Synergy of ferroelectric polarization and oxygen vacancy to promote CO2 photoreduction

Solar-light driven CO2 reduction into value-added chemicals and fuels emerges as a significant approach for CO2 conversion. However, inefficient electron-hole separation and the complex multi-electrons transfer processes hamper the efficiency of CO2 photoreduction. Herein, we prepare ferroelectric Bi3TiNbO9 nanosheets and employ corona poling to strengthen their ferroelectric polarization to facilitate the bulk charge separation within Bi3TiNbO9 nanosheets. Furthermore, surface oxygen vacancies are introduced to extend the photo-absorption of the synthesized materials and also to promote the adsorption and activation of CO2 molecules on the catalysts’ surface. More importantly, the oxygen vacancies exert a pinning effect on ferroelectric domains that enables Bi3TiNbO9 nanosheets to maintain superb ferroelectric polarization, tackling above-mentioned key challenges in photocatalytic CO2 reduction. This work highlights the importance of ferroelectric properties and controlled surface defect engineering, and emphasizes the key roles of tuning bulk and surface properties in enhancing the CO2 photoreduction performance.


Comment1
The author mentioned that this material structure is nanosheets. So, is there any possibility that bandgap dependent on the number of layers? Besides, is nanosheets stacking effects on the activity and ferroelectric properties of Bi3NbTiO9 material? What does the author think about related CO2 reduction activity? Besides, did you try to get the single layer of this nanosheet?
Comment2 Figure 1b shows that the perovskite structure sandwiched between two nanosheets. So, was there any issue of stability for the material? In the same figure 1g why is not much difference between BNT-OV2 and BNT-OVP magnetic field?

Comment3
In the case of photocatalytic activity batch reactor has used, and a source of CO2 was NaHCO3/H2SO4. So here, How has the author determined the specific concentration of CO2 evolved from NaHCO3 at a specific time?
The CO2 reacted at 1atm on the surface of photocatalyst; however, at this pressure, the CO2 adsorption it might not be good enough for photoreduction of CO2. So How authors deal with it?
During the photoreduction from Sodium bicarbonate, there might be the possibility of forming bi-dent or mon-dent carbonate interaction. Any similar author observed as if yes then add an explanation about it either not also.

SO4-ion it might find catalyst surface poising with Sulphur ion
The author has to show detailed optimization of a number of moles or concentration of CO2 from a specific amount of NaHCO3 and it.
In batch reactor at a specific amount of NaHCO3 will produce a specific concentration of CO2 and at a point after attaining the equilibrium, then there not be any more CO2. So How the author has shown that 4hrs reaction time. Because every hour's author has taken CO2 gas from the batch reactor, at the time of initial reaction, the concentration of CO2 will be different than during the 4hrs. As we can understand, after introducing the 1.3g NaHCO3 and 5ml H2SO4 after 1hr irradiation time, a certain amount of CO2 will be there but how we could say that the after 1hr also the same amount of CO2 evolution is there.
Please do experiment with the 1.3g NaHCO3 and 5ml H2SO4 fresh reaction mixture every hour and compared with current results.
Moreover, perform the TPD analysis for carbonate ion interaction with a photocatalyst, because here source of CO2 is NaHCO3

Comment4
The author has not explained about the selectivity of photoreduction Bi3TiNbO9. The after reaction between 1.3g NaHCO3 and 5ml H2SO4 produces the H2O in the reaction mixture and CO2 to CO and low amount of CH4. Thus, it can conclude that there might be the possibility of high HER. Nonetheless, HER showed low, how it has been controlled reaction process and what is a specific reason, please added relevant explanation.
Moreover, in figure 3C in all samples, the H2 evolution is the same if we compared the CO and CH4 evolution, then corresponding why did not change in the HER activities? Please add a relevant explanation.

Comment5
In figure S3, there is not much difference in the specific area, and the case of BNT showed significant low activity concerning BNT-OV2 for CO evolution or even for CH4. However, the specific area difference is not much. So, here the author focused surface engineering. So why is this contradiction? Please add a relevant brief description.

Comment6
As we mentioned earlier, in figure S15 cycling test, CO2 concentration optimization, please add this part. Moreover, please include the information about the regeneration of catalyst after each cycle. Moreover, what happened after cycle 3, is that catalyst absolutely showed deactivation or was activated. If it showed the absolute deactivation, please add the relevant information and even included the SO4-poising because of Sulphur ion's consistent contact.

Comment7
In Mott-Schottky's analysis, figure 4h and S25, we are surprised that so all flat band potentials are similar to all samples? Please add an explanation. Furthermore, both Mott-Schottky analysis at different frequencies 500 and 800HZ for figure 4h and S25 respectively, but both analyses show the similar flat band potential. Please add specify the reason.

Comment8
Please check the possibility to analysis time-resolved photoluminescence spectroscopy to calculate the charge separation time for ferroelectric materials.

Comment9
In the figure, S26 add VB and CB specific values calculated from the Mott-Schottky analysis, which might explain the reaction mechanism. Moreover, write relatively reaction mechanism for CO, CH4 and HER in brief.

Comment10
If possible, DFT calculation, then please perform DOS analysis for change in fermi level relatively changes in the oxygen vacancies. This will also help understand the reaction mechanism because as mentioned, selectivity for CO, CH4 and HER can be proven.

Comment11
Please perform the electrochemical active surface area (ECSA) analysis and calculate the capacitance (Cs. μF/cm2). Moreover, add relevant information about the change in the electrochemical doublelayer capacitance with relativity changes in the ferroelectric material's oxygen vacancies.

Comment12
The Bi3NbTiO9 these are nanosheets so please perform the Raman spectroscopic analysis for charge density wave and magnetic transition in nanosheets.

Comment 13
Regarding OV, study the following related papers and cite them -"Sustained, Photocatalytic CO2 Reduction to CH4 in a Continuous Flow Reactor by Earth-Abundant Materials: Reduced Titania-Cu2O Z-Scheme Heterostructures", Applied Catalysis B: Environmental, 279 (2020) 119344.
-"A novel N-doped graphene oxide enfolded reduced titania for highly stable and selective gas-phase photocatalytic CO2 reduction into CH4: An in-depth study on the interfacial charge transfer mechanism", Chemical Engineering Journal (2020) https://doi.org/10.1016/j.cej.2020.127978 Reviewer #3 (Remarks to the Author): In this manuscript, the authors prepared ferroelectric Bi3TiNbO9 (BNT) nanosheets via a mineralizerassisted soft-chemical route. Surface oxygen vacancies were introduced to extend the photoabsorption and promote the adsorption and activation of CO2 molecules on the catalysts' surface. Corona poling was further adopted to strengthen the ferroelectric internal polarization to facilitate bulk charge separation within the catalysts. Synergistic effects of surface oxygen vacancies and ferroelectric polarization were demonstrated in CO2 photo-reduction. This manuscript is well organized with thorough characterization, therefore I recommend its publication in Nature Communications after some revisions as follows: 1. The authors stated that SEM images show the thickness of nanosheet to be 10-30 nm. Yet, it is difficult to confirm. Some additional evidence is suggested to be provided. 2. Will the corona poling process fill a part of the oxygen vacancies? 3. How does corona poling of BNT increase CO2 absorption and thus lead to increased CO2 reduction? 4. The authors stated "CO2 is adsorbed as carboxylate (CO2− 221, 1298 cm−1), bidentate carbonate (b-CO3 2−, 1381 and 1602 cm−1), bicarbonate (HCO3 − 222, 1205 and 1418 cm−1), *HCOO (2883 cm−1) and bidentate (2940 and 2981 cm−1) that are eventually converted to CO and CH4 upon illumination". How is CO2 activated into these intermediates? What is the conversion route of these intermediates to CO or CH4? Why CO is predominated in this study? 5. What is the isosurface level in Figure 3 e-g? It should be the same for comparison. The color of the atoms in these figures is suggested to be labeled. In addition, the authors stated that "In comparison with pristine BNT, the formation of OVs results in a higher charge accumulation in close proximity to the defect." However, obvious charge depletion was seen on the oxygen vacancy site in Figure 3e. 6. What is the mechanism on oxygen vacancy retarding the back switching of domains after poling? The authors stated that oxygen vacancy can retard the back switching of domains, thus leading to a higher remnant polarization. How does oxygen vacancy affect the origin ferroelectricity of BNT, as the literature has revealed that the existence of oxygen vacancies in ferroelectrics reduce the ferroelectric properties? 7. In Figure 5d, it is the un-poled sample, the "polarization" should be removed. 8. Much progress in CO2 conversion has been made in recent years via adopting other methods on catalysts, reaction systems, etc. (e.g., Adv. Funct. Mater. 2020, 2005983; Chemical Engineering Journal 322, 22-32), which can be discussed.

Dear Editor and Referees,
Thanks a lot for your comments and suggestions on our manuscript! Those suggestions are constructive and very helpful for revising and improving our paper. We have studied the comments carefully and made corrections accordingly. Please find below our point-by-point response.
Thanks for your kind consideration! I am looking forward to hearing from you.
With best regards, 1. In Figure 5 it is shown that the ferroelectric polarization is in the in-plane direction. This is the direction in which the electron-hole separation is boosted by the ferroelectric polarization and With regard to the oxygen vacancies, it is found from the TEM images that they mainly appear at the edges of BNT-OVP nanosheets (the direction parallel or anti-parallel to the ferroelectric polarization), as shown in the Figure 1f. For further confirmation, we conduct the density functional theory (DFT) calculations on the formation energy (ΔE f,vac ) of oxygen vacancies at different sites in Bi 3 NbTiO 9 , which reveals that the oxygen vacancies are easier to be formed in the direction parallel to the ferroelectric polarization than that in the direction perpendicular to the ferroelectric polarization. The theoretical results will be detailedly discussed in our answers to the following two questions. Oxygen vacancies can exert a pinning effect on the domain and hinder the switching back of the domain after corona poling, which allow BNT-OVP to remain a larger polarization intensity for promoting the separation of electron-hole pairs in this direction. Please find below updated Figure 1 and corresponding discussions were added in the revised main manuscript.

The synergy between polarization direction and thermodynamic favorability of oxygen vacancies
has not been discussed. There are several fundamental theory works on this subject in the literature. Response: Thank you so much for this constructive comment and the suggested reference! We have carefully studied this reference, which inspires us to conduct the following investigations. The ferroelectric materials can attract electrically charged species from the ambient environment on their surfaces to screen the spontaneous electric field for the sake of charge neutrality. Different atomic vacancies are generated on both sides of the positive and negative polarization surface to keep the material stable, and the negative polarization surface tends to form oxygen vacancies. As   and the adsorption energy of CO 2 on the surface of Bi 3 NbTiO 9 and Bi 3 NbTiO 9 with oxygen vacancies; the simulation is based on their energy-converged structural model with dipole moment correction by minimizing the surface Gibbs free energy. The much smaller ΔE f,vac of oxygen vacancies in the direction parallel to the polarization (1.092 eV) than that in the direction perpendicular to the polarization (2.955 eV) demonstrates that the formation of oxygen vacancies parallel to the polarization direction is more thermodynamically favorable (Figure 5d, e), reflecting the stable phase, which is consistent with our experimental results. After dipole moment correction, it is notable that the oxygen vacancies and constituent atoms in Bi 3 NbTiO 9 with oxygen vacancies shift slightly, while their composition and arrangement remain unchanged. It indicates that slight surface reconstruction occurs in this polar material, which will not obviously affect the photocatalytic activity. Besides, the adsorption energy of CO 2 on the surface of Bi 3 NbTiO 9 is calculated to be -0.990 eV, which is smaller than that on Bi 3 NbTiO 9 with oxygen vacancies (-1.808 eV), revealing that oxygen vacancies largely promote the adsorption of CO 2 molecules on the surface of this polar material.

Reviewer #2
Generally Response: Thanks a lot for these good comments! For the nanosheet photocatalysts, the number of layers has a great influence on the bandgap, ferroelectric properties, photocatalytic activity.
Compared with the bulk materials, thin-layer nanosheets always show a blue-shifted absorption edge and a larger bandgap. In addition, the thin-layer structure can increase the ferroelectricity by enhancing the polyhedral distortion, which strengthens the polarization electric field to promote the charge separation for enhancing the photocatalytic activity for CO 2 reduction. Thin-layer structure with more exposed surfaces can also provide abundant active sites for further enhancing the photocatalytic activity. 1,2 Now, the corresponding explanations of the relationship between layer number and the physiochemical property of the materials were added in the revised manuscript together with relevant literatures.
It is indeed a good suggestion to synthesize single-layer nanosheet. However, we tried but failed to achieve this. It is because the high reaction temperature restricts the use of surfactants in the hydrothermal process of Bi 3 NbTiO 9 , where surfactant-induced method is the most established strategy reported. We will keep trying to find a suitable synthetic route to obtain monolayer of Bi 3 NbTiO 9 for achieving higher photocatalytic performance in the future.
3. In the case of photocatalytic activity batch reactor has used, and a source of CO 2 was

Moreover, perform the TPD analysis for carbonate ion interaction with a photocatalyst, because
here source of CO 2 is NaHCO 3 .

Response:
We are very thankful for these good comment and concerns! The photocatalytic CO 2 reaction is conducted in a custom-designed reactor cell. As shown in Supplementary Figure 38, the as-prepared photocatalyst is dispersed evenly on the glass pane and placed on a triangle glass support in the 400 mL reactor cell; 5ml H 2 SO 4 reacts with 1.3g NaHCO 3 to generate 400 mL CO 2 as the reaction source at the bottom of reactor cell. Here, the H 2 SO 4 and NaHCO 3 do not contact the photocatalyst dispersed on the glass pane. The XPS results in Figure R1 show that there is no S peak detected in BNT-OVP, which excludes the presence of SO 4 2on the surface of the catalyst. In the process of photocatalytic CO 2 reduction, 1 mL gas was extracted from the reactor cell per hour (4 ml in total in 4 hours). The volume of reactor is 400 mL, and the pressure in the reactor is only reduced by 1% after 4 hours, which is negligible. We also conduct the experiment of photocatalytic CO 2 reduction with the 1.3g NaHCO 3 and 5ml H 2 SO 4 fresh reaction mixture every hour. As shown in Figure R2, the amount of evoluted CO and CH 4 from CO 2 (umol/h) shows no obvious difference with that with the 1.3g NaHCO 3 and 5ml H 2 SO 4 every 4 hours. According to the previous references, the pressure of photocatalytic CO 2 reduction reactors in the gas-solid reaction system is almost kept at 1 atm. [3][4][5] In addition, the photocatalytic CO 2 reduction experiment by using high-purity CO 2 gas and 5 ml H 2 O as reactants instead of 1.3g NaHCO 3 and 5ml H 2 SO 4 is also performed. The results show that the amount of evoluted CO and CH 4 from CO 2 shows no evident difference for the different reactants ( Figure R3). Furthermore, the CO 2 -TPD test is conducted to investigate the interaction of CO 2 on BNT, BNT-P and BNT-OV2, BNT-OVP samples. As shown in Figure. R4, there are three major peaks, which can be assigned to weak (<200 ), moderate (200 -400 ), and strong (>400 ) basic sites of catalysts, respectively. In general, the moderate and strong adsorption sites are active for CO 2 conversion, while the weak adsorption sites are inactive for CO 2 conversion. In comparison with BNT, BNT-P and BNT-OV2, BNT-OVP show stronger chemisorption of CO 2 and more moderate and strong basic sites, indicating that corona poling process and creating oxygen vacancies allow the CO 2 molecules to strongly interact with the catalyst and to be easily activated by the photoexcited electrons.     Response: The schematic illustration for electron/proton transport process and the formation of product is proposed according to the results of in-situ FT-IR experiments (Supplementary Figure 19) to discuss the reaction mechanism of CO 2 reduction to CO and CH 4 , 6,7 which is added in Supplementary Figure 20. CO 2 molecules are easily adsorbed on the surface of Bi 3 TiNbO 9 nanosheets, and then accept electrons and H + to be converted into CO. The corona poling process enhances the ferroelectric properties of the samples, and the band bending allows the photo-generated electrons a higher reducing ability. Some CO molecules can accept two additional electrons, leading to carbon residue on the surface. These radicals can subsequently combine with up to four electrons and H + eventually forming CH 4 .
It is a reaction occurring at the gas/solid interface, and water vapor is the reactant in the process of photocatalytic CO 2 reduction and the corresponding oxidation product is O 2 . There is no substantial liquid-phase water in the reaction system (vapour instead), which is the reason for low HER. 8 Due to the low performance of HER, the change is not very obvious in Figure 3c. But actually, the H 2 evolution amount has increased from 0.07 μmol g −1 h −1 for BNT to 0.19 μmol g  figure S3, there is not much difference in the specific area, and the case of BNT showed significant low activity concerning BNT-OV2 for CO evolution or even for CH 4 . However, the specific area difference is not much. So, here the author focused surface engineering. So why is this contradiction? Please add a relevant brief description.

Response:
We are sorry for the misleading, and we have clarified our finding in the revised manuscript. Specifically, the surface engineering means creating oxygen vacancies, but not to mainly alter the specific surface area. Introduction of oxygen vacancies on the surface of Bi 3 TiNbO 9 not only extends its photo-responsive range and tremendously promotes separation of photoinduced charge carriers, but also produces unsaturated bonds and dangling bonds for promoting the adsorption and activation of CO 2 molecules on the surface of the catalyst, thus greatly enhancing the photocatalytic activity. It is actually good that these samples have similar specific surface area, which can exclude the influence of specific surface area on the photocatalytic activity, further confirming the advantageous role of oxygen vacancies in promoting the photocatalytic CO 2 reduction. Corresponding clarifications were added in the updated version. figure S15  Response: Thank you for these good suggestions! We have performed the photocatalytic CO 2 reduction tests after three cycles. As shown in Figure R5, the CO and CH 4 evolution amount shows no notable decrease, which indicates that the photocatalyst still maintains high catalytic activity.

As we mentioned earlier, in
After each cycling test, the photocatalyst after reaction was re-dispersed evenly on the glass pane and dried at 60 for 6 h for measurement. There is no special photocatalyst regeneration process.
The XPS results in Figure R1 show that there is no S peak detected over BNT-OVP, which excludes the presence of SO 4 2on the surface of the catalyst.  which might explain the reaction mechanism. Moreover, write relatively reaction mechanism for CO, CH 4

and HER in brief.
Response: Thank you for this advice. The schematic illustration for electron/proton transport process and the formation of product has been added in Supplementary Figure 20 to discuss the reaction mechanism of CO 2 reduction to CO and CH 4 .

If possible, DFT calculation, then please perform DOS analysis for change in fermi level
relatively changes in the oxygen vacancies. This will also help understand the reaction mechanism because as mentioned, selectivity for CO, CH 4 and HER can be proven.

Response:
We have added DOS of Bi 3 TiNbO 9 and Bi 3 TiNbO 9 with oxygen vacancies calculated by DFT in Supplementary Figure 32. The introduced oxygen vacancies lead to a new defect level in the band gap, which is beneficial to photoexcitation and charge separation. In addition, the schematic illustration for electron/proton transport process and the formation of product are added in Supplementary Figure 20 to discuss the reaction mechanism of CO 2 reduction to CO and CH 4 . Response: According to your suggestion, we have conducted the related electrochemical tests to analyze the electrochemical active surface area (ECSA) and calculate the capacitance (Cs. μF/cm 2 ).
As shown in Figure R6 and Figure Figure 4, the thickness of BNT-OVP is approx. 16 nm, which is consistent with the SEM results.
Supplementary Figure 4. AFM images and the corresponding height of BNT-OVP.

Will the corona poling process fill a part of the oxygen vacancies?
Response: Compared with BNT and BNT-OV2, the EPR signals of the poled BNT-P and BNT-OVP show no obvious change as shown in Figure 1g. Thus, the corona poling process will not fill the oxygen vacancy. In addition, there is no obvious change in surface state of elements ( Figure   2a, b), further confirming that the corona poling process will not fill a part of the oxygen vacancies.

How does corona poling of BNT increase CO 2 absorption and thus lead to increased CO 2 reduction?
Response: The ferroelectric materials with polarized charges can attract electrically charged species from the ambient environment onto their surfaces to screen the spontaneous polarization electric field for the sake of charge neutrality. The corona poling endows BNT a stronger remnant polarization and more polarized charges, which gives rise to a better adsorption performance. As shown in Figure 3d, BNT-P and BNT-OVP show the stronger adsorption of CO 2 . Furthermore, the  Response: Thank you for these good comments! The schematic illustration for electron/proton transport process and the formation of product is added in Supplementary Figure 20 to discuss the reaction mechanism of CO 2 reduction to CO and CH 4 Figure 3 e-g? It should be the same for comparison. The color of the atoms in these figures is suggested to be labeled. In addition, the authors stated that "In comparison with pristine BNT, the formation of oxygen vacancies results in a higher charge accumulation in close proximity to the defect." However, obvious charge depletion was seen on the oxygen vacancy site in Figure 3e.

What is the isosurface level in
Response: Thank you for this good comment. The isosurface level is 0.001. The color of the atoms in these figures has been labeled in our revised manuscript. The charge accumulation and depletion is in blue and in yellow, respectively. These corrections have made in the revised manuscript in Figure 5 d-g.

6.
What is the mechanism on oxygen vacancy retarding the back switching of domains after poling?
The authors stated that oxygen vacancy can retard the back switching of domains, thus leading to a higher remnant polarization. How does oxygen vacancy affect the origin ferroelectricity of BNT, as the literature has revealed that the existence of oxygen vacancies in ferroelectrics reduce the ferroelectric properties?
Response: Ferroelectric materials have attracted a great deal of attention for potential applications in ferroelectric random access memory, actuators and microwave electronic components. Fatigue problems of ferroelectric materials, such as retention loss and imprint, refer to the reduction of switchable polarization after repetitive electrical cycling, which limit their commercial applications.
Oxygen vacancies as typical defects in ferroelectric films can cause domain pinning, change domain structure and decrease reversible domains. 5,6 Therefore, the presence of oxygen vacancies should be avoided in traditional ferroelectric thin films. In our work, we applied a high voltage on the samples by corona poling process to make the ferroelectric domains aligned. After removing the applied voltage, the domains of ferroelectric photocatalysts without oxygen vacancies are easier to deflect to near the initial position. Nevertheless, the introduction of oxygen vacancies will hinder this domains back switching process, which eventually retain greater remnant polarization. Figure 5d, it is the un-poled sample, the "polarization" should be removed.

In
Response: Thank you for this advice. Even if there is no corona poling process, the ferroelectric materials also have spontaneous polarization electric field, which will lead to band bending. So we think that it is necessary to keep the "polarization". 7 Response: Great thanks for your suggestion. The above important references have been cited in our revised manuscript. We have also add comparisons between our current work and these literautres in termas of performance and catalytic reaction mechanism.