (l ¼ 1.54056Å). Electron spin resonance (ESR) spectra were recorded on a Bruker ESP350E spectrometer. Elemental analyses of the samples were performed using an inductively coupled plasma-atomic emission spectrometer (ICP-AES; P-4010, Hitachi). UV-visible absorption spectra were obtained by the diffuse reection method using a UV-2550 spectrometer (Shimadzu). The morphologies of the prepared Nb x Ti 1Àx O 2 nanocomposites were investigated by scanning electron microscopy (SEM) using a Hitachi SU-8000 apparatus. Transmission electron microscopy (TEM) images were observed by the Hitachi HF-2000 instrument using acceleration voltage at 200 kV. The specic surface areas of the samples were determined from the nitrogen absorption data at liquid nitrogen temperature using the Barrett-Emmett-Teller (BET) technique. The samples were degassed at 200 C and the pressure was kept below 100 mTorr for a minimum of 2 h prior to analysis using a Micromeritics VacPrep 061 instrument. Surface compositions and valence band X-ray photoelectron spectra were studied by X-ray photoelectron spectroscopy (XPS; model 5600, Perkin-Elmer). The binding energy data were calibrated with reference to the C 1s signal at 284.5 eV.

Evaluation of photocatalytic properties
The photocatalytic activities of the prepared nanocomposites were evaluated by monitoring the decomposition of gaseous 2-propanol (IPA) under visible-light irradiation. A Xe lamp (LA-251Xe, Hayashi Tokei) equipped with L-42, B-47, and C-40C glass lters (Asahi Techno-Glass) was used as a source of visible light (420-530 nm, 1 mW cm À2 ). The light intensity was measured using a spectrum-radiometer (USR-45D, Ushio Co.) and was adjusted to 1 mW cm À2 . A 500 mL cylindrical glass vessel was used as the photocatalysis reactor. To perform the photocatalytic experiments, 300 mg photocatalyst powder was evenly spread on the bottom of a circular glass dish (area of 5.5 cm 2 ) that was mounted in the middle of the vessel reactor. The vessel was sealed with a rubber O-ring and a quartz cover, evacuated, and lled with fresh synthetic air. To eliminate organic contaminants from the sample surface, the vessel was illuminated with a Xe lamp (LA-251Xe) until the CO 2 generation rate was less than 0.02 mmol per day. The vessel was then evacuated and relled with fresh synthetic air. The pressure inside the vessel was kept at $1 atm. To begin the photocatalytic measurement, 300 ppmv ($6 mmol) of gaseous IPA was injected into the vessel, which was then incubated in the dark for 12 h to achieve the absorption/desorption equilibrium of IPA on the photocatalyst surfaces. During this period, the IPA concentration rst decreased and then remained constant, demonstrating that adsorption/desorption equilibrium had been reached. During the equilibration process, no acetone or CO 2 was detected, demonstrating that the IPA molecules were not decomposed by the photocatalysts under dark conditions. The vessel was then irradiated with light, and 1 mL gaseous samples were periodically extracted from the reaction vessel to measure the concentrations of IPA, acetone, and CO 2 using a gas chromatograph (model GC-8A; Shimadzu Co., Ltd.).

Results and discussion
Nb x Ti 1Àx O 2 nanocomposites were prepared by a simple impregnation method and heat treatment using commercial rutile TiO 2 and NbCl 5 as starting materials. The total amount of Nb in the prepared samples was measured by ICP-AES and was found to be nearly equal to the initial value used in the preparation process (Table S1 †). The graing of Cu(II) nanoclusters onto Nb x Ti 1Àx O 2 was performed by a simple impregnation method. 5 XRD analysis showed that the prepared samples maintained a pure rutile TiO 2 crystalline structure (JCPDS card no. 21-1276) aer Nb doping or surface graing of Cu(II) nanoclusters ( Fig. 2a and S1 †), and an aggregated Nb 2 O 5 phase was not detected at any of the examined doping concentrations (Fig. S1 †). Notably, the diffraction peaks of the Nb-doped samples, Nb x Ti 1Àx O 2 and Cu(II)-Nb x Ti 1Àx O 2 , were shied towards smaller angles compared with that of the undoped sample (Fig. 2b). This nding indicates that Nb was substitutionally introduced into the TiO 2 lattice at the Ti site, as the ionic radii of Nb 5+ and Nb 4+ , which are 0.69 and 0.74Å, respectively, are larger than the 0.68Å radius of Ti 4+ . 11,12 The morphologies of the prepared nanocomposites were examined by SEM ( Fig. 3 and S2 †), which revealed that all samples were composed of uniformly distributed nanoparticles with an average grain size of approximately 200 nm. Introduction of Cu(II) nanoclusters on the TiO 2 surface as well as the doping of Nb into the TiO 2 lattice did not change the morphology or the particle size of the obtained Cu(II)-Nb x Ti 1Àx O 2 samples. Therefore, the effects of morphology and particle size on the photocatalytic activity can be excluded in the present study. 13 A TEM image clearly shows that the Cu(II) clusters, in the size of $2 nm, are well dispersed on the surface of Nb x Ti 1Àx O 2 particles (Fig. S3 †). The good attachment of Cu(II) clusters to the Nb x Ti 1Àx O 2 surfaces was observed from the corresponding high-resolution TEM (HRTEM) image. Point analysis of energy dispersive X-ray spectroscopy (EDS, Fig. S3c †) and surface analysis of XPS (Fig. S4 †) proved that these clusters consist of copper compound. Further, BET surface area analysis revealed that the introduction of Nb and Cu(II) ions did not alter the surface area of TiO 2 (Table S2 †).
The surface composition and elemental chemical states of the prepared nanocomposites were examined by XPS (Fig. 4, S4 and S5 †). For bare TiO 2 , only Ti and O were detected. The additional peaks associated with Nb were clearly observed in the spectrum of Nb x Ti 1Àx O 2 , indicating that Nb ions were successfully introduced into the TiO 2 lattice (Fig. S4 †). For Cu(II)-Nb x Ti 1Àx O 2 , a signal attributable to Cu was clearly detected, conrming that Cu(II) nanoclusters were graed onto the TiO 2 surface (Fig. S4 †). Fig. 4a and b show the Ti 2p and O 1s core-level spectra, respectively, of the nanocomposite samples. No obvious differences between the chemical states of elemental Ti and O were observed, 14 demonstrating that neither the graed Cu(II) nanoclusters nor doped Nb ions affected the bonding structure between titanium and oxygen. Further, no shoulders associated with Ti 3+ were observed in the Ti 2p core-level spectra ( Fig. 4a and S5 †), indicating that the density of Ti 3+ was below the detection limit of the XPS analysis. The low density of Ti 3+ in Nb x Ti 1Àx O 2 was conrmed by ESR analysis (Fig. S6 †). Hitosugi et al. 15a studied the microstructure of Nb 5+ -doped TiO 2 using XPS and found that the incorporation of Nb 5+ into the TiO 2 lattice resulted in the formation of minor Ti 3+ components that maintain the charge balance. However, their XPS results also indicated that Nb 5+ is reduced to Nb 4+ at high annealing temperatures. Consistent with this nding, Khoviv et al. 15b also showed that doped Nb ions in TiO 2 mainly existed as Nb 4+ aer high temperature treatment. 15b The formation of Nb 4+ would not induce the generation of Ti 3+ species. 15,16 Fig. 4c shows the Nb 3d core-level XPS spectra of the nanocomposite samples. For Nb x Ti 1Àx O 2 and Cu(II)-Nb x Ti 1Àx O 2 , well-dened photoelectron signals located at 206.5 and 209.3 eV were observed in the obtained spectra, whereas no feature was observed in the spectrum of the undoped sample, indicating that Nb was successfully doped into TiO 2 and was presented in the Nb 4+ oxidation state. 15 In addition, the Nb signals in Nb 2 O 5 and physically mixed TiO 2 and Nb 2 O 5 located at 207.5 and 210.4 eV ( Fig. S7 †), indicating that Nb 5+ was not detected in Nb x Ti 1Àx O 2 samples. Some previous reports indicated that the Nb doped TiO 2 sample is possible to show Nb 4+ in the bulk and Nb 5+ on the surface because of the aggregation of Nb on the surface. 15b, 17 Our XRD data clearly demonstrated the single rutile phase of the obtained Nb x Ti 1Àx O 2 samples, revealing the well distribution of Nb in the bulk TiO 2 . Thus, Nb 5+ could not be detected in our Nb x Ti 1Àx O 2 samples and the doped Nb existed as Nb 4+ in the samples. On the other hand, previous reports demonstrated that the formation of Nb 5+ and Ti 3+ would result in the up-shi of the energy. 11c-f However, our valence band (VB) XPS data clearly showed that the value of Nb x Ti 1Àx O 2 was similar to that of pure TiO 2 (Fig. S8 †). Further, no shoulder associated with Ti 3+ was observed in the Ti 2p core-level spectra ( Fig. 4a and S5 †), revealing that the density of Ti 3+ was below the detection limit of the XPS analysis. Based on these results, it can be safely concluded that the doped Nb was well distributed in the sample and existed as the Nb 4+ oxidation state.
Aer modication of Nb x Ti 1Àx O 2 with Cu(II) nanoclusters, the Cu 2p3/2 core-level XPS signal was observed at 932.3 eV  ( Fig. 4d), a value that is consistent with the results from our previous studies that used a combination of X-ray absorption ne structure (XAFS) measurements, XPS analysis, and TEM observation to characterize the Cu(II) state in a Cu(II)-TiO 2 system. 5a,b Based on these analyses of the local crystal structure, we conrmed that Cu(II) nanoclusters were graed onto the surface of TiO 2 as distorted amorphous CuO-like structures with a ve-coordinated square pyramidal form and a particle size of less than 3 nm. 5d Our present results indicate that the chemical state and the environment of Cu(II) nanoclusters in the present Cu(II)-TiO 2 and Cu(II)-Nb x Ti 1Àx O 2 nanocomposites are identical to those of the previously characterized Cu(II)-TiO 2 system.
The light absorption properties of the prepared samples were investigated by UV-visible spectroscopy (Fig. 5). The spectra presented in Fig. 5a clearly show that Nb doping into TiO 2 increased the visible-light absorption in the range of 420 to 550 nm. In addition, the graing of Cu(II) nanoclusters onto the surface of TiO 2 enhanced the light absorption of the resulting nanocomposite in the 420-550 and 700-800 nm wavelength regions. The increase in the shorter wavelength region can be assigned to the IFCT of VB electrons to surface Cu(II) nanoclusters, and the longer wavelength region is attributable to the d-d transition of Cu(II). 5a,b According to the band-gap estimation using the Kubelka-Munk function (Fig. S9 †), introduction of Nb ions into the TiO 2 lattice does not narrow the band gap of TiO 2 , as Nb ions are predicted to exist as isolated states in the forbidden gap. XPS (Fig. 4a) and ESR analyses (Fig. S6 †) suggested that Ti 3+ ions were not presented in the Nb x Ti 1Àx O 2 nanocomposite. On the basis of these results, the visible-light absorption of the Nb x Ti 1Àx O 2 nanocomposite is mainly attributable to the orbital of the doped Nb 4+ ions. A comparison of the difference absorption spectra of Cu(II)-TiO 2 , Nb x Ti 1Àx O 2 , and Cu(II)-Nb x Ti 1Àx O 2 versus bare TiO 2 revealed that the graed Cu(II) nanoclusters and doped Nb ions similarly increased the visible-light absorption of these systems between 420 and 550 nm (Fig. 5b). This nding indicates that doped Nb ions and graed Cu(II) nanoclusters have similar energy levels, and that the enhanced light absorption of Cu(II)-Nb x Ti 1Àx O 2 is mainly due to the doped Nb ions.
To specify similar energy levels between doped Nb ions and graed Cu(II) nanoclusters, it is necessary to examine the detailed electronic structure of Nb-doped TiO 2 . It is generally considered that excess electrons are generated when pentavalent Nb ions are substituted for tetravalent Ti ions. The generated electrons could delocalize from Nb ions to the neighboring Ti ions, resulting in the formation of Ti 3+ species, which introduce donor levels below the CB of TiO 2 . 10,18 However, the localized states of Nb ions in the band gap differ between the anatase and rutile forms; in the former, doped Nb ions exist as Nb 5+ , because a relatively low temperature is required for the formation of the anatase structure. Thus, the excess electrons from Nb 5+ ions could be transferred to neighboring Ti 4+ ions, resulting in a shi of the Fermi level toward the CB and an enhancement of the metallic behavior of anatase TiO 2 . 19 In other words, the transfer of excess electrons from Nb 5+ ions to Ti 4+ ions induces the generation of Ti 3+ ions, which leads to the formation localized energy states at 0.3-0.8 eV below the CB. 20 In contrast to the anatase form, Nb cations appear to substitute for Ti cations in rutile TiO 2 and exist as Nb 4+ . 15,16,21,22 The state of Nb in rutile single crystals and single crystal thin lms, and rutile polymorph particles and polymorph particle thin lms has been extensively analyzed using XRD, ESR, XPS, TEM, reection high-energy and low-energy electron diffraction (RHEED and LEED), X-ray photoelectron diffraction (XPD), and scanning tunneling microscopy (STM). 16,21 The results of these various analyses conrmed that Nb exists as Nb 4+ in rutile TiO 2 . 16,21 Investigation of the phase diagram of Ti-Nb-O also indicates that TiO 2 and NbO 2 may form a solid solution of Nb x Ti 1Àx O 2 with 0 < x < 0.85 for the normal rutile phase and x > 0.85 in a deformed rutile phase at room temperature. 12a The formation of Nb 4+ is speculated to be because excess electrons remain on the Nb ions and form donor levels in the TiO 2 lattice. Nb-doped rutile TiO 2 exhibits semiconductor behavior, rather than the metallic behavior observed in the anatase form, and its Fermi level is located in the band gap. 19 Theoretical and experimental studies have shown that Nb 4+ (4d 1 ) energy levels lie $0.12-0.22 eV below the bottom of the CB as a partially lled state. 22 These ndings are consistent with recent theoretical calculations that indicate Nb would form deep states in anatase and shallow states in rutile TiO 2 . 23 Although doped Nb in rutile TiO 2 has also been reported to remain as Nb 5+ ions, 24 whose 4d states would overlap with the O2p VB states of TiO 2 , our present UV-vis spectra, XPS, ESR, and VB XPS results (Fig. 4, 5, and S4-S9 †) clearly indicate that doped Nb exists as Nb 4+ in the band gap. Thus, we can conclude that the doped Nb existed as Nb 4+ in our rutile TiO 2 sample and produced a shallow energy level at $0.12-0.22 eV below the bottom of the CB, a value that matches the redox potential of Cu 2+ /Cu + (0.16 V vs. SHE, pH ¼ 0). 5a,b The photocatalytic performance of the nanocomposites prepared in the present study was evaluated by the visible-light induced decomposition of IPA, which was used as a representative gaseous volatile organic compound (VOC) and is a serious pollutant of indoor air. 5 IPA can be completely decomposed to CO 2 and water by photocatalytic oxidation. 25 For the photocatalytic tests, the light intensity was set to 1 mW cm À2 , which corresponds to an illuminance of 300 lux and is comparable to the intensity of white uorescent light and LED light, the wavelength of the irradiation light ranged from 400 to 530 nm (Fig. S10 †), and the initial IPA concentration was 300 ppmv ($6 mmol). Under these conditions, the complete decomposition of IPA would result in a CO 2 concentration of 900 ppmv ($18 mmol), which is three times the initial IPA concentration (CH 3 CHOHCH 3 + 9/2O 2 / 3CO 2 + 4H 2 O). A representative curve of the change in gas concentration during the decomposition of IPA by the Cu(II)-Nb x Ti 1Àx O 2 sample is shown in Fig. 6a. Under dark conditions, the IPA concentration initially decreased and then remained constant, demonstrating that adsorption equilibrium had been established. In addition, acetone and CO 2 were not detected, indicating that IPA was not decomposed by Cu(II)-Nb x Ti 1Àx O 2 under these conditions. With the onset of light irradiation, the IPA concentration decreased rapidly and the amount of acetone increased sharply until reaching a peak at approximately 300 ppmv, aer which, the acetone concentration started to decrease. Accompanying the decrease of acetone, the concentration of CO 2 increased quickly. The observed reaction prole is consistent with the photocatalytic decomposition of IPA proceeding via the formation of acetone as an intermediate, followed by the decomposition of acetone to the nal products CO 2 and H 2 O. 25 Aer 200 h of irradiation, the concentration of CO 2 in the reaction vessel reached approximately 900 ppmv ($18 mmol), which was nearly 3 times the amount of the initially injected IPA (300 ppmv), indicating that IPA was completely decomposed.
Comparative studies of the photocatalytic activities of Nb x Ti 1Àx O 2 , Cu(II)-TiO 2 and Cu(II)-Nb x Ti 1Àx O 2 samples under the same visible light source are shown in Fig. 6b. Aer the doping of Nb ions into TiO 2 , the generated Nb x Ti 1Àx O 2 nanocomposite exhibited visible-light sensitivity (Fig. 5), but had relatively low photocatalytic performance, because the photogenerated charge carriers were not efficiently separated and transferred to the surface. 3 The Cu(II)-TiO 2 nanocomposites exhibited high visible-light activity, owing to the IFCT from the VB of TiO 2 to the surface Cu(II) nanoclusters and the efficient multi-electron reduction of oxygen by these nanoclusters. 5-7 Irie et al. 5b investigated the role of Cu(II) nanoclusters in electrontrapping by performing in situ XAFS analysis under visible light in the presence of IPA and absence of oxygen and found that Cu(I) was generated under these conditions, but was converted back to Cu(II) upon exposure to oxygen. However, theoretical and experimental studies have revealed that the doping of Nb into TiO 2 induces the formation of an energy level approximately 0.12-0.22 eV below the CB, 22 which is within the range of the reported redox potential of Cu 2+ /Cu + , 0.16 V (vs. SHE, pH ¼ 0). 5a,b These results indicate that graed Cu(II) nanoclusters and doped Nb ions have closely matched energy levels.
The enhanced photocatalytic performance of Cu(II)-Nb x Ti 1Àx O 2 compared with those of Cu(II)-TiO 2 and Nb x Ti 1Àx O 2 demonstrates that efficient energy level matching is achieved between the graed Cu(II) nanoclusters and doped Nb ions. Based on the observed photocatalytic activities of the prepared photocatalysts, the QE for CO 2 generation was calculated using the following equation: QE ¼ R r p /R a p ¼ 6R CO 2 /R a p , where R r p is the reaction rate of photons involved in CO 2 generation, R CO 2 is the CO 2 generation rate, and R a p is the absorption rate of incident photons. The details of this calculation are described in the literature 5b and ESI (Fig. S11 †), and the data used in the calculations are summarized in Table 1. Under the same light irradiation conditions, the Nb x Ti 1Àx O 2 sample exhibited a high absorption rate of incident photons, but a low CO 2 generation rate of only 0.015 mmol h À1 , indicating that this photocatalyst has a low charge separation efficiency. Although the Cu(II)-TiO 2 nanocomposites displayed a high QE (27.7%), indicating that efficient IFCT and a multi-electron reduction reaction proceeded on the surface, 5-7 the visible-light absorption rate of this material was relatively low and can be attributed to the limited light absorption by the IFCT process (Table 1). Interestingly, the graing of Cu(II) nanoclusters onto the surface of Nb x Ti 1Àx O 2 resulted in strong visible-light absorption by the synthesized Cu(II)-Nb x Ti 1Àx O 2 nanocomposite, which also exhibited a high QE of 25.3%. The high reaction rate of Cu(II)-Nb x Ti 1Àx O 2 is due to the efficient light absorption by doped Nb ions and electron transfer between the doped Nb and surface-graed Cu(II) nanoclusters, as well as the efficient multi-electron reduction of oxygen on the surface Cu(II) nanoclusters. 5-7 The QE of Cu(II)-Nb x Ti 1Àx O 2 is markedly higher than that of Cu(II)-graed, Ti 3+ self-doped TiO 2 , 5d indicating that efficient charge transfer proceeds between the dopants and surface Cu(II) nanoclusters because of similar energy levels. Due to these excellent properties, the Cu(II)-Nb x Ti 1Àx O 2 nanocomposites exhibited a CO 2 generation rate of 0.20 mmol h À1 , which is much higher than those of Nb x Ti 1Àx O 2 and Cu(II)-TiO 2 under the same visiblelight irradiation conditions.
We also investigated the photocatalytic activities of TiO 2 modied with Cu(II) surface nanoclusters and various metal ions as dopants (Fig. 6c). In addition to Nb dopants, we have also  checked the photocatalytic performances of Fe and W doped TiO 2 with various doping densities (Fig. S12 †). The results showed that 0.1 wt% was the optimal amount for Fe and W doped TiO 2 . The synthesized nanocomposites were all single phases of the rutile TiO 2 crystal structure. Among the examined metal dopants, Cu(II)-Nb x Ti 1Àx O 2 nanocomposites exhibited the best performance, indicating that well energy level matching occurred between doped Nb ions and graed Cu(II) nanoclusters. Further, the visible-light activity of Cu(II)-Nb x Ti 1Àx O 2 was superior to that of TiO 2Àx N x , which is widely recognized as one of the most efficient visible-light photocatalysts reported to date ( Fig. 6d and Table 1). 3a TiO 2Àx N x exhibited a rather low QE of 3.9% and required a much longer incubation period, over 300 h, to completely decompose the gaseous IPA. The low activity of this system originates from the lower oxidation power of the photogenerated holes in the nitrogen levels than those in the VB. 4 Finally, we attempted to optimize the experimental conditions for enhancing the activity and found that 0.1 wt% was the optimal amount for both doped Nb and surface-graed Cu(II) nanoclusters ( Fig. S13 and S14 †). A good junction between the surface-graed Cu(II) nanoclusters and bulk-doped Nb ions was also critical for the efficient charge transfer. Notably, if a thin layer was introduced between the surface Cu(II) nanoclusters and doped Nb ions (Fig. S15 †), the visible-light activity was markedly reduced (Fig. S16 †). The Cu(II)-Nb x Ti 1Àx O 2 sample was also very active under UV-light irradiation. Taken together, these ndings suggest that Cu(II)-Nb x Ti 1Àx O 2 nanocomposites are promising visible-light-sensitive photocatalysts for practical applications.

Conclusions
Efficient visible-light-sensitive TiO 2 photocatalysts were developed based on the concept of energy level matching between surface-graed Cu(II) nanoclusters and bulk-doped Nb ions. Bulk-doped Nb ions produce energy levels below the CB of TiO 2 , which matches well with the redox potential of Cu 2+ /Cu + in surface-graed Cu(II) nanoclusters. Both graed Cu(II) nanoclusters and doped Nb ions induce similar increases in light absorption in the wavelength region from 420 to 550 nm. In this photocatalytic system, Ti ions were substituted for doped Nb ions that existed in the Nb 4+ oxidation state, which avoided the generation of Ti 3+ species. The doping of Nb ions enhanced the visible-light absorption of TiO 2 , whereas the graing of Cu(II) nanoclusters retained the high QE of this system. The present Cu(II)-Nb x Ti 1Àx O 2 nanocomposites exhibited strong visiblelight absorption and maintained a high QE, leading to high visible-light photocatalytic performance for the decomposition of gaseous organic compounds. Thus, our ndings demonstrate that Cu(II)-Nb x Ti 1Àx O 2 is a suitable visible-light-sensitive photocatalyst for practical applications, and that the concept of energy level matching is an effective approach for the construction of advanced visible-light photocatalysts.