1 Introduction

The energy crisis and environmental pollution caused by the use of non-renewable energy sources, such as coal, oil and other fossil fuels, have become two major issues, which restrict the sustainable development of human beings. It is highly desirable to seek for green and reproduction energy sources [1, 2]. Hydrogen, with the merits of zero emissions and high energy density values per unit mass, is one of the promising candidates for green energy sources [3]. The current hydrogen production technology mainly includes methanol steam reforming reaction, hydrogen production from natural gas, electrocatalysts for water splitting, etc. [4]. Among various hydrogen production technologies, photocatalytic water splitting is considered to be one of the most promising and environmental-friendly techniques [5, 6]. Since the discovery of photoelectrochemical (PEC) water splitting in 1972 [7], TiO2 has attracted intensive attention as photocatalytic materials for its appropriate energy band position, high chemical stability, low cost and non-toxicity [8,9,10,11,12,13,14,15,16]. However, the intrinsic shortcomings of TiO2, e.g., slow electron transfer and rapid recombination of photogenerated charge carriers drastically, have limited their practical applications [17].

To improve the photocatalytic activity, co-catalysts are usually introduced/decorated onto the surface of semiconductors/supports (e.g., TiO2) [3, 18,19,20]. Nowadays, the noble metal Pt is reported as a state-of-the-art co-catalyst for the photocatalytic hydrogen evolution [21,22,23]. However, the application of Pt-based co-catalysts is still restricted by its high cost and scarcity. Size reduction of co-catalyst to nano-scale and even single atoms was widely applied to improve utilization efficiency of noble elements for cost reduction and catalytic performance promoting simultaneously [24,25,26,27]. For example, in 2014, Xing et al. [28] reported that 6–13-fold enhancement of the photocatalytic hydrogen evolution was achieved when the Pt clusters were down to single atoms on the TiO2 support. Meanwhile, non-noble metal (e.g., Fe, Co, Ni, Cu) co-catalysts are another way to reduce cost and facilitate the practical applications [29,30,31,32]. For instance, as Ni atoms were introduced on the supports (e.g., TiO2 [33] or C3N4 [34]), the photocatalytic performance of the systems has been improved significantly. Aside from the co-catalysts, designing/optimizing the support materials is also employed as an alternative route to promote the catalytic performance. For support materials, surface area increasing through specific nano-morphology and electronic structure modulation via doping or defects engineering seems to be efficient approaches [17, 18, 35,36,37,38]. For example, the hierarchical TiO2 with large specific surface areas of 170 m2·g−1 shows excellent photocatalytic performance than the commercial P25 (~ 45 m2·g−1) [39]. This is because the enlarged surface area would result in more exposure active sites and solid interfaces which play vital roles on the photocatalytic reactions.

To promote the photocatalytic reaction, the photogenerated electrons and holes must be separated and transferred from the bulk phase to the active sites to drive redox reactions [40,41,42]. Actually, photocatalytic activity is generally charge separation and transfer dependence [43]. Metal clusters (especially ultra-fine clusters or fully exposed cluster catalysts (FECC) [44]) are found to have better performance in tuning the surface electronic properties and subsequently accelerate the charge separation and transfer process [45,46,47,48]. Wang et al. [45] reported that the photocatalytic reaction was boosted by Pd (containing both single atoms and clusters) loaded onto TiO2. It was suggested that the Pd clusters promote the photogenerated charge separation and provide the optimum active sites for H2 evolution. Interestingly, Yan et al. [47] found that the charge was transferred from Pt single atoms to the sulfur-doped carbon supports, but contrarily, for the ultra-small Pt clusters (~ 1.5 nm), the transfer direction was reversed, i.e., from the support to Pt clusters. In the last few years, Cu species co-catalysts (both nanoparticles and single atoms) have drawn wide attention due to their competitive performance, low-cost and natural abundance [3, 43]. In 2019, Lee et al. [49] designed and synthesized highly active TiO2 decorated with Cu single atoms for photocatalytic hydrogen evolution. Seadira et al. [50] found that Cu-doped TiO2 material prepared by impregnation method had better hydrogen evolution activity than Co, Cr, Ag and Ni, respectively. However, there are fewer reports concerned about the ultra-fine Cu clusters co-catalysts for photocatalytic hydrogen evolution.

In this study, a hydrangea-like TiO2 support with enhanced photocatalytic performance is synthesized via a simple hydrothermal method. An improved hydrogen evolution rate (1.8 mmol·h−1·g−1) of this hydrangea-like TiO2 is achieved, ~ 10.7 times higher than that of P25 (168 μmol·h−1·g−1). Then, ultra-fine (~ 1 nm) Cu clusters were further loaded onto the hydrangea-like TiO2 supports (denoted as Cu–TiO2). A maximal hydrogen evolution value of 3.7 mmol·h−1·g−1 is obtained in the optimized system of 6Cu–TiO2 (6 wt% Cu), which is about twofold enhancement on the base of the bare hydrangea-like TiO2. The experimental and theoretical analysis stress that the enhanced hydrogen production should be attributed to the promoted charge separation and transfer, which originate from the surface electronic properties tuned by the ultra-fine Cu clusters. These results not only fill the gaps of Cu ultra-fine clusters in the study of Cu-based co-catalysts, but also provide guidance for the development of non-noble metal co-catalysts.

2 Experimental

2.1 Materials and chemicals

Titanium butoxide (Ti(OBu)4), acetic acid (CH3COOH) and cupric acetate monohydrate (Cu(CH3COO)2·H2O) purchased from Shanghai Aladdin Bio-Chem Technology Co., Ltd., were of analytical grade.

2.2 Materials synthesis of hydrangea-like TiO2 and XCu–TiO2 (where X is mass percent of Cu)

The synthesis procedure of the hydrangea-like Cu–TiO2 is as following. 30 ml CH3COOH and a certain amount Cu(CH3COO)2·H2O (0.0176 g, 0.0352 g, 0.0528 g, 0.0704 g) were added into a 50 ml Teflon-lined stainless steel autoclave with stirring. Subsequently, 1 ml Ti(OBu)4 was added drop-wise to acetic acid solution followed with 5-min stirring. Then, the autoclave was kept at 140 °C for 24 h. The obtained powder samples were washed several times with water and alcohol and dried at 60 °C overnight. The samples were further annealed at 450 °C for 3 h. Finally, the hydrangea-like TiO2 with Cu decoration were obtained and denoted as 3Cu–TiO2, 6Cu–TiO2, 9Cu–TiO2, 12Cu–TiO2, respectively. The pure hydrangea-like TiO2 was prepared under the same condition without the presence of Cu(CH3COO)2·H2O.

2.3 Material characterization

X-ray diffraction (XRD) patterns were carried out using a PANalytical X'Pert apparatus (Netherlands) operated at Cu Kα radiation to study the crystal structures of the samples. Raman spectroscopy was collected on a LabRAMM HR evolution imaging microscope system. Transmission electron microscopy (TEM, Thermoscientific Talos F200S) and scanning electron microscopy (SEM, Thermoscientific Quattro S) equipped with energy-dispersive X-ray spectroscopy (EDS) were utilized to investigate the microstructure (crystal structure, morphology, chemical composition) of the samples. The high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image was collected on a probe-corrected FEI Titan Themis microscope at 300 kV. The Brunner–Emmet–Teller (BET) surface areas were analyzed by N2 adsorption on Unicube (Elementar). X-ray photoelectron spectroscopy (XPS, Thermoscientific ESCALAB250Xi, Al Kα) was performed to explore the surface chemical states, where the C 1s peak at 284.8 eV (i.e., the surface adventitious carbon) was used for calibration. Ultraviolet–visible (UV–Vis) spectra were collected on Shimadzu UV-3600 spectrometer (BaSO4 as a reference background, absorbance was calculated by the Kubelka–Munk method). Time-resolved fluorescence emission spectra (TRPL) were collected on a fluorescence spectrophotometer (FLS1000). A steady-state fluorescence spectrometer (Shimadzu RF-6000) was utilized to measure the photoluminescence (PL) spectra.

2.4 Photocatalytic hydrogen evolution

Photocatalytic hydrogen evolution reactions were performed on a Pyrex vessel connected to a glassed closed circulation and evacuation system (Beijing Perfect Light Company). A 300 W Xe lamp was used as the light source. As for the reaction, 20 mg photocatalysts were dispersed into 100 ml 10 vol% aqueous solution methanol. First, the slurry in the reactor was vacuumed for 30 min to eliminate the dissolved gas molecular and provide an anaerobic condition for the reaction. During the photocatalytic performance tests, the catalyst was kept in suspension by continuous magnetic stirring at 25 °C (controlled by cooling circulating water). The evolved H2 was measured by chromatograph (GC7900, Techcomp) with a thermal conductivity detector (TCD) using Ar as the carrier gas.

2.5 Photoelectrochemical measurements

The electrochemical impedance spectroscopy (EIS), Mott–Schottky (MS) and photocurrent were measured by electrochemical workstation (Zahner CIMPS-2). All the measurements were performed on a standard three-electrode system using Pt plate as the counter electrode, Ag/AgCl electrode as reference electrode and 0.5 ml Na2SO4 as the electrolyte. The working electrode (glassy carbon electrode) was made as follows: 5 mg above photocatalyst was dispersed into the mixture of 50 μl Nafion and 950 μl ethanol with ultrasonication for about 30 min to form a uniform dispersion of the catalysts. Then, 10 μl of the solution was deposited on the glassy carbon electrode and subsequently dried in air at room temperature.

2.6 Density functional theory calculations

The calculations were carried out based on density functional theory (DFT) utilizing the Vienna ab initio Simulation Package (VASP) [51, 52] with spin-polarized Perdew–Burke–Ernzerh generalized gradient approximation (GGA-PBE) function [53, 54]. The kinetic cutoff energy was set as 500 eV for all calculations, 25% larger than the recommended largest value of component species, sufficiently ensuring calculation accuracy. The DFT-D3 method was used to describe the van der Waals interactions between the adsorbed species and the catalyst [55]. The slab model of anatase surface containing 3 TiO2 layers and 216 atoms was constructed along z-direction. The thick vacuum space was set to 2 nm to avoid interactions between the adjacent models. Moreover, the bottom two atomic layers were fixed, while the uppermost layers and adsorbed cluster molecules were relaxed during all calculations. An Γ-centered 1 × 2 × 1 grid was used, and the energy convergence was chosen as 1 × 10–4 eV. The slab model was fully relaxed until the residual forces of each atom less than 0.2 eV·nm−1. To investigate the effect of ultra-fine Cu clusters on the hydrangea-like TiO2, difference charge density analysis was performed. The difference in charge density is defined as \(\Delta \rho = \rho ({\text{system}}) - \rho ({\text{substrate}}) - \rho ({\text{Cu}}_{13}),\) where (system) denotes the charge density of the whole system, (substrate) denotes the charge density of TiO2, and (Cu13) denotes the charge density of Cu13 removing substrate.

3 Results and discussion

3.1 Structure characterization and chemical composition analysis

SEM images (Fig. 1a, d) show that the obtained samples have hydrangea-like sphere structure, with diameter of ~ 5 μm. Each sphere is made up of ultra-thin 2D nanosheets radiating out from its center. The hydrangea-like structure of Cu–TiO2 can be almost maintained after Cu decoration, as the morphology of 6Cu–TiO2 shows in Fig. 1b, e. SEM–EDS mapping analysis shows that the Cu clusters are uniformly dispersed over TiO2 (Fig. 1c), which is further confirmed by HAADF observation (Fig. 1h). TEM was further conducted to uncover the microstructure of pure TiO2 and 6Cu–TiO2. High-resolution transmission electron microscopy (HRTEM) images (Fig. 1f, g) reveal that the TiO2 nanosheets in the hydrangea sphere have a typical size of several tens of nanometers with good crystallinity. The measured d-spacing of lattice fringe (0.35 nm) could be ascribed to the (101) planes of anatase TiO2. Furthermore, HAADF-STEM image (Fig. 1h) shows the uniformly dispersed ultra-fine Cu clusters with average size of ~ 1 nm. Additionally, these Cu clusters exhibit non-crystalline structure with no regular fringes observed.

Fig. 1
figure 1

a, d SEM images of pure TiO2; b, e SEM images of 6Cu–TiO2 samples; c SEM–EDS mapping of 6Cu–TiO2; f, g TEM images of pure TiO2; h STEM-HAADF image of 6Cu–TiO2 (several Cu clusters are marked by red circles)

Full size image

As shown in Fig. 2a, all XRD patterns of the samples can be well indexed to anatase TiO2 phase (JCPDS card No. 21–1272) without detection of any impurities. No diffraction peaks of Cu crystals are observed due to (1) the low loading amount and (2) the non-crystalline structure of the ultra-fine Cu clusters (as confirmed by the above HAADF analysis). The anatase structure is further confirmed by Raman spectroscopy, as shown in Fig. 2b. The major Raman peaks appeared at 143.0, 196.6, 395.9, 515.9 and 639.6 cm−1 can be definitely assigned to Eg(1), Eg(2), B1g(1), A1g + B1g(2) and Eg(3) of anatase TiO2, respectively [29, 56, 57]. BET adsorption–desorption isotherms and pore size distribution of P25, the hydrangea-like TiO2 and 6Cu–TiO2 samples were further studied, as shown in Fig. 2c, d. Type IV isotherms are observed for the pure hydrangea-like TiO2 and 6Cu–TiO2 catalysts, indicating that there is mesoporous in both samples. The detailed values of the BET surface area, pore volume and average pore diameter of P25 and the hydrangea-like samples are presented in Table 1. Similar BET surface areas between the hydrangea-like TiO2 and 6Cu–TiO2 (both larger than that of P25) indicate the almost unchanged morphology and surface structure of the hydrangea-like TiO2 after Cu decoration. However, the pore volume and average pore diameter of the 6Cu–TiO2 are slightly increased compared with the pristine hydrangea-like TiO2.

Fig. 2
figure 2

a XRD patterns and b Raman spectra of samples; c N2 adsorption/desorption isotherm and d corresponding pore size distribution curves of TiO2 and 6Cu–TiO2

Full size image
Table 1 Specific surface areas, average pore sizes and pore volumes of P25, pure TiO2 and 6Cu–TiO2 samples
Full size table

XPS technique was conducted to investigate surface chemical states of the samples. Figure 3a presents the high-resolution XPS spectrum of Ti 2p. The peaks located at 458.75 and 464.48 eV, suggesting that the chemical state of Ti is in the form of Ti4+ [43], i.e., TiO2. Figure 3b shows the O 1s spectrum, consisting of a doublet at binding energies of 529.95 and 531.73 eV, which can be deemed as the lattice oxygen (Ti–O–Ti) and Ti–OH, respectively [58]. Basically, 6Cu–TiO2 exhibits similar characteristic peaks to the pure TiO2 in the Ti 2p and O 1s. In addition, the Cu peaks can be also observed in 6Cu–TiO2, as shown in Fig. 3c. The two peaks, i.e., Cu 2p3/2 at 933.29 eV and Cu 2p1/2 at 953.40 eV, can correspondingly be assigned to the mixture of Cu+ and Cu0 due to the similar (almost indistinguishable) binding energies between these two species [43].

Fig. 3
figure 3

XPS spectra of TiO2 and 6Cu–TiO2: a Ti 2p, b O 1s and c Cu 2p; d Mott–Schottky plots of TiO2 and 6Cu–TiO2

Full size image

4 Charge separation and transportation

The separation efficiency and migration of photogenerated charge carriers are closely related to the photocatalytic activity. To understand the effect of the ultra-fine Cu clusters on the TiO2, the carrier concentration (Nd) was evaluated by the MS test (Fig. 3d), which can be calculated from the slope of MS curves according to the following Eq. (1) [33, 43, 59]:

$$ N_{{\text{d}}} = \frac{2}{{e_{0} \varepsilon_{0} \varepsilon_{\text{r}} }}({\text{d}}C^{ - 2} /{\text{d}}V)^{ - 1} $$
(1)

where e0 is the elemental charge (1.60 × 10–19 C), ε0 is the permittivity of vacuum (8.85 × 10–12 F·m−1), and εr is the dielectric constant (48.0). It is assumed that the e0, ε0 and εr are the same for both TiO2 and 6Cu–TiO2. The obtained Nd value of the 6Cu–TiO2 (6.40 × 1015) is nearly 3.4 times larger than that of the pure hydrangea-like TiO2 (1.90 × 1015). The higher Nd in the 6Cu–TiO2 means better electric conductivity and faster charge transfer, which promotes the hydrogen evolution.

The PL and TRPL spectral analyses were carried out for understanding the influence of ultra-fine Cu clusters decorated on the TiO2. Figure 4a shows PL emission spectra, in which the intensity of Cu–TiO2 catalysts decreases compared to the pure TiO2, and the smallest one is found in the 6Cu–TiO2 system. Generally, the lower PL intensity suggests lower recombination rate of photogenerated charge carriers. Thus, it is deduced that the decoration of Cu clusters helps restrain the recombination of photogenerated charge carriers. In addition, the TRPL spectra were measured to probe the photophysical properties of TiO2 and 6Cu–TiO2 (Fig. 4b). Biexponential fittings were applied to analyze the luminescence decay curves (inset in Fig. 4b). The decay kinetics of 6Cu–TiO2 exhibits a short average lifetime (1.89 ns) as compared to the pure TiO2 (2.26 ns), which implies the existence of effective electron transfer channels between the TiO2 support and Cu clusters. Furthermore, the electron transfer rate (kET) and efficiency (ηET) can be obtained by Eqs. (2, 3) [60]:

$$ k_{{{\text{ET}}}} = \frac{1}{{\tau_{{{\text{ave (6Cu}} - {\text{TiO}}_{{2}} )}} }} - \frac{1}{{\tau_{{{\text{ave (TiO}}_{{2}} )}} }} $$
(2)
$$ \eta_{{{\text{ET}}}} = 1 - \frac{{\tau_{{{\text{ave (6Cu}} - {\text{TiO}}_{{2}} )}} }}{{\tau_{{{\text{ave (TiO}}_{{2}} )}} }} $$
(3)
Fig. 4
figure 4

a PL spectra of samples; b TRPL decay curves for TiO2 and 6Cu–TiO2 and (inset) parameters of time-resolved photoluminescence decay curve determined by biexponential fitting, where τ1 and τ2 are values of the lifetime of each decay component and Rel represent the normalized amplitudes (or weighing factors); c photocurrent responses of samples, d Nyquist plot of EIS (Z’: real part of impedance; -Z″: imaginary part of impedance)

Full size image

The kET and ηET of 6Cu–TiO2 are 8.67 × 107 s−1 and 16.37%, respectively, compared to that of the pure TiO2, indicating that ultra-fine Cu clusters can promote the charge transfer between TiO2 support and Cu clusters effectively. 6Cu–TiO2 shows larger photocurrent intensity (about 8 times) than the pure TiO2 (Fig. 4c), confirming that the TiO2 has better charge generation and migration capabilities after ultra-fine Cu clusters decoration which is further proved by EIS (Fig. 4d). The semicircle in the Nyquist plot at high frequency represents the charge transfer process, in which the smaller arc of the semicircle means smaller charge resistance and higher electron mobility. As shown in Fig. 4d, the Nyquist plot of 6Cu–TiO2 at high frequency shows a smaller arc of the semicircle than that of TiO2, which implies a much lower charge transfer resistance of 6Cu–TiO2. The above results demonstrate that the charge carrier transfer and separation are promoted by the introduction of ultra-fine Cu clusters onto TiO2.

4.1 Photocatalytic H2 evolution activity

In this work, the pure hydrangea-like TiO2 shows excellent performance (1.8 mmol·h−1·g−1) for hydrogen production which is ~ 10.7 times that of P25 (168 μmol·h−1·g−1), as illustrated in Fig. 5a. Moreover, to explore the relationship between the amount of ultra-fine Cu clusters and the photocatalytic activity in the Cu–TiO2 systems, photocatalytic reduction of H2 evolution reactions was conducted. Generally, the photocatalytic activity was promoted in the Cu–TiO2 systems (Fig. 5a), suggesting that the introduction of ultra-fine Cu clusters is beneficial for the photocatalytic H2 evolution. Initially, the H2 evolution rate increases from 1.8 mmol·g−1·h−1 (TiO2) to 3.7 mmol·g−1·h−1 (6Cu–TiO2), about twofold enhancement with the loading amount of ultra-fine Cu clusters increasing from 0 wt% to 6 wt%. In contrast, with the further increase in Cu loading amount to 12 wt% (12Cu–TiO2), the activity of the catalyst begins to slightly decline, which might be attributed to the decrease in active sites. Furthermore, the activity of photocatalyst is merely decreased slightly after four cycles (Fig. 5b). XRD pattern (Fig. S1) and SEM image (Fig. S2) of the catalyst (6Cu–TiO2) after cycles of photocatalytic hydrogen evolution activity test were collected. No obvious changes neither in structure nor morphology were found, suggesting good stability of the Cu–TiO2 systems. While the slight decrease in the photocatalyst performance may be attributed to the interaction disruption between the ultra-fine Cu cluster and the surface of TiO2.

Fig. 5
figure 5

a H2 evolution rate of samples; b cycle stability of 6Cu–TiO2; c free energy versus reaction coordinates of TiO2 and 6Cu–TiO2 and (inset) 3D differential charge density of 6Cu–TiO2; d 2D differential charge density of 6Cu–TiO2, where yellow and blue colors represent positive and negative, respectively

Full size image

4.2 Photocatalytic mechanism

DFT calculations were performed for analyzing Gibbs free energy (Fig. 5c), the difference of charge density between TiO2 and 6Cu–TiO2 (Fig. 5d) and band structures (Fig. S3). Firstly, a higher charge density at the Cu cluster, particularly near the interface of Cu cluster and TiO2 support, is found in Cu–TiO2 systems, suggesting that the electrons tend to transfer from TiO2 to Cu cluster. Therefore, the enhanced charge transfer efficiency could be expected, which is consistent with the above results of higher Nd in 6Cu–TiO2 than in pure TiO2. Secondly, with the decoration of ultra-fine Cu clusters, newly generated bands appear in the vicinity of the Fermi level, as seen in the band structure of TiO2 and 6Cu–TiO2 in Fig. S3. Finally, as shown in Fig. 5c, the hydrogen adsorption (ΔG) of 6Cu–TiO2 is −0.16 eV, manifesting much more favorable H2 evolution activity than the pure TiO2 (−0.25 eV). The existence of Cu ultra-fine cluster may contribute to the optimal ∆G by tuning the electronic structures of 6Cu–TiO2. These theoretical results further suggest that the improved charge separation and transfer can be achieved after loading Cu clusters, which gives rise to the excellent photocatalytic hydrogen production.

UV–Vis absorption spectra in Fig. 6a reveal that after Cu clusters loading onto the surface of TiO2, absorbance in the range of 400–800 nm is observed, which means enhancement of the visible light absorption. The bandgap of 6Cu–TiO2 is determined to be 3.26 eV as obtained by Tauc plots (Fig. 6b). The valence-band (VB) position of 6Cu–TiO2 is determined to be 2.66 eV from XPS spectra (Fig. 6c). The Tauc plots and XPS valence-band spectra of TiO2 sample are shown in Figs. S4, S5. Accordingly, the energy band diagram and the photocatalytic mechanism can be drawn, as shown in Fig. 6d. During the reaction process, semiconductor material (support) is stimulated by light to produce photogenerated carriers, and the electrons transfer from the conduction band (CB) of TiO2 to the ultra-fine Cu clusters, resulting in the separation of the photogenerated electrons and holes and thus reducing H2O to H2. Notably, ultra-fine Cu clusters as co-catalyst on the surface of TiO2 will perform as active sites to capture the electrons through modulating interfacial charge transfer process and lead to excellent photocatalytic hydrogen evolution performance.

Fig. 6
figure 6

a UV–Vis absorption spectra of samples (enlarged view as inset); b Tauc plots and c XPS valence-band spectra of 6Cu–TiO2; d energy band diagram of 6Cu–TiO2

Full size image

5 Conclusion

In summary, a promising hydrangea-like TiO2 support with enhanced photocatalytic performance was obtained by a simple hydrothermal method. The hydrangea-like TiO2 (1.8 mmol·h−1·g−1) shows ~ 10.7 times higher photocatalytic H2 evolution efficiency than P25 (168 μmol·h−1·g−1). The photocatalytic hydrogen evolution is further improved after ultra-fine (~ 1 nm) Cu clusters loaded onto the surface of the TiO2. The optimized system 6Cu–TiO2 (6 wt%) shows a maximal value of 3.7 mmol·h−1·g−1 for hydrogen production, which is approximately twofold higher than that of the bare hydrangea-like TiO2. The enhanced H2 production is attributed to the surface electronic structure/properties tuned by the ultra-fine Cu clusters. Both experimental and theoretical results reveal that the charge separation and transfer have been improved after Cu clusters decoration. In the Cu–TiO2 systems, the photogenerated electrons tend to transfer from CB of TiO2 to Cu clusters, which could effectively promote the separation and transfer of photogenerated carriers. Subsequently, the electrons accumulate on the Cu clusters and accelerate the reduction of H2O to H2, resulting in high efficiency of photocatalytic hydrogen evolution. More importantly, this work offers a novel route to design high-performance TiO2 support for high-efficiency photocatalysis.