Photocatalysis has become one of the most promising energy technologies on account of its potential to provide sustainable solutions to environmental and energy problems [1-6]. Semiconductor photocatalysis has been widely applied in various fields, such as environmental remediation, clean energy production, carbon fixation and chemical synthesis [1-6]. Alternatively, the newly emerging class of plasmonic photocatalysts based on noble metals has the potential to significantly boost the photocatalytic performance of semiconductors. When the incident frequency of the photons matches the oscillating frequency of the conduction-band electrons in metallic nanoparticles, the nanoparticles undergo the localized surface plasmon resonance (LSPR) effect. This enhances the local electromagnetic fields around the metal surface to improve its optical trapping capability and photoelectric conversion rate, thus enabling some metals to perform as photocatalysts [7-10]. Both noble metals (Au, Ag, Pd) and non-noble metals (Cu) exhibit LSPR characteristics under visible-light excitation [8, 11-13]. It has been found that the optical response of some non-noble metals (Bi and Al) can be tuned from ultraviolet (UV) light to the visible-light region [14, 15].

Bismuth, the most abundant non-precious semimetal element on earth, has recently been found to demonstrate impressive and stable photocatalytic activity towards the removal of NO under 280 nm light irradiation, representing a new, direct plasmonic photocatalyst [14]. In combination with appropriate semiconductors such as C₃N₄ and (BiO)₂CO₃, metallic Bi as a cocatalyst can promote their photocatalytic activity by surface plasmon resonance (SPR)-based enhancement of their light-harvesting ability, and by the SPR-mediated formation of electromagnetic fields, promoting the separation of photogenerated charge carriers. In addition, Bi can perform as an electron trap to facilitate charge-carrier separation, similar to the role of noble metals in metal/semiconductor composites [16-18].

However, no study has yet been reported concerning the use of pristine elemental Bi to enhance the efficiency of plasmonic photocatalysis. Naked SiO₂ is not a viable photocatalyst because of its large band gap, which only enables it to absorb shortwave ultraviolet light [19, 20]. Based on its stability and high specific surface area, SiO₂ is usually employed as a support or composite component to enhance the thermal stability of the photocatalyst or elevate its adsorption capacity for reactants [21-23]. To date, there has been no report on the use of SiO₂ to enhance the plasmonic photocatalytic ability of metallic Bi. Moreover, the reaction pathway of the photocatalytic oxidation of NO by SiO₂@Bi remains unknown.

Herein, Bi spheres modified by SiO₂ nanoparticles were developed for the first time by a facile method. The addition of SiO₂ considerably enhanced the plasmonic photocatalytic activity of metallic Bi towards the removal of NO under 280 nm light irradiation, and ensured high chemical stability. The enlarged surface area and the formation of Bi-O-Si bonds played key roles in improving the photocatalytic performance of SiO₂@Bi. Using the Bi-O-Si bonds as a transportation channel, the photogenerated electrons in metallic Bi were rapidly transferred to the surface of SiO₂@Bi, promoting the separation and transfer of photoexcited carriers, as well as the formation of reactive radicals for oxidation. The reaction mechanism of photocatalytic NO oxidation by SiO₂@Bi was revealed with electron spin resonance (ESR) and in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS), and involved the chain reaction NO → NO₂ → NO₃^- with •OH and •O₂^- radicals as the main reactive species. A new mechanism for plasmonic photocatalysis by SiO₂@Bi composites involving Bi-O-Si bonds as transportation channels was also proposed.

All chemicals used in this study were analytical grade and were used without further purification. In a typical synthesis, 0.364 g of Bi(NO₃)₃·5H₂O was first dissolved in 10 mL of HNO₃ (1 mol/L). The mixture was then transferred into a Teflon-lined autoclave (100 mL) with stirring for 30 min. After dissolution, ethylene glycol (55 mL) was added and stirred for 30 min. 0.6 g of polyvinylpyrrolidone (PVP, molecular weight 24000) was subsequently added into the solution and stirred until dissolution. Next, 0.011 g of commercial SiO₂ nanoparticles was added into the above solution and stirred for 10 min. The aqueous suspension was then heated to 160 ℃ for 24 h. After the reaction, the black solid product was filtered, washed three times with distilled water and absolute ethanol to remove impurities, and then dried at 60 ℃ for 12 h to obtain the final SiO₂/Bi products. For comparison, Bi spheres were prepared according to the above method without the addition of SiO₂ nanospheres.

The crystal phases of the samples were analyzed by X-ray diffraction (XRD) with Cu K_α radiation (model D/max RA, Rigaku Co., Japan). X-ray photoelectron spectroscopy (XPS) with Al K_α X-rays (Thermo ESCALAB 250, USA) was used to investigate the surface properties. The microstructure of the samples was examined with a scanning electron microscope (SEM, JEOL model JSM-6490) and a transmission electron microscope (TEM, JEM-2010, Japan). The UV-vis diffuse-reflectance spectra were obtained for the dry-pressed disk samples utilizing a scanning UV-vis spectrophotometer (UV-vis DRS, UV-2450, Shimadzu, Japan) equipped with an integrating sphere assembly. Photoluminescence (PL) spectroscopy (F-7000, HITACHI) was employed to probe the charge-separation properties of the samples. Nitrogen adsorption-desorption isotherms were obtained on a nitrogen adsorption apparatus (ASAP 2020, USA). The samples were degassed at 90 ℃ prior to measurements. Time-resolved fluorescence decay spectra were taken with a fluorescence spectrophotometer (Edinburgh Instruments, FLSP-920). ESR signals of radicals spin-trapped by 5, 5-dimethyl-1-pyrroline N-oxide (DMPO) were recorded on a JES FA200 spectrometer. Samples for ESR measurement were prepared by mixing the samples in a DMPO solution (40 mmol/L) tank and irradiation with UV light at 280 nm. The aqueous dispersion was used for DMPO-•OH detection and the methanol dispersion for DMPO-•O₂^- detection.

The photocatalytic activity of the samples was investigated by the removal of NO at 1 x 10^-3 μg/mL in a continuous-flow reactor at room temperature. The detailed specifications of the reactor system can be found in our previous publications [16-18]. For photocatalytic activity testing, the light source was an 8-W ultraviolet (280 nm) lamp placed vertically outside the reactor.

In situ DRIFTS measurements were conducted using a Bruker Tensor Ⅱ Fourier transform infrared spectroscopy (FTIR) equipped with an in situ diffuse-reflectance cell (Harrick) as shown in Scheme 1 and Fig. 1. Photocatalyst samples were put into the cell. First, He gas (50 mL/min) was used to remove residual hydrocarbons, H₂O and CO₂. The real-time FTIR spectrum obtained after ventilation was utilized as the background. Then, the reaction mixtures (25 mL/min NO, 25 mL/min O₂) were introduced into the cell. The NO was allowed 20 min to adsorb onto the catalysts. Next, the photocatalysts were illuminated by the UV light source (MUA-165) for 40 min. The real-time FTIR spectra were recorded at intervals of 8 min. The gas fluxes were kept constant throughout (25 mL/min NO, 25 mL/min O₂). Finally, after turning off the light source, the FTIR spectra were recorded at intervals of 2 min with the same gas fluxes. The full IR scanning range was 4000-600 cm^-1. To analyze the photocatalytic oxidation process on SiO₂/Bi, the relevant section of this range was 1900-1200 cm^-1.

In the XRD patterns of SiO₂@Bi shown in Fig. 2(a), all the diffraction peaks can be exactly indexed to the rhombohedral phase of elemental Bi (JCPDS PDF card 85-1329). The signature diffraction peaks for SiO₂ cannot be detected because of the low content of SiO₂ in the composites. In Fig. 2(b), the peak around 582 cm^-1 can be ascribed to the Bi-O bond, as elemental Bi is easily oxidized by O₂ in air [14].The resulting bismuth oxide layer on the surface protects Bi from further oxidation. In the XRD patterns (Fig. 2(a)), only the diffraction peaks of rhombohedral metallic Bi can be detected, which implies that the bismuth oxide layer on the surface is amorphous [14]. For the SiO₂@Bi composites, the peak arising from the Bi-O bond disappears. However, a new peak at 783 cm^-1 appears. The peaks at 466 and 1095 cm^-1 can be assigned to the symmetrical stretching vibration and anti-symmetrical stretching vibration, respectively, of Si-O-Si [24-26]. The peak at 1614 cm^-1 is from the bending vibration of O-H bonds in H₂O. Considering the composition of the catalyst, the new peak at 783 cm^-1 can probably be assigned to the Bi-O-Si bond in the SiO₂@Bi composites.

The SEM images in Fig. 3(a) and (b) show that the overall morphology of the SiO₂@Bi sample consists of microspheres and nanoparticles. The microspheres, of about 1-2 μm in size, are formed of Bi, and are surrounded by a large amount of SiO₂ nanoparticles, as shown in Fig. 3(c). Fig. 3(d) exhibits an HRTEM image of SiO₂@Bi at one edge. The lattice spacing of 0.227 nm corresponds to the (110) plane of rhombohedral Bi (JCPDS 85-1329). The lattice spacing of 0.386 nm can be ascribed to the (111) plane of SiO₂ (JCPDS 85-0419). The intimate contact between SiO₂ and Bi can be clearly observed in the SiO₂@Bi samples, which provides further evidence for the formation of Bi-O-Si linkages. The successful construction of SiO₂@Bi can also be confirmed by the EDX mapping in Fig. 4, demonstrating the uniform distribution of Bi and Si elements in the sample.

The addition of SiO₂ enlarges the specific surface area of the Bi spheres from 16 to 27 m²/g, as determined from the N₂ adsorption-desorption isotherms (Fig. 5(a)). The pore volumes are also increased upon the addition of SiO₂ (Table 1). The mesopores in SiO₂@Bi allow light to be scattered within their pore channels, and thus enhance the light-harvesting efficiency (Fig. 5(b)). The enlarged specific surface area of SiO₂@Bi increases the material's adsorption capacity for NO and thus promotes its photocatalytic activity. The specific surface area (A_BET), pore volume, peak diameter and NO removal ratio of the Bi spheres and SiO₂@Bi samples are shown in Table 1.

XPS was used to investigate the chemical state of SiO₂@Bi. Fig. 6 depicts the XPS spectra of the elements C, O, Bi and Si, which are consistent with the EDX results (Fig. 4). The peak of C is ascribed to adventitious carbon (Fig. 6(a)). Before etching, the O 1s peaks centered at 529.6 and 530.8 eV (Fig. 6(b)) can be assigned to the binding energies of Bi-O-Si and Si-O, respectively [27]. In comparison with the Si-O bond, the binding energy of Bi-O-Si is lower, so the shift of O 1s to lower energy after etching indicates that Bi-O-Si is present. After etching, an additional peak of O 1s around 533.1 eV (Fig. 6(b)), arising from SiO₂, can also be observed, indicating that some of the SiO₂ particles are incorporated into the interior of SiO₂@Bi [28, 29]. The Bi 4f spectra before and after etching have distinct differences, as shown in Fig. 6(c). The characteristic peaks centered at 156.8 and 162.1 eV can be attributed to metallic Bi, suggesting the presence of Bi-Bi bonds. The peaks around 158.9 and 164.2 eV are the characteristic peaks of Bi³⁺ ions, originating from the bismuth oxide layer formed on the surface of the SiO₂@Bi composites [18].For comparison, the intensity of the Bi-Bi peaks is significantly increased after etching, which demonstrates that the surface of the elemental Bi is easily oxidized by O₂ in air, forming a thin, amorphous bismuth oxide layer that protects Bi from further oxidation. As depicted in Fig. 6(d), the intensity of the Si 2p peak is increased after etching, which further confirms the presence of SiO₂ particles in the interior of SiO₂@Bi.

The UV-vis DRS results of the Bi and SiO₂@Bi samples are shown in Fig. 7(a). Both the samples show strong UV light absorption (250-290 nm), which can be mainly assigned to the characteristic SPR absorption of metallic Bi. The light-harvesting ability of SiO₂@Bi is enhanced in the range from 290 to 400 nm after the addition of SiO₂. The photophysical properties of the photoexcited charge carriers in Bi and SiO₂@Bi were further analyzed using ns-level time-resolved fluorescence decay spectroscopy, as illustrated in Fig. 7(b) and (c). The decay spectra can be fitted to obtain the radiative lifetime of the samples. The fitting results are given in Table 2, showing the lifetime components (τ₁) and the corresponding relative percentages of charge carriers. The lifetime (τ₁) of metallic Bi is 7.76 ns. After the addition of SiO₂, the lifetime increases to 8.34 ns, i.e., the carrier lifetime is increased by the formation of SiO₂@Bi. The prolonged lifetime of the charge carriers can be attributed to the formation of Bi-O-Si bonds. Benefiting from SPR, the photogenerated electrons in metallic Bi are rapidly transferred to the surface of SiO₂ via the Bi-O-Si bonds, which serve as hot-electron transportation channels. This promotes the separation and transfer of the photoexcited carriers, as well as the formation of reactive radicals for oxidation [30]. The prolonged lifetime of the charge carriers increases their probability of participating in photocatalytic reactions before recombination. Moreover, the increased formation of radicals enhances the photocatalytic oxidation strength of the material.

To determine which reactive radical species take part in photocatalysis, trapping experiments for superoxide (•O₂^-) and hydroxyl radicals (•OH) were performed. As depicted in Fig. 8, for both •O₂^- and •OH radicals, the signals of SiO₂@Bi are much stronger than those of Bi spheres under 280 nm light irradiation. This indicates that the addition of SiO₂ enhanced the production of both species of active radicals, presumably by increasing the charge separation via the formation of Bi-O-Si bonds.

In situ DRIFTS was then performed to further study the mechanism of photocatalytic NO oxidation by SiO₂@Bi. Fig. 9 shows the time evolution of the IR spectra during 40 min of photocatalytic treatment of NO over SiO₂@Bi with UV-light irradiation. The spectra obtained before switching the light source on and after switching it off are also presented for comparison. Table 3 summarizes the assignments of the observed bands. After introducing the mixture of NO and O₂, the NO interacted with reactive sites on the surface of SiO₂@Bi, generating surface nitrosyl species. At the same time, the surface reactions of NO with oxidative species (•O₂^- and •OH) resulted in the production of nitrates [31].The main IR peaks of the adsorbed NO₃^- species can be found in the regions of 1507-1576, 1636 and 1749 cm^-1. The IR absorptions of NO₂^- are located at 1419-1473, 1653 and 1684 cm^-1 [31-37]. After 8 min of UV-light illumination, the IR peaks in the regions corresponding to NO₂⁻ and NO₃^- become markedly stronger, which can be attributed to the fast electron transfer induced by the Bi-O-Si bonds. Fast electron transfer promotes the formation of greater amounts of •O₂^- and •OH radicals, generating nitrate/nitro species (NO + •O₂^- → NO₃^-, NO + 2•OH → NO₂ + H₂O, NO₂ + •OH → NO₃^- + H⁺). From 16 to 24 min, the peaks of both NO₂^- and NO₃^- species decline slightly, which is related to the increasing saturation of the active sites. From 32 to 40 min, the bands corresponding to the final product NO₃^- are stronger than those of the other nitrate species. The peaks for NO₂^- can be formed by two routes (NO₃^- + O^2- + 2NO → 3NO₂^- and NO₃^- + 2NO → NO⁺ + 2NO₂^-) [35]. A trace amount of NO (with peaks at 1772 and 1869 cm^-1 from 32 to 40 min) is formed by the desorption and decomposition of the main adsorbed nitrates [34-36, 38].After turning off the light for 8 min, the IR intensities for the NO₂^- and NO₃^- species decline because of the cessation of the photocatalytic reaction.

The photocatalytic activity of SiO₂@Bi was evaluated for the removal of NO in air under UV light irradiation. Fig. 10(a) shows the time evolution of the NO concentration (C/C₀) during irradiation over elemental Bi and SiO₂@Bi. C₀ and C are the initial and real-time concentration, respectively. The figure also shows the corresponding data for SiO₂, which, as an insulator, shows no photocatalytic activity for NO removal under light irradiation. Significantly, a high NO removal ratio of 53.6% is achieved in the presence of SiO₂@Bi photocatalyst, whereas Bi exhibits a relatively low NO removal ratio of 36.9%. Note that the amorphous bismuth oxide is inactive under light irradiation [14, 18]. The photoinduced electrons can pass through the thin layer of amorphous bismuth oxide via the well-known quantum tunneling effect [18]. The ability of SiO₂@Bi and elemental Bi to remove NO under photocatalytic conditions can be ascribed to the direct plasmonic photocatalytic effect of the semimetal Bi [14, 39-41]. The greatly enhanced activity of SiO₂@Bi can be attributed both to the enlarged surface areas and to the increased production of active species, benefiting from the presence of polar Bi-O-Si bonds [30]. It is noteworthy that the photocatalytic activity of SiO₂@Bi under UV light irradiation actually increases slightly after five cycles (Fig. 10(b)). Its high photocatalytic activity and photochemical stability indicate the great potential of SiO₂@Bi as a novel Bi-metal-based plasmonic photocatalyst.

The mechanism of photocatalytic NO oxidation over metallic Bi is shown in Fig. 11(a). Through surface plasmon resonance in metallic Bi, •OH and •O₂^- radicals are produced together with hot holes and hot electrons. These two species of reactive radicals are responsible for the oxidation of NO [14]. However, further improvement is necessary to suppresses the recombination of holes and electrons, which terminates the photocatalytic activity. The hybridization of metallic Bi with SiO₂ results in the formation of Bi-O-Si bonds. As indicated by the HRTEM and time-resolved PL results, the Bi-O-Si bonds are beneficial for charge separation and electron transfer, thus promoting the production of •OH and •O₂^- radicals, leading to enhanced photocatalytic efficiency of metallic Bi (Fig. 11(b)).

The first plasmonic photocatalyst to be based on SiO₂@Bi microspheres was developed by a facile method. The SiO₂@Bi nanohybrid exhibited an impressive and stable activity for the direct plasmonic photocatalytic removal of NO in air under UV light irradiation. The enhanced activity of SiO₂@Bi can be ascribed to the enlarged specific surface area and to the promotion of electron transfer by the Bi-O-Si bonds. The mechanism of plasmonic photocatalytic NO oxidation was revealed with in situ DRIFTS and ESR trapping, and involved the chain reaction NO → NO₂ → NO₃^-. The present work provides new insights into the mechanistic understanding of plasmonic photocatalysis using metallic Bi and may aid the design of high-performance Bi-based plasmonic photocatalysts.