Recently, the rapid consumption of fossil fuel has led to serious environmental issues, especially global warming and haze. It is urgent to find clean energy sources to replace the traditional fossil energy [1-5]. Photocatalytic water splitting was regarded as a promising way to convert solar energy into green and renewable energy. As a semiconductor for spontaneous photocatalysis, there are three requirements should be fulfilled: (1) the band gap should be higher than water decomposition voltage (1.23 eV); (2) the band edge positions should stride the hydrogen and oxygen redox potential, and (3) it ought to be stable against photo corrosion during photocatalysis [6]. However, the half-reaction of water oxidation is tough because of the complex four-electrons redox process and high activation energy for O–O bond formation [7]. Therefore, huge efforts have been made to increase the efficiency of the water oxidation on a variety of catalysts, especially on TiO₂, which is perhaps one of the most important materials in photocatalysis due to its natural abundance, low cost, non-toxicity and superior photostability [8]. However, TiO₂'s photocatalytic efficiency is severely limited because of the low visible light utilization and the high recombination rate of charge carriers [9]. It is known that H₂ reduction could introduce Ti³⁺ [10]. The resultant reduced TiO₂ had a narrower band gap [11], which exhibited high activity in visible–light–driven water oxidation [12, 13].

The combination of noble metals (e.g., Au and Ag) with TiO₂ is another promising route to extend the light absorption of the photocatalysts to longer wavelength light due to surface plasmon resonance (SPR) effects of noble metal nanoparticles [14]. Whereas, the price of noble metals limits their commercialization, thus call for low-cost metals as substitute [15]. Most recently, semimetal bismuth (Bi) has been demonstrated to be an ideal substitute for noble metals due to its obvious SPR effect in the visible or even NIR range [16]. By in situ reduction routes such as photo-reduction, chemical reduction and hydrothermal reduction, respectively, Bi-based powder can be facilely obtained with significant role of photocatalyst performed in further application [17]. For example, Bi/g-C₃N₄ showed outstanding visible photocatalytic capability [15]. The enhancement of visible light absorption was considered to be caused by the SPR effect of Bi metal. In addition, the electrical conductivity of these Bi-coupled nanocomposites increases so that the transfer rate of photogenerated carriers can be accelerated by coupling with Bi metal [18]. However, it is still a great challenge to directly deposit Bi metal nanoparticle onto the surface of semiconductors through the in situ deposition route.

In this paper, we report a MOF-derived strategy that metallic Bi particles are in situ deposited inside and outside of reduced TiO₂ microspheres (Bi@R-TiO₂) via a simple and low energy cost hydrothermal method. In the hydrothermal procedure, ethylene glycol (EG) as a reductant in the reduction of Bi³⁺ and Ti⁴⁺ to metallic Bi and Ti³⁺, respectively. In the case of the Bi@R-TiO₂ composite, the main role of Bi metal is taken as follows: (1) Bi metal behaves as an "electron acceptor" to accelerate the charge carrier transfer from TiO₂ to Bi and (2) the composite strongly absorbs visible and NIR light because of the SPR effect of Bi metal. Meanwhile, the generation of Ti³⁺ further narrows the band gap of TiO₂. Based on this, the Bi@R-TiO₂ composite demonstrates a superior full solar spectrum photocatalytic oxygen evolution.

In a typical process, stoichiometric amounts of urea, titanium butoxide and bismuth nitrate pentahydrate in a molar ratio of 3:1:1 were added into 60 mL of EG. To form clean solution, the solution should be stirred at room temperature. After being stirred for 2 h, the mixed suspension was transferred into a 100 mL Teflon-coated autoclave at 160 ℃ for 6 h. The as-prepared Bi-Ti-MOFs were washed thoroughly with ethanol, and then dried at 60 ℃ for 12 h under vacuum. Finally, the obtained products were calcined at temperatures ranging from 200 to 350 ℃ for 5 h in Ar atmosphere at a ramping rate of 10 ℃ min^–1 to form the Bi@R-TiO₂ composites. For distinguishing the products, Bi@R-TiO₂ obtained upon calcination at 200, 250, 300 and 350 ℃ in Ar atmosphere is labeled as Bi@R-TiO₂-200, Bi@R-TiO₂-250, Bi@R-TiO₂-300 and Bi@R-TiO₂-350, respectively.

X-ray powder diffraction (XRD) patterns were recorded with a D/Max2500PC X-ray diffractometer with Cu-K_α1 radiation (λ = 0.15406 nm) operated at 40 kV. Scanning electron microscopy (SEM) was performed with a FEI NanoSEM 450 instrument with an energy dispersive X-ray spectroscopy (EDX). Transmission electron microscopy (TEM) and high resolution TEM (HRTEM) were carried out with a FEU G20 transmission electron microscope with an accelerating voltage of 200 kV. X-ray photoelectron spectroscopy (XPS) was performed using a Theromo ESCALAB 250Xi spectrometer using Al K_α radiation at a voltage of 1486.6 eV and this instrument was calibrated with the C 1s binding energy of 284.8 eV. The Pore size distribution was measured using the Brunauer Emmett Teller (BET) method as-examined on a Micromeritics ASAP2020 nitrogen adsorption-desorption apparatus. The UV-vis-NIR diffuse reflectance spectra (DRS) of the samples were tested on a UV-vis-NIR spectrophotometer (Hitachi UV-3101) with an integrating sphere attachment within 200–1800 nm range. The photoluminescence (PL) spectra were acquired at room temperature with a FLS920 fluorescence spectrometer under the ultraviolet excitation of 325 nm.

The photocatalytic activity of the prepared samples was evaluated toward the oxygen evolution in a quartz reactor containing 100 mL of deionized water in the presence of sacrificial agent. The photocatalytic tests were tested with 20 mg catalyst dispersed with a mechanical stirring and argon was used as a carrier gas. Deionized water was used as oxygen source and silver nitrate was added as sacrificial agent and electron acceptors with the purpose of exclusively discuss the oxygen evolution reaction. The reaction products were measured on-line by thermal conductivity detectors on a microgas chromatography GC-7920 instruments.

Based on the SEM, XPS and FTIR spectra of Bi-Ti-MOFs (Figs. S1–S3), the preparation process is schematically illustrated in Scheme 1. The Bi-Ti bimetal organic frameworks (Bi-Ti-MOFs) are firstly prepared using a hydrothermal method, which present uniform sphere structures (Fig. S1). Herein, EG is not only utilized as a solvent but also as both a reductive and organic complexing agent. During the solvothermal reaction process, partial Bi³⁺ are gradually reduced to elemental Bi⁰ and partial Ti⁴⁺ are reduced to Ti³⁺ to catalyze the growth of Bi-Ti-MOFs, which are confirmed by XPS analyses of Bi-Ti-MOFs shown in Fig. S2. As shown in Fig. S3, peaks centered at 2853.5 and 2920 cm^–1 correspond to the symmetry (ν_s) and anti-symmetry (ν_as) stretching vibration of C–H bonding in the group of –CH₂–. The peaks centered at 3361, 1455, 1069 and 884 cm^–1 can be assigned, respectively, to the ν_OH, δ_OH, ν_CO and γ_OH of C–O and O–H bonding in the group of –CH₂–OH. Subsequently, the Bi nanoparticles are uniformly dispersed on the inside and outside of reduced TiO₂ microspheres (Fig. S4) after annealing of Bi-Ti-MOFs at 200 to 350 ℃ in Ar.

The powder XRD measurement can be used to characterize the phase information (Fig. 1). In Bi-Ti-MOFs, the peak at 11.63° can be attributed to graphite oxide. The metal-organic network layers interact with each other through the hydrogen bonding and intermolecular interaction to form a graphite-like layered structure. Upon annealing at 200 ℃ (curve (b)), only the interlayer spacing is slightly enlarged to some extent owing to hydrate evaporation within the Bi-Ti-MOF layer. When the annealing temperature is increased to 250 ℃ (curve (c)), the Bi-Ti-MOF structure starts to decompose into Bi@R-TiO₂. All peaks can be perfectly indexed to anatase phase of TiO₂ (JCPDS No. 21-1272) and rhombohedral structure of Bi (JCPDS No. 05-0519) [19]. However, there is no diffraction peak of Ti₂O₃ or Ti₃O₅ in the Bi@R-TiO₂-250 samples, which may be ascribed that the low amount of reduced TiO₂. Further increasing the annealing temperature, the XRD diffraction peaks still belong to TiO₂ and Bi.

In order to better understand the growth of the Bi@R-TiO₂ microspheres, XPS of Bi@R-TiO₂ microspheres was performed. The survey spectrum in Fig. S5 indicates that the dominant elements are Ti, C, O and Bi in the Bi@R-TiO₂ composite. The high-resolution XPS of Bi 4f reveals that the two tiny peaks center at 156.9 and 162.3 eV, which can be assigned to Bi³⁺ [20]. The two strong peaks at 158.6 and 164.0 eV can be ascribed to the Bi 4f_7/2 and Bi 4f_5/2 of metallic Bi⁰, respectively [21]. The probable reason for this phenomenon is probably due to the fact that the amount of oxidation of metallic Bi⁰ to Bi³⁺ was very small. This result further demonstrates that the oxide layer exclusively exists on the Bi sphere surface, which could prevent the Bi metal from further oxidation [21]. Similar observation had been reported by Dong et al. [22], which could be attributed to the easy surface oxidation of Bi element in air to form a thin layer of Bi₂O₃. Fig. 2(b) shows the high resolution XPS scans over Ti 2p peaks. Two XPS peaks at 458.6 and 464.7 eV could be assigned to Ti⁴⁺ 2p_3/2 and Ti⁴⁺ 2p_1/2 [23]. Besides, the existence of peaks at 458.3 and 464.2 eV can prove the formation of Ti³⁺ [24], which is the result of reduction of Ti⁴⁺ with the treatment of EG. Because of the presence of Ti³⁺, the band gap of TiO₂ will be narrowed. The asymmetric O 1s signal (Fig. 2(c)) indicates several oxygen species coexist on the surface of the sample. After deconvolution, the O 1s XPS spectrum can be fitted with four peaks corresponding to Bi–O bands (529.5 eV), Ti–O bands (529.9 eV) and surface hydroxyl oxygen (531.5 eV) [25]. For C 1s spectra (Fig. 2(d)), the two strong peaks located at 284.8 and 288.2 eV can be attributed to graphitic –C–C and –C–O functionalities, respectively [26]. The MOFs precursor is calcined at 300 ℃ to form carbon structures.

The morphologies of the prepared samples were characterized by SEM, TEM, and HRTEM. As is shown in Fig. 3(a), the Bi@R-TiO₂-300 still consists of nearly spherical-sharped nanoparticles with the average size of 400 nm. However, the Bi metal can not be directly detected in the SEM images due to its small particle size. To obtain further information about the microstructures of these Bi nanoparticles, TEM analysis was employed. Figs. 3(b) and S4 clearly demonstrates that many uniformly distributed dots with the size of several nm were dispersed on the inside and outside of spheres, which are probably Bi nanoparticles. The lattice fringe spacings of 0.328 and 0.35 nm (Fig. 3(c)), which can be attributed to the (012) plane of Bi and (101) plane of TiO₂ [27]. It further indicates that the metallic Bi has successfully distributed on the reduced TiO₂ microspheres to form the effective heterojunctions. The EDX mapping of Bi@C-TiO_x-300 as shown in Fig. S6 suggests that the Bi, O, C and Ti elements are distributed uniformly in Bi@R-TiO₂-300.

To further analyze the textural properties of the samples, the isotherms and the pore size distributions are measured by N₂ adsorption-desorption measurement (Fig. 4). All the samples show the typical type IV isotherms with H3 hysteresis loops, suggesting the existence of mesopores [28]. As is shown in the Table S1, The BET surface area of the Bi@R-TiO₂-300 is up to 156.5 m² g^–1, which is higher than that of other samples. Therefore, such a high specific surface area of Bi@R-TiO₂-300 with mesoporous structure would be beneficial for O₂ evolution because of its abundant active edge sites to the reactants.

The adsorption properties of the samples were analyzed by UV-vis-NIR absorption spectroscopy. As is shown in Fig. 5, compared with pure TiO₂, the photoabsorption of Bi@R-TiO₂ samples undergo distinct enhancement in the entire visible and NIR light region which is due to SPR effect of the loading Bi and introduction of Ti³⁺. The Bi@R-TiO₂-200 and Bi@R-TiO₂-250 show poor light absorption due to the low amount of Bi [29]. The Bi@R-TiO₂-300 exhibits the best light absorption, which should be attributed to the suitable loading amount of Bi [30]. However, the Bi@R-TiO₂-350 exhibits a relatively poor visible-light absorption compared with Bi@R-TiO₂-300 due to the much bigger Bi particles. The band gap energy of pure TiO₂ could be simulated from the formula (αhν)^1/2∝hν − E_g, where α, h, ν, and E_g are absorption coefficient, Planck's constant, light frequency, and band gap energy, respectively (Fig. S7(a)). The fitted band gap energy of pure TiO₂ is about 3.1 eV. The band gap energy of pure TiO₂ XPS valence band spectra is performed to determine the valence band energy. As is shown in Fig. S7(b), the valence band (VB) potential of pure TiO₂ (V = +2.79 eV) is below the water oxidation levels (V=+1.23 eV) (Fig. S7(c)). On the basis of the band gaps obtained from the UV-vis-NIR diffuse reflection spectra, the conduction band (CB) edges can be estimated to be –0.31 eV for pure TiO₂ (Fig. S7(c)).

The photoelectrochemical (PEC) analysis has been carried on investigating the charge carrier separation and transfer process of the samples, which are closely related to the photocatalytic performance of photocatalysts [31]. As is shown in Fig. 6(a), it is clear that the Bi@R-TiO₂-300 exhibits the higher photocurrent density than that of TiO₂ under light illumination, which suggests that the introduction of metallic Bi has a significant effect on the electronic properties of Bi@R-TiO₂-300. This result indicates that the electron-hole transfer and separation process will be significantly enhanced in the Bi@R-TiO₂-300. Fig. 6(b) illustrates the comparison of the electrochemical impedance (EIS) measurements of the Bi@R-TiO₂-300 and TiO₂. As previously reported, a smaller arc size on an EIS Nyquist plot indicates a smaller charge-transfer resistance on the electrode surface and a higher separation efficiency of electron-hole pairs [32]. Clearly, the impedance values of composites in the dark are obviously larger than those under light illumination, which is ascribed to the enhanced electron conductivity under light irradiation [33]. Furthermore, the arc radius of Bi@R-TiO₂-300 are smaller than that of TiO₂ under light illumination, implying that the Bi@R-TiO₂-300 possesses the lower charge transfer resistance and the higher electron mobility. The reason is that the existence of metallic Bi plays a key role as an electron trap in facilitating the separation of photogenerated electrons and holes [21]. Then the recombination of photogenerated electron-hole pairs is inhibited and the photocatalytic activity is improved finally.

To investigate the separation of photoinduced charge carriers, steady-state and time-resolved PL spectroscopy is carried out. In general, a lower steady-state PL intensity implies a longer lifetime and higher separation efficiency of photo-generated charge carriers, leading to higher photocatalytic activity [34]. As shown in Fig. 7(a), pure TiO₂ has a fast recombination of photoinduced electron-hole pairs. When the metallic Bi load on the reduced TiO₂ spheres, the PL is drastically quenched (Fig. 7(a)). The recombination of the photo-induced charge carriers can be restrained by the SPR effect of the Bi nanospheres. An increased lifetime of charge carriers can be observed in Fig. 7(b), which is in agreement with the steady-state PL measurement results. The intensity-average lifetimes (τ) of Bi@R-TiO₂-300 is 2.59 ns, which is higher than that of TiO₂ (2.43 ns). It is reasonable that the increased lifetime of Bi@R-TiO₂-300 composites is related to the charge-separation efficiency.

The photocatalytic O₂ evolution over pure TiO₂, Bi-Ti-MOFs and Bi@R-TiO₂ synthesized at different temperatures without using any noble-metal co-catalyst was tested under simulated solar light irradiation. As is shown in the Fig. 8(a), pure TiO₂ and Bi-Ti-MOFs have only little O₂ evolution activity. As expected, metallic Bi deposition on the reduced TiO₂ spheres further improved the photocatalytic activity of the composite. The photocatalytic activity increases with the increasing annealing temperature from 200 to 300 ℃, which can be attributed to the improved light absorption (Fig. 5). However, with the annealing temperature increasing to 350 ℃, the photocatalytic activity decreased, which is probably due to the formation of Bi₂O₃ covering the active sites (Bi) of Bi@R-TiO₂, which block the light absorption and electron transfer [35]. The Bi@R-TiO₂-300 exhibits the most efficient photocatalytic performance (4728.709 μmol h^–1 g^–1), which is ∼5.9 and 9.4 times higher than that of pure TiO₂ and Bi-Ti-MOFs (Fig. 8(b)). This evidence apparently demonstrates that the photocatalytic activities of Bi@R-TiO₂ are highly improved by the modification of metallic Bi. Thus, the discovery that metallic Bi could behave as a superior non-noble-metal-based cocatalyst is of great significance. The photocatalytic O₂ evolution activity of Bi@R-TiO₂-300 is measured under a prolonged simulated solar light irradiation for 15 h (Fig. S8), and no obvious decay of photocatalytic activity is discerned, suggesting exceptional photocatalytic stability during the photocatalytic O₂ evolution. The high-resolution peaking-fitting XPS spectra of Bi 1s of Bi@R-TiO₂-300 after cycling under simulated solar light irradiation are measured. As shown in Fig. S9, there is almost no change in Bi³⁺ (Bi₂O₃) XPS peaks of Bi@R-TiO₂-300 before cycling and after cycling, which further confirms the excellent stability of Bi metal in Bi@R-TiO₂-300.

On the basis of the results and discussion above, a possible photocatalytic mechanism for the Bi@R-TiO₂ composites was proposed and schematically exhibited in Scheme 2. First, the SPR effect of metallic Bi propels the Bi@R-TiO₂ samples to absorb more visible and NIR light, as demonstrated in UV-vis-NIR DRS (Fig. 5). Second, Ti³⁺ ions in the Bi@R-TiO₂ composites form sublevel states below the CB of TiO₂, which reduce the band gap of TiO₂ and make the electrons of Bi@R-TiO₂ composites can be excited by the visible light. Third, the Fermi level (vs. NHE) of metallic Bi is approximately –0.17 eV [36], which is more positive than the CB of pure TiO₂ (–0.31 eV). The metallic Bi was deposited on the inside and outside of TiO₂ microspheres and functioned as an electron acceptors, which played an important role in the electron transfer [37, 38]. The electrons transfer from CB of TiO₂ to metallic Bi, which reduces the recombination of electron-hole pairs and increases the lifetime of charge carriers in TiO₂. The holes in the VB can react with H₂O to form O₂. The equation is as follows:

(1)

In summary, Bi cocatalysts are in situ deposited on the inside and outside of reduced TiO₂ microspheres (Bi@R-TiO₂ composites) with outstanding full solar spectrum photocatalytic oxygen evolution activity are fabricated via using a bimetallic MOF-derived synthesized strategy by adjusting the synthesizing temperature. During the hydrothermal process, Ti⁴⁺ is reduced to Ti³⁺ to form reduced TiO₂ microspheres and Bi³⁺ is reduced to metallic Bi nanoparticles by EG and then deposited on the reduced TiO₂ microspheres. The introduction of Ti³⁺ reduces the bandgap of TiO₂ and extends the visible light absorption of TiO₂. Metallic Bi also has a positive effect on the morphology as well as the optical and electronic properties of TiO₂. SPR effects of Bi nanoparticles enhance the visible and NIR light absorption of Bi@R-TiO₂. The electrons excited by reduced TiO₂ would transfer to metallic Bi, which promotes the separation of charge carriers. Among the as-synthesized samples, the Bi@R-TiO₂-300 sample with an appropriate amount of Bi nanoparticles exhibits the most efficient full solar spectrum photocatalytic oxygen evolution activity (4728.709 μmol h^–1 g^–1), which is 5.9 and 9.4 times higher than that of pure TiO₂ and Bi-Ti-MOFs. This work not only provides new insights into the in situ fabrication of Bi/semiconductor nanocomposites but also opens a new avenue for the modification of photocatalysts with non-noble metal as cocatalyst to achieve a highly enhanced performance.

Acknowledgments

The authors are thankful for fundings from the National Natural Science Foundation of China (51872173 and 51772176), Taishan Scholarship of Young Scholars (tsqn201812068), Natural Science Foundation of Shandong Province (ZR2017JL020), Taishan Scholarship of Climbing Plan (tspd20161006), and Key Research and Development Program of Shandong Province (2018GGX102028).