Highly Selective Photocatalytic CO2 Reduction to CH4 by Ball-Milled Cubic Silicon Carbide Nanoparticles under Visible-Light Irradiation

The ultimate goal of photocatalytic CO2 reduction is to achieve high selectivity for a single product with high efficiency. One of the most significant challenges is that expensive catalysts prepared through complex processes are usually used. Herein, gram-scale cubic silicon carbide (3C-SiC) nanoparticles are prepared through a top-down ball-milling approach from low-priced 3C-SiC powders. This facile mechanical milling strategy ensures large-scale production of 3C-SiC nanoparticles with an amorphous silicon oxide (SiOx) shell and simultaneously induces abundant surface states. The surface states are demonstrated to trap the photogenerated carriers, thus remarkably enhancing the charge separation, while the thin SiOx shell prevents 3C-SiC from corrosion under visible light. The unique electronic structure of 3C-SiC tackles the challenge associated with low selectivity of photocatalytic CO2 reduction to C1 compounds. In conjugation with efficient water oxidation, 3C-SiC nanoparticles can reduce CO2 into CH4 with selectivity over 90%.


INTRODUCTION
Semiconductor-based photocatalysis has attracted worldwide attention due to its great potential in tackling environmental and energy issues. 1,2 Environmental photocatalysis can remove organic pollutants from water through deep oxidation, while energy photocatalysis aims to generate hydrogen through water splitting. 3−5 Recently, solar reduction of CO 2 into C 1 compounds (e.g., CO, CH 4 , CH 3 OH, and HCOOH) is considered as a new green and ideal process to address the global warming effect and simultaneously generate renewable fuels under ambient conditions. 6−11 In light of this fact, significant efforts have been devoted to synthesizing transition−metal complexes or semiconductors for photocatalytic CO 2 reduction. Unfortunately, photocatalytic CO 2 reduction usually necessitates an electron donor as the scavenger to quench photoinduced holes due to the limited water oxidation capability of many semiconductors. Meanwhile, controlling the selectivity for a desirable C 1 product is difficult as the CO 2 reduction is generally associated with complex CO 2 deoxygenation or hydrogenation processes. 12 For instance, the formation of CH 4 with lower reduction potential should be thermodynamically more favorable than the CO formation. 13 However, in reality, CH 4 necessitates the transfer of eight electron/proton pairs, thus having higher kinetic barriers than the two-electron/proton reduction of CO 2 to CO. 14 Since both the thermodynamics and kinetics of CO 2 reduction are closely related to the energy bands of photocatalysts, the development of efficient photocatalysts with desirable band structures represents a significant challenge.
To enable complete and efficient solar-driven CO 2 reduction, the chosen semiconductor band gap should be smaller than 3.00 eV and better larger than 1.23 eV, so that conduction (CB) and valence bands (VB) will respectively satisfy the thermodynamic requirement for CO 2 reduction and water oxidation. This essential criterion can be met by cubic silicon carbide (3C-SiC), a representative metal-free semiconductor with broad applications due to its suitable band gap, good thermal conductivity, high stability, and low cost. Particularly, 3C-SiC has a small band gap of 2.36 eV that enables the harvest of the visible sunlight and ideal band positions that straddle CO 2 reduction and water oxidation potential. 15−18 All of these properties endow 3C-SiC great potential for photocatalytic CO 2 reduction. As far as we are concerned, seldom studies have been devoted to investigating 3C-SiC for photocatalytic CO 2 reduction, which is essentially a cheap and high abundance material in nature.
In this work, we prepared 3C-SiC nanoparticles through a top-down ball-milling approach from 3C-SiC crystalline powders. This facile mechanical milling strategy ensures the production of 3C-SiC nanoparticles on a gram scale. The surface structure and composition analysis indicated the formation of surface states and a silicon oxide (SiO x ) shell.
The influence of the surface states and the SiO x shell on the charge-carrier dynamics and photocatalytic CO 2 reduction under visible light is comprehensively discussed.

RESULTS AND DISCUSSION
Ball milling is a typical top-down technique for the large-scale production of nanomaterials. 19,20 This process is associated with breaking and reducing the bulk solid materials into nanoscale particles through high-energy balls (Figure 1a).
According to the scanning electron microscope (SEM) images, the as-grown 3C-SiC crystalline powder we adopted contained micron-sized particles. After the ball-milling process, the 3C-SiC powder was transformed into nanosized particles (Figures 1b,c and S1). The phase structure of the as-prepared 3C-SiC nanoparticles was then investigated by X-ray diffraction (XRD), whose diffraction peaks at 35.6, 41.4, and 60.0°c orresponded well with the crystal faces (111), (200), and (220) of the cubic SiC structure, also called as β-SiC (PDF#29-1129) (Figure 1d). The Brunauer−Emmett−Teller (BET) specific surface area of the as-grown 3C-SiC (0.39 m 2 / g) was increased to 15.14 m 2 /g after the ball-milling process, according to the adsorption−desorption isotherms of N 2 ( Figure 1e).
Raman spectra revealed two vibration modes for the asgrown 3C-SiC: the transverse optical mode at 796.6 cm −1 and the longitudinal optical mode at 973.2 cm −1 . 21 Transverse optical and longitudinal optical modes are the phonon modes generated through interaction between the light and 3C-SiC. Due to the polarity of the crystalline 3C-SiC, the phonon modes can propagate either in longitudinal or transverse direction within 3C-SiC. Compared with the as-grown 3C-SiC, 3C-SiC nanoparticles showed additional Raman modes between 300 and 700 cm −1 , due to the presence of the surface Si−O species (Figure 2a). 22 This result was confirmed by the X-ray photoelectron spectroscopy (XPS) measurement. According to the high-resolution XPS Si 2p spectra, the asgrown 3C-SiC powder displayed a dominant peak at around 100.4 eV corresponding to the Si−C bond (Figure 2b). In contrast, 3C-SiC nanoparticles showed a remarkable shoulder peak at around 102.0 eV (Figure 2b). We believe that the new shoulder peak should be indexed to the surface amorphous Si− O x species, rather than crystallized SiO 2 , whose Si−O bond will appear at around 103.0 eV in the XPS spectra. 23−25 This assumption was consistent with the XRD result, as no diffraction peaks were detected for SiO 2 . Moreover, compared to that of the 3C-SiC powder, the intense peaks at 100.28 eV (typical of Si−C bond) of 3C-SiC nanoparticles shifted to lower binding energies ( Figure 2b). The intriguing bindingenergy shift was due to the localized charge decrease, possibly due to newly formed surface states, including the surface defects and dangling groups. 26−28 This is clear, as the highenergy ball-milling process could result in surface defects while the surface layer might also be oxidized due to the high temperature during the process. The amorphous SiO x was directly visualized through high-resolution transmission electron microscopy (TEM). The interplanar distance of 0.25 nm matches with the (111) plane of 3C-SiC, and an amorphous thin SiO x shell was formed on the 3C-SiC surface (Figures 2c and S2). These results demonstrated that along with the ball-milling process, an amorphous Si−O x shell was in situ formed on the nanosized 3C-SiC (inset of Figure 2c). The optical absorption of the as-prepared 3C-SiC nanoparticles was then evaluated through UV−vis absorption spectra. In contrast to commercial P25-TiO 2 with a band-edge absorption at around 390 nm, 3C-SiC nanoparticles showed broader absorption with an edge at around 520 nm and an absorption tail extending to 800 nm ( Figure S3). The extended absorption tail is also an evidence of the new surface states. The Tauc plot shown in the inset of Figure S3 reveals the band gap of 3C-SiC  nanoparticles is at around 2.36 eV. The schematic illustration of the band structure of the 3C-SiC nanoparticles is shown in Figure 2d. Based on CO 2 reduction and redox potentials of water oxidation, 3C-SiC nanoparticles can energetically reduce CO 2 into CH 4 or CH 3 OH in conjugation with O 2 evolution reaction under visible light. 6,18,29 We then collected and analyzed the CO 2 reduction products through gas chromatography (GC) to test our hypothesis. P25 was used as the reference sample, which selectively reduces CO 2 into CO with 63% selectivity ( Table 1). As for both asgrown 3C-SiC powder and nanosized 3C-SiC, CH 4 became the dominant product; however, the nanosized 3C-SiC showed over four times an enhanced linear CH 4 generation rate (4.9 μmol g −1 h −1 ) than that of the as-grown 3C-SiC powder ( Figure 3a). Compared with other single-component and sacrificial reagent-free photocatalytic CO 2 reduction systems, 3C-SiC nanoparticles display either comparable or much higher CH 4 generation rates (Table S1). The low cost and large yield for synthesizing 3C-SiC nanoparticles are promising for practical visible-light-driven CO 2 reduction (Table S1).
Meanwhile, along with CH 4 generation, O 2 simultaneously evolved, suggesting that CO 2 would react with H 2 O through two half-reactions with photoinduced electrons and holes, respectively, reducing CO 2 and oxidizing H 2 O. Moreover, the evolved O 2 amount from 3C-SiC nanoparticles was almost 2.1 times that of CH 4 , close to the theoretical value of 2.0 (CO 2 + 2H 2 O → CH 4 + 2O 2 ), exhibiting the high selectivity of 3C-SiC nanoparticles for CO 2 reduction (Figure 3b). To confirm that the reduced CO 2 was not from 3C-SiC decomposition, we then adopted isotopic 13 CO 2 as the carbon source. It was found that 13 CH 4 with a mass/charge ratio (m/z) of 17 was the only product, indicating the high stability of 3C-SiC nanoparticles (Figures 3c and S4). As observed in 3C-SiC nanoparticles, the unexpected high selectivity could be explained by its unique band structure, as shown in Figure  2d, which demonstrates that CH 4 is the most likely and thermodynamically possible CO 2 reduction product. To evaluate the stability and charge separation efficiency of the as-prepared 3C-SiC, we monitored the photocurrent response of the as-prepared 3C-SiC photoanodes under visible-light irradiation. Nanosized 3C-SiC showed a much higher photocurrent density than that of the as-grown 3C-SiC powder, consistent with its higher reactivity for CO 2 reduction ( Figure  3d). Moreover, the transient decay of the 3C-SiC powder shows a large spike in the initial stage, followed by a multiexponential decrease. However, the spike was suppressed in 3C-SiC nanoparticles, exhibiting more efficient chargecarrier separation and transfer within the 3C-SiC nanoparticles, as well as higher stability (Figure 3d). 15,30 The efficient chargecarrier separation and transfer can be further supported by electrochemical impedance spectroscopy, which revealed that the 3C-SiC nanoparticles had a much smaller impedance arc radius than that of the 3C-SiC powder (Figure 3e).
Enlightened by the above results and discussion, it is reasonable to propose that surface states are responsible for the enhanced activity of the photocatalytic CO 2 reduction. At the same time, the amorphous SiO x shell prevents 3C-SiC from oxidation. Room-temperature photoluminescence (PL) spectroscopy was conducted to determine the possible surface states and charge-carrier dynamics of 3C-SiC nanoparticles. For the 3C-SiC powder, it displayed a negligible near-bandedge emission (∼520 nm) due to the nature of its indirect band gap (Figure 4a). However, after the high-energy ballmilling process, the 3C-SiC nanopowders exhibited a much broader (from 450 to 700 nm) and more substantial emission peak centered at 550 nm, which was located at the lower energy side of the band gap of 3C-SiC and thus could not be related to the quantum confinement. This broad emission band was thus attributed to the radiative recombination through surface states. 31 In particular, given that the high-energy ballmilling process could result in many surface defects, the radiative recombination through these surface states explained the significantly increased emission band compared to the low photoluminescence emission of the as-grown 3C-SiC powder. We proposed that the photogenerated electrons could be easily trapped in the surface states below the conduction band, whose recombination with the photogenerated holes in the valence band was responsible for the defect emission (Figure 4b). To further confirm these assignments, we measured the timeresolved PL. The decaying of the PL emission band at 550 nm of 3C-SiC nanoparticles was apparently much slower than that of the 3C-SiC powder (Figure 4c). Both of the decaying curves can be fitted by a double-exponential model as illustrated in Table 2, including a fast decay component (τ 1 ) and a much slower component (τ 2 ), which are attributed to the radiative recombination from the conduction band to the valence band and the radiative recombination through the surface states, respectively. 32,33 The band-to-band recombination dominated the decaying curve of the as-grown 3C-SiC powder, based on the higher proportion of the fast decay component (τ 1 ,  68.33%). As for the 3C-SiC nanoparticles, the high proportion of long-living surface defects (τ 2 , 63.15%) suggested the efficient trapping of electrons on the surface states, resulting in a much longer τ 2 than that of as-grown 3C-SiC. As a result, 3C-SiC nanopowders show an average lifetime of 25.02 ns, almost nine times longer than that of the as-grown 3C-SiC powder ( Table 2). The surface states of 3C-SiC nanoparticles act as a trapping center of the photogenerated electrons, thus suppressing the charge recombination from the conduction band to the valence band and enhancing the charge separation ( Figure 4b). These results agree well with the transient photocurrent response and electrochemical impedance spectroscopy results. The stability test for multiple CH 4 generations revealed that the photoreactivity of 3C-SiC nanoparticles did not decay, while the 3C-SiC powder gradually lost 60% of its original photoreactivity ( Figure S5). Photocorrosion is a common phenomenon for conventional SiC due to the surface oxidation (SiC + 4H 2 O + 8h + → SiO 2 + CO 2 + 8H + ). 16,18,34 The corrosion was directly reflected by rapidly decreased photoreactivity and photocurrent response (Figures 3d and S5). Our previous studies pointed out that only if its surface was protected by cocatalysts or graphene would 3C-SiC achieve high stability for photocatalytic applications. 16,18,34 Here, the high stability of 3C-SiC nanoparticles was due to the protection of the in situ formed SiO x layer because the nanosized 3C-SiC was inclined to suffer from self-corrosion under light irradiation (Figure 4e). 3C-SiC nanoparticles maintained their original phase structure after multicycle CO 2 photoreduction according to the XRD pattern and Raman spectra ( Figure S6a,b). Meanwhile, the SiO x shell still remained amorphous ( Figure S6c). To further clarify the photocatalytic mechanism, we first evaluated the CO 2 adsorption capacity of the 3C-SiC powder and nanoparticles. According to CO 2 adsorption isotherms, 3C-SiC nanoparticles showed a much better ability to capture CO 2 than the 3C-SiC powder (Figure 4d). The enhanced CO 2 adsorption was not only due to the increased surface area but also related to the surface states that served the CO 2 binding sites, as evidenced by the CO 2 -temperature-programmed desorption (TPD) profiles. As compared with the 3C-SiC powder, 3C-SiC nanoparticles displayed two prominent CO 2 desorption peaks at 326 and 336°C, indicating that the surface defects served as the chemical CO 2 adsorption and activation sites (Figure 4e).
Meanwhile, we performed in situ Fourier transform infrared (FTIR) spectroscopy to explore the intermediates associated with CO 2 reduction to CH 4 (Figure 4f). Once the light was turned on, we were able to detect a dominant peak at 1622 cm −1 ascribed to the *COOH species in the presence of saturated CO 2 . *COOH is widely accepted as a critical intermediate for CO 2 -to-CH 4 reduction, formed through a proton-coupled electron transfer process (CO 2 + e − + H + → *COOH). 35,36 Another two absorption bands at 1042 and 1166 cm −1 were the feature bands of *CH 3 O, while the absorption band at 1089 cm −1 was the characteristic band of *CHO. Both *CH 3 O and *CHO are the important intermediate products of CO 2 -to-CH 4 reduction. 13,37,38 A shoulder peak next to *COOH at 1683 cm −1 was the asymmetric stretching of *HCO 3 . 35 Based on the in situ FTIR analysis, the selective CO 2 reduction to CH 4 was likely to proceed via the following steps   The data in the decaying curve could be correctly fitted by a doubleexponential function of I(t) = A 1 exp(-t/τ 1 ) + A 2 (-t/τ 2 ). I(t) is the intensity of the PL signal, A i are the pre-exponential factors, and τ i are the characteristic lifetimes. We calculated the average lifetime (τ average ) through τ average = B 1 × τ 1 + B 2 × τ 2 . TheB i is the relative concentration in the double-exponential decay (B i = A i /(A 1 + A 2 )).  (Figure 2d). Therefore, CH 3 OH is unlikely the selective CO 2 reduction product of 3C-SiC. For a better understanding of the selectivity of 3C-SiC for photocatalytic CO 2 reduction, preparation of 3C-SiC nanoparticles with distinct exposed facets coupled with the density functional theory calculation is highly desirable, which is one of our research goals in the future.

CONCLUSIONS
In conclusion, gram-scale 3C-SiC nanoparticles were prepared through a top-down ball-milling approach from low-priced 3C-SiC crystalline powder. This facile mechanical milling strategy ensures large-scale production of 3C-SiC nanoparticles and simultaneously induces the formation of abundant surface states and an amorphous SiO x shell. The surface states are shown to remarkably trap the photogenerated electrons and thus suppress the electron−hole recombination, while the SiO x shell prevents 3C-SiC from corrosion under visible light. Due to the unique electronic structure of 3C-SiC, it is capable of reducing CO 2 into CH 4 with a high selectivity of over 90% in conjugation with efficient water oxidation. We believe this work provides a new strategy to synthesize nanosized SiC and other nonmetallic carbides with good chemical stability and high catalytic efficiency for practical photocatalytic applications.

EXPERIMENTAL SECTION
4.1. Catalyst Preparation. 3C-SiC nanoparticles were prepared through a ball-milling process from 3C-SiC crystalline powders grown by chemical vapor deposition in a typical high-energy ball-milling machine. The whole milling process was carried out under ambient conditions for 5 h at a speed of 500 rpm. Then, the 3C-SiC nanoparticles were carefully collected and thoroughly washed with water and ethanol, and finally dried at 120°C in the air.
4.2. Photocatalytic CO 2 Reduction. We adopted a 300 W Xenon lamp with a 420 nm filter as the light source to simulate visible light. For CO 2 reduction, we uniformly dispersed 0.05 g of 3C-SiC onto a glass reactor (volume: 200 mL). The reaction setup was vacuum-treated six times before high-purity CO 2 was pumped to achieve ambient pressure. Then, we injected 1 mL of Milli-Q water into the reaction setup and simulated a gas−solid heterogeneous reaction condition. In the end, the visible light was turned on. During CO 2 reduction, we took 1 mL of gas each time using a silicone rubber septum for gas chromatography (GC: Agilent HP 6890) analysis. For an isotopic labeling experiment, 13