Light Control-Induced Oxygen Vacancy Generation and In Situ Surface Heterojunction Reconstruction for Boosting CO2 Reduction

The weak adsorption of CO2 and the fast recombination of photogenerated charges harshly restrain the photocatalytic CO2 reduction efficiency. The simultaneous catalyst design with strong CO2 capture ability and fast charge separation efficiency is challenging. Herein, taking advantage of the metastable characteristic of oxygen vacancy, amorphous defect Bi2O2CO3 (named BOvC) was built on the surface of defect-rich BiOBr (named BOvB) through an in situ surface reconstruction progress, in which the CO32− in solution reacted with the generated Bi(3−x)+ around the oxygen vacancies. The in situ formed BOvC is tightly in contact with the BOvB and can prevent the further destruction of the oxygen vacancy sites essential for CO2 adsorption and visible light utilization. Additionally, the superficial BOvC associated with the internal BOvB forms a typical heterojunction promoting the interface carriers’ separation. Finally, the in situ formation of BOvC boosted the BOvB and showed better activity in the photocatalytic reduction of CO2 into CO (three times compared to that of pristine BiOBr). This work provides a comprehensive solution for governing defects chemistry and heterojunction design, as well as gives an in-depth understanding of the function of vacancies in CO2 reduction.


Introduction
The over-reliance on fossil fuels has boosted the industrialization of the world during the past hundred years; however, it also caused the emission of a large amount of greenhouse gas carbon dioxide (CO 2 ) [1,2]. Controlling or reducing the concentration of CO 2 in the atmosphere is very important for addressing mentioned environmental problems [3]. Several prevalent strategies, such as electrochemical CO 2 reduction [4], CO 2 hydrogenation [5,6], and photocatalytic CO 2 reduction [7,8], were developed as potential solutions for future CO 2 capture and conversion. In 1978, Halmann et al. first reported the photocatalytic reduction of CO 2 to produce chemical fuels by using a semiconductor photocatalyst. The photocatalytic CO 2 reduction has attracted more and more attention. Up to now, many semiconductor-based materials, such as ZnO, GaN, ZrO 2 , Bi 2 WO 6 , TiO 2 , and C 3 N 4 , have been developed as functional catalysts for photocatalytic CO 2 reduction, proving that CO 2 can be converted into CO, CH 4 , methanol, and other valued chemicals using H 2 O sacrificial agent [9][10][11][12][13][14][15][16]. After decades of exploration and development, many excellent achievements have been obtained, but the practical application of photocatalytic CO 2 reduction is severely limited by the low activity and the poor stability of the existing Molecules 2023, 28 catalysts. Theoretically, the final efficiency of solar energy utilization is determined by three steps: light capture and carrier generating, migration and separation of electron-hole pairs, and surface reduction at the active sites. Accordingly, the ameliorated efficiency of the above-mentioned steps synergistically is the emphasis on photocatalyst design. Bismuth oxyhalides BiOX (X = Cl, Br, and I), as sillén structure materials containing [Bi 2 O 2 ] layer interleaved between two X layers, have attracted worldwide focus in the photocatalysis field recently given rise to prominent properties, including composition adjustability, chemical stability, low toxicity, and inexpensiveness [17][18][19][20]. Nevertheless, the photocatalytic performance of BiOX catalysts is still limited due to the fast recombination of carriers and the lacking catalytic active sites [21,22]. Currently, many approaches have been explored to improve the separation efficiency of photogenerated carriers of BiOBr catalysts, such as doping with other metal or nonmetallic atoms, surface vacancy designing, morphologies adjustment, heterojunction construction, and cocatalyst modification, etc. [23]. The internal mechanism for enhancing the photocatalytic activities in the above approaches can be typically explained in three facts: extending the light absorption, promoting the separation of the carriers, and building more active reaction sites, which are also considered as main challenges in highly efficient catalyst designing and future practical application. After the photocatalysts were excited by a certain wavelength of light, photo electron holes were produced. While the migration distance of the generated carriers is usually limited, and this means not all the carriers can migrate to the surface of catalysts. Most of the photo-generated electron-hole pairs are recombined during transmission. What is more, the further going on of the reaction needs appropriate reaction sites. The final finish of the photocatalytic reaction must combine all the above steps. In a word, photocatalytic reactions are complicated, and either of the steps can be the rate-determining step. The designing of highly active catalysts is a systematic project. Among the BiOX catalysts, BiOBr shows visible light response-ability and has a proper band gap position compared to the BiOCl and BiOI, which are good candidates for CO 2 reduction. Recently, lots of effort have been performed to improve the activity of the BiOBr. Wu et al. [24] prepared a kind of Gd 3+ doped BiOBr material, and they found that the doping of Gd 3+ can widen visible light response compared to the pure BiOBr. Additionally, the Gd 3+ doped BiOBr more negative conduction band position, which is beneficial to CO 2 reduction. The enhanced light response was considered the main reason for the improved performance. Mi et al. designed a series of BiOBr nanosheets with exposed different sizes and crystal facets. [25] Due to the surface energy difference of different facets, an internal electric field is formed between the facets, which can force the migration and separation of the photo electron-hole pairs. As a result, the activity of the BiOBr nanosheet was improved. Constructing heterojunction is the widely used approach for facilitating interface carrier separation. Giving rise to the potential bandgap differences, the formed internal electric field at the surface can separate the carriers [23,26]. For instance Guo's team reported a novel Bi/BiVO 4 /V 2 O 5 and the properties of the ternary catalyst in water oxidation were studied. The optimized Bi/BiVO 4 /V 2 O 5 exhibited a much better activity than BiVO 4 catalyst. The authors prove that the enhanced performance was attributed to the synergistic effect of the formed Bi/BiVO 4 /V 2 O 5 heterojunction structure, which can greatly enhance the separation efficiency of the photogenerated carriers [27]. In another recent research, a Z-scheme Bi 4 [28]. What is more, the oxygen vacancies are helpful for the adsorption and activation of carbon dioxide and are proven to the forming of COOH* intermediate.
All the above strategies are ingenious in building transmission channels or creating activation sites; however, how to systematically integrate the above advantages used in different tactics is still a huge challenge and rarely reported. Recently, the importance of surface reconstruction theory was proposed and developed to design highly efficient catalysts. The surface reconstruction theory also helps to understand the true catalytic active site of catalysts. Kibria and the co-authors made use of the surface reconstruction route in the preparation of the CO 2 electroreduction catalyst [29]. Using CuCl as the precursor, a Cu-based catalyst owing to the advantages of oxidation state and morphology was constructed through a wet-oxidation method, which helps the tuning of C 2+ selectivity in CO 2 reduction. Li's group prepared an oxygen-doped BiSI catalyst containing rich sulfur vacancies utilizing the surface reconstruction route. The surface BiSI was oxidized slightly by controlling the reaction conditions, which caused the generation of an O-doped BiSI layer. As a result, a special BiSI/O-doped BiSI catalyst was constructed and showed an enhanced Cr(VI) reduction activity because of the formed tight contact interface, which can hugely boost the migration of the photogenerated carriers and help the adsorption of the Cr(VI) on the surface [30].
Here, in this work, based on the chemical nature of vacancies in BiOBr material and the surface reconstruction strategy, a novel BO v B/BO v C photocatalyst was prepared using BiOBr as raw material through an in-situ surface reconstruction induction progress. In detail, oxygen vacancies rich BiOBr was first prepared through a UV light irradiation method. Under the irradiation of UV light, the deep-level electrons were excited, and some of the Bi 3+ atoms were reduced to a lower valance state which induced the formation of oxygen vacancies. During the photocatalytic CO 2 reduction progress, the defect sites were attacked by CO 3 2− in solution and generated amorphous BO v C, which has a mass of vacancies. This kind of formed heterojunction was caused by in situ phase-changing progress, which contains a tight interface and benefits the transferring of electrons. The amorphous BO v C contains amounts of oxygen vacancies that are pivotal for the adsorption and activation of carbon dioxide. This study offers a thorough understanding of how to design advanced photocatalysts with synergistic defect and heterojunction engineering advantages.

Structural Characterization and Morphological Analysis
The morphologies changing process were investigated with the Scanning electron microscope (SEM) and Transmission electron microscope (TEM). Figure 1a-c and d-f presents the SEM and TEM images of pristine BOB, BO v B, and BO v B/B 2 O v C-5 photocatalysts, respectively. From the SEM results, it is found that the pristine BOB sample is composed of micro sheets with smooth surfaces. After the irradiation treatment, much fragmentation occurred on the surface BO v B, which is due to the morphology structure destruction derived from the stirring process. Through the final reaction in saturated CO 2 solution, nanoflakes formed on the surface of the micro sheets, and the surface transformation maybe is caused by the conversion of BOBr to Bi 2 OCO 3 . Similar results are also observed in the TEM images, and it can be concluded from Figure 1f that BO v B/BO v C are composed of a shaggy shell and crystalline core, which is entirely different from the pristine BOB ( Figure 1d) and BO v B ( Figure 1e). Additionally, the surface morphologies of all BO v B samples were presented in Figure S1, and it is clearly observed that there is much more fragmentation occurred on the surface BO v B with the prolonged irradiation time.
The crystal structure of the prepared BOB sample was characterized, and the results were presented in Figure 2. As shown in the XRD patterns, both the pristine BOB and UV light-treated samples have intense and distinct diffraction peaks, which indicate the purity and good crystallinity of the samples. It also means the forming of oxygen vacancies didn't destroy the major structure of the BOB. The series of peaks at around 2θ degree of 10.9, 21.9, 25.2, 32.2, 39.4, and 46.2 correspond to the (0 0 1), (0 0 2), (1 0 1), (1 1 0), (1 1 2), and (2 0 0) planes, respectively, which response to the BOB (JCPDS No. 09-0393) [31]. In addition, the diffraction intensity of (1 1 0) gets weaker with the prolong of the irradiation time, which can be due to the replacement of the oxygen atoms by oxygen vacancies, which weakens crystallinity. In addition, the BO v B/BO v C-5 sample obtained after a photocatalytic reaction has been performed the XRD test. As shown in Figure 2, it should be noted that we did not find the diffraction peaks of Bi 2 OCO 3 after CO 2 reduction progress, and this can be due to the amorphous properties of the formed Bi 2 OCO 3 . The crystal structure of the prepared BOB sample was characterized, and the results were presented in Figure 2. As shown in the XRD patterns, both the pristine BOB and UV light-treated samples have intense and distinct diffraction peaks, which indicate the purity and good crystallinity of the samples. It also means the forming of oxygen vacancies didn't destroy the major structure of the BOB. The series of peaks at around 2θ degree of 10.9, 21.9, 25.2, 32.2, 39.4, and 46.2 correspond to the (0 0 1), (0 0 2), (1 0 1), (1 1 0), (1 1 2), and (2 0 0) planes, respectively, which response to the BOB (JCPDS No. 09-0393) [31]. In addition, the diffraction intensity of (1 1 0) gets weaker with the prolong of the irradiation time, which can be due to the replacement of the oxygen atoms by oxygen vacancies, which weakens crystallinity. In addition, the BOvB/BOvC-5 sample obtained after a photocatalytic reaction has been performed the XRD test. As shown in Figure 2, it should be noted that we did not find the diffraction peaks of Bi2OCO3 after CO2 reduction progress, and this can be due to the amorphous properties of the formed Bi2OCO3.  The crystal structure of the prepared BOB sample was characterized, and the results were presented in Figure 2. As shown in the XRD patterns, both the pristine BOB and UV light-treated samples have intense and distinct diffraction peaks, which indicate the purity and good crystallinity of the samples. It also means the forming of oxygen vacancies didn't destroy the major structure of the BOB. The series of peaks at around 2θ degree of 10.9, 21.9, 25.2, 32.2, 39.4, and 46.2 correspond to the (0 0 1), (0 0 2), (1 0 1), (1 1 0), (1 1 2), and (2 0 0) planes, respectively, which response to the BOB (JCPDS No. 09-0393) [31]. In addition, the diffraction intensity of (1 1 0) gets weaker with the prolong of the irradiation time, which can be due to the replacement of the oxygen atoms by oxygen vacancies, which weakens crystallinity. In addition, the BOvB/BOvC-5 sample obtained after a photocatalytic reaction has been performed the XRD test. As shown in Figure 2, it should be noted that we did not find the diffraction peaks of Bi2OCO3 after CO2 reduction progress, and this can be due to the amorphous properties of the formed Bi2OCO3.

Analysis of UV-Vis Absorption Spectra
As it is known that the forming of oxygen vacancies will induce the generation of defect states, the presence of defect states will fabricate an intermediate energy level near the conduction band [32]. Theoretically, the intermediate energy level can accept the electrons excited from the valance band, in turn causing the broadening of the light absorption range. To further illustrate the influence of oxygen vacancy defects for enhanced photocatalytic performance, we investigate the optical properties of pristine and vacancies-rich samples through UV/Vis diffuse reflectance spectra. As shown in Figure 3, both vacancies-rich BO v B-5 and BO v B/BO v C-5 present strong absorption in the range of the visible light region compared to the pure BOB. While the absorption intensity of BO v B/BO v C-5 gets weaker compared with the vacancies-rich BO v B-5. From the optical properties, we can conclude that the existence of oxygen vacancy does affect light absorption properties and widen the light response region. The weakened light absorption intensity of the BO v B/BiO v C-5 sample indicates the consumption of defects by the CO 2− 3 .
As it is known that the forming of oxygen vacancies will induce the generation of defect states, the presence of defect states will fabricate an intermediate energy level near the conduction band [32]. Theoretically, the intermediate energy level can accept the electrons excited from the valance band, in turn causing the broadening of the light absorption range. To further illustrate the influence of oxygen vacancy defects for enhanced photocatalytic performance, we investigate the optical properties of pristine and vacancies-rich samples through UV/Vis diffuse reflectance spectra. As shown in Figure 3, both vacanciesrich BOvB-5 and BOvB/BOvC-5 present strong absorption in the range of the visible light region compared to the pure BOB. While the absorption intensity of BOvB/BOvC-5 gets weaker compared with the vacancies-rich BOvB-5. From the optical properties, we can conclude that the existence of oxygen vacancy does affect light absorption properties and widen the light response region. The weakened light absorption intensity of the BOvB/BiOvC-5 sample indicates the consumption of defects by the CO .

Raman and EPR Analyses
The generation and vacancies concentration in the catalysts were further characterized using Raman spectroscopy and EPR spectra tests, as shown in Figure 4. In the Raman spectra results (Figure 4a), the peaks located at around 91 and 113 cm −1 are assigned to the signal of the A1g internal Bi-Br stretching mode, whereas the weak peak at 162 cm −1 is related to the Eg internal Bi-Br stretching mode [32,33]. It apparently regularly weakens the Raman peaks by prolonging the irradiation time, which can be attributed to the gradual distortion of the crystal structure after the inducing of oxygen vacancies. To further prove the relation between oxygen vacancies generation and the irradiation operation, electron paramagnetic resonance (EPR) analyses tests were given, as shown in Figure 4b, and the signals significantly enhanced at around g = 2.003 as the prolonging of the irradiation time, which means the increase of the vacancy's concentration [33,34].

Raman and EPR Analyses
The generation and vacancies concentration in the catalysts were further characterized using Raman spectroscopy and EPR spectra tests, as shown in Figure 4. In the Raman spectra results (Figure 4a), the peaks located at around 91 and 113 cm −1 are assigned to the signal of the A1g internal Bi-Br stretching mode, whereas the weak peak at 162 cm −1 is related to the Eg internal Bi-Br stretching mode [32,33]. It apparently regularly weakens the Raman peaks by prolonging the irradiation time, which can be attributed to the gradual distortion of the crystal structure after the inducing of oxygen vacancies. To further prove the relation between oxygen vacancies generation and the irradiation operation, electron paramagnetic resonance (EPR) analyses tests were given, as shown in Figure 4b, and the signals significantly enhanced at around g = 2.003 as the prolonging of the irradiation time, which means the increase of the vacancy's concentration [33,34].

XPS Characterization
The surface chemical composition change progress during the reaction was further characterized through X-ray photoelectron spectroscopy (XPS) technology, and the spectrum results are presented in Figure 5. In the C1s spectrums (Figure 5a), the existing single peak at 284.6 eV excludes the influence of carbon impurity on the surface of the pristine

XPS Characterization
The surface chemical composition change progress during the reaction was further characterized through X-ray photoelectron spectroscopy (XPS) technology, and the spectrum results are presented in Figure 5. In the C1s spectrums (Figure 5a), the existing single peak at 284.6 eV excludes the influence of carbon impurity on the surface of the pristine BOB. After irradiation 5 h, there is one obviously raised peak at around 288 eV, and this peak is attributed to the surface absorbed CO 2 [35][36][37]. As is known, the oxygen vacancies at the material surface are metastable and can be oxidized or occupied by other anions, and based on this rule; the vacancies-rich BO v B-5 was treated in the saturated CO 2 solution. From the results, it can be seen that two peaks at 285.9 and 289.1 eV appeared, which responded to the binding energy of C-O and C=O groups of the CO 2− 3 [31]. The insertion of CO 2− 3 can also be confined in the O1s spectrums in Figure 5b, three similar peaks occurred at around 520, 531, and 532 eV in both BOB, BO v B-5, and BO v B/BiO v C-5 samples, which corresponded to the lattice oxygen, and oxygen vacancies, and surface adsorbed oxygen species, respectively [38]. It is worth noting that the peak intensity of BO v B and BO v B/BiO v C at 532.1 eV was much more enhanced than the BOB sample, indicating the higher intensity of oxygen vacancy. In addition, the BO v B/BiO v C-5 sample owned a stronger surface adsorbed oxygen peak, indicating the insertion of CO 2− 3 [36,39]. In addition, the oxygen vacancy intensity of BO v B/BiO v C is also enhanced compared with BiO v C-5, which may be caused by the amorphous property of surface Bi 2 O v CO 3 . The low valance Bi 3−x signal peak in the Bi 4f spectrum of BOB-5 sample (Figure 5c) also illustrates the formation of oxygen vacancies. The binding energy around 68.2 and 69.3 eV is related to Br 3d5/2 and 3d3/2 respectively (Figure 5d), which is assigned to the monovalent oxidation state Br [40]. The XPS results elucidate the forming progress of oxygen vacancies and heterojunction structure.

Researches on Photocatalytic Performance and CO2 Reaction Path
The photocatalytic CO2 reduction performance of the prepared catalysts was evaluated in a quartz reactor containing saturated CO2 under visible light irradiation (λ > 420

Researches on Photocatalytic Performance and CO 2 Reaction Path
The photocatalytic CO 2 reduction performance of the prepared catalysts was evaluated in a quartz reactor containing saturated CO 2 under visible light irradiation (λ > 420 nm), and the temperature of the quartz reactor was steadily kept at 15 • C. Figure 6a is the results of CO yield in 4 h, and it was found that the activity was gradually enhanced with the increasing intensity of oxygen vacancies, and the BO v B/BO v C-5 shows the best CO 2 reduction activity of 0.518 µmol/g, which is nearly 3 times of the pristine BOB (0.175 µmol/g). As mentioned, the formation of BO v C relied on the generation of oxygen vacancies, which can provide a mass of low-valance Bi 3−x to react with the CO 2− 3 and form BO v C. The CO 2 adsorption isotherms were performed under ambient conditions (298 K), and the results are shown in Figure S2. It could be observed that adsorption capacity is linearly related to the oxygen vacancy concentrations, which also illustrates the critical role of vacancies in the BO v B-X. The decay of activity was owing to formed recombination centers caused by the existence of excess oxygen vacancy. The enhanced activity indicates the success of the surface modification strategy. The stability of the photocatalyst was also investigated, and the results are presented in Figure 6b. In the three cycles test, the activities have no obvious change, proving the good stability of the catalyst. To investigate the internal mechanism of the CO 2 reduction reaction, the transient photocurrent and electrochemical impedance spectra (EIS) tests were carried out to confirm the generation and separation properties of the carriers. As shown in Figure 6c, the BO v B/BO v C-5 exhibits a higher photocurrent response compared with the pristine BOB and oxygen defect BOB. The EIS results (Figure 6d) indicate the interfacial charge transfer efficiency and the smaller arc radius of the EIS Nyquist plots means smaller charge transfer resistance. As presented, BO v B/BO v C-5 shows the best separation efficiency of the carriers. smaller arc radius of the EIS Nyquist plots means smaller charge transfer resistance. As presented, BOvB/BOvC-5 shows the best separation efficiency of the carriers. To deeply understand the possible paths of CO2 reduction, the in situ FTIR spectra were used for the signals collection of the reaction intermediates, as shown in Figure 7. As the reaction went on, the characteristic absorption peaks of HCO3 − (1095 cm −1 and 1360 cm −1 ), m-CO2 − (1215 cm −1 ), CO2 − (1670 cm −1 ), and COOH* (1452 cm −1 ) were clearly identified in the spectra results. From the in situ FTIR results, it can be concluded that the CO2 molecules were fixed onto the surface of the catalyst and formed into HCO3 − . Then, the photogenerated electrons were captured by HCO3 − and CO2 − was produced. The generated m-CO2 − was further transferred to COOH*, which is the key intermediate for CO evolution. The in situ FTIR results can give clear proof for the CO evolution path [41,42]. To deeply understand the possible paths of CO 2 reduction, the in situ FTIR spectra were used for the signals collection of the reaction intermediates, as shown in Figure 7. As the reaction went on, the characteristic absorption peaks of HCO 3 − (1095 cm −1 and 1360 cm −1 ), m-CO 2 − (1215 cm −1 ), CO 2 − (1670 cm −1 ), and COOH* (1452 cm −1 ) were clearly identified in the spectra results. From the in situ FTIR results, it can be concluded that the CO 2 molecules were fixed onto the surface of the catalyst and formed into HCO 3 − . Then, the photogenerated electrons were captured by HCO 3 − and CO 2 − was produced. The generated m-CO 2 − was further transferred to COOH*, which is the key intermediate for CO evolution. The in situ FTIR results can give clear proof for the CO evolution path [41,42].

Mechanism
Based on the foregoing experimental results, a possible mechanism using the BOvB/BOvC heterojunction for the photocatalytic CO2 reaction is proposed in Figure 8. Under visible light irradiation, BOvB is also excited to produce photo-generated electrons (e − ) and holes (h + ). The photo-generated electron transfers to the conduction band minimum (CBM), leaving a hole in the valence band maximum (VBM). The left hole could directly oxidize water molecules giving rise to O2 and protons. In addition, the photogenerated electrons on CBM flow to the CB of BOvC, which leads to the effective separation of photon-generated carriers. The e − on the CB of BOvC would reduce CO2 into CO. The origin of this enhancement of the photocatalytic CO2 reduction rate is the result of the effective separation of electron-hole pairs and the improvement of CO2 adsorption capacity derived from the oxygen vacancy.

Mechanism
Based on the foregoing experimental results, a possible mechanism using the BO v B/ BO v C heterojunction for the photocatalytic CO 2 reaction is proposed in Figure 8. Under visible light irradiation, BO v B is also excited to produce photo-generated electrons (e − ) and holes (h + ). The photo-generated electron transfers to the conduction band minimum (CBM), leaving a hole in the valence band maximum (VBM). The left hole could directly oxidize water molecules giving rise to O 2 and protons. In addition, the photo-generated electrons on CBM flow to the CB of BO v C, which leads to the effective separation of photongenerated carriers. The e − on the CB of BO v C would reduce CO 2 into CO. The origin of this enhancement of the photocatalytic CO 2 reduction rate is the result of the effective separation of electron-hole pairs and the improvement of CO 2 adsorption capacity derived from the oxygen vacancy.

Mechanism
Based on the foregoing experimental results, a possible mechanism using the BOvB/BOvC heterojunction for the photocatalytic CO2 reaction is proposed in Figure 8. Under visible light irradiation, BOvB is also excited to produce photo-generated electrons (e − ) and holes (h + ). The photo-generated electron transfers to the conduction band minimum (CBM), leaving a hole in the valence band maximum (VBM). The left hole could directly oxidize water molecules giving rise to O2 and protons. In addition, the photogenerated electrons on CBM flow to the CB of BOvC, which leads to the effective separation of photon-generated carriers. The e − on the CB of BOvC would reduce CO2 into CO. The origin of this enhancement of the photocatalytic CO2 reduction rate is the result of the effective separation of electron-hole pairs and the improvement of CO2 adsorption capacity derived from the oxygen vacancy.

Materials
KBr, Na 2 SO 4 , and Bi(NO 3 ) 3 ·5H 2 O were purchased from the Sinopharm Chemical Reagent Corporation (Shanghai, China). All materials were analytical grade and without further purification in the experimental. All used materials are analytical reagents.

Synthesis of BiOBr and Defect-Rich BiOBr Photocatalysts
Pristine BiOBr (BOB) was synthesized through the following steps: 2 mmol KBr was dispersed into 70 mL deionized water, and then, 2 mmol Bi(NO 3 ) 3 ·5H 2 O was added into the solution and continually stirred for 0.5 h at ambient temperature. Subsequently, the precursor suspension was transferred to a 100 mL autoclave and maintained at 160 • C for 12 h in an oven. The obtained product has been washed with absolute ethanol and deionized water, respectively. At last, the obtained BiOBr sample was dried at 60 • C for 6 h in an oven.
The defect-rich BiOBr was prepared via in situ photo-induced method and 0.3 g BiOBr was dispersed into 100 mL H 2 O. The 300 W Xe arc lamp was used as a light source to irradiate the above solution for 1 h, 3 h, 5 h, and 7 h, respectively, for obtaining the defect BiOBr of different oxygen vacancy content. The solutions of different irradiation periods were filtered and washed several times with deionized water, ultimately dried at 60 • C for 6 h in a vacuum oven; the obtained defect-rich BiOBr samples were marked as BO v B-1, BO v B-3, BO v B-5, and BO v B-7.

Characterization
The phase structures of samples were investigated by power X-ray diffraction (XRD) with Cu Kα radiation (λ = 0.154056 nm) on a Bruker AXS D8 advance power diffractometer, and XRD spectra were measured in the range of 2θ = 10-80. The morphologies and composition of the samples were observed by SEM and EDS using a Hitahi S-4800 microscope (Hitachi Limited, Tokyo, Japan) with an accelerating voltage of 7.0 kV. Raman spectra of the samples were recorded on the LABRAM-HR800 system with laser excitation of 532 nm. The spectra were recorded in a shift range of 50-600 cm −1 . High-resolution transmission electron microscopy (HRTEM) measurements were performed by a JEOL-2100 microscope (Japan Electronics Co., Ltd. (JEOL) Tokyo, Japan) at an acceleration voltage of 200 kV. The preparation process of this test sample is as follows: A small amount of sample was added to 1 mL of ethanol, ultrasonic dispersion for 2 min, and then an appropriate amount of suspension was added to the net copper surface, drying with an infrared lamp. X-ray photoelectron spectroscopy (XPS) was obtained on a Thermo Fisher Scientific, Waltham, MA, USA (ESCALAB 250) spectrometer with the multichannel detector, and C 1s as a signal-calibration standard of binding-energy values at 284.6 eV. Ultraviolet-visible (UV-vis) absorption spectra were recorded from 800-200 nm by a Shimadzu UV-2600 spectrophotometer and using Ba 2 SO 4 as the reflectance standard sample. The CO 2 adsorption isotherms were carried out by A Micromeritics ASAP 2020 analyzer (Beijing Builder electronic technology Co., Ltd., Beijing, China). The in situ FT-IR was carried out using FT-IR 4200 Jasco spectrometer (Tokyo, Japan) equipped with a diffuse reflectance accessory. The spectrum was recorded in the wavenumber range of 2200-1000 cm −1 . Photocurrent and Electrochemical impedance spectroscopy were investigated by CHI660E electrochemical, using 0.5 M Na 2 SO 4 aqueous solution as an electrolyte solution, Pt as a counter electrode, and Ag/AgCl as reference electrodes. The photocatalysts were deposited on ITO conductive glass to be applied as the working electrode. The preparation method of the working electrode is as follows: a suitable amount of photocatalyst was first mixed with a small amount of ethanol solution. The obtained mixed suspension was ground for 15 min, then a proper amount of supernatant was taken out and spin-coated on ITO glass using the Spin Coater (KW-4A, Institute of Microelectronics, Chinese Academy of Sciences, Beijing, China). At last, the obtained working electrode was dried at 60 • C for 2 h in a vacuum oven. The used light source was a 300 W xenon lamp (PLS-SEX300, Beijing Trusttech CO., Ltd., Beijing, China) (wavelength > 420 nm). A short photocurrent density measurement was performed during the ON/OFF cycle for 110 s.

Photocatalytic CO 2 Reduction
The photocatalytic CO 2 test is carried out using a quartz reactor. First, 100 mg of the sample was mixed with 100 mL of deionized water. Subsequently, we sealed it and continuously bubbled high-purity CO 2 into the reactor for 15 min. During the whole reaction process, the reactor was kept at 15 • C by using cooling water circulation equipment. The used light source was a 300 W Xe arc lamp (PLS-SEX300, Beijing Trusttech Co., Ltd.) (wavelength > 420 nm). At the one-hour interval, the gas samples were obtained using needle tubing. And the reaction products have been analyzed by Varian CP-3800 gas chromatograph (FID detector, Porapak Q column, and the N 2 gas was used as the carrier gas). The stability of the photocatalyst was also carried out according to the above method.

Conclusions
Defects chemistry has been proven efficient strategy to provide active sites and accelerate the catalytic activity. For BiOX materials, oxygen vacancy was usually considered to enhance CO 2 adsorption and widen the optical response range in CO 2 reduction. Here, taking advantage of the metastable property, the defect-rich BO v B/BO v C photocatalyst was prepared through the reaction of Bi 3−x and CO 2− 3 . After the surface reconstruction progress, the photocatalyst was composed of oxygen vacancy-rich BO v B and surface amorphous BO v C. The formed heterojunction catalyst achieves multiple functions: the oxygen vacancy realizes better visible light absorption of the BiOBr and CO 2 activation; BO v C was generated through an in situ phase changing progress, and this kind of tight contact interface is beneficial for carriers' migration; the formed BO v C layer will provide protection and avoid the oxidization of vacancies by the O 2 . As a result, the defect-rich BO v B/BO v C shows better activity and good stability in photocatalytic CO 2 reduction. This study provides a new view for the design of highly efficient photocatalysts which collaborate defect and heterojunction advantages. Data Availability Statement: Data will be made available on request.