With increasing attention on serious air pollution problems, efficient reduction of nitrogen oxides in air has long been a hot topic in the field of photocatalysis [1]. Among various potential technologies, heterogeneous photocatalysis using semiconductor materials, such as bismuth-based layered semiconductors, has displayed promising results for environmental governance and solar energy conversion [2-12]. BiOI, which has a layered structure consisting of [Bi₂O₂]²⁺ layers sandwiched between two sheets of I ions, is a visible-light photocatalyst that can be activated by irradiation with visible light due to the narrow band gap [13, 14]. However, serious recombination of photogenerated charges prohibits the photocatalytic performance of BiOI.

Since visible light comprises the major part of solar light, a number of approaches have been developed to boost the visible light activity of photocatalysts, involving doping, metal loading, and compositing [15-25]. In particular, many studies have been conducted to prove that defects, especially oxygen vacancies, are the most reactive sites on the catalyst surface [26]. In addition, the photo-response range can be expanded by the generation of oxygen vacancies that could narrow the band gap by formation of oxygen-vacancy states located within the band gap. Moreover, oxygen defects could also serve as photoinduced charge traps to prevent electron-hole recombination and adsorption sites where the charges were transferred to adsorbed species [27]. Thus, the introduction of oxygen defects in oxide semiconductors should be conducive to improvement in visible photocatalytic activity.

To further enhance the visible light photocatalytic activity of BiOI, various modification methods such as metal deposition or doping [28-31] are applied. Integration of nanoscale plasmonic metals into photocatalysts has been utilized as an effective approach to construct visible-light-induced plasmonic composite photocatalysts [32]. Compared with noble metals such as Au or Ag, non-noble metals such as bismuth (Bi) have significant advantages of high efficiency, low cost, easy availability, and facile synthesis. UV and visible-light active semi-metal Bi with surface plasmon resonance (SPR) from the collective oscillation of its surface electrons [33, 34] enables hot carriers to transfer from the plasmonic metal to BiOI, thus enhancing the charge carrier separation of semiconductor composites [22, 35, 36].

Recently, Dong et al. [37] found that the SPR effect and oxygen defect of the semiconductor photocatalysts can significantly improve the photocatalytic performance. However, the synergistic effect of SPR and the multiple intermediate levels of energy excited by oxygen defects have not been revealed, which inspired the construction of heterojunctions with SPR and oxygen defects, simultaneously, to achieve dramatically enhanced photocatalytic performance. BiOI is a typical semiconductor photocatalyst, while pristine BiOI undergoes limited visible light photocatalytic activity. In addition, it is appealing to explore the mechanism of the synergistic integration of elemental Bi and defects for promoting the photocatalysis efficiency of BiOI.

In the present study, metallic Bi was deposited in situ on BiOI nanosheets by the reduction of partial Bi³⁺ on the surface of BiOI along with the creation of oxygen defects. Further experimental and theoretical investigations confirmed that the obtained catalysts consisted of BiOI phase, elemental Bi phase, and oxygen defects. The synergy effect of metallic Bi and oxygen defects could cause BiOI to be highly active under visible irradiation for NO removal. The detailed mechanisms for the synergistic integration of metallic Bi and defects on BiOI under visible light irradiation were proposed based on in situ DRIFTS, EPR, and DFT calculations. The conversion pathway of photocatalytic NO oxidation was also revealed. The present work could provide a new strategy for the development of high-performance photocatalysts and the understanding of the gas-phase reaction mechanisms for environmental and energetic applications.

The chemical reagents used in this experiment were analytical grade. The preparation process is as follows: 4 mmol (1.949 g) of bismuth nitrate (Bi(NO₃)₃·5H₂O) and 25 mL of distilled water were added in a 50-mL beaker and kept stirring for 30 min to dissolve. Then, 4 mmol (0.664 g) of KI was dissolved in distilled water (10 mL) and added to bismuth nitrate solution and kept stirring for 30 min. The mixture was aged in the water bath under 80 ℃ for 2 h. After the precipitate was cooled to room temperature, it was washed (twice with water and twice with ethanol and dried at 55 ℃ to obtain BiOI. Then, the prepared BiOI sample (2 mmol) was added into 20-mL distilled water and kept stirring for 20 min. At the same time, the NaBH₄ solution was prepared by adding NaBH₄ solid (0.038, 0.076, 0.113, and 0.189 g) into 40-mL distilled water. The NaBH₄ solution was dropped into the BiOI suspension and kept stirring for 20 min. The obtained sample was washed twice with water and twice with ethanol to obtain the final products, labeled Bi/BiOI-1, Bi/BiOI-2, Bi/BiOI-3, and Bi/BiOI-5.

The crystal structure of the samples was analyzed by X-ray diffraction with Cu K_α (XRD, Model D/Max RA, Rigaku Co., Japan). An X-ray photoelectron spectrometer (XPS, Thermo ESCALAB 250, USA) analyzed the chemical composition of the samples. A scanning electron microscope (SEM, JEOL Model JSM-6490, Japan) characterized the morphology of the samples. A N₂ adsorption-desorption instrument (ASAP 2020, USA) measured the specific surface area (BET) and pore structure of the samples. UV-vis DRS (UV2550, Shimadzu) and photoluminescence spectra (PL, F-7000, Hitachi, Japan) were utilized to analyze the optical properties of the samples. The electron paramagnetic resonance (EPR) measurements of the samples were performed at –196 ℃ (FLsp920, England). Electron spin resonance (ESR) spectra of radicals that were spin-trapped using 5, 5-dimethyl-1-pyrroline-n-oxide (DMPO) was employed to verify the formation of superoxide (•O₂^–) and hydroxyl radicals (•OH).

The photocatalytic activity was evaluated by monitoring the NO concentration in a continuous stream reactor at room temperature. From the prepared sample, 0.1 g was ultrasonically dispersed in a 12-mm diameter glass disk and dried at 60 ℃. After natural cooling, the mixture was placed in a rectangular reactor with a capacity of 4.5 L (30 cm × 15 cm × 10 cm), covered with a layer of quartz glass, and a 150-W commercial halogen tungsten lamp placed 20 cm above the reactor. For visible light catalytic activity tests, filters were used to remove light with wavelengths less than 420 nm from the beam. A low concentration of NO was prepared using standard air and NO standard gases with an original concentration of 100 ppm. The air flow rate was 2.4 L/min and the NO gas flow rate was 15 mL/min. The air flow and the NO gas stream were mixed by means of a three-way valve to obtain an initial concentration of 600 ppb of NO. The diluted NO was passed into the reactor, the NO concentration was stabilized, and the lamp was turned on. The NO_x analyzer (Thermo Scientific, 42i-TL, USA) was sampled every 1 min and the concentrations of NO, NO₂, and NO_x (NO + NO₂) were recorded. The removal ratio (η) of NO can be calculated as η = (1 − C/C₀) × 100%, in which C represents the NO concentration at the outlet of the reactor after turning on the lamp and C₀ represents the NO concentration when the adsorption-desorption equilibrium is reached before turning on the lamp.

Spin-polarized DFT-D2 [38] calculations were carried out utilizing the "Vienna ab initio simulation package" (code VASP5.4) [39, 40], using a universal gradient correlation function [41]. A plane-wave basis set with energy cut off at 400 eV within the framework of the projector-augmented wave method was adopted. The Gaussian smearing width was set to 0.2 eV [42, 43]. The Brillouin zone was sampled at 3 × 3 × 1 K points. All atoms were allowed to converge to 0.01 eV Å⁻¹. A 12 × 8 × 18 supercell of BiOI, including 24 oxygen atoms, 18 bismuth atoms, and 12 I atoms was employed. Then, 10 Bi atoms were supported on the surface of BiOI [44, 45].

In situ infrared test equipment was composed of a Tensor Ⅱ FTIR spectrometer (Bruker) and in situ reflection reaction (Fig. 1). The catalyst was loaded in the reaction chamber. The real-time FT-IR spectrum after heat treatment was utilized as background. Because of the detection sensitivity of the in situ FT-IR measurement, the concentration in the activity test (ppb level) is too low to be detected. Therefore, 50 ppm of NO was introduced into the reactor. The reaction gas (50 mL/min NO, 50 mL/min O₂) was then passed into the reaction chamber. After 20 min of adsorption, the catalyst was irradiated with visible light for 20 min. During the process of adsorption and reaction, the IR spectrum data were recorded every 2 min. From the infrared scanning range of 4000–600 cm⁻¹, 2200–800 cm⁻¹ can be used to analyze the process of Bi-BiOI adsorption and photocatalytic oxidation of NO.

Fig. 2 shows the XRD patterns of BiOI and Bi/BiOI-X (X = 1, 2, 3, 5). It can be observed that the diffraction peaks of BiOI belong to tetragonal phase BiOI (JCPDS-ICDD Card No. 10-0445). With the increased addition of the NaBH₄ reductant, the diffraction peaks of the Bi element (JCPDS-ICDD Card No. 44-1246) begin to appear and the intensity is gradually intensified. These results indicate that part of the Bi³⁺ in BiOI is in situ reduced to elemental Bi by NaBH₄ in the solvent thermal treatment process. The metallic Bi content is increased with increasing reductant quantity. The decrease in the diffraction intensity of Bi/BiOI-X can be ascribed to the partial transformation of BiOI to metallic Bi.

Fig. 3(a) shows the XPS spectra of the samples, which indicate that elemental Bi, O, C, and I are contained on the surface of Bi/BiOI. Fig. 3(b) shows that the binding energy at 164.5 eV (Bi 4f_5/2) and 159.2 eV (Bi 4f_7/2) can be attributed to Bi³⁺ [46], and the binding energy of 161.9 eV (Bi 4f_5/2) and 156.8 eV (Bi 4f_7/2) belong to elemental Bi [27, 47]. According to the XPS results, the molar ratio of metallic Bi is 2.12%. This result is consistent with the XRD results, further confirming that elemental Bi is formed when the reductant is used. In Fig. 3(c), two dominant peaks located at 630.9 and 619.5 eV can be assigned to I 3d. In Fig. 3d, the peak at 530.3 eV belongs to the Bi–O bond [48, 49].

Figs. 4(a) and 4(b) show the SEM images of BiOI and Bi/BiOI-2. Fig. 4(a) shows that BiOI is composed by stacked nanosheets. It can be clearly observed from Fig. 4(b) that Bi/BiOI-2 is hierarchically assembled by nanosheets, forming a superstructure. Fig. 4(c) shows that the self-assembled Bi/BiOI has a sponge-like 3D stereostructure. The TEM image (Fig. 4(d)) further demonstrates that Bi/BiOI is self-assembled by thin-layer nanosheets. Figs. 4(e) and 4(f) show the HRTEM image of a single nanosheet. The lattice spacing (0.152 and 0.099 nm) can be observed, corresponding to the (0 0 6) crystal plane of BiOI and the (0 0 12) plane of Bi [50, 51].

Figs. 5(a) and 5(b) show the N₂ adsorption-desorption isotherms and corresponding pore sizes of BiOI and Bi/BiOI-X. According to the Brunauer-Deming-Deming-Teller (BDDT) classification, the above samples belong to isotherm Ⅳ, indicating the presence of mesopores in the sample (Fig. 5(a)). The hysteresis loop of the adsorption-desorption isotherms can be classified as H3 (classified by IUPAC), indicating the presence of fractured pores from the stacking of nanosheets, consistent with SEM results (Fig. 4). From the pore-size distribution curve (Fig. 5(b)), the pore sizes of the samples BiOI and Bi/BiOI-X are all located at 2 to 100 nm, which further confirms the formation of mesoporous/microporous structure (2–100 nm). As the reductant quantity is increased to 2 mmol, the hierarchical stacking of nanosheets accounts for the formation of large pores and enlarged specific surface areas. Table 1 shows the specific surface area and pore volume of the samples; they were increased from 2.3–7.2 m²/g and 0.007–0.043 cm³/g, respectively, as the reductant quantity was increased to 2 mmol. However, when more than 2 mmol was used, the specific surface area and pore size of the samples were slightly reduced.

Fig. 6(a) displays the UV-vis diffuse reflectance spectra of BiOI and Bi/BiOI-X. We can observe that the BiOI already shows extended light absorption up to 650 nm. Furthermore, the band gap of the BiOI estimated from the intercept of the tangent to the (αhν)^1/2 vs. photoenergy (Fig. 6(b)) plot is 1.80 eV, which causes BiOI to easily absorb visible light in a wide range. Solid-state EPR spectra are recorded to demonstrate the existence of the oxygen vacancies. As shown in Fig. 6(c), BiOI has a strong signal peak at g = 2.001 in the dark, which is attributable to the existence of oxygen vacancies [52, 53]. At the same time, the signals of Bi/BiOI-2 become stronger, indicating the increased production of oxygen vacancies. The production of oxygen vacancies in BiOI can be ascribed to the reductive effect of NaBH₄ [40]. Under visible light irradiation, the intensity of the EPR signal on BiOI and Bi/BiOI-X is enhanced, indicating higher electron mobility and lower electron recombination rates [54, 55]. The density of states (DOS) of intrinsic BiOI and defective BiOI (Fig. 6(d)) are calculated to further reveal the role of oxygen vacancies. Significant intermediate levels appeared under the conduction band of the BiOI, implying that the oxygen vacancies contribute to the formation of multiple intermediate levels, which could promote electron transition from the valence band (VB) to conduction band (CB) through intermediate levels under visible light irradiation [28, 56]. With the increase in reductant quantity, the reduction ability of NaBH₄ is enhanced, and thus, a greater amount of elemental Bi and oxygen vacancies is produced [57-59]. As a result, Bi/BiOI-X exhibits full absorption in the visible region owing to the SPR effect of elemental Bi, consistent with the change in color of the sample from brick-red to black [18, 20, 28].

Fig. 7(a) shows the photoluminescence (PL) spectra for all the samples. The PL intensity of Bi/BiOI-X is significantly lower than that of BiOI, because elemental Bi in Bi/BiOI-X induces the electromagnetic field through the SPR effect, which could promote the separation of photogenerated charge carriers [60]. Fig. 7b shows the visible light photocatalytic NO removal ratio by BiOI and Bi/BiOI-X under visible light. BiOI exhibits little NO removal ratio after 30-min irradiation because of rapid recombination of photogenerated charge carriers [61, 62]. When the reductant quantity is increased from 1 to 5 mmol, the NO removal ratio of Bi/BiOI-2 is increased to 40.8%, which is much more efficient than that of BiOI (5.0%) without 3D structure [28]. Although Bi/BiOI-1 has a higher BET surface area, the electronic structure of Bi/BiOI-1 has not been fully optimized; therefore, it does not exhibit higher photocatalytic performance. The highly enhanced visible light photocatalysis of Bi/BiOI-2 can be ascribed to the synergistic effects of oxygen vacancies and metallic Bi. On the one hand, oxygen vacancies favor the transition of electrons from the valence band to an intermediate level and then to the conduction band under visible light irradiation [28, 56]. Multiple intermediate levels can be conducive to the transition of electrons and the separation of photogenerated electron-hole pairs (Fig. 7(a)). On the other hand, the SPR effect of elemental Bi can improve visible light absorption efficiency (Fig. 6(a)) and promote the separation of photogenerated electron-hole pairs. Bi/BiOI-2 exhibits strong light absorption and higher photogenerated charge separation efficiency. However, the specific surface area decreased and visible light photocatalytic performance declined when more than 2 mmol of reductant NaBH₄ was added, through the collapse of the sponge-like 3D structure of Bi/BiOI-X.

To further recover the vital roles of elemental Bi in Bi/BiOI-2, we used the DFT to calculate the electronic structure of the catalysts. Fig. 8(a) shows the charge difference distribution of elemental Bi on BiOI. Blue and yellow stand for charge accumulation and depletion, respectively. Evidently, the Bi-O layer of BiOI obtains the electrons from elemental Bi and becomes the photo-active sites. The electronic location function (ELF) (Fig. 8(b)) demonstrates the existence of a strong covalent interaction between the elemental Bi and the Bi-O layer, forming a channel for hot electrons of elemental Bi to be transferred from metallic Bi to BiOI. To explore how the excited electrons of Bi can be easily transferred to the Bi-O layer, the potentials of Bi/BiOI and BiOI were calculated. As is shown in Fig. 8(c), the potential of elemental Bi is higher than that of the Bi-O layer. The energy barrier between the Bi layer and Bi-O layer is 17.80 eV. Charge transfer from the Bi layer to Bi-O layer is easier than the transfer from the Bi-O layer to Bi layer. Under visible light irradiation, excited hot electrons from elemental Bi can easily cross the barrier and transfer to the adjacent Bi-O layer. From another aspect, we calculate the density of states (DOS) of Bi/BiOI and BiOI (Figs. 8(d) and 8€) and the partial density of state (PDOS) of elemental Bi (Fig. 8(f)). Clearly, the valence-band maximum (VBM) of Bi/BiOI is significantly extended by the contribution of the orbital from elemental Bi. This may be ascribed to the SPR effect of elemental Bi, driving the electrons of BiOI to a higher excited state to accumulate at the VBM [63]. This is beneficial for promoting the generation of photogenerated charge carriers.

Fig. 9(a) shows the hydroxyl radical (•OH) and superoxide radical (•O₂^–) signals in BiOI and Bi/BiOI-2, which is produced from water oxidation by electron-holes and oxygen reduced by free electrons. The •OH and • O₂^– signals of Bi/BiOI-2 are higher than that of pure BiOI because of enhanced charge separation and transfer. The increased production of reactive radicals is responsible for the highly enhanced photocatalytic activity of Bi/BiOI. Notwithstanding the amount of radical production, the band structure and optical absorption of Bi/BiOI-2 have been improved by oxygen vacancy and metallic Bi. Thus, the photocatalytic performance of Bi/BiOI-2 is significantly higher than that of BiOI.

Based on various experimental and theoretical results, the corresponding photocatalytic mechanism of metallic Bi@defective BiOI was proposed as shown in Fig. 10. The oxygen vacancies in BiOI induce the formation of multiple intermediate levels (Fig. 6(d)), allowing direct transition from VB to intermediate levels under visible light irradiation. This also enhances defective BiOI with visible photocatalytic activity, because the electrons from the intermediate levels can be further excited to the CB of BiOI.

Moreover, the introduction of elemental Bi in BiOI exerts a significant promotion effect on the photocatalytic performance in visible light (Fig. 7(b)). First, the SPR effect of elemental Bi produces an electromagnetic field and promotes defective BiOI with electrons in a higher excited state, accumulating at the VBM, which facilitate the separation of the photogenerated electron-hole pairs. Next, the hot electrons of elemental Bi transfer to the Bi-O layer, and the electrons from the intermediate levels transfer to elemental Bi simultaneously to maintain neutrality of elemental Bi, promoting the electron-hole pair separation of BiOI and radical production. Hence, by in situ reduction of Bi³⁺ to elemental Bi and the creation of defects on the hierarchical microsheet structure of BiOI, photocatalytic performance can be significantly improved.

Fig. 11 displays the in situ FT-IR spectra of NO adsorption in the dark and photocatalytic reaction processes on the surface of Bi/BiOI-2 under visible light irradiation. Table 2 summarizes the assignment of the corresponding absorption bands. Under dark conditions, the absorption peak of nitrates (1048 cm⁻¹) is noted after adsorption equilibrium. This could be attributed to oxygen vacancies in BiOI, which activate adsorbed O₂ to form superoxide radical • O₂^–, oxidizing the NO to nitrates [64]. A proposed workable mechanism of NO adsorption is illustrated in Scheme 1.

After visible light irradiation, enhanced absorption bands of the NO₂ polymer (N₂O₄), nitrites (at 1174 and 1327 cm⁻¹), and nitrates (at 809, 841, 1000, 1015, 1048, 1445, and 1482 cm⁻¹) can be detected. This is indicative of the active species generated under visible light, which induces the photocatalytic NO oxidation, consistent with the transient decrease in NO concentration under visible light for the fi ~2 min in Fig. 7(c). As the photocatalytic reaction progresses, the absorption peaks of different forms of nitrites and nitrates are generated, which may be originated from the NO adsorbed at diverse locations on the surface of the photocatalyst (as shown in Scheme 2). In addition, nitrites and nitrates tend to be oxidized to the more stable state by superoxo species [65]. Hence, the absorption peak at 1015 cm⁻¹ (bidentate NO₃^–) gradually shifts to 1000 cm⁻¹ (bridging NO₃^–) with illumination time (as shown in Scheme 3) [68-72].

The plasmonic metallic Bi@defective BiOI with 3D hierarchal structure was constructed by reduction of partial Bi³⁺ in BiOI. The photocatalytic removal of NO through visible light on plasmonic metallic Bi@defective BiOI was greatly improved. This is primarily attributed to the synergistic effects of oxygen vacancies and metallic Bi. The electrons from the valence band tend to accumulate at vacancies as well as intermediate levels so that the increasing charge density would cause oxygen accumulated at those sites to easily form superoxide radicals and hydroxyl radicals, which are the main active species that oxidize NO into final products. The surface plasmon resonance effect of elemental Bi enables the improvement in visible light absorption efficiency and the promotion of the charge carriers. The reaction process of photocatalytic NO oxidation was systematically studied by in situ FT-IR. NO was adsorbed on the surface of photocatalysts at different locations and would be oxidized to produce diverse forms of nitrite and nitrate. Combining DFT calculations and the reaction process, the relevant reaction mechanism of photocatalytic NO oxidation on metallic Bi@defective BiOI was proposed involving the synergistic roles of multiple intermediate levels excited by oxygen defects and metallic Bi. This study provides a new method to understand the photocatalytic reaction mechanism of heterojunction photocatalyst composites by elemental semi-metals and semiconductors, as well as the mechanisms of the SPR effect and multiple intermediate levels in cooperatively enhancing photocatalytic performance.

Acknowledgments

The authors acknowledge the AM-HPC in Suzhou, China for computational support.