1. 套磁信 I am an undergraduate student. I want to write an email to a Professor in MIT working on large language models to sell my experience and ask him if he is willing to recruit me as a PhD student in next year.

2. 催审稿 I am a graduate student. I want to write an email to the reviewer of Nature journal to ask him to accelerate the review process of my submitted paper.

Dr. Davide Esposito
Chief Editor
Nature Catalysis
No.: NATCATAL-24020271
Manuscript Title: Low-coordinate Cu1–N2 sites in single-atom catalyst for highly efficient propylene epoxidation with molecular oxygen
Dear Dr. Davide Esposito,
We would like to express our deep appreciation for the helpful suggestions and have carefully revised the manuscript following the comments. We have included additional discussions in response to the reviewers’ comments. This letter is our detailed point-to-point response to the reviewers. Additionally, an uploaded copy of the manuscript displays the revisions. We believe that the concerns raised by the reviewer have been fully addressed in our revised manuscript.



Reviewers' comments:

Reply to Reviewer #1:

Figure R1 (Supplementary Figure 35) | AIMD simulation results for Cu–N2 and Cu–N4 active sites at 550 K. (a) Energy fluctuations of the Cu–N2 active site in C3N4–L during a 30 ps AIMD simulation, along with its representative structure. (b) Energy fluctuations of the Cu–N4 active site in C3N4 during a 30 ps AIMD simulation, along with its representative structure. The atoms Cu, N, C, and H are represented by the orange, blue, grey, and white balls, respectively.


Supplementary Figure 11 | Characterizations of various catalysts. In situ DRIFTS of CO adsorption of Cu1–N2, Cu1–N3 and Cu1–N4 SACs.


Supplementary Figure 13 | Schematic illustration of the plausible formation mechanism of low-coordinated Cu1–N2 centres.

Supplementary Figure 20 | Catalytic performance of propylene epoxidation with O2. Product formation rate at different reaction temperatures over Cu1–N2 SAC. Reaction conditions: C3H6: O2: N2 = 10: 5: 85 vol.%, GHSV = 36,000 h-1, 50.0 mg catalyst.

General Comment: The authors demonstrated that they prepared Low-coordinate CuN2 active site in C3N4-L sheets, and the antithetical CuN3 and CuN4 actives sites in traditional C3N4 sheets through a pyrolysis based synthesis approach from similar classic beginning chemicals. More important, the CuN2 active site in C3N4-L possesses eminent catalytic performance for direct epoxidation of propylene into propylene oxide (PO) with molecular oxygen. Two important highlights in this study is the attained remarkable catalytic performance in the DEP reaction, including an impressive 78.0% selectivity for propylene oxide (PO) on the as-prepared sample, a high propylene conversion rate of 5.7% and the observed PO formation rate reaches 5.9 mmol gcat-1 h-1.
Generally, the result in this manuscript is very fascinating and its experiment data is very compelling. In contrast, the theoretical derivation at the active sites and the catalytic mechanisms need further to be clarified. However, there are some uncertainty in the deduction from the experiment results and the origin of structure activity. The responding data and the proposed active structure need to be clarified and confirmed before its final publication.
Response: We sincerely thank Reviewer #1 for their thoughtful and positive comments on the significance and quality of our work. We are pleased to hear that the result in this manuscript is very fascinating and its experiment data is very compelling. Futhermore, the responding data have been confirmed. To facilitate the review process, we have placed the revised figures at the end of this document in the order they appear in the main text. 

Comment#1. The work temperature for its eminent catalytic performance for direct epoxidation of propylene is 275 oC, hence the future work to check the AIMD stability at 550 K is a must. I hope the embedded positions for the CuN2 active site in C3N4-L and CuN4 active site in C3N4 sheets should be kept without obviously migration.
Response: Thank you for pointing out the importance of evaluating the stability of the active sites at elevated temperatures. We acknowledge the importance of evaluating the AIMD (Ab Initio Molecular Dynamics) stability of the Cu–N2 active site in C3N4–L and the Cu–N4 active site in C3N4 sheets at elevated temperatures, such as 550 K, to ensure their stability and performance under realistic conditions. We agree that this analysis is crucial for a comprehensive understanding of the catalytic performance.
To address this, we performed AIMD simulations using the NVT ensemble and controlled the temperature with a Nosé-Hoover thermostat at 550 K, employing a time step of 1 fs for a total simulation duration of 30 ps. The energy fluctuations during the simulation are presented in Figure R1. The structural stability of the active sites is illustrated by showing a representative structure after 20 ps of simulation.
Our results demonstrate that both CuN2 and CuN4 active sites remained stable during the simulation, with no significant migrations of the Cu atoms or active site distortions observed. This indicates that the Cu–N2 active site in C3N4–L and the Cu–N4 active site in C3N4 sheets exhibit strong structural integrity at elevated temperatures, which thereby supports their potential for practical catalytic applications under these conditions.
We appreciate your insightful comments and concur that this analysis makes a significant contribution to a comprehensive understanding of the thermal stability and robustness of these active sites.

Comment#2. The cutoff energy of 400 eV and the Hellman-Feynman forces at 0.05 eV/A is not enough in the VASP simulation to study the catalytic mechanism, I suggest the cutoff value is set at 500 eV and 0.02.
Response: Thank you for your valuable feedback regarding the computational parameters in our VASP simulations. As per your suggestion, we have raised the plane-wave cutoff energy from 400 eV to 500 eV, resulting in a more precise representation of the electronic wavefunction. In addition, we have tightened the Hellman–Feynman force convergence criteria for optimized structures from 0.05 eV/Å to 0.02 eV/Å, thus achieving a more precise optimization of the atomic geometries.
Although these adjustments led to minor changes in some of the calculated energy values, our main conclusions remain unchanged. Notably, for the Cu1–N2 catalyst, the rate-limiting step changed from TS1 to TS2 under the revised computational parameters. Meanwhile, the activation barriers for the second PO formation on Cu1–N2 and Cu1–N4 single-atom catalysts were still comparable, being 0.83 eV and 0.79 eV, respectively. Despite this, the energy barriers for the initial PO formation showed significant differences (0.56 eV vs. 1.86 eV). Furthermore, both of these barriers are lower than those for the α C–H bond abstraction on these catalysts.
We appreciate your insightful comments that have helped us improve the accuracy and reliability of our study. The manuscript has been updated accordingly with the revised results and discussions. If you have any additional questions or suggestions, please let us know.

Comment#3. The catalytic product of CO2 in these materials are very high, sometimes the percent is over 50%. The catalytic active sites haven't been mentioned in the manuscript. Meanwhile, the competition between CO2 and PO also is missing. Hence, the authors should give rational explanation for CO2 production.
Response: Thank you for the insightful comments from the reviewer. Our detailed and rational response to dispel the reviewers' doubts is provided in the following content.
As reported in most of the published literature and discussed in our manuscript, copper-based catalysts have been widely utilized in DEP reactions and exhibit exceptional catalytic performance. It is generally recognized that the Cu+ species possesses a high efficiency in DEP reactions with high PO selectivity because the oxygen species activated on Cu+ sites have relatively strong electrophilic properties, which is conducive to attacking the C=C double bond of propylene. However, the oxygen species activated on Cu2+ sites exhibit nucleophilicity, tending to attack the α C–H bond of propylene, leading to the formation of a deeply oxidized product (Refer to J. Mol. Catal. Chem. 228, 27–33 (2005); J. Catal. 268, 165–174 (2009)). In our study, we have characterized the oxidation states of Cu in Cu1–N2 SAC using XAS, XPS Cu 2p and Cu LMM characterizations in detail (Figs. 2c, d and Supplementary Figure 12). The results indicated that the oxidation state of Cu species was nearly +1 (XANES) and the Cu+:Cu2+ ratio in Cu1–N2 SAC was nearly 9:1. According to the above results, Cu1–N2 SAC possessed an intrinsic activity which was suitable for the production of PO with high selectivity.
Regarding the catalytic performance of DEP in our manuscript, we only presented the propylene conversion, the selectivity of various products and PO formation rate to evaluate the catalytic performance of the Cu-based catalysts. This led to misunderstandings among the reviewers. We sincerely apologize for our negligence and have conducted additional experimental analysis to better depict the catalytic performance. We have added the Supplementary Fig. 20 depicting the formation rates of various products including PO, CO2, acrolein, acetaldehyde and aceton at different reaction temperatures. It can be observed that from 200 oC to 275 oC, the formation rate of PO increased rapidly, while the formation rate of CO2 increased slowly., The selectivity of PO was primarily impeded by thermodynamic energy barriers at this stage, and the rate of complete oxidation was influenced by adsorption/desorption processes (Refer to J. Colloid Interface Sci. 611, 564–577 (2022)). Furthermore, the complete oxidation was more favorable according to the relevant data of thermodynamics (Refer to Appl. Surf. Sci. 355, 660–666 (2015); ACS Catal. 10, 13415–13436 (2020)). When the reaction temperatures were above 275°C, several factors contributed to the significant increase in CO2 formation. First, the excessively high reaction temperatures led to aggregation and oxidation of Cu species to Cu2O or CuO, which have been proved by the XRD and XPS Cu 2p characterizations (Supplementary Figs. 23b and 25a). The lattice oxygen species excited at high temperatures exhibit nucleophilicity, accelerating the extraction of α C–H bond of propylene, converting more C3H6 into by-product and further to CO2. Besides, the PO generated will be susceptible to isomerization at high temperatures for CO2 formation (Refer to J. Am. Chem. Soc. 111, 4553–4561 (1989); Surf. Sci. Rep. 76, 100524 (2021)). The carbon nitride, which serves as the support, can also promote the conversion of propylene into acrolein and CO2 at high temperatures (Supplementary Fig. 17). The generation rate of CO2 increased significantly due to these factors working together.
In summary, we believe that the Cu+–N2 species served as the real active-site for the production of PO, while Cu2+ favored the generation of CO2. Within a moderate temperature range, an increase in the reaction temperature was conducive to accelerating the efficient production of PO, compared to other products. However, the agglomeration/oxidation of copper species, the isomerization of PO, and the over-oxidation of propylene will be greatly exacerbated by the excessively high reaction temperatures.

Comment#4. There is no data on C3N4-L in supplementary Fig. 6. The author missed these data.
Response: Thank you for carefully reviewing our manuscript. We express our sincere apologies for our oversight. The correct content for Supplementary Fig. 6 should be “Physicochemical properties of carriers. (a) N2 adsorption/desorption isotherms and (b) pore size analysis of Cu1–N2, Cu1–N3 and Cu1–N4 SACs”. Besides, the relevant data on C3N4–L can be found in Supplementary Fig. 14h and i.

Comment#5. The simulation box with the integral structures must be presented in Computational models for (d) Cu1-N2 (e) Cu1-N3 and (f) Cu1-N4 SACs in supplementary Figure 10.
Response: Thank you for your detailed review and for highlighting the need to present the simulation box with the integral structures in our Computational Models section. Your attention to this crucial aspect of our study is greatly appreciated.
To address your concern, we have included the simulation boxes for the (d) Cu1–N2, (e) Cu1–N3, and (f) Cu1–N4 single-atom catalysts (SACs) in Supplementary Figure 10. This addition will ensure that the structural configurations and computational setup are clearly presented and can be accurately reproduced.
We appreciate your efforts in helping us refine our manuscript and anticipate any additional feedback you may have.

Comment#6. The peak for in situ DRIFTS of CO adsorption of Cu1-N2, Cu1-N3 and Cu1-N4 SACs in Supplementary Figure 11 seems not correct. Please double check these peaks and tagged with its vibrational modes from what bonding.
Response: We thank you for carefully reviewing our manuscript. Based on our rigorous academic approach, we have carefully checked and further retested this controversial measurement of all the samples to study the scientific results. Except the peaks around 2115 cm-1 and 2175 cm-1 assigned to gaseous CO in condition, an extra signal at 2054 cm-1 (Cu1–N2 SAC), 2057 cm-1 (Cu1–N3 SAC) and 2060 cm-1 (Cu1–N4 SAC) representing the CO linear adsorption (COL) on Cu atoms was found for all Cu based SACs. Besides, the absence of bridge bonded CO band in the range of 1700-1900 cm-1, ruling out the presence of multi-atom Cu species (Refer to Angew. Chem. Int. Ed. 62 (2023), e202217220, Angew. Chem. Int. Ed. 126, 4956–4961 (2014), Nat. Commun. 12, 6022 (2021), Appl. Catal. B Environ. 339, 123119 (2023), Chem. Eng. J. 146870 (2023), J. Catal. 405, 333–345 (2022), Appl. Catal. A Gen. 505, 334–343 (2015), J. Phys. Chem. C 113, 10689–10695 (2009) and J. Catal. 268, 367–375 (2009)). The in situ DRIFTS of CO adsorption on Cu1–N2, Cu1–N3 and Cu1–N4 SACs at 30 °C in our previous experiments showed only two peaks due to the CO adsorption time was too long and the CO concentration was too high, resulting in the overlap of the two peaks (around 2055-2060 cm-1 and 2115 cm-1). The revised manuscript has corrected the details of CO DRIFTS finally. Thanks again for the reviewer's comments and the relevant content has been corrected in the revised manuscript.

Comment#7. There some irrational expression or some errors are listed here:
1."with the Cu1-N2 SAC showing a significant number micropores compared to Cu1-N3 and Cu1-N4 "... in the end of Page 6.
2. "The presence of high- angle-shifted (002) peaks at 2θ of 27.5° was characteristic of carbon nitride supports." in the middle of Page 7. Here, the peaks at 2θ of 27.5° should be the obvious characteristic of pi-pi packing with ~ interlayer distance around 3.35 Å, since the angle at 2θ of 27.5° exists in C3N4 and graphite.
3. "revealed the co-existence of triazine and pyridine structures within the porous C3N4-L" in the middle of Page 11. In my understanding, the authors want to mention pyrimidine-containing fragments not pyridine fragments, since the novel aroma fragment is C4N2 nor C5N1.
4. For the paragraph "about the development of synthetic methodology about fabricated a SAC featuring low-coordinated Cu1-N2 anchored onto a C3N4–L framework via the ACAS technique." in page 11.1 suggest the author introduce a scheme to make it much clear in Supporting information.
Response: We greatly thank you for carefully reviewing our manuscript. We have carefully checked the entire manuscript and corrected the irrational expressions or errors in the revised manuscript.
1.According to the reviewer's suggestion, this part has been amended to "and the Cu1–N2 SAC demonstrated a significantly greater number of micropores compared to Cu1–N3 and Cu1–N4 SACs".
2.According to the reviewer's suggestion, this part has been amended to "The presence of high-angle-shifted (002) peaks at 2θ of 27.5° was characteristic of pi-pi packing with the interlayer distance around 3.35 Å, since the angle at 2θ of 27.5° exists in C3N4 and graphite".
3.According to the reviewer's suggestion, this part has been amended to "revealed the co-existence of triazine and pyrimidine-containing fragments within the porous C3N4–L".
4.According to the reviewer's suggestion, we have refined the scheme in Supplementary Fig. 13 to make the formation mechanism of Cu1–N2 SAC through the ACAS strategy clearer and more visualized. The schematic illustration included the amidation reaction, the self-condensation reaction, the metal-heteroatom coordination, and the dehydration reaction, etc. The following are the details.
Supplementary Fig. 13 specifically illustrates the synthetic steps for Cu1–N2 SAC utilizing the ACAS strategy. First of all, the hydrolysis of CuCl2 and the modifier L-cysteine creates an acidic medium that facilitates the hydrolysis of melamine. This step leads to the formation of Compound I through the reaction between L-cysteine and melamine. The abundant terminal –NH2 of melamine would coordinate with free Cu2+ species to form Compound II, as evidenced by the XRD characterization (Supplementary Fig. 14a). Secondly, during the solvothermal reaction process, melamine underwent further hydrolysis to generate cyanuric acid units. The in-situ formed cyanuric acid and melamine spontaneously assembled into a supramolecular structure through hydrogen bonding (Refer to Chem. Eng. J. 405, 126685 (2021); Nano Res. 11, 3462–3468 (2018)). Due to the in-situ modification of nitrogen-containing units, the low-coordinate Cu–N structures can form (Compound III) during the assembly process, which further stabilizes the anchoring of Cu–N species. After that, the cyanuric acid in Compound III released the ammonia gas and left melamine to self-condense into heptazine units (Compound IV) (Refer to ACS Appl. Mater. Interfaces 11, 10651–10662 (2019); Small 20, 2309032 (2024); Langmuir 30, 447–451 (2014)). Subsequently, the modified moiety within Compound III underwent cyclization to form Compound V by releasing NH3 and H2S molecules, followed by dehydration to generate Compound VI during pyrolysis. These steps were iteratively and periodically repeated, ultimately forming the low-coordinate Cu1–N2 SAC. Throughout these above processes, the assembly of the supramolecular structure and the formation of asymmetric melon configurations created favorable conditions for the formation of high-loading and low-coordinate copper single-atom sites.




Reply to Reviewer #2:


Fig. 3 | Catalytic performance of propylene epoxidation with O2. Product selectivity and C3H6 conversion at different reaction temperatures over (a) Cu1–N2, (b) Cu1–N3 and (c) Cu1–N4 SACs. Reaction conditions: C3H6: O2: N2 = 10: 5: 85 vol.%, GHSV = 36,000 h-1, 50.0 mg catalyst. (d) Performance comparisons of various IB group catalysts in propylene epoxidation including PO formation rate, PO selectivity and propylene conversion. The digital characters from 1 to 16 represent as Cu7Ce, LaCu0.5Mn0.5O3, Cl-RD-Cu2O, Ag-MoO3/ZrO2, Au/TS-1-KOH, Cu/SiO2, RD-Cu2O, VCe0.5Cu0.5-NaCl, (KAc)Cu/SiO2, K+-CuOx/SBA-15, Ag/CaCO3-NaCl, Ag/Y2O3-K2O/α-Al2O3, CuNP/C3N4-L, Cu1–N4 SAC, Cu1–N3 SAC and Cu1–N2 SAC, respectively. (e) Stability test over Cu1–N2, Cu1–N3 and Cu1–N4 SACs. Reaction conditions: C3H6: O2: N2 = 10: 5: 85 vol.%, T = 275 oC, GHSV = 36,000 h-1, 50.0 mg catalyst.


Supplementary Figure 5 | Characterizations of various catalysts. FT-IR spectra of C3N4, C3N4–L, Cu1–N2, Cu1–N3 and Cu1–N4 SACs.


Supplementary Figure 10 | Characterizations of various catalysts. EPR pattern of Cu1–N2, Cu1–N3 and Cu1–N4 SACs. 
 

Supplementary Figure 11 | Characterizations of various catalysts. In situ DRIFTS of CO adsorption of Cu1–N2, Cu1–N3 and Cu1–N4 SACs.


Supplementary Figure 20 | Catalytic performance of propylene epoxidation with O2. Product formation rate at different reaction temperatures over Cu1–N2 SAC. Reaction conditions: C3H6: O2: N2 = 10: 5: 85 vol.%, GHSV = 36,000 h-1, 50.0 mg catalyst.


Figure R1 | Catalytic performance. PO selectivity and C3H6 conversion over Cu1–N2 SAC. Gas type: C3H6 (0-75 min) mix gas (75-150 min, C3H6: O2: N2 = 10: 5: 85 vol.%). Reaction conditions: GHSV = 36,000 h-1, 50.0 mg catalyst, T = 275 oC.

General Comment: This article reports on epoxidation of propylene with molecular oxygen over Cu single atom catalysts comprised of Cu atoms dispersed in C3N4 prepared from pyrolysis of various precursors. This is an extraordinarily thorough article as pertains to characterization of the materials, and the computational aspects. The actual catalytic testing felt relatively limited compared to all of the characterization section, it reads at times like a materials science article rather than a catalysis article.
The article provides significant evidence that Cu atoms coordinated to only two nitrogen atoms are more reactive than Cu bound to 3 or 4 atoms, and Cu in nanoparticles. These Cu-N2 sites are active and selective for epoxidation, giving 78% PO selectivity at initial time on stream and around 6% propylene conversion. This selectivity decreases significantly in under 4 h, yet the article shows that the structure of the materials is relatively unchanged after reaction. So, it is unclear why selectivity changes, and the claims of high selectivity are relatively hindered by the limited stability of the selective material.
This is an interesting article, it's extraordinarily thorough in many aspects, and it could be appropriate for publication in Nature Catalysis after only relatively minor revisions.
Response: We thank you for the highly positive comment of “This is an extraordinarily thorough article as pertains to characterization of the materials, and the computational aspects” and “This is an interesting article, it's extraordinarily thorough in many aspects, and it could be appropriate for publication in Nature Catalysis after only relatively minor revisions”. It is common knowledge that the catalysts are at the core of most of catalytic reactions, determining the orientation of reaction pathways, product selectivity, kinetic processes, and more. In the past few years, a broad range of nanocatalytic materials have been achieved, which has improved the corresponding catalytic performance. Despite these advances, further development of nanocatalytic materials has been hindered by inefficient active site design, unclear active site properties, and the structure-performance relationship. Thus, the design, characterization, and optimization of the true active sites serve as the foundation for the further advancement of catalytic science. Among these aspects, regulating the fine coordination structure in atomically dispersed catalysts is of key importance. This is why we spared no effort to analyze the fine structure of the as-prepared catalysts, providing a solid foundation and specific approach for in-depth exploration of the catalytic reaction mechanism. Based on these, we have carefully revised our manuscript based on your comments.

Comment#1. Could the authors add some context for the claim of 5.9 mmol gcat-1 h-1 being an excellent PO formation rate? How does this compare to other rates reported in literature for this reaction? At what time on stream was this rate measured, and how stable is it with time on stream?
Response: We are thankful for the reviewer’s comments and admire the reviewer's rigorous academic attitude. 
The following is background supplementation for the DEP reaction. The DEP route has been considered as an ideal method for PO production because of the abundant availability and cost-effectiveness of molecular oxygen. However, during the DEP reaction process, both the α C–H in propylene and the PO produced have a high level of reactivity. Additionally, the DEP reaction is exothermic and the actual heat release for the DEP reaction with 50.0 mg Cu1–N2 SAC achieved about 270 mW tested via the C80 micro calorimeter technique. Additionally, other DEP catalysts exhibited more significant exothermic effects. The self-exothermic characteristics of the DEP reaction is unfavorable for the chemical equilibrium movement especially at high reaction temperatures, leading to an inevitable conflict between the reaction kinetics and thermodynamics. As a result, increasing the formation rate of PO at high propylene conversion remains an insurmountable challenge. As is the case with the DEP performance of the reported catalysts, both the PO formation rate and the propylene conversion remain at an extremely low level, severely impeding the industrialization process of this crucial catalytic reaction. 
Through extensive experimental exploration and theoretical studies, we ultimately identified the low-coordinated Cu–N single-atom center as the optimal active site for the DEP reaction. Furthermore, thanks to the implementation of our proposed ACAS strategy, a uniform loading of the Cu1–N2 sites on the surface of the carbon nitride has been achieved. Consequently, among all reported IB group metal catalysts, only Cu1–N2 SAC can achieve the ultra-high PO formation rate (5.9 mmol gcat-1 h-1) with high propylene conversion and high PO selectivity at relatively mild conditions at 275 oC. Considering the industrial value and separation costs, and choosing the reported DEP catalysts with ≥60% PO selectivity as a reference, the PO formation rate of 5.9 mmol gcat-1 h-1 over Cu1–N2 SAC has been increased by 1,073%~14,565%. The corresponding results have been presented in detail in Fig. 3d and Supplementary Table 6. This is a significant breakthrough in terms of the DEP catalytic reaction, which has great scientific and industrial value. The relevant content has been added in the revised manuscript.
As suggested, the stability results of the PO formation rate over Cu1–N2 SAC at 275 oC were provided in the revised manuscript. As shown in Fig. 3e, the PO formation rate over Cu1–N2 SAC exhibited steady performance. Over a 300-minute period test, the formation rate of PO for Cu1–N2 SAC remained at approximately 80% of its initial value. By extending the reaction time, only a slight decrease in the PO formation rate can be observed due to a slight increase of by-product acrolein selectivity. 

Comment#2. In the description of the FTIR spectra of the materials themselves, the shift in NH2 stretching vibration from 3165 to 3430 cm-1 seems remarkably large. What evidence do the authors have that this is from the same vibration, and have such large shifts been observed previously?
Response: We are thankful for the reviewer’s comments and admire the reviewer's rigorous academic attitude. 
The remarkable shift of –NH2 stretching vibration, moving from 3165 to 3430 cm–1, can be attributed to various synthesis methods especially the atmosphere during the pyrolysis process of carbon nitride carriers and Cu-based SACs. As shown in the Methods section in Supplementary information, the C3N4 and C3N4–L were preprared via a pyrolysis step in the muffle furnace, resulting in the introduce of water molecules from the air during the polymerization and subsequent cooling processes. This led to the appearance of a strong and wide absorption peak caused by intermolecular hydrogen bonds around 3000-3500 cm–1 in the FT-IR spectrum, specifically attributed to O–H bonds. Moreover, the C3N4 and C3N4–L exhibit a high degree of polymerization and a limited proportion of terminal –NH2 in the absence of Cu coordination, which leads to a weak signal of –NH2 stretching vibration in the FT-IR spectrum, causing the shift of IR peaks and shield by O–H bonds. By contrast, the Cu-based SACs were synthesized under a protective gas atmosphere, without the interference from complex components in the air. Besides, this above phenomenon can be also observed in other similar reported C3N4 materials (Refer to Bioresour. Technol. 372, 128677 (2023), Chem. Sci. 13, 754–762 (2022), and Sep. Purif. Technol. 308, 122955 (2023)).
With a rigorous scientific approach, a new batch of the carbon nitride carrier (termed as C3N4–L (5% H2/Ar)) have been synthesized by the same procedure to that of C3N4–L, except that the atmosphere was changed to 5% H2/Ar and re-tested the surface chemical property using FT-IR measurements. As shown in Supplementary Fig. 6, the presence and characteristic of the –NH2 stretching vibration on C3N4–L (5% H2/Ar) can be observed at nearly 3450 cm–1 in the FT-IR spectrum. After the single-atom Cu coordination, the vibrational spectral analysis also revealed a slight blue shift in the peaks corresponding to the stretching vibrations of the –NH2 groups, shifting from 3450 to 3430 cm–1. What's interesting is that the above phenomenon matches a blue shift in the bending vibrations of the triazine units shifting from 800 to 778 cm–1. These above conclusions correct the previously inadequate descriptions and provide credible evidence for the shift of the –NH2 group after the coordination anchoring of single-atom Cu species. 
We sincerely appreciate your comment again, and we have revised the manuscript to provide a clearer and more accurate expression.

Comment#3. In the discussion of the Raman data, how do the authors know that a material with many N vacancies necessarily has a Cu1-N2 configuration? Couldn't this just mean there are more sites with no Cu at all, and then the only sites with Cu have Cu-N4 coordination (a combination of Cu free sites and Cu-N4 sites seems similar in terms of number of N vacancies to the idea of Cu-N2 sites?)
Response: We thank you for carefully reviewing our manuscript. In order to explain the reviewer's confusion more scientifically, we have now added more experimental details in the modified manuscript, which makes our work more readable. 
The Raman pattern of Cu1–N2, Cu1–N3 and Cu1–N4 SACs (Supplementary Figure 9a) showed that the decrease of ID/IG ratio from Cu1–N4 SAC (ID/IG = 0.919) to Cu1–N3 SAC (ID/IG = 0.947) and to Cu1–N2 SAC (ID/IG = 0.980). Besides, a combination of C3N4 with Cu-free-site and Cu1–N4 SAC show a similar ID/IG ratio with Cu1–N4 SAC, indicating that the properties of Cu–N structure are the key determinants of its defect concentration. The relationship between the concentration of N vacancies and unpaired electrons was further confirmed by electron paramagnetic resonance (EPR) spectra (Supplementary Figure 9b). The Cu1–N4 SAC showed an extremely slight paramagnetic signal with a g value at about 2.004 to the unpaired electrons, suggesting that Cu1–N4 SAC exhibited a high degree of Cu–N coordination saturation with a handful of N vacancy. By comparison, the corresponding EPR peak intensity at g = 2.004 increases in Cu1–N3 SAC. Furthermore, the trend of EPR peak intensity variation became more evident in Cu1–N2 SAC, consistent with the Raman analysis in terms of the increase of N vacancies and the formation of low-coordinated Cu1–N2 single-atom sites. These results indicate that Cu1–N2 SAC has the highest defect concentration and the most unpaired electrons, thereby possessing the highest surface energy, which is conducive to the formation of unsaturated coordinated metal sites (Refer to Nat. Commun. 12, 6806 (2021)). 
In order to clearly reveal the relationship between the N vacancy and the Cu–N single-atom configuration, we re-tested the XAFS data and supplemented the quantitative fitting of the EXAFS at k-space. The relative results revealed that the copper species in Cu1–N2, Cu1–N3 and Cu1–N4 SACs all showed atomically dispersed, with Cu–N coordination numbers of 2.2, 2.8 and 3.7, respectively. Additionally, the ICP-OES, EDS and XPS results show that the SACs maintained the similar content of Cu and N species (Supplementary Figure 4 and Supplementary Table 1). These above results indicate a consistent coordination environment for the Cu single-atom sites within all as-prepared Cu based SACs. In conclusion, the difference in N vacancy between Cu1–N2, Cu1–N3 and Cu1–N4 SACs was directly derived from the difference in the coordination structure of Cu species. It can be seen that the ID/IG ratio of SACs is negatively correlated with the Cu-N coordination number.
We agree with the reviewer's suggestion that this is not direct evidence for demonstrating the coordination saturation of single atoms. To avoid improper guidance, we have made appropriate revisions in the revised manuscript, hoping to address the reviewers' concerns.


Comment#4. Were the CO peak centers the same for all samples? Doesn't it seem surprising that the CO vibration would not be sensitive to the proposed differences in Cu structure that apparently sufficient to change the electronics of adsorbed O2?
Response: Thank you for your important comment. Based on the rigorous academic attitude, we have carefully checked and further retested this controversial measurement of all the samples to study the scientific results. 
Except the peaks around 2115 cm-1 and 2175 cm-1 assigned to gaseous CO in condition, an extra signal at 2054 cm-1 (Cu1–N2 SAC), 2057 cm-1 (Cu1–N3 SAC) and 2060 cm-1 (Cu1–N4 SAC) representing the CO linear adsorption (COL) on Cu atoms was found for all Cu based SACs. It is worth noting that the COL peak has shifted to the lower frequency direction from Cu1–N4 SAC to Cu1–N3 SAC and to Cu1–N2 SAC. This is attributed to the difference of the oxidation state of Cu, which corroborates with the results of XANES and XPS. Moreover, the absence of bridge bonded CO band in the range of 1700-1900 cm-1, which rules out the presence of multi-atom Cu species (Refer to Angew. Chem. Int. Ed. 62 (2023), e202217220, Angew. Chem. Int. Ed. 126, 4956–4961 (2014), Nat. Commun. 12, 6022 (2021), Appl. Catal. B Environ. 339, 123119 (2023), Chem. Eng. J. 146870 (2023), J. Catal. 405, 333–345 (2022), Appl. Catal. A Gen. 505, 334–343 (2015), J. Phys. Chem. C 113, 10689–10695 (2009) and J. Catal. 268, 367–375 (2009)). The in situ DRIFTS of CO adsorption on Cu1–N2, Cu1–N3 and Cu1–N4 SACs at 30 °C in our previous experiments showed only two peaks due to the CO adsorption time was too long and the CO concentration was too high, resulting in the overlap of the two peaks (around 2055-2060 cm-1 and 2115 cm-1). In addition, the different intensity of CO linear adsorption peak also indicates that Cu1–N2 SAC contains a higher level of Cu+ species than Cu1–N3 and Cu1–N4 SACs. This indicates that the CO adsorption is sensitive to the Cu–N fine coordination structure, which is sufficient to activate the adsorbed oxygen and change their electronic properties. 
In order to further explore the above results, DFT calculations have also been carried out to analyze the Bard charge on the Cu atoms and the oxygen species. The Bard charge on the Cu atom in Cu1–N2 SAC is + 0.74 e, which is much lower than that of Cu1–N3 and Cu1–N4 SACs (+ 0.81 e and + 0.84 e, respectively). This indicates that the low-coordinate Cu1–N2 site is more likely to donate electrons. According to the Bard charge of activated intermediate oxygen species on copper species, the Bard charge of the intermediate oxygen species O2- shows obviously electrophilic compared with the Cu1–N3 sites, Cu1–N4 sites and traditional DEP catalysts (such as Cu nanoparticles or Cu2O), which tended to attack the electrically rich C=C double bond in propylene to generate PO. These results reflect the key role of isolated Cu1–N2 sites in changing the electronic structure of adsorbed O2 into electrophilic oxygen species. 
In the revised manuscript, the details have been corrected finally. Thanks again for the reviewer's comments.

Comment#5. What is the explanation for selectivity increasing with temperature? Typically, this increases rates of C-H scission/H abstraction in selective oxidation reactions, leading to decreased selectivity with increasing temperature. How do the authors explain the decreasing selectivity to PO with time on stream, despite all the characterization data that shows the materials are structurally stable during reaction?
Response: Thank you for the insightful comments from the reviewer. 
As reported in most of the published literature and discussed in our manuscript, copper-based catalysts have been widely utilized in DEP reactions and exhibit exceptional catalytic performance. It is generally recognized that the Cu+ species possesses a high efficiency in DEP reaction with high PO selectivity because the oxygen species activated on Cu+ sites have relatively strong electrophilicity, which is conducive to attacking the C=C double bond of propylene. However, the oxygen species activated on Cu2+ species exhibit nucleophilicity, tending to attack the α C–H bond of propylene, leading to the C-H scission/H abstraction and the formation of an over-oxidation product (Refer to J. Mol. Catal. Chem. 228, 27–33 (2005); J. Catal. 268, 165–174 (2009)). In our study, we have characterized the Cu oxidation states in Cu1–N2 SAC using XAS, XPS Cu 2p and Cu LMM characterizations in detail (Figs. 2c, d and Supplementary Figure 12). The results indicated that the oxidation state of Cu species was nearly +1 (XANES) and the Cu+:Cu2+ ratio in Cu1–N2 SAC was nearly 9:1. The above results indicated that Cu1–N2 SAC possessed an intrinsic activity that was suitable for the production of PO with high selectivity.
Regarding the catalytic performance of DEP in our manuscript, we only presented the propylene conversion, the selectivity of various products and PO formation rate to evaluate the catalytic performance of the Cu-based catalysts. This led to misunderstandings among the reviewer. We sincerely apologize for our negligence and have carried out additional experimental analysis to better display the catalytic performance. We have added the Supplementary Fig. 20 depicting the formation rates of various products including PO, CO2, acrolein, acetaldehyde and aceton at different reaction temperatures. It can be observed in Supplementary Fig. 20 that the formation rate of PO increased rapidly from 200 oC to 275 oC, while the formation rate of CO2 increased slowly. At this stage, the selectivity of PO was primarily hindered by thermodynamic energy barriers, while the rate of complete oxidation was influenced by adsorption/desorption processes (Refer to J. Colloid Interface Sci. 611, 564–577 (2022)). Furthermore, the complete oxidation is more favorable according to the relevant data of thermodynamics (Refer to Appl. Surf. Sci. 355, 660–666 (2015); ACS Catal. 10, 13415–13436 (2020)). Thus, the elevation in temperature from 200 oC to 275 oC exhibits tiny influence on adsorption/desorption processes, but instead promotes the activation of O2/C3H6 by Cu+–N single-atom sites. The experimental data indicate that the ratio of the formation rates of PO to CO2 increases approximately from 1.0 (200 oC) to 4.5 (275 oC), leading to an enhancement in PO selectivity. 
When the reaction temperatures was above 275°C, several factors contributed to the significant increase in CO2 formation. First, the excessively high reaction temperatures lead to aggregation and oxidation of Cu species to Cu2O or CuO, which have been proved by the XRD and XPS Cu 2p characterizations (Supplementary Figs. 23b and 25a). The lattice oxygen species excited at high temperatures exhibit nucleophilicity, accelerating the extraction of α C–H bond of propylene, converting more C3H6 into by-product and further to CO2. Besides, the generated PO will be susceptible to isomerization at high temperatures for CO2 formation (Refer to J. Am. Chem. Soc. 111, 4553–4561 (1989); Surf. Sci. Rep. 76, 100524 (2021)). The carbon nitride, serving as the support, can also promote the conversion of propylene into acrolein and CO2 at high temperatures (Supplementary Fig. 17). These factors collectively result in a substantial increase in the generation rate of CO2.
In summary, we believe that the Cu+–N2 species serves as the real active-site for the production of PO, while Cu2+ favors the generation of CO2. Within a moderate temperature range, an increase in the reaction temperature is conducive to accelerating the efficient production of PO, compared to other products. However, the excessively high reaction temperatures will significantly exacerbate the agglomeration/oxidation of copper species, the isomerization of PO, and the over-oxidation of propylene.
In the "Migration of Cu in the Cu1–N2 SAC" section, the EXAFS results demonstrate the durability of the Cu1–N2 coordination structure. Additionally, Supplementary Figs. 23, 24 and 25 also confirm the high dispersion and valence stability of the copper species in Cu1–N2 SAC. Concerning the decrease in PO selectivity over time on stream, the nature of the carbon nitride support caught our attention. As known, the carbon nitride material shows a two-dimensional layered characteristic with interlayer van der Waals forces. Prolonged exposure at high-temperatures and an oxygen-enriched environment leads to interlayer peeling and is easily oxidized. This was accompanied by a substantial reduction in BET surface area and the emergence of numerous pores, which reduces the structural order and stability of the support (Supplementary Table 4 and Supplementary Fig. 24a, Refer to Adv. Funct. Mater. 22, 4763–4770 (2012) and J. Colloid Interface Sci. 553, 530–539 (2019)). Defected and oxidized carbon nitride structures can increase the activation of molecular oxygen into nucleophilic oxygen species, leading to the formation of by-products and a decrease inPO selectivity (as demonstrated in Appl. Catal. B Environ. Energy 353, 124022 (2024)).
We sincerely appreciate your comment, and we have revised the manuscript to provide a clearer and more accurate expression. Thank you again for your helpful feedback.

Comment#6. The high selectivity initially that then decreases could result from formation of surface species that consume PO. Have the authors measured the selectivity as a function of propylene conversion, or co-fed PO to observe whether it is increasingly consumed with time on stream when co-fed during propylene epoxidation?
Response: 

Comment#7. In the "implication of reaction mechanism for DEP" section, if none of these peaks are observed when propylene is fed first, does that mean that if you first feed propylene, then feed propylene and oxygen, this catalyst does not make PO? Or is this observation just saying that if you feed propylene first, then O2 alone second, no PO is formed?
Response: We greatly thank you for carefully reviewing our manuscript. We conducted an analysis of the reaction mechanism by altering the feedstock introduction sequence. The results showed that the dispersed single atoms with low-coordinate Cu–N coordination structure avoided the co-adsorption of propylene and oxygen, and followed the sequential activation mechanism. This is obviously different from the traditional DEP reaction mechanism, which involves adsorbing oxygen first and activating propylene. Therefore, when propylene gas was first introduced into the test system, all the active sites were occupied by the propylene molecules. The non-oxidation of propylene was caused by the lack of active sites on the catalyst surface that could absorb the molecular oxygen that was introduced later. When the propylene gas was first fed, and then the propylene/oxygen mixed gas was fed, the results are as follows (Figure R1). In the initial stage, we could not observe the formation of any product through propylene oxidation, because the propylene molecule that was adsorbed on the surface had not been purged. However, a large number of products appeared shortly thereafter, indicating that the competitive adsorption of molecular oxygen in the mixed gas on low-coordinate Cu–N coordination sites was dominant. The activation of oxygen molecules resulted in the intermediate reactive oxygen species attacking the propylene molecule. 
These above results strongly support the conclusion that the excellent DEP performance via Cu1–N2 SAC primarily stems from the special reaction mechanism, that is, adsorption and activation of oxygen before the bridged adsorption of propylene. Finally, the results above have been appropriately included in the revised manuscript. 

Comment#8. The computational part was so thorough, it was great. Not every paper takes the time to determine pathways on the control/non-optimal samples (like the Cu(111) in this study). Nice work!
Response: We thank you for the highly positive comment. We hope this study will provide in-depth and meaningful insights for readers of Nature Catalysis.

Reply to Reviewer #3:


Fig. 2 | Structural characterizations. (a) Cu K-edge EXAFS of Cu foil, Cu2O, CuO, Cu1–N2, Cu1–N3 and Cu1–N4 SACs. (b) Wavelet transform of EXAFS of Cu foil, CuO, Cu1–N2, Cu1–N3 and Cu1–N4 SACs. (c) Cu K-edge XANES of Cu foil, Cu2O, CuO, Cu1–N2, Cu1–N3 and Cu1–N4 SACs. (d) Cu 2p XPS spectra of Cu1–N2, Cu1–N3 and Cu1–N4 SACs.


Supplementary Fig. 8 | XPS spectra of various catalysts. N 1s XPS spectra of C3N4, Cu1–N2, Cu1–N3 and Cu1–N4 SACs.


Supplementary Figure 13 | Schematic illustration of the plausible formation mechanism of low-coordinated Cu1–N2 centres.


Supplementary Figure 25 | Characterizations of spent catalysts. (a) Cu 2p XPS spectra and (b) Cu LMM spectra of spent Cu1–N2, Cu1–N3 and Cu1–N4 SACs. (c) Cu 2p XPS spectra of CuNP/C3N4–L and spent CuNP/C3N4–L.

Supplementary Table 4 | FWMH of C3N4 and C3N4–L for C 1s XPS.
Sample	FWMH (ev)
	C−C/C=C	N−C=N	C−NHX
C3N4	1.30	1.21	1.30
C3N4-L	1.30	1.26	1.30

Supplementary Table 4 | FWMH of C3N4, Cu1–N2 SAC, Cu1–N3 SAC and Cu1–N4 SAC for N 1s XPS.
Sample	FWMH (ev)
	C−N=C	C−NHx	N−(C)3	Cu−N
C3N4	1.35	1.30	1.31	/
Cu1–N2 SAC	1.35	1.35	1.25	1.28
Cu1–N3 SAC	1.40	1.25	1.40	1.40
Cu1–N4 SAC	1.40	1.40	1.20	1.37

Supplementary Table 4 | FWMH of Cu1–N2 SAC, Cu1–N3 SAC, Cu1–N4 SAC, CuNP/C3N4–L and their spent samples for Cu 2p XPS.
Sample	FWMH (ev)
	Cu+	Cu2+
Cu1–N2 SAC	1.84	1.83
Cu1–N3 SAC	1.85	1.85
Cu1–N4 SAC	1.85	1.85
Spent Cu1–N2 SAC	1.80	1.85
Spent Cu1–N3 SAC	1.80	1.85
Spent Cu1–N4 SAC	1.80	1.85
CuNP/C3N4–L	1.85	1.85
Spent CuNP/C3N4–L	1.85	1.85


Supplementary Figure 14 | Characterizations and physicochemical properties of carriers. (a) XRD pattern and (b) FT-IR spectra of melamine, L-cysteine and precursor. (c) FT-IR spectra of precursor in different steps. (d) C 1s XPS spectra of C3N4 and C3N4–L. (e) 13C/1H CP-MAS NMR spectra of C3N4 and C3N4–L. SEM images of (f) C3N4 and (g) C3N4–L. (h) N2 adsorption/desorption isotherms and (i) pore size analysis of pristine C3N4 and C3N4–L.


General Comment: The manuscript (Ref. NATCATAL-24020271) describes the synthesis of a series of Cu-based catalysts deposited on g-C3N4 and their subsequent use in the epoxidation of propylene in gas phase using molecular oxygen as sole oxidant. The manuscript reads very well in general and presents some very interesting insights of the interplay between synthetic parameters and catalytic performance. However, there are some fundamental flaws that significantly hinder the potential impact of the manuscript that must be addressed by the authors to deliver a more solid and robust work. I can only recommend publication of the manuscript after major revisions are made to the original text. The authors are kindly asked to revise their manuscript following these aspects, making the appropriate changes. 
Response: We thank you for the highly positive comment of “The manuscript reads very well in general and presents some very interesting insights of the interplay between synthetic parameters and catalytic performance”. We also greatly appreciate your constructive and meaningful suggestions on our manuscript. These suggestions are very important to us and we have made every effort to refine our study and carefully revised our manuscript based on your comments.

Comment#1. Concerning English usage, despite the manuscript very clearly and concisely written is not entirely devoid of flaws. Some grammatical and typographical errors may still be found which should be revised for a better reading. The authors are kindly asked to proof-read their manuscript in order to come up with a more robust and readable text. The following are just a few illustrative examples to support this point:
- In the Introduction section "For above discussion, the low-coordinate" should read "From the above discussion, the low-coordinate"
- In the section "Formation process of Cu1-N2 SAC" several Figure numbers are incorrect, please revise all Figure referencing and mention them accordingly
- "Understanding the migration of active metal species during catalysis is crucial for investigating the underlying mechanismss" should read "Understanding the migration of active metal species during catalysis is crucial for investigating the underlying mechanisms"
- "This was followed by the interaction of the remaining O- species (M2) with second propylene to create an oxametallopropylene intermediate" should read "This was followed by the interaction of the remaining O- species (M2) with a second propylene molecule to create an oxametallopropylene intermediate" 
Response: We thank the reviewer for carefully reviewing our manuscript. We have taken your comments seriously and have carefully proofread the manuscript to address these issues. We have completed the revisions and have highlighted the changes in yellow in the manuscript for your reference. We sincerely apologize for any inconvenience or confusion caused by these errors and appreciate your understanding and patience.

Comment#2. The authors mention that the figure shown in SF6 are Type I isotherms. This is a somewhat optimistic assumption. The isotherms are barely Type I, and they appear to be type II. Did the authors perform BET analyses to check the accessible surface area? The plots on PSD are simply not informative. The obtained surface areas do not correspond to microporous solids. Thus, the discussion following this identification is fundamentally not true.
Response: We thank this useful suggestion from the reviewer. In the isothermal adsorption process, the BET equation can be expressed in terms of adsorption volume (V) and relative pressure (P/P0) can be expressed as follows:

Where P0 represents the saturated vapor pressure of the adsorbed substance at the adsorption temperature, Vm represents the saturated adsorption amount of the adsorbed substance at the single molecular layer, C is the constant of BET theory.
According to the BET equation, the value of C has a significant influence on the characteristic of the curve. The value of C in the BET equation plays a crucial role in determining how P/P0 influences the gas adsorption when Vm reaches the saturation point. When V = Vm, the BET equation can be simplified to: 

Based on the BET analyses to check the accessible surface area, the experimental details are shown as follows: The Brunauer-Emmett-Teller surface area and Barrett-Joyner-Halenda (BJH) porosity of samples were analyzed by the N2 adsorption/desorption isotherms by using Micromeritics (ASAP 2460) at -196 °C, then dried at 100 °C for 6 h under vacuum conditions. According to the above equations and the obtained results, the C values of the Cu1–N2, Cu1–N3 and Cu1–N4 SACs have been calculated at 5.978, 5.977 and 5.985, respectively, which is fundamentally scientific. In addition, C > 2 is the characteristic of an adsorption isotherm conforming to type II or type IV. As shown in Supplementary Fig. 6, the type H3 hysteresis loop appears in the middle of the adsorption isotherms, which corresponds to the capillary condensation of the porous adsorbent. This is mainly due to the nonrigid aggregates of two-dimensional carbon nitride sheets. Based on the IUPAC classification, it appears that the as-prepared SACs have type IV adsorption isotherms.
We would like to express our sincere gratitude to you again for your valuable comments.

Comment#3. While the discussion is apt, the XPS analysis does not appear entirely correct. The FWHM values should be similar for all deconvoluted peaks, yet this does not appear to be the case. The authors should re-analyze the XPS spectra.
Response: Thank you for the careful and rigorous review. Based on the principles of XPS and the physical significance of FWHM, the same element should exhibit similar FWHM values under different chemical environment. Thanks to the reviewer's recommendations, we have performed a re-analysis and peak-fitting of the XPS spectra to ensure rigorousness in our revised manuscript. We sincerely appreciate your comment again. In case there are any other aspects you deem necessary for us to further consider, we are willing for the further revision.

Comment#4. Compared to pristine C3N4, the sp3- hybridized nitrogen peaks in all SACs shifted to a higher binding energy, suggesting an increased tendency for pyridinic N species to coordinate with Cu atoms." This observation is hardly visible from SF8. The shiftings should be marked for a better identification.
Response: Thank you for your thorough review of our manuscript and for bringing this important error to our attention. We deeply regret the typo mistake in our manuscript, where we incorrectly marked "sp3-hybridized nitrogen" instead of "sp2-hybridized nitrogen." This error has led to an inaccurate conclusion in our interpretation of the data presented in Supplementary Fig. 8. The N 1s spectrum showed a new peak at 397.9 eV which can be assigned to the characteristics of Cu–N coordination, further corroborating the interaction between Cu and N atoms. Furthermore, the peaks at 399.0 eV can also be observed, corresponding to sp2-hybridized nitrogen in the triazine ring (C–N=C). We acknowledge that the binding energy of the sp2-hybridized nitrogen peaks has shifted towards higher binding energies, as indicated in the corrected version, suggesting that the pyridinic N species tend to coordinate with Cu atoms. For better identification, the shifts have been marked in Supplementary Fig. 8 in the revised manuscript.

Comment#5. What do the arrows indicate in SF14a?
Response: We thank you for carefully reviewing our manuscript. In the original Supplementary Figure 14a, the arrows indicate the XRD diffraction peaks of melamine, demonstrating that the mixture obtained after Step I retained the original structure of the melamine precursor. To gain deeper insights into the products of Step I, we conducted a meticulous analysis of the XRD spectra. In the updated illustration, melamine structures were marked in black and the XRD diffraction peaks can be visible at 2θ of 17.6°, 21.6°, 22.1°, 26.2°, and 28.8°, which can be characteristic of (-111), (210), (-211), (-301), and (-311) of melamine (JCPDS 75-1288). Additionally, the XRD diffraction peaks at 2θ of 9.9°, 19.8°, 21.0°, 27.5°, 29.3°, 30.0°, and 46.3° can be assigned to (010), (020), (121), (200), (0-22), (014), and (304) of Cu(C2H7N5)2Cl2·2H2O (JCPDS 24-1654) , which came from the coordination complex between Cu2+ and melamine and was highlighted in red. The blue signs denote the copper oxychloride crystals arising from the hydrolysis of CuCl2 and the XRD diffraction peaks at 2θ of 16.2°, 32.3°, and 39.7° can be attributed to (-101), (210), and (220) of Cu2(OH)2Cl (JCPDS 86-1391). As shown in Supplementary Figure 14b, all the above-mentioned XRD diffraction peaks are well indexed, proving the reliability of the conclusion. We sincerely appreciate your thoughtful comments, as they have allowed us to provide a more detailed explanation. If there are additional points you would like us to address, we would be glad to incorporate further insights into the revised manuscript.

Comment#6. The decreased absorbance at 3564 cm-1, ascribed to the desorption of H2O, verifies the above point." This observation does not appear very realistic. The intensity of the broad band around 3500 cm-1 is very similar for the three samples.
Response: We thank you for carefully reviewing our manuscript and apologize for the oversight of not presenting the relevant information clearly and adequately. To better illustrate the decreased absorbance at 3564 cm-1, the spectral region between 3000 cm⁻¹ and 4000 cm⁻¹ range was appropriately magnified to facilitate detailed observation. As can be clearly seen in Supplementary Figure 14e, the decreased absorbance specifically at 3564 cm⁻¹ can be attributed to the desorption of H₂O. The evident result confirmed our aforementioned point, that is a significant dehydration process occurred due to the reaction between melamine and L-cysteine during the Step I in Supplementary Figure 14a. We apologize for our oversight and would like to express our sincere gratitude to you once again for your thorough review. 

Comment#7. NMR spectroscopy does not reveal, to the best of my perception, a clear signal for Carbons 1, 2, or 3. H NMR does show the differences in the NH2 signal, but other differences are practically invisible.
Response: We thank the reviewer for the insightful commen. As shown in the 13C CP-MAS NMR spectroscopy (Supplementary Figure 14e), both C3N4 and C3N4–L exhibit resonant signals at 156.2 ppm and 163.9 ppm, respectively, which can be attributed to the C4 atom in the N=C–N2 structure and the C5 atom in the N=C–N(NH2) structure of the heptazine unit (Refer to Angew. Chem. Int. Ed. 49, 441–444 (2010); Appl. Catal. B Environ. 254, 128–134 (2019); Chem. Eng. J. 383, 123132 (2020)). XPS C1s spectra indicates that the ratio of C5 to C4 has increased from 2.24 (C3N4) to 2.54 (C3N4–L), which aligns with the conclusions derived from the through further analysis (Refer to Adv. Sci. 11, 2408293 (2024)). Therefore, the successful modification by ACAS strategy was validated. The addition of the L-cysteine modifier during the synthesis of C3N4–L was limited to only 0.5% of the melamine content to ensure the feasibility of modification and that the periodicity of the carrier structure remains intact. This resulted in weak resonant signals for C1, 2 and 3 in the NMR spectra. Magnifying the spectral regions around 151.4, 158.7, and 169.4 ppm allows us to better illustrate this variation. We extend our heartfelt gratitude again to the reviewers for their incisive and profound comments.

Comment#8. SF13b is not a mechanism, merely the insertion of Cu (II) atoms in the polymerized structure deducted by the authors from their findings.
Response: We appreciate your thorough review of our manuscript and we also acknowledge the reviewer’s point that Supplementary Fig. 13b is not a mechanism. Thank to this suggestion, which has prompted us to refine the scheme to make the formation process of Cu1–N2 SAC through ACAS strategy clearer and more visualized. Specifically, the processes of the functionalization modification for C3N4 carrier and the Cu single-atom site anchoring were integrated, including the amidation reaction, the self-condensation reaction, the metal-heteroatom coordination, and the dehydration reaction, etc. The details are as follows.
Supplementary Fig. 13 specifically illustrates the synthetic steps for Cu1–N2 SAC utilizing the ACAS strategy. First of all, the hydrolysis of CuCl2 and the modifier L-cysteine creates an acidic medium that facilitates the hydrolysis of melamine. Compound I is formed by the reaction between L-cysteine and melamine in this step. The abundant terminal –NH2 of melamine would coordinate with free Cu2+ species to form Compound II, as evidenced by the XRD characterization (Supplementary Fig. 14a). Secondly, during the solvothermal reaction process, melamine underwent further hydrolysis to generate cyanuric acid units. The in-situ formed cyanuric acid and melamine spontaneously assembled into a supramolecular structure through hydrogen bonding (Refer to Chem. Eng. J. 405, 126685 (2021); Nano Res. 11, 3462–3468 (2018)). Due to the in-situ modification of nitrogen-containing units, the low-coordinate Cu–N structures can form (Compound III) during the assembly process, which further stabilizes the anchoring of Cu–N species. After that, the cyanuric acid in Compound III released the ammonia gas and left melamine to self-condense into heptazine units (Compound IV) (Refer to ACS Appl. Mater. Interfaces 11, 10651–10662 (2019); Small 20, 2309032 (2024); Langmuir 30, 447–451 (2014)). Subsequently, the modified moiety within Compound III underwent cyclization to form Compound V by releasing NH3 and H2S molecules. This was followed by dehydration to generate Compound VI during pyrolysis. The low-coordinate Cu1–N2 SAC was formed by repeating these steps iteratively and periodically. Throughout these above processes, the assembly of the supramolecular structure and the formation of asymmetric melon configurations created favorable conditions for the formation of high-loading and low-coordinate copper single-atom sites. 
We would like to express our sincere gratitude to you again for your valuable comments. The relevant content has been updated in the revised manuscript and the Supplementary information.

Comment#9. I apologize, but I find that the conclusions concerning electrophilic oxygen species is not entirely correct. Electrophilic species must inevitably possess a diminished electron density (delta+) for them to behave as electrophilic species, which would then target electron-rich domains (such as a double bond). Any species with a (delta minus) is usually considered Nucleophile and would thus never target a double bond. What electrophilic oxygen species have the authors identified?
Response: Thank you for your insightful comments regarding the electrophilic and nucleophilic nature of oxygen species. We value the opportunity to address this question in detail as it is a crucial aspect of understanding the DEP reaction mechanism.
Firstly, we appreciate your emphasis on the distinction between electrophilic and nucleophilic oxygen species. The classification of oxygen species based on their electron density and reactivity is supported by extensive research in the context of oxidation reactions. In particular, oxygen species adsorbed on metal sites typically undergo a series of electron transfers from the metal, progressing towards the lattice oxygen state. Among these species, singlet oxygen (~1O2) as well as molecular and atomic ion-oxygen species (O2-, O-, O0) are widely recognized for their strongly electrophilic properties. In contrast, lattice oxide ions (O2-) are typically nucleophilic reagents with the strongest nucleophilicity.
Reactions involving oxygen can therefore be divided into two groups based on the nature of the oxygen species: (1) electrophilic oxidation, which involves the activation of oxygen to its electrophilic forms and their attack on electron-rich domains of the reacting molecule, and (2) nucleophilic oxidation, where nucleophilic oxygen species add to a molecule that has been activated to make such addition favorable. For instance, in the case of hydrocarbons, electrophilic oxidation often leads to C–C/C=C bond cleavage, whereas nucleophilic oxidation typically tends to attack the allylic or partial oxidation. The alignment of this framework with Boreskov's early ideas on associative and stepwise oxidation mechanisms is reflected in contemporary studies (React. Kinet. Catal. L. 35, 369–379 (1987); Catal. Rev. 19(1), 1–41 (1979)). 
Furthermore, there is a plethora of literature that shows how electrophilic oxygen species are involved in olefin oxidation reactions(J. Am. Chem. Soc. 144, 7693–7708 (2022); ACS Catal. 9, 6262–6275 (2019)). In our previous work, we have similarly highlighted the critical role of electrophilic oxygen species in controlling the PO selectivity and catalytic performance of DEP reactions (J. Am. Chem. Soc. 142, 14134–14141 (2020)). 
In the present study, we have identified the electrophilic O2- species on Cu1–N2, Cu1–N3, and Cu1–N4 SACs through density functional theory calculations. The Bader charge analysis revealed that these oxygen species carry a charge of -0.08 |e| to -0.33 |e|, which is significantly less negative compared to oxygen species dissociated on Cu (111) surfaces or lattice oxygen in Cu2O (110) surface. Specifically, on Cu (111) surfaces, molecular oxygen is easily dissociated, and the resulting oxygen atoms have a significant negative charge of -0.90 |e|, which is typicalof nucleophilic oxygen species. For lattice oxygen in Cu₂O at elevated temperatures, the Bader charge reaches -0.97 |e|, and these nucleophilic oxygen species were found to require surmounting a significant barrier (0.87 eV) for oxygen insertion reactions. These results highlight the distinct reactivity of nucleophilic oxygen species, which is not conducive to the activation of C=C double bonds in propylene. In addition, the above results also emphasize the critical role of electrophilic oxygen species in the DEP reaction. Furthermore, the low-coordinate Cu–N single-atom sites effectively suppress the formation of strongly nucleophilic oxygen species, enabling precise control of selectivity in the DEP process. 
Your thoughtful comments have enabled us to provide a more detailed explanation and strengthen our conclusions, which we sincerely appreciate. If there are any additional points you want us to address, we are more than happy to incorporate them into the revised manuscript.

Comment#10. Concerning the catalytic data, while the authors have prepared a catalyst that performs quite well, the PO yield, propylene conversion and PO selectivity fall below the benchmark values which are considered the thresholds for their commercial implementation (as reviewed by Khatib and Oyama in S.J. Khatib, S.T. Oyama, Catal. Rev. 57 (2015) 306.) Furthermore, in terms of these parameters, the reported values fall below previously published papers (see for example J. García-Aguilar et al. J. Catal. 386 (2020) 94-105) which use significantly simpler and more cost-effective catalysts. The impact of this novel research needs to be further justified. While the PO production rate values are interesting, the other parameters are not impressive.
Response: Thank you for your thoughtful comments and the valuable references you have provided. They have helped us to better showcase the advantages of our study compared to the reported literature. We fully acknowledge that the propylene conversion, and PO selectivity reported in our study fall below the benchmark values required for commercial implementation, as highlighted in the review by Khatib and Oyama (Catal. Rev. 57, 306 (2015)). According to the reference of Catal. Rev. 57, 306 (2015), a possible target for commercialization of PO production is a PO selectivity of 70% at a propylene conversion of 10%. This was based on comparison to ethylene oxide (EO) processes, which currently operate with EO selectivity of 80-90% and ethylene conversion of 7-15%. It is worth noting that, our study was based on the direct epoxidation of propylene route with molecular oxygen alone, which is believed to be an ideal route for PO production and is significantly different from the traditional technology. Even though the DEP process struggles with the trade-off between activity and selectivity, the as-prepared Cu1–N2 SAC has achieved the optimal performance among the reported DEP catalysts and demonstrated highly competitive catalytic activity when compared with other catalysts using other PO production processes. The potential of this catalyst system is demonstrated in our study and provides a valuable foundation for future improvements and optimizations. To better illustrate the differences between García-Aguilar’s work (J. Catal. 386, 94–105 (2020)) and our work, a comparison was made in the following Table R1.
Table R1. The contrast between García-Aguilar’s work and our work.
	García-Aguilar’s work	This work
Goal	Study Ni/Ti-SiO2 catalysts for propylene epoxidation using H2/O2, with the aim of substituting noble metals with cost-effective alternatives.	Propose an ACAS strategy to synthesize low coordinate Cu1–N2 SAC for propylene epoxidation using O2 alone, significantly boosting the catalytic performance of copper-based catalysts.
Catalyst structure	Ni nanoparticles loaded on Ti-SiO2
(Ni/Ti-SiO2)	Cu1–Nx single-atom sites coordinated on C3N4 nanosheets (x=2, 3, or 4)
Research Strategy	Performing the catalytic reaction and characterizations over the Ni/Ti-SiO2 catalysts. The active site and mechanism were suggested based on the catalytic performance and structure characterization. The author demonstrated that Ni can serve as a substitute for Au in the epoxidation of propylene using H2/O2 processes.	Considering the particularity of the DEP reaction, which involves the use of molecular oxygen as the sole oxidant and the difficulty of the DEP reaction, we have developed an ACAS strategy to facilitate the stabilization of low-coordinate Cu1–N2 species. Cu1–N2 SAC has an excellent formation rate for PO with a high selectivity for PO and a high propylene conversion. Unveiling the intricate structure-activity relationship of Cu–N coordination via in-situ characterization and theoretical calculations.
Catalytic performance and reaction condition	Propylene epoxidation to PO with H2 and O2: 6 % propylene conversion, 85% selectivity to PO and 1.92 mmolPO gcat-1 h-1 PO formation rate. Reaction condition: WHSV=20000 h-1, C3H6:H2:O2=1:1:1 (vol.), 200 oC.	Propylene epoxidation to PO with only O2: 5.7 % propylene conversion, 78%  PO selectivity and 5.87 mmolPO gcat-1 h-1 PO formation rate. Reaction condition: GHSV=36000 h-1, C3H6:O2=2:1 (vol.), 275 oC.
Relationship between catalytic performance and catalyst structure	The metal species loaded on Ti-SiO2 determine its catalytic performance. The well-dispersed Ni particles and tetrahedral Ti species enhanced the catalytic performance, demonstrating high PO selectivity and stable catalytic activity, which can be comparable to some noble metal-based catalysts.	The low-coordinate Cu1–N2, with the dynamic regulation of valence state of Cuδ+ (where 1<δ<2) center,  facilitated the formation of electrophilic oxygen species, which targeted the electronrich C=C bond in propylene for efficient PO production. In contrast, Cu1–N4, characterized by higher Cu–N coordination numbers and Cu oxidation valence, delivered moderate catalytic performance. Cu1–N3 sites, on the other hand, allowed the co-adsorption of C3H6 and O2, resulting in poor performance for DEP reaction.
We are confident that through multi-dimensional comparisons between García-Aguilar’s work and our work, we can allay your concerns and gain your recognition for our study.

Comment#11. Concerning literature revision, I find it surprising that reference works (such as the papers mentioned in the previous point) have been seemingly overlooked by the authors. I strongly advise to check recent literature references on DEP to further strengthen the discussion and make the manuscript of a broader appeal.
Response: Thank you for your constructive feedback on our manuscript. We have thoroughly reviewed the latest literature on propylene epoxidation reaction in response to your advice, with a particular focus on advancements in DEP reaction. Our discussion has been enhanced by the additional literature search, which provides a broader context for our work.
Furthermore, to disperse your doubts directly, we have updated our manuscript to include a comprehensive table that systematically compares the catalysts and their performance, specifically highlighting only those studies that involve the direct epoxidation of propylene using molecular oxygen. The innovative nature and significant breakthrough of our work are clearly demonstrated in this table. According to the manuscript, the Cu1–N2 SAC (5.9 mmol gcat-1 h-1) has a PO formation rate that is 5 to 400 times higher than that of other catalysts. The PO formation rate over Cu1–N2 SAC has been increased by 1,073%~14,565%, based on the reported DEP catalysts with ≥60% PO selectivity. The catalytic performance of Cu1–N2 SAC has reached and even surpassed the performance of propylene epoxidation catalysts through H2/O2. Our catalyst's effectiveness and superiority in PO production rate are demonstrated by this unprecedented performance, which has great scientific and industrial value. 
We believe that these revisions have further strengthened our manuscript and will make it more appealing to a broader audience. Thank you once again for your valuable suggestions, which have undoubtedly improved the quality and comprehensiveness of our work.

Table R2. Comparison between Cu1–N2 SAC, Cu1–N3 SAC, Cu1–N4 SAC, CuNP/C3N4–L and other reported IB group metal catalysts for propylene epoxidation with O2.
No.	Catalyst	Conv. (%)	PO select. (%)	PO formation rate
(mmol gcat-1 h-1)	Temp. (oC)	Flow rate (mL min-1)	Pressure (MPa)	Ref.
1	Ag3/Al2O3	/	85	/	110	/	/	Science. 339, 1590-1593 (2013)
2	Ag-MoO3/ZrO2	0.6	58	0.236	350	62.5	0.1	J. Mol. Catal. A-Chem. 232, 165-172 (2005)
3	Ag-CaCO3-NaCl	3.7	45	1.24	260	30.0	0.3	Appl. Catal. Gen. 302, 283–295 (2006)
4	Ag/Y2O3-K2O/α-Al2O3	4	47	0.303	245	16.7	0.1	Catal. Lett. 119, 185–190 (2007)
5	Ni1Ag0.4/SBA-15	/	70.7	0.015	220	20.0	0.1	Appl. Catal. B-Environ. 243，304-312 （2019）
6	Au/TS-1-KOH	0.88	52	0.19	200	20.0	0.1	Angew. Chem. Int. Ed. 48, 7862-7866 (2009)
7	Cubic Cu2O	0.80	10	/	225	50.0	0.1	Angew. Chem. Int. Ed. 126, 4956–4961 (2014)
8	Octahedral Cu2O	0.67	3	/	200	50.0	0.1	
9	Rhombic dodecahedra Cu2O	0.2	50	0.057	250	50.0	0.1	
10	Rhombic dodecahedra Cl-Cu2O	0.05	95	0.02	150	50.0	0.1	J. Am. Chem. Soc. 142, 14134-14141 (2020)
		1.0	63	0.5	200	50.0	0.1	
11	Cu/SiO2	0.25	53	0.014	225	50.0	0.1	J. Catal. 236, 401-404 (2005)
12	VCe0.5Cu0.5-NaCl	0.26	33	0.165	250	60.0	0.1	J. Catal. 211, 552-555 (2002)
13	c-Cu2O-27	0.06	82	0.050	110	50.0	0.1	Nat. Commun. 12, 5921 (2021)
14	Cu7Ce	0.68	95	0.078	80	50.0	0.1	J. Phys. Chem. C 112, 7731–7734 (2008)
15	KAc-Cu/SiO2	1.3	61	/	275	100.0	0.1	J. Catal. 241, 225–228 (2006)
16	LaMn0.5Cu0.5O3	0.02	74	0.239	150	30.0	0.1	Catal. Sci. Technol. 12, 2426–2437 (2022)
17	LaCo0.8Cu0.2O3−δ-NaCl	10	12	1.04	250	30.0	0.1	ACS Sustain. Chem. Eng. 9, 794–808 (2021)
18	Cu/TiO2	0.3	24	0.22	250	50.0	0.1	ACS Catal. 14, 10172–10180 (2024)
19	Ag8Cu1/Cs2O-α-Al2O3	5.5	48.5	0.272	160	20.0	0.1	Rare Met. 34, 477–490 (2015)
20	AgCuCl/BaCO3-N2	0.45	85.29	0.5	275	30.0	0.1	ACS Appl. Nano Mater. 6, 9687–9696 (2023)
21	K+-1% CuOx/SBA-15	2.1	21	2.1	350	60.0	0.1	J. Catal. 241, 225–228 (2006)
22	CuNP/C3N4–L	1.6	8	0.163	325	30.0	0.1	This work
23	Cu1–N4 SAC	5.5	44	3.023	300	30.0	0.1	
24	Cu1–N3 SAC	0.4	/	/	350	30.0	0.1	
25	Cu1–N2 SAC	5.7	78	5.866	275	30.0	0.1	

Table R3. Comparison between Cu1–N2 SAC, Cu1–N3 SAC, Cu1–N4 SAC, CuNP/C3N4–L and other reported IB group metal catalysts for propylene epoxidation H2 and O2.

No.	Catalyst	Conv. (%)	PO select. (%)	PO formation rate
(mmol gcat-1 h-1)	Temp. (oC)	Flow rate (mL min-1)	Pressure (MPa)	Ref.
1	Au/TiO2	0.6	96	0.54	50	33.3	0.1	J. Catal. 178, 566–575 (1998)
2	Au/TS-1(48)–Na1	7.4	85	2.05	200	20.0	0.1	Appl. Catal. B Environ. 95, 430–438 (2010)
3	Au(mIWI)/TS-1	2.8	91.5	1.61	200	35.0	0.1	ChemCatChem 12, 5993–5999 (2020)
4	Au/TS-1PT	1.3	89	0.39	200	66.7	0.1	Angew. Chem. Int. Ed. 60, 18185–18193 (2021)
5	Au20Pt1/TS-1-B	11	88	6.14	200	35.0	0.1	AIChE J. 68, e17416 (2022)
6	Au/HTS-1(NIMG)	6	91.2	2.59	200	35.0	0.1	ACS Sustain. Chem. Eng. 10, 9515–9524 (2022)
7	Au/TS-1-B	3.99	91.7	2.34	200	35.0	0.1	Appl. Catal. B Environ. 319, 121837 (2022)
8	Au/TS-1-CTES	13.3	92.7	2.02	180	20.0	0.1	ACS Sustain. Chem. Eng. 11, 7042–7052 (2023)
9	Au1&Aun	0.29	75.6	0.2	200	35.0	0.1	Nat. Commun. 15, 3249 (2024)
10	Cu1–N4 SAC	5.5	44	3.023	300	30.0	0.1	This work
11	Cu1–N3 SAC	0.4	/	/	350	30.0	0.1	
12	Cu1–N2 SAC	5.7	78	5.866	275	30.0	0.1	



Reply to Reviewer #4:


Fig. 1 | Schematic illustration and characterizations. (a) Scheme for the synthesis of Cu1–N2 SAC. TEM images of (b) Cu1–N2, (c) Cu1–N3 and (d) Cu1–N4 SACs. Aberration-corrected HAADF-STEM images of (e) Cu1–N2, (f) Cu1–N3 and (g) Cu1–N4 SACs. Element mapping images of (h) Cu1–N2, (i) Cu1–N3 and (j) Cu1–N4 SACs.


Fig. 2 | Structural characterizations. (a) Cu K-edge EXAFS of Cu foil, Cu2O, CuO, Cu1–N2, Cu1–N3 and Cu1–N4 SACs. (b) Wavelet transform of EXAFS of Cu foil, CuO, Cu1–N2, Cu1–N3 and Cu1–N4 SACs. (c) Cu K-edge XANES of Cu foil, Cu2O, CuO, Cu1–N2, Cu1–N3 and Cu1–N4 SACs. (d) Cu 2p XPS spectra of Cu1–N2, Cu1–N3 and Cu1–N4 SACs.


Fig. 4 | Dynamic changes of single-atom sites during propylene epoxidation with O2. (a) The first derivatives of XANES spectra of Cu foil, Cu2O, CuO and Cu1–N2 SAC after the successive feeding of O2 and C3H6. (b) The dynamic changes in mean chemical valence of Cu1δ+–N2 sites according to the edge position and the first derivatives of XANES spectra. (c) Fitting of k3-weighted EXAFS data of Cu1–N2 SAC after the successive feeding of O2 and C3H6. In situ DRIFTS spectra were recorded in the mixed flow of C3H6 and O2, after the successive feeding of C3H6 and O2 and after the successive feeding of O2 and C3H6 over (d) Cu1–N2, (e) Cu1–N3 and (f) Cu1–N4 SACs. (g) In situ DRIFT spectroscopy of Cu1–N2 SAC for different wavenumber intervals.

Supplementary Figure 10 | The corresponding fitting results and atomic configurations. The corresponding EXAFS fitting results of Cu1–N2, Cu1–N3 and Cu1–N4 SACs of (a) R-space and (b) k-space. Computational models for (c) Cu1–N2 (d) Cu1–N3 and (e) Cu1–N4 SACs. Orange: Cu atoms, blue: N atoms, grey: C atoms and white: H atoms.


Supplementary Figure 12 | Structural analysis of various catalysts. (a) The first derivatives of XANES spectra of Cu foil, Cu2O, CuPc, Cu1–N2, Cu1–N3 and Cu1–N4 SACs. (b) The dynamic changes in mean chemical valence of Cu1δ+ sites according to the edge position and the first derivatives of XANES spectra.


Supplementary Figure 13 | Schematic illustration of the plausible formation mechanism of low-coordinated Cu1–N2 centres.


Supplementary Figure 26 | Structural characterizations. Cu K-edge XANES of Cu foil, Cu2O, CuO, Cu1–N2 SAC and Cu1–N2 SAC-O2-C3H6. 


Supplementary Figure 28 | Structural characterizations. Cu K-edge EXAFS of Cu foil, Cu2O, CuO, Cu1–N2 SAC and Cu1–N2 SAC-O2-C3H6.

Supplementary Table 3 | EXAFS fitting data of various catalysts.
Sample	Shell	N	R (Å)	σ2 (10-3Å2)	E (eV)	R-factor
Cu foil	Cu–Cu	12.0	2.55	-	-	-
CuO	Cu–O	4.0	1.95	-	-	-
Cu2O	Cu–O	2.0	1.85	-	-	-
Cu1–N2 SAC	Cu–N	2.2	1.95	6.7	9.0	0.67%
	Cu–C	2.4	3.20	3.0	9.0	
Cu1–N2 SAC	Cu–N	2.2	1.91	5.9	2.7	2.97%
	Cu–Cu	0.2	2.49	3.0	2.7	
Cu1–N3 SAC	Cu–N	2.8	1.90	6.7	2.6	1.73%
	Cu–C	1.1	3.21	3.0	2.6	
Cu1–N3 SAC	Cu–N	2.8	1.90	7.5	1.7	3.81%
	Cu–Cu	0.1	2.40	3.0	1.7	
Cu1–N4 SAC	Cu–N	3.7	2.04	12.3	13.1	1.98%
	Cu–C	1.2	3.24	3.0	13.1	
Cu1–N4 SAC	Cu–N	3.7	2.12	28.8	9.6	2.92%
	Cu–Cu	0.8	2.40	3.0	9.6	
Cu1–N2 SAC-O2-C3H6	Cu–N	2.1	1.92	7.1	3.8	1.29%
	Cu–C	0.7	2.97	3.0	3.8	
Cu1–N2 SAC-O2-C3H6	Cu–N	2.1	2.04	3.0	-49.0	4.30%
	Cu–Cu	0.1	2.25	9.0	-49.0	
N, coordination number; R, distance between absorber and backscatter atoms; σ2, DebyeWaller factor to account for thermal and structural disorders; ∆E0, threshold Energy Correction; R factor indicates the goodness of the fit.


General Comment: This manuscript is another attempt to produce propylene oxide (PO) by the direct epoxidation of propylene with molecular oxygen (a green process), where at least the catalytic results (high selectivity and stable conversion) suggest success. This manuscript suggests that for isolated low coordinated Cu sites the formation of electrophilic oxygen species could be formed that direct the insertion of electrophilic oxygen species in the C=C bond of propylene. As no nucleophilic oxygen species are present that lead to unwanted CO2 and acrolein formation, high selectivities were obtained. No experimental evidence however is provided on the presence or absence of either oxygen species.
Response: We thank you for the positive comment of “the catalytic results (high selectivity and stable conversion) suggest success”. We are sorry for the confusion caused by the incomplete description of oxygen species in the manuscript, and we would be grateful if we could address this question in detail.
Based on the in situ DRIFTS results presented in our study, the vibration peaks of the functional groups of the epoxidation intermediates can only be observed under the condition that oxygen was fed prior to propylene. In contrast, when propylene was introduced first followed by oxygen, no IR peaks corresponding to the reaction intermediates were detected. Moreover, comprehensive characterizations including the X-ray absorption fine structure (XAFS), X-ray photoelectron spectroscopy (XPS), and X-ray diffraction (XRD) consistently demonstrated that the low-coordinate Cu1–N2 SAC possessed an oxygen-free structure and maintained its stability at a reaction temperature of 275 oC. Nucleophilic lattice oxygen species can effectively be eliminated by the absence of Cu–O coordination structure and Cu2O/CuO crystal phases. Collectively, these above results clearly show that the chemisorbed molecular oxygen on the catalyst surface can be activated into the intermediate oxygen species, which play a crucial role in promoting the DEP reaction. 
As we all know, the electrophilic and nucleophilic nature of oxygen species is indeed a key aspect of understanding the DEP reaction mechanism. A substantial amount of research on this issue in oxidation reactions supports the classification of oxygen species based on their electron density and reactivity. When oxygen species are adsorbed on metal sites, they usually go through a series of electron transfers from the metal and move towards the lattice oxygen state. Among these species, molecular and atomic iono-oxygen species (O2-, O-, O0) are widely recognized for their strongly electrophilic properties. In contrast, lattice oxide ions (O2-) are typically nucleophilic reagents with the strongest nucleophilicity. In addition, numerous studies have shown that electrophilic oxygen species play a role in olefin epoxidation reactions. (J. Am. Chem. Soc. 144, 7693–7708 (2022); ACS Catal. 9, 6262–6275 (2019)). In our previous work, we have similarly highlighted the critical role of electrophilic oxygen species in controlling the PO selectivity and catalytic performance of DEP reactions (J. Am. Chem. Soc. 142, 14134–14141 (2020)). 
Based on the above, in our study, we used density functional theory calculations to identify the electrophilic O2- species on Cu1–N2, Cu1–N3, and Cu1–N4 SACs. The Bader charge analysis revealed that these oxygen species carry a charge of -0.08 |e| to -0.33 |e|, which is significantly less negative compared to oxygen species dissociated on Cu (111) surfaces or lattice oxygen in Cu2O (110) surface. This electrophilic oxygen species is suitable for activating C=C double bonds in propylene.  
Your thoughtful comments have enabled us to provide a more detailed explanation and strengthen our conclusions, and we are grateful for them. Incorporating additional insights into the revised manuscript is something we would happily do if there are any additional points you would like us to address.

Major comment#1. What is the catalytic active site?
In your results section you state: progression of Cu into NP with a size of 15 nm: This is something completely different as isolated Cu atoms. Please discuss the implication of these findings. Also, the XRD results suggest the presence of metallic Cu NPs larger than 5 nm: what role do these NP play in the catalysis? Why are these NPs not visible in the TEM (although Figure 1b suggest you have any)? EXAFS fitting results are missing (also missing the data, i.e. the k-space spectra), but already a look at the FT EXAFS suggest that there could be Cu-Cu contributions present (Furthermore, XAFS by itself is not a proof of the presence of isolated Cu atoms: https://pubs.acs.org/doi/full/10.1021/acscatal.3c01116). This has large implication on how data are presented: for example, the interpretation of the Cu-N coordination number obtained from EXAFS is based on your assumption that all Cu atoms are isolated. This is clearly not the case and with that the assignment of Cu-N2, Cu-N3 and Cu-N4 incorrect and thus assumptions that went into the DFT calculations do not hold up.
Response: Your insightful comments are greatly appreciated. We sincerely apologize for the inappropriate of expression in our manuscript, which has led to your misunderstanding of our experimental results.
In this study, the Cu-based catalysts with ~15 nm nanoparticles (NPs) loading are the controlled catalysts (namely CuNP/C3N4–L) and present the 0.20 nm lattice fringes that correspond to the (111) plane of crystalline Cu. While the Cu1–N2, Cu1–N3 and Cu1–N4 SACs are atomically dispersed without any Cu NPs. Fig. 1b presents the TEM image of Cu1–N2 SAC. As can be seen in Fig. 1b, no nanoparticles were observed in this TEM image and the white regions represent the pores on the surface of the two-dimensional sheets of the catalysts because this image was taken under bright field imaging mode. To better understand, we have added the clear signs in the revised manuscript to mark the porous structure. Besides, we cannot agree the reviewer’s point that the XRD results indicate the presence of metallic Cu NPs larger than 5 nm. The XRD patterns showed only high-angle-shifted (002) peaks at 2θ of 27.5° belonging to pi-pi packing with the interlayer distance around 3.35 Å in all SACs. The controlled sample CuNP/C3N4–L is clearly different from Cu-based SACs because of the uncontrollable proximity of Cu active sites, showing the XRD diffraction peaks belong to Cu (JCPDS 04-0836)
To dispel the reviewer's doubts, we have re-tested the XAFS characterizations of all Cu-based SACs samples and re-fitted the data, including both R-space and k-space. The revised manuscript and supporting information have been updated with the relevant results. First, only a prominent peak at ~1.43 Å in EXAFS (Fig. 2a) was observed in these Cu-based SACs, indicative of the first coordination sphere of Cu–N interactions instead of Cu–O interactions. Focusing on the second-shell, the absence of Cu–Cu scatterings confirmed that the copper species was atomically dispersed. Furthermore, the FT-IR, XPS and other characterizations indicated that the formation of the Cu–N coordination structure in the Cu-based SACs was reliable. According to the CO DRIFTS test and HAADF-STEM images, the copper species in the Cu-based SACs were highly dispersed and did not aggregate. In order to prove our viewpoint more scientifically and effectively, the corresponding fitting of the EXAFS data in both R-space and k-space have been carried out, and the fitting results are presented in Supplementary Table 3. The quantitative fitting results elucidated that the average N coordination numbers of the central Cu atom were 2.2, 2.8 and 3.7 for Cu1–N2, Cu1–N3 and Cu1–N4 SACs, respectively. This confirmed that one Cu atom was coordinated with approximately two (Cu1–N2 sites), three (Cu1–N3 sites) and four (Cu1–N4 sites) nitrogen atoms of different Cu based SACs. Next, we also conducted a systematic fitting of the second-shell. By comparing the fitting results using Cu–Cu path or Cu–C path, it was observed that there are no Cu–Cu interactions exist in the Cu-based SACs, while the weak Cu–C interactions can be proved. According to the computational models optimized by DFT calculations, the Cu–C interactions arising from the interaction between the Cu–N centre and the C atoms in the carbon nitride structure.
Based on the above results, we are certain that the copper species in the as-prepared Cu-based SACs are atomically dispersed. Additionally, other experimental results and DFT calculations are both scientifically and reliably. We have updated the manuscript accordingly, thanking the reviewer for this enhancement.

Major comment#2. What is the role of the support?
One recent example of direct propylene epoxidation, where no catalyst was needed, is the that proposed by Trunschke et al (Nature Communications 13, 2022) over boron nitride and silicon dioxide. Please also consider this manuscript and the influence of the support, which is not zero. Raman (and maybe XPS) suggests you have undercoordinated N sites, what could be the potential role of the support alone?
Response: We sincerely appreciate the time and effort you have dedicated to reviewing our manuscript. They have helped us to better showcase the advantages of our study compared to the reported literature. According to the reference of Catal. Rev. 57, 306 (2015), a possible target for commercialization of PO production could be achieved with a PO selectivity of 70% at a propylene conversion of 10%. Furthermore, the PO selectivity is a vital industrial evaluation metric, that significantly influences the utilization of raw material and the cost of product separation. As far as another important industrial indicator of the reaction temperature is concerned, the PO production reaction is exothermic and the high reaction temperatures are unfavorable for the chemical equilibrium movement, leading to an inevitable conflict between the reaction kinetics and thermodynamics. The implementation of the lowest attainable reaction temperature serves as a critical measure to suppress thermal runaway phenomena. It is worth noting that, the as-prepared Cu1–N2 SAC has achieved the optimal performance among the reported DEP catalysts. And highly competitive catalytic activity at a relatively low reaction temperature was achieved when compared to other catalysts using other PO production processes. Our study demonstrated the potential of this catalyst system, which provides a valuable foundation for future improvements and optimizations. 
Although the Trunschke’s work (Nat. Commun. 13:7504, (2022)) has scientifically and objectively demonstrated a new route for the direct oxidation of propane to produce PO, the issues such as high reaction temperatures and low PO selectivity still remain. We believe that the Trunschke’s study is highly influential and provides us with meaningful insights. To better illustrate the differences between Trunschke’s work and our work, a comparison was made in the following Table R1.
Response: Table R1. The contrast between Trunschke’s work and our work.
	Trunschke’s work	This work
Goal	Examine the process of directly oxidizing propane to propylene and propylene oxide using boron nitride and silicon dioxide, which can reduce CO2 emissions and energy consumption associated with traditional industrial processes.	Propose an ACAS strategy to synthesize low coordinate Cu1–N2 SAC for propylene epoxidation using O2 alone, significantly boosting the catalytic performance of copper-based catalysts.
Catalyst structure	Boron nitride, silica, and silicon carbide	Cu1–Nx single-atom sites coordinated on C3N4 nanosheets (x=2, 3, or 4)
Research Strategy	By optimizing process conditions, it was possible to synthesize propylene oxide from propane. The comparison was made between the performance of various inert fillers in propane oxidation. 
The techno-economic viability of producing propylene oxide directly from propane was evaluated through kinetic experiments, microkinetic simulations, and economic efficiency assessments.	Considering the particularity of the DEP reaction, which involves the use of molecular oxygen as the sole oxidant, we have developed an ACAS strategy to facilitate the stabilization of low-coordinate Cu1–N2 species. Cu1–N2 SAC has an excellent PO formation rate with a high PO selectivity and a high propylene conversion. Unveiling the intricate structure-activity relationship of Cu–N coordination via in situ characterization and theoretical calculations.
Catalytic performance and reaction condition	Propene oxidation to PO: about 40.0 % propene conversion, 11%  PO selectivity.
Reaction condition: C3H8:O2:He= 6:3:11, T=490oC, flow rate=3.3-20 mL min-1, mass was 0.666 g for SiO2.	Propylene epoxidation to PO with only O2: 5.7 % propylene conversion, 78% PO selectivity and 5.87 mmolPO gcat-1 h-1 PO formation rate.
Reaction condition: GHSV=36000 h-1, C3H6:O2:N2=2:1:17, T=275 oC, mass was 0.050 g for Cu1–Nx SAC.
Relationship between catalytic performance and catalyst structure	The fast reaction at non-specific interfaces facilitate the synthesis of products prone to overoxidation via  gas-phase reaction.
The gas-phase oxygen itself was essential for the initiation and propagation of the radical chain reactions by generating radical precursors like •OH, affecting reaction activity.	The low-coordinate Cu1–N2, with the dynamic regulation of valence state of Cuδ+ (where 1<δ<2) center,  facilitated the formation of electrophilic oxygen species, which targeted the electronrich C=C bond in propylene for efficient PO production. In contrast, Cu1–N4, characterized by higher Cu–N coordination numbers and Cu oxidation valence, delivered moderate catalytic performance. Cu1–N3 sites, on the other hand, allowed the co-adsorption of C3H6 and O2, resulting in poor performance for DEP reaction.
Downsizing the active center to the single-atom scale and decreasing the metal-heteroatomic coordination number can promote the special DEP reaction mechanism, that is, adsorption and activation of oxygen before the bridged adsorption of propylene, as well as the enhanced charge transfer dynamics. This finally improves the electrophilicity of intermediate oxygen species, which selectively activates the C=C double bond of propylene for PO formation. However, as the particle size and metal-heteroatomic coordination number decrease, the inherently high surface energy of SACs would rapidly increase and result in cluster formation. The underlying support plays a decisive role in affecting the metal-support interaction and structure-activity relationship (Refer to Nat. Chem. 3(8), 634–641 (2011), and Coordin. Chem. Rev. 505:215693 (2024)). Therefore, to anchor isolated and low-coordinate Cu sites, functionalized C3N4 with plenty of undercoordinated N sites and a strong enough bond is a crucial and ideal support. The detailed process of preparing low-coordinate Cu1–N2 SAC based on our proposed ACAS strategy can be found in the revised manuscript. As we can see in Supplementary Figure 17, the metal-free carbon nitride (support alone) demonstrated the negligible capacity to activate propylene, and predominantly formed side products, such as acrolein and CO2. And across all the investigated temperatures, no PO selectivity towards the DEP reaction was observed. Once the undercoordinated N sites interacted with Cu atoms to form the low-coordinate Cu1–N2 single-atom sites, the Cu1–N2 SAC outperformed all other catalysts in the DEP reaction, especially in achieving a state-of-the-art PO formation rate at relatively mild temperatures. In conclusion, there is a large number of undercoordinated N atoms in functionalized C3N4 with inherent tunability that can trap low-coordinate and high content single-atom metal centers, making it a more favorable support for precision catalysis. The sp2 hybridization in the C3N4 structure also produce a highly conjugated electronic structure, which improves the catalytic stability of SACs.
We hope this clarification addresses the concerns raised.

Minor comments:
The term single-atom catalyst is not correct, you have many atoms... and apart from isolated Cu sites also larger copper nanoparticles present.
Response: We appreciate the reviewer’s observation. The concept of "single-atom catalysis" was first introduced in 2011 by Tao Zhang, Jun Li, and Jingyue Liu. They reported the Pt single atoms loading on FeOx and defined the single atom catalysts (SACs) as heterogeneous catalysts in which all metal species are atomically dispersed (Refer to Nat. Chem. 3(8), 634–641 (2011)). SACs have attracted considerable interest because they offer 100% metal atom utilization and exhibit excellent catalytic performance compared to traditionally supported nanoparticles.
In our study, we have developed an innovative asymmetric cavity-anchoring single-atom (ACAS) strategy to address this above challenge. This approach involved introducing amino acids with carbon skeletons to generate a multitude of asymmetric anchoring sites on functionalized C3N4, thus facilitating the stabilization of low-coordinate Cu1–N2 species. The XRD patterns showed only high-angle-shifted (002) peaks at 2θ of 27.5° belonging to pi-pi packing with an interlayer distance of around 3.35 Å in all SACs. The appearance of distinct XRD peaks that represent metallic copper can be only observed in CuNP/C3N4–L. The CO DRIFTS showed an IR peak at 2054 cm-1 (Cu1–N2 SAC), 2057 cm-1 (Cu1–N3 SAC) and 2060 cm-1 (Cu1–N4 SAC) representing the CO linear adsorption (COL) on Cu atoms. It is worth noting that the absence of a bridge bonded CO band in the range of 1700-1900 cm-1 rules out the presence of multi-atom Cu species (Refer to Angew. Chem. Int. Ed. 62 (2023), e202217220, Angew. Chem. Int. Ed. 126, 4956–4961 (2014), Nat. Commun. 12, 6022 (2021), J. Catal. 405, 333–345 (2022), Appl. Catal. A Gen. 505, 334–343 (2015), J. Phys. Chem. C 113, 10689–10695 (2009) and J. Catal. 268, 367–375 (2009)). According to the TEM and HAADF-STEM images, the copper species were atomically dispersed and did not aggregate in as-prepared SACs. 
Most importantly, we have re-tested the XAFS data and re-fitted the EXAFS data of all Cu-based SACs samples. The relevant results have been updated in the revised materials. First, only a prominent peak at ~1.43 Å in EXAFS (Fig. 2a) was observed in these Cu-based SACs, which indicates the first coordination sphere of Cu–N interactions. Focusing on the second-shell, the absence of Cu–Cu scatterings confirmed that the copper species was atomically dispersed. The corresponding fitting of the EXAFS data in both R-space and k-space have been carried out, and the fitting results are presented in Supplementary Table 3. The quantitative fitting results revealed that the average N coordination numbers of the central Cu atom were 2.2, 2.8 and 3.7 for Cu1–N2, Cu1–N3 and Cu1–N4 SACs, respectively. This confirmed that one Cu atom was coordinated with approximately two (Cu1–N2 sites), three (Cu1–N3 sites) and four (Cu1–N4 sites) nitrogen atoms of different Cu based SACs. Next, we also conducted a systematic fitting of the second-shell. By comparing the fitting results using Cu–Cu path or Cu–C path, it was observed that no Cu–Cu interactions exist in the Cu-based SACs, while the weak Cu–C interactions can be proved. According to the computational models optimized by DFT calculations, the Cu–C interactions arise from the interaction between the Cu–N centre and the C atoms in the carbon nitride structure. As we can understand from the reaction mechanism for DEP in in situ DRIFTS spectra, O2 can be preferentially adsorbed and activated on the low-coordinate Cu1–N2 site and the propensity for C3H6 bridged adsorption happen later, favoring the oxygen insertion to producing PO. The single-atom center is responsible for promoting this specific reaction pathway.
The atomic dispersion of copper species in the as-prepared Cu-based SACs is confirmed by the above results. We hope this clarification satisfactorily addresses the reviewer’s concerns and further reinforces the conclusions drawn in our manuscript.

Minor comments:
Abstract, last sentence: you do not optimize metal oxide catalysts, you optimize the coordination environment of isolated metal centers.
Response: Thank the reviewer for pointing this out. As known to us all, the traditional DEP catalysts are generally copper-based oxides, such as Cu2O nanocrystals, which have Cu–O active sites and are prone to causing the deep oxidation of propylene. In contrast, in this study, the Cu–N active sites were introduced and effectively prevented the side reactions. Furthermore, the precise manipulation of metal-nitrogen (M–N) coordination environments shed light on the coordination sensitivity of low-coordinate active-sites. 
Our apologies for the misunderstandings caused by our negligence and inappropriate language. This part has been revised to be more reasonable based on this comment. “These insights pave the way for a new avenue in the optimization of the fine structure of isolated active sites through the precise manipulation of metal-nitrogen (M–N) coordination environments.”

Minor comments:
Introduction: why is the interatomic distance between copper atoms important when you claim that isolated Cu atoms do the job?
Response: We thank you for this comment and admire the reviewer's rigorous academic attitude. As known, the formation of surface oxygen species with eletrophilicity is crucial for the DEP reaction. Taking traditional DEP catalysts (Cu nanoparticals and Cu2O nanocrystals with continuous Cu–Cu or Cu–O active sites) as an example, at the reaction temperature, C3H6 and O2 molecules would both tend to co-adsorb on their surface due to the unseparated Cu sites. This promoted the Mvk mechanism and induced the formation of nucleophilic oxygen species. (Refer to Angew. Chem. Int. Ed. 53, 4856–4861 (2014), Nat. Commun. 12, 5921 (2021), and J. Am. Chem. Soc. 142, 14134–14141 (2020)). Specifically, the Bader charge analysis revealed that molecular oxygen is easily dissociated on Cu (111) surfaces, and the resulting oxygen atoms have a significant negative charge of -0.90 |e|, which is typical of nucleophilic oxygen species. For lattice oxygen in Cu₂O at elevated temperatures, the Bader charge reaches -0.97 |e|, and these nucleophilic oxygen species were found to require surmounting a significant barrier (0.87 eV) for oxygen insertion reactions. These results highlight the distinct reactivity of nucleophilic oxygen species, which is not conducive to activation of the C=C double bond in propylene. 
In our study, the single-atom Cu1–Nx center has been designed, boosting the special DEP reaction mechanism which is different from the Mvk mechanism on Cu nanoparticals and Cu2O nanocrystals. The reaction began with the adsorption of molecular oxygen at the single-atom center to yield O2- species with a charge of -0.08 |e| to -0.33 |e|, which subsequently underwent an oxygen insertion into the C=C bond of propylene, leading to the formation of PO. This finding highlights the crucial role played by the atomic dispersion state of copper in controlling the reaction mechanism and product selectivity for the DEP process. 

Minor comments:
Pyrolysis is typically an uncontrolled synthesis method; how do you manage to maintain isolated Cu sites (XRD/TEM suggests you do not)? Can you show the repeatability of this synthesis method? Are you sure that all S is out of the final structure, or could this influence the support chemistry as well? 
Response: We appreciate your insightful comment. The structural characteristics of SACs determine that the interaction between the atomic dispersed metal and the coordination atoms (such as N and others) on the support is crucial for the stabilization of metal single atoms. Various strategies based on coordination geometry engineering, confinement effects or chemical bonding interactions can be used to reduce the surface free energy and kinetically inhibit metal-atom aggregation (Refer to Nat. Synth. 3, 1427–1438 (2024), Angew. Chem. Int. Ed. 60, 15248–15253 (2021), and Science 353, 150–154 (2016)). Among the above strategies, the metal-support interaction (MSI) is often considered to be the key to stabilizing metal single-atom sites in synthesis and catalysis of SACs. Thus, the key to regulating the MSI lies in the rational design of the C3N4 support. Because of its polymer characteristics, easy synthesis, abundant N content and rich chemical properties, the C3N4 support can be designed at the molecular level to anchor isolated metal atoms with different coordination environments.
 In this study, we have developed an innovative asymmetric cavity-anchoring single-atom (ACAS) strategy. This approach involved introducing amino acids with carbon skeletons to generate a multitude of asymmetric anchoring sites on functionalized C3N4, thus facilitating the stabilization of low-coordinate Cu1–N2 species. Supplementary Fig. 13 specifically illustrates the synthetic steps for Cu1–N2 SAC utilizing the ACAS strategy. First of all, the hydrolysis of CuCl2 and the modifier L-cysteine creates an acidic medium that makes it easier to hydrolyze melamine. The reaction between L-cysteine and melamine results in the formation of Compound I. The abundant terminal –NH2 of melamine would coordinate with free Cu2+ species to form Compound II, as evidenced by the XRD characterization (Supplementary Fig. 14a). Secondly, during the solvothermal reaction process, melamine undergoes further hydrolysis to generate cyanuric acid units. A supramolecular structure is formed by the spontaneous assembling of in-situ formed cyanuric acid and melamine through hydrogen bonding (Refer to Chem. Eng. J. 405, 126685 (2021); Nano Res. 11, 3462–3468 (2018)). Due to the in-situ modification of nitrogen-containing units, Cu2+ forms low-coordinate structures (Compound III) during the assembly process, which further stabilizes the anchoring of Cu–N bnn species. Finally, the cyanuric acid in Compound III undergoes pyrolysis, releasing ammonia gas and leaving melamine to self-condense into heptazine units (Compound IV) (Refer to ACS Appl. Mater. Interfaces 11, 10651–10662 (2019); Small 20, 2309032 (2024); Langmuir 30, 447–451 (2014)). Meanwhile, the modified moiety within Compound III undergoes pyrolysis and cyclization to form Compound V, which is then dehydrated to yield Compound VI. These steps are iteratively and periodically repeated, ultimately forming the low-coordinate Cu1–N2 SAC. Throughout these above processes, the assembly of the supramolecular structure and the formation of asymmetric melon configurations create favorable conditions for the formation of high-loading and low-coordinate copper single-atom sites.
As the reviewer mentioned, pyrolytic strategies commonly used for synthesizing SACs generally suffer from aggregation especially at high metal loadings. In the extensive preliminary experiments conducted in this study, we identified similar issues and solved them. An improper ratio of metal precursor, nitrogen-containing precursor and L-cysteine, along with factors such as the temperature, and reaction time of the solvothermal process, as well as the temperature, time, atmosphere, and heating rate of the pyrolysis process, could all lead to aggregation or collapse of the catalyst structure. The following table lists some of the influence of the synthesis parameters. 

	Parameters	Catalysts state	Cu species
CuCl2:melamine:L-cysteine
(mass ratio)	0.78:2.00:0.01	Nanosheet	Single atom
	0.60:2.00:0.015	Nanosheet	Single atom
	0.95:2.00:0.00	Nanosheet	Single atom
	1.00:2.00:0.00	Nanosheet	Nanopartical
	0.78:1.60:0.05	Nanotube	Nanopartical
	1.20:2.00:0.01	Collapse	-
	0.78:2.00:0.30	Nanotube	Nanopartical
	More……
Temperature, time (solvothermal process)	220 oC, 4 h	Nanosheet	Single atom
	220 oC, 6 h	Collapse	-
	220 oC, 3 h	Nanosheet	Single atom
	220 oC, 2 h	Nanosheet	Nanopartical
	220 oC, 1 h	Nanosheet	Nanopartical
	180 oC, 4 h	Nanosheet	Nanopartical
	160 oC, 4 h	Collapse	-
	More……
Temperature, time, atmosphere, and heating rate (pyrolysis process)	550 oC, 3 h, 5%H2/Ar, 8 oC/min	Nanosheet	Single atom
	550 oC, 3 h, Ar, 8 oC/min	Nanosheet	Single atom
	520 oC, 3 h, Ar, 5 oC/min	Nanosheet	Nanopartical
	550 oC, 3 h, Air, 8 oC/min	Nanosheet	Nanopartical
	550 oC, 3 h, 5%H2/Ar, 3 oC/min	Nanosheet	Nanopartical
	550 oC, 3 h, Air, 10 oC/min	Collapse	-
	More……
Although the EDS, XPS, and XAS results indicate that the low-coordinate Cu1–N2 SAC contains no S element, the presence of extremely trace amounts of residual S cannot be ruled out. In light of the detailed mechanistic insights into ACAS strategy presented above, its universality was further explored. As shown in Supplementary Fig. 14, in addition to L-cysteine, glycine and L-arginine, which do not contain the S element, can also be used for the functionalization of C3N4 support. Moreover, the structural features including the crystal structures, infrared signatures, and coordination environments of the prepared copper-based single-atom catalysts shows no significant difference compared to that of Cu1–N2 SAC, resulting in the similarly excellent DEP performance (Supplementary Fig. 14). This result effectively rules out the interference of the S element and highlights the significance of the ACAS strategy proposed in our study.
We hope this clarification addresses the concerns raised and provides a deeper understanding of ACAS strategy for SACs preparation.



Minor comments:
The synthesis approach taken here is inspired by that published by Yang et al. (reference 33) where similarly low coordinated Cu sites were held responsible for high ORR activity. To obtain further stability, amino acids were used that should lead to stronger anchoring sites. Related to this, the catalyst synthesis procedure proposed here will be difficult to scale up, please comment.
Response: We sincerely thank you for your thoughtful review of our work. It is important to note that the fresh Cu-N-C SAC prepared by Yang et al. (reference 33) using a two-step approach generated only Cu–N4 coordination structure, which was anchored on a N-doped carbon support without displaying the characteristics of Cu–N2 and Cu–N3. Only under ORR working conditions, the dynamic evolution of Cu–N4 to Cu–N3 and further to HO–Cu–N2 can be driven by the applied potential. Comprehensively considering the structural characteristics of Cu-N-C SAC, there are obvious differences between Cu-N-C SAC and ours. In our study, we developed an innovative asymmetric cavity-anchoring single-atom (ACAS) approach for fabricating the low-coordinate Cu-based SAC. This approach involved introducing amino acids with carbon skeletons to generate a multitude of asymmetric anchoring sites on functionalized C3N4. This facilitated the stabilization of high content of low-coordinate Cu1–N2 species without the need for additional reaction conditions. The fitting of the EXAFS data elucidated that the average N coordination numbers of central Cu atom were 2.2 and 2.1 for on fresh and spent Cu1–N2 SAC, respectively. This confirmed the strong bonding between one Cu atom and two N atoms, which facilitates that the Cu1–N2 sites can be activated under the oxygen adsorption, simultaneously promoting the dynamically stability of Cu1–N2 sites can during the DEP reaction.
Overall, the preparation method of Cu1–N2 SAC was not inspired by that published by Yang et al, but referred to their descriptions of the low-coordinate active centers. We are sincerely grateful for your valuable insights once again, which have significantly propelled our research forward. Benefiting from the development of the ACAS strategy, the fine coordination structure of low-coordinate Cu1–N2 single-atom centers can be ultimately stabilized on functional carrier (C3N4–L). 
Considering the easy availability and low cost of raw materials such as the amino acids, melamine and metal precursors, and the theoretical synthesis mechanism underlying the ACAS method has been elucidated in our work, the scaled up synthesis of Cu1–N2 SAC can be achievable. We expanded the synthesis scale of Cu1–N2 SAC by 2 fold (100 mL), 4 fold (200 mL) and even 20 fold (1000 mL), respectively. It was found that the fill rates of the reaction materials in the solvothermal reactor significantly affected the pressure of solvothermal process, which was the key factor influencing the synthesis of Cu1–N2 SAC. Based on the variable regulation and optimization of the synthesis experiments, the fill rate of 55% was determined as the most excellent parameter. Eventually, the stable synthesis of ten-gram-scale Cu1–N2 SAC has be achieved. As shown in Supplementary Fig. 21, there are no conspicuous differences in the structure and the DEP performance of the catalysts prepared on different scales, demonstrating the practicality of the ACAS strategy.
 We declared that the main purpose of our work was to assess the structure-performance relationship between fine coordination structure of Cu1–Nx SACs and the DEP reaction systematically. We overlooked whether the catalysts could be synthesized on a large scale or not and we apologize for this. We have updated the manuscript accordingly, thanking the reviewer for this enhancement.


Minor comments:
Results: Not sure how a significant number of micropores is beneficial to electron transfer between active sites and oxygen species. Please elaborate. Also not clear how the Cu1-N2 leads to more micropores compared to Cu1-N3, what is done differently in the synthesis procedure?
Response: Thank the reviewer for pointing this out. Microporous structures in catalysts which can be beneficial for electron transfer, are is traceable and widely proven. 
As can be seen in Cao’s study (Refer to Nat. Commun. 14:2494, (2023)), they discussed in detail the beneficial effects of the microporous structure on electron transfer in single-atom catalysis for enhanced alkene epoxidation. The main functions of the microporous structure are as follows. (1) Modifying electronic structure: charge redistribution regulated by the confinement effect of micropores can alter the electron cloud of SACs. This creates an optimal electronic structure, facilitating electron transfer efficiency between olefins, intermediates and active sites, and promoting alkene epoxidation as well as selectivity. (2) Boosting charge transfer: the microporous micro-environment can lower the reaction activation energy and stabilize the reaction intermediates by interacting with reactants and active sites. This is crucial for accelerating electron transfer, and improving alkene epoxidation reaction rate.
Meanwhile, Zhu’s work (Refer to Nat. Catal. 6, 574–584 (2023)) showed that micropores can shorten the transport path of photo- generated excitons, enabling efficient separation of electron- hole pairs, ensuring high-efficiency electron transfer. The active sites within the micropores can utilize the energy of excitons to drive the reaction. Micropores structure can restrict the diffusion and adsorption behavior of reactant and product molecules within them. Additionally, Li’s study (Refer to Angew. Chem. Int. Ed. 64, e202415691 (2025)) discussed that the interconnected meso- and micropores structure in porous Fe-N/C materials provides higher specific surface area and efficient pathways for electron transfer during the oxygen reduction reaction. In addition, Luo’s study (Refer to Nat. Commun. 16:800, (2025)) proved that a microporous-dominated structure with a large specific surface area and pore volume can be conducive to the exposure of active sites and the diffusion of reactants. Besides, the microporous structure can also benefit electron transfer, resulting in the rapid conversion of PMS. This leads to an improvement in the catalytic performance..
To sum up, the microporous structure facilitates electron transfer during the catalytic process through multiple mechanisms, such as exposure of more active sites, generation of a confinement effect, influence of the electronic structure, enhancement of the contact between reactant and active sites, and optimization of charge transfer. This, in turn, improves the catalytic performance.
The synthesis methods of Cu1–N2 and Cu1–N3 SACs can be learned in detail from our manuscript and supplementary information. Cu1–N2 SAC was synthesized according to our proposed ACAS strategy. This approach involved introducing amino acids with carbon skeletons to generate a multitude of asymmetric anchoring sites on functionalized C3N4, thus facilitating the stabilization of low-coordinate Cu1–N2 species. As can be seen in Supplementary Figure 13, during the solvothermal reaction process, melamine underwent hydrolysis to generate cyanuric acid unit. The in-situ formed cyanuric acid and melamine spontaneously assembled into a supramolecular structure through hydrogen bonding. Subsequently, the introduction of L-cysteine resulted in the shear stress breaking the stacking interactions between layers and cutting the stacked structure into ultrathin layers. The modified moiety within Compound III was cyclized to form Compound V by releasing NH3 and H2S molecules, followed by dehydration to generate Compound VI during pyrolysis. Porous and ultra-thin nanosheets were formed due to the escape of these gas molecules. By contrast, Cu1–N3 SAC was synthesized according to the scalable two-step annealing method referring to Lu’s work (Refer to Nat. Nanotechnol. 17, 174–181 (2022)). Initially, bulk carbon nitride was synthesized. Subsequently, single-atom copper sites were anchored via impregnation and calcination procedures. Consequently, Cu1–N3 SAC has a relatively thick nanosheet morphology and is devoid of a porous architecture, leading to a limited specific surface area. 
We are sincerely grateful for your valuable insights once again. They have significantly propelled our research forward, enabling us to gain deeper in this study.

Minor comments:
Results XPS: Please explain what you mean with: 'It was important to recognize that the prevalence of Cu-N coordination and the relative quantities of other nitrogen species displayed an inverse relationship, underscoring a more unsaturated coordinative environment for the Cu1-N2 SAC' This is absolutely no proof for undersaturated sites.
Response: Thank you very much for your valuable comment. We agree with the reviewer’s suggestion that the prevalence of Cu–N coordination is not the evidence for undersaturated sites. The Cu–N peaks in the N 1s XPS spectrum can only corroborate the interaction between Cu and N atoms. We hope this clarification satisfactorily addresses the reviewer’s concerns and further reinforces the conclusions drawn in our manuscript.

Minor comments:
EXAFS, please show the full EXAFS spectra somewhere. Where are the fitting results of the EXAFS spectra? Also, I could not find any information anywhere on the fitting approach taken. This is not good practice.
Response: We appreciate the reviewer’s critical observation. First of all, we would like to express our apologies for any possible oversights. Secondly, the EXAFS data and the fitting results of the EXAFS spectra in the R-space have been both presented in our manuscript and Supplementary information (We apologize again for the misunderstanding caused by the inappropriate presentation format). To clear up any uncertainties, we have retested the XAFS characterizations of all SACs and re-fitted the data, covering both R-space and k-space. The relevant EXAFS spectra and the fitting results have been updated in the revised manuscript (Fig. 2a) and Supporting information (Supplementary Figure 10 and Supplementary Table 3). Fig. 2a showed only a prominent peak at ~1.43 Å in EXAFS in these Cu-based SACs, which was indicative of the first coordination sphere of Cu–N interactions. The quantitative fitting results elucidated that the average N coordination numbers of the central Cu atom were 2.2, 2.8 and 3.7 for Cu1–N2, Cu1–N3 and Cu1–N4 SACs, respectively. This confirms that one Cu atom is coordinated with approximately two (Cu1–N2 sites), three (Cu1–N3 sites) and four (Cu1–N4 sites) nitrogen atoms of different Cu based SACs. The reliability of our study can be demonstrated by the consistency of the conclusions drawn from the updated results with the initial ones. In addition, the EXAFS fitting approach has been supplemented in the Supplementary information and highlighted in yellow. Thank you again for your thoughtful comments, and if there are any additional points you would like us to address, we would be glad to incorporate further insights into the revised manuscript.

Minor comments:
As a matter of fact, most information on the analysis methods applied is missing in both the manuscript and SI.
Response: Thank the reviewer for pointing this out. As required, we have meticulously supplemented the information on all analysis methods item by item in the revised Supplementary information from page 3 to page 9.

Minor comments:
EXAFS: Not sure how you would get bond angle information out of EXAFS fitting.
Response: We sincerely apologize for any misunderstanding that may have occurred due to the inappropriate expressions in our manuscript. In fact, the bond angle information of Cu1–N2, Cu1–N3 and Cu1–N4 SACs was obtained from the optimized configurations through computational studies. The EXAFS fitting merely proves that the experiment and the theoretical models are in agreement. The corrected manuscript has been amended accordingly. We appreciate your feedback once more.

Minor comments:
XANES analysis, your spectra do not have the distinct feature from Cu(1) and also not from Cu(0), probably as linear combination fitting will show (The CuO and Cu2O reference spectra are not good references for Cu surrounded by N). And thus, I doubt the assignment to be that of Cu(1) (what is the fitting error of linear combination fits? Please have a look at the literature (XANES spectra of Cu in SSZ-13 for NOx oxidation for example) which is the closest to single site Cu XANES in environments closest to what you have here. The assignment of Cu(1) species further does not corroborate with the XPS/AES results and the observation of metallic Cu peaks in the XRD. 
Response: Thank you for your thoughtful comment. Your comment is truly insightful, and we are grateful for the chance to clarify this point. We have read J. Phys. Chem. Lett. 9, 3035-3042 (2018) and other references carefully. Furthermore, we also re-tested the XAFS characterizations of all SACs in order to ensure the reliability and reproducibility of the data. The updated Cu K-edge XANES analysis shows that the as-prepared SACs exhibit distinct features from Cu+. The obvious shoulder feature appearing in all Cu-based SACs around 8983 eV can be assigned to the 1s → 4pxy transition, indicative of a Cu+ cation in a linear configuration. There is no visible Cu2+ feature at about 8986 eV. Interestingly, compared to Cu1–N3 and Cu1–N4 SACs, the intensity of the peak around 8983 eV in Cu1–N2 SAC decreases and the 1s → 4pz transition around 9004 eV shifts the edge position to a lower-energy position. These results show that the linear configuration in Cu1–N2 SAC is broken, which is in agreement with the optimized configurations (Refer to Nature 622, 754-760 (2023) and J. Phys. Chem. Lett. 9, 3035-3042 (2018)). 
In view of the doubt from reviewer that the CuO and Cu2O reference spectra are not good references for Cu surrounded by N, we added the CuPc standard sample which is characterized by Cu–N coordination structure. As shown in Supplementary Fig. 12, based on the XANES spectra of Cu foil and CuPc, the valences of Cuδ+ in Cu1–N2, Cu1–N3 and Cu1–N4 SACs were estimated to be ~0.70, 0.93, and 0.83 according to our linear fitting results (R=0.99). This is consistent with the Cu 2p XPS/AES analysis results. Among them, the valence state of Cu1–N3 SAC is different from that of Cu 2p XPS. The reason is that the surface carbon atomic content in the Cu1–N3 SAC was less than that observed in the other two catalysts, suggesting a replacement of carbon atoms by copper in the Cu1–N3 structure, as confirmed by the distinct reduce of C–N group related vibration bands intensity, which affects the electron density of the copper centers (Refer to Nat. Catal. 5, 818-831 (2022) and Nat. Commun. 12:6022, (2021)). As we can see in Supplementary Figure 7, the metallic Cu XRD peaks can only be observed in CuNP/C3N4–L. The XRD patterns showed only high-angle-shifted (002) peaks at 2θ of 27.5° belonging to pi-pi packing with an interlayer distance of around 3.35 Å in all SACs. These indicate that the Cu species in Cu1–N2, Cu1–N3 and Cu1–N4 SACs are atomically dispersed.
We want to express our sincere appreciation to you once more for your valuable comments. The revised manuscript and Supplementary information now have the updated relevant content.

Minor comments:
Not clear at all how Figure 4d is obtained, i.e. oxidation state changes during reaction, is the feed changed? These results are based on linear combination fitting of XAS spectra using references which do not represent the species at hand (and thus lead to very approximate fits with large fit errors). How this is then direct proof for the formation of electrophilic oxygen species is really not clear. Also, the suggestion that a reduction in the FT EXAFS peak intensity at a certain bond distance suggests a reduction in electron density at the isolated Cu sites is not correct. When O species are formed, this might lead to a contribution at a different distance potentially cancelling out the Cu-N contributions. But, as stated before, as XAFS is a bulk averaging method and you have metallic Cu nanoparticles and Cu-Nx species, there might be other processes leading to a reduction of the intensity in the FT EXAFS spectra. 
Response: Figure 4d was obtained by the in situ DRIFT spectroscopy of Cu1–N2 SAC recorded in the mixed flow of C3H6 and O2, after the prior feeding of C3H6 and then feeding O2 and after the prior feeding of O2 and then feeding C3H6. The DRIFTS analysis highlighted that the molecular oxygen tends to be preferentially activated on Cu1–N2 sites and then the propylene molecule can be activated via bridged adsorption, favoring the oxygen insertion to produce PO. We apologize for the potential confusion caused by our imprecise description and think that you mentioned Figure 4b instead of Figure 4d.
In view of the doubt from reviewer that the references using for the linear combination fitting of XAFS spectra may be unbefitting, we have retested the XAFS data of all SACs and the in situ XAFS of Cu1–N2 SAC (Cu1–N2 SAC-O2-C3H6). Besides, we added the CuPc reference which is characterized by Cu–N coordination structure. As shown in Fig. 4b, the valences of Cuδ+–N2 in fresh and in situ Cu1–N2 SACs were estimated to be nearly 0.70 and 0.90 according to our linear fitting results based on the first derivatives of XANES spectra. The dynamic nature of Cu1–N2 site migration was also corroborated by Bader charge analysis, with charge states oscillating between +0.73e and +0.87e (Supplementary Fig. 27) during the DEP reaction progress. We agree with the reviewer’s suggestion that this cannot be direct proof for the formation of electrophilic oxygen species. Based on the special reaction mechanism that occurs at the low-coordinate Cu1–N2 center, that is, adsorption and activation of oxygen before the bridged adsorption of propylene, as well as the enhanced charge transfer dynamics, we have identified the electrophilic O2- species on Cu1–N2 SAC through density functional theory calculations. The Bader charge analysis revealed that these oxygen species carry a charge of -0.33 |e|, which is significantly less negative compared to oxygen species dissociated on Cu (111) surfaces or lattice oxygen in Cu2O (110) surface. Specifically, on Cu (111) surfaces, molecular oxygen readily dissociates, and the resulting oxygen atoms carry a substantial negative charge of -0.90 |e|, characteristic of nucleophilic oxygen species. For lattice oxygen in Cu2O at elevated temperatures, the Bader charge reaches -0.97 |e|, and these nucleophilic oxygen species were found to require surmounting a significant barrier (0.87 eV) for oxygen insertion reactions. Thus, evidence for electrophilic oxygen species originates from DFT calculations, and the theoretical reaction pathway strictly follows the evidence obtained from in situ spectroscopic characterizations. The misunderstanding caused by the inappropriate expression has been corrected in the revised manuscript.
Indeed, in conjunction with the EXAFS fitting data including the in situ results (Fig. 4c and Supplementary Table 3), it can be concluded that Cu1–N2 SAC remained atomically dispersed with stable low-coordinate Cu–N coordination structure (such Cu–N coordination numbers and Cu–N bond distances) without Cu–O centers. We also conducted a systematic fitting of the second-shell. By comparing the fitting results using Cu–Cu path or Cu–C path, it was observed that there are Cu–C interactions instead of Cu–Cu interactions existing in Cu1–N2 SAC during the DEP reaction, indicating that the Cu–N interactions are stable and cannot be oxidized easily to form the Cu–O coordination structure. Compared with the XAFS data of the spent Cu1–N2 SAC in our original manuscript, the more rigorous added in situ XAFS experimental results indicate that there is no contribution at different distances that could potentially interfere with the Cu–N coordination structure during the DEP reaction. This more objectively and scientifically demonstrates that the electron density at the isolated Cu sites did not decrease significantly during the DEP reaction. The above results also indicate that the DEP reaction does not disrupt the Cu–N coordination environment of the single-atom sites, but merely affects the electronic state of the Cu species. 
Moreover, the XRD patterns (Supplementary Fig. 7 and Supplementary Fig. 23) showed only high-angle-shifted (002) peaks at 2θ of 27.5° belonging to pi-pi packing with an interlayer distance of around 3.35 Å in all SACs. The diffraction XRD peaks of metallic copper are only present in the control sample CuNP/C3N4–L. Next, as can be seen in Fig. 1b, no nanoparticles were observed and the white regions in this TEM image represent the pores on the surface of the ultra-thin nanosheets of Cu1–N2 SAC because this image was taken under bright field imaging mode. To better understand, we have added the clear signs in the revised manuscript. Therefore, we believe that the above explanations can be sufficient to confirm the absence of Cu nanoparticles or other metal-heteroatom coordination structures in SACs. 
We sincerely appreciate your thoughtful comments, as they have allowed us to provide a more detailed explanation and strengthen our conclusions. If there are additional points you would like us to address, we would be happy to incorporate further insights into the revised manuscript.

Comments:
This brings me than to the last point: if all of your EXAFS data interpretation is correct, i.e. you have only isolated and undercoordinated Cu atoms, than the formation of a Cu-O bond should be very visible in the FT EXAFS spectra at larger bond distance as Cu-N. Please consider this. 
Response: Thank you for the insightful comments from the reviewer. In order to eliminate your doubts and ensure the authenticity of our experimental data, we have retested the XAFS data of all SACs and analyzed them in more detail. 
The relevant results, including figures and tables, have been updated in the revised materials. First, as shown in Fig. 2a, only a prominent peak at ~1.43 Å in EXAFS was observed in these Cu-based SACs, which indicates the first coordination sphere of Cu–N interactions. The absence of Cu–O scatterings at ~1.50 Å excluded the possibility of Cu–O coordination structure formation. Besides, the corresponding fitting of the EXAFS data in both R-space and k-space have been carried out, and the fitting results (Supplementary Table 3) also elucidated that one Cu atom was coordinated with approximately two (Cu1–N2 sites), three (Cu1–N3 sites) and four (Cu1–N4 sites) nitrogen atoms of different Cu based SACs. Next, we also conducted a systematic fitting of the second-shell. By comparing the fitting results using Cu–Cu path or Cu–C path, it was observed that there are Cu–C interactions instead of Cu–Cu interactions existing in the Cu-based SACs, which arise from the interaction between the Cu–N centre and the C atoms in the heptazine units on carbon nitride structure. Furthermore, the in situ XAFS (Cu1–N2 SAC-O2-C3H6) results also determine that the Cu–N interactions are stable and cannot be oxidized easily to form the Cu–O coordination structure, which strongly confirms our conclusion. 
We are truly grateful for your meticulous and insightful comments. In case there are any other aspects you deem necessary for us to further consider, we are willing for the further revisions.

Here, all of changes were highlighted in yellow in revised files. We hope the revised manuscript could meet the publication standards of Nature Catalysis.

Sincerely yours,
Gang Fu, Qingbiao Li, and Jiale Huang

College of Chemistry and Chemical Engineering, 
State Key Laboratory of Physical Chemistry of Solid Surfaces,
Xiamen University, Xiamen, 361005, China

This is a typical template for responding to reviewers, assuming you are a native English speaker and Chemist, please assist me in responding to reviewers' comments.
 Later, I give you the relevant questions from the reviewers, and you respond to the questions I give, modeling on the corpus I provide you. For the reviewer comments I give, you can just give a direct response.
 I would prefer not to use a letter format, but rather a paragraphed description.
 The language should not be wordy and the brief should be dry and polite.I may give you more than one question at a time, so please answer each one in turn.The overall response idea is [thanking the reviewer for pointing out the problem] [apologizing for the problem] [a revision or explanation of this problem] [thanking the reviewer again].At the same time I would like you as an expert to give me the changes I need to make to the text in response to this comment.I'll give you ideas for your answer, and you'll expand on them with modifications that take into account your status as an expert.

The original version of my manuscript is as follows:"Circumventing Radical Generation on Fe-V Atomic Pair Catalyst for Robust Oxygen Reduction and Zinc-Air Batteries
Lan Ran†[a], Yichen Zhang†[b], Wenming Tong[c], Long Chen[a], Maoyu Wang[e,f], Hua Zhou[f], Pau Farràs*[c],Shanyong Chen*[a,d], Xiaoqing Qiu*[a]
[a]	L. Ran, L. Chen, Dr. S. Chen, Prof. X. Qiu
College of Chemistry and Chemical Engineering
Central South University
410083, Changsha, China
E-mail: shanyongchen@csu.edu.cn, xq-qiu@csu.edu.cn
[b]	Y. Zhang 
Institute of Applied Chemistry, Department of Chemical Engineering
Tsinghua University
100084, Beijing, China
[c]	Dr. W. Tong, Dr. P. Farràs
School of Biological and Chemical Sciences, Ryan Institute
University of Galway
H91 TK33, Galway, Ireland
[d]	Dr. S. Chen
Guangdong Key Laboratory of Environmental Pollution and Health, School of Environment
Jinan University
511443, Guangzhou, China
[e]	Dr. M. Wang
Shanghai Synchrotron Radiation Facility, Shanghai Advanced Research Institute
Chinese Academy of Sciences
201204, Shanghai, China.
[f]	Dr. M. Wang, Prof. H. Zhou
X-ray Science Division
Argonne National Laboratory
60439, Lemont, IL, United States
          
 
Abstract: Iron-nitrogen-carbon (Fe-N-C) catalysts are considered the most active platinum-free alternative for oxygen reduction reaction (ORR), yet the generated reactive oxygen species (ROS) from general mechanistic pathway rapidly impair the ORR activity and stability of Fe-N-C. Herein, we establish and report an ORR pathway-switching strategy to circumvent ROS generation and fundamentally improve the activity and stability of Fe-N-C via DFT guided catalyst design. The constructed Fe-V atomic pair catalyst (Fe1V1-NC) with N2Fe-N2-VN2 configuration enables side-on adsorption of O2 and subsequent direct-breaking of the O=O bond to form O*, thereby avoiding the formation of ROS radicals. Importantly, there is inter-site electron interactions between FeN4 and VN4, which further boost the ORR activity. Consequently, Fe1V1-NC exhibits outstanding ORR activity with onset and half-wave (E1/2) potentials at 1.02 and 0.89 V vs. RHE, respectively, in 0.1 M KOH. Record-high stability is achieved on Fe1V1-NC with a minimal decay in E1/2 by 16 mV over 50,000 cycles, surpassing Fe-N-C counterpart and most of the catalysts reported to date. The Fe1V1-NC-based zinc-air battery reported here demonstrates exceptional durability up to 400 h at 10 mA·cm-2. This work identifies the intrinsic correlation between ORR pathway, activity, and stability, advancing development of stable catalytic systems.
Introduction
Developing cost-effective oxygen reduction reaction (ORR) electrocatalysts is crucial for the next-generation energy conversion and storage devices, such as metal-air batteries and fuel cells, that are environmentally friendly[1-3]. Currently, platinum-group-metal-based (PGM) catalysts have been recognized as having optimal electrocatalytic activity for cathodic ORR reaction. However, their scarcity and economic viability seriously limit the large-scale applications [4-7]. Great efforts have been devoted to exploiting efficient PGM-free alternatives [8-10]. Wherein, transition metal-nitrogen-carbon (M-N-C) catalysts exhibit prominent initial ORR activity, among which Fe-N-C has been widely reported as one of the most promising alternatives to PGM catalysts[11-14]. However, Fe-N-C catalysts are widely proven to show unsatisfactory durability, typically experiencing 50-70% degradation over tens of hours during a long-term ORR operation[15-18]. Notably, the reactive oxygen species (ROS), including ·OH and ·OOH generated during an ORR process, are regarded as one of the most probable causes of degradation for Fe-N-C[19-21]. The ROS species could ferociously attack carbon carriers to promote demetallation or even direct destruction of the Fe-Nx active sites in Fe-N-C, leading to a sharp decline in its ORR activity and stability[22-24]. This has, therefore, become one of the great challenges in further developing practical applications for the utilization of Fe-N-C.
Previous efforts to improve the durability of Fe-N-C have mainly focused on introducing scavengers[25-26], which could trap the generated ROS radicals and, thus, mitigate degradations in the ORR activity[27-28]. PtFe alloys, when implanted into the Fe-N-C matrix, have been shown to effectively lower H2O2 production and harmful oxygen-containing radicals[29]. Ta-TiOx nanoparticle as additives to Fe-N-C has led to the suppression of H2O2 yield by 51% at 0.7 V and an enhancement in the stability of the corresponding fuel cell with a current density decay of only 3% at 0.9 ViR-free, demonstrating an excellent ROS scavenging ability.[30] Our recent study[31] also reported Mn single atom tuning Fe-N-C catalysts, where the generated ROS can be scavenged on Mn sites. Clearly, the introduction of radical scavenging sites appears to improve ORR stability of Fe-N-C to a certain degree. Nevertheless, the ROS radicals, once generated, will inevitably poison neighboring Fe active sites as illustrated in Figure 1a[32-34]. That is to say, radical scavengers play a limited role in improving the intrinsic ORR stability. Moreover, some of the inactive oxides employed as radical scavengers may even reduce the mass-specific activity of Fe-N-C[29, 35]. Undoubtedly, more fundamental and effective strategies are needed to circumvent ROS generation during ORR to improve the intrinsic activity and stability. Mechanistically, an ORR process typically follows an associative pathway that includes formation of OOH* intermediates[36-37] with generations of many ROS radicals. In contrast, direct cleavage of O=O bond via the dissociative pathway could potentially circumvent the step where the formation of ROS occurs [12, 38]. Such direct engineering of ORR mechanistic pathway for Fe-N-C would lead to an avoidance of ROS generation at the source level and, thereby, fundamentally improving ORR activity and stability. To the best of our knowledge, few studies have noticed the effect of ORR pathway on operational stability, which is theoretically valid but remains unexplored.
In this work, we propose an ORR pathway-switching strategy via constructing Fe-V atomic pairs in Fe-N-C matrix to enhance its intrinsic stability. As depicted in Figure 1b, the adsorption of O2 on the Fe-V atomic pair with a side-on orientation would result in a direct cleavage of O=O bonding to form O* and, hereby, preventing the formation of H2O2 and OOH* intermediates. Unlike the general strategy that relies on external scavengers to trap ROS radicals generated from H2O2 and OOH* intermediate, the proposed unique ORR pathway on Fe-V atomic pairs enables avoiding the step that proceed to generate ROS radicals. We first construct three theoretical models, which are then evaluated by DFT calculations under various scenarios according to the respective mechanistic ORR pathways. We then experimentally validate the proposed ORR pathway-switching strategy. The synthesized catalysts, such as Fe1V1-NC and Fe1-NC, are investigated by a series of advanced characterization techniques and thoroughly assessed electrochemically for catalytic properties toward ORR. We employ a range of in situ and ex situ techniques to confirm that the improved ORR activity and stability are due to the successful suppression of ROS radical generation. We ultimately show the ORR activity and stability of Fe1V1-NC at the device level by applying it to a zinc-air battery as the cathode catalyst.
 
 
  
Figure 1. Illustration of the strategies to improve the ORR activity and stability of Fe-N-C in (a) previous reports and (b) this study.
 
Results and Discussion
Theoretical analysis on ORR pathway effect
 
Density functional theory (DFT) calculations were initially employed to evaluate the impact of 4e- ORR pathways (associative and dissociative) on catalytic behavior. As shown in Figures S1-3, these theoretical models were constructed with consideration of axial OH* adsorption, which is attributed to the spontaneous adsorption behavior as reported in previous studies [39]. Conventional Fe1-NC catalyst (Figure S3) follows the associative ORR pathway, where O2 is adsorbed in an end-on configuration at the Fe sites, undergoes protonation prior to O=O bond cleavage, and proceeds to subsequent conversion. The specific reaction pathway is: O2→O2*→OOH*→O*→OH*→H2O (Figure 2a), during which the generated OOH* intermediates tend to desorb to become ·OOH radicals, which can further transform into ·OH radicals. These ·OOH and ·OH radicals, as oxidative species, readily cause severe corrosion to Fe sites and carbon substrate, leading to active site damage and reduced stability of Fe1-NC. In contrast, the Fe1V1-NC dual-metal sites shown in Figure 2b follow the dissociative ORR pathway, where O2 is adsorbed via a side-on bridging configuration between Fe and V sites, and the O=O bond cleaves directly before the subsequent protonation. The specific reaction pathway is: O2→O2*→O*-O*→OH*-O*→OH*-OH*→OH*→H2O. This process circumvents the formation of OOH* intermediates, thereby preventing the generation of ·OOH radicals and further transformation. The different reaction pathways in ORR may lead to distinct reaction kinetics and activity.
We further investigated the detailed processes of ORR pathway on different models, with the corresponding free energy diagrams shown in Figures 2c-2d. The V sites show thermodynamically favorable strong adsorption of O2 (ΔGO2* = -2.31 eV) but high barrier for the desorption of OH* (ΔGOH* = 1.19 eV), resulting in the formation of axial OH* adsorption configuration on V sites. Interestingly, the axial OH* adsorption configuration on V sites (*OH-V-N4) induces a greatly reduced ΔG for the potential-determining step (PDS) compared to pristine V-N4 sites, while the axial OH* adsorption configuration on Fe sites generates finite effects. The PDS on Fe1-NC is the transformation of OH* into H2O, with a ΔG value of 0.804 eV. The PDS for Fe1V1-NC is the step of OH*-O* to OH*-OH*, which exhibits a lower ΔG value of 0.493 eV. The reduced ΔG for Fe1V1-NC indicates an enhanced ORR activity. Additionally, we calculated the projected density of states (PDOS) for the d-orbitals of Fe1-NC, V1-NC, and Fe1V1-NC respectively (Figure 2e). Fe1V1-NC exhibits a broader PDOS near Fermi level than that of Fe1-NC and V1-NC, suggesting enhanced charge transfer capability [40]. Moreover, the d-band centers of Fe (-0.754 eV) and V (0.240 eV) in Fe1V1-NC are significantly lower than those in Fe1-NC (-0.428 eV) and V1-NC (1.425 eV). The electron-rich property of Fe1V1-NC facilitates O2 activation and more efficient charge transfer, promoting ORR process. The adsorption energies of OH* radical on the metal sites of Fe1-NC, V1-NC, and Fe1V1-NC were further investigated to evaluate their ·OH formation process. Figure 2f shows the adsorption energies of OH* on the Fe and V sites in Fe1V1-NC are stronger than those on Fe1-NC and V1-NC. This indicates that OH* intermediates are more stably adsorbed on Fe1V1-NC, which facilitates the subsequent ORR process rather than the intermediates desorbing to form ·OH as occurs with Fe1-NC. In summary, the catalyst with Fe1-V1 atomic pair presents unique ORR features. Fe1V1-NC catalyzes the dissociative ORR pathway to enhance intrinsic ORR stability by avoiding the formation of ROS radicals, and generates a site-electron interaction that boosts ORR activity.
 
 
Figure 2. Schematic diagram of (a) associative pathway for Fe1-NC and (b) dissociative pathway for Fe1V1-NC. Free energy diagram of ORRs on (c) Fe or V single site and (d) Fe-V dual sites. (e) The PDOS of central metal atoms in Fe1-NC, V1-NC and Fe1V1-NC. (f) Adsorption energy of OH* on the metal sites of Fe1-NC, V1-NC and Fe1V1-NC. 
Synthesis and Characterization
The Fe-V atomic pair in Fe1V1-NC catalyst was synthesized via the approach shown in Figure 3a. Fe and V were anchored onto zeolitic imidazole framework (ZIF-8), and the resulting Fe,V-ZIF-8 were subsequently pyrolyzed under N2 atmosphere at 910 oC. During the pyrolysis, Fe and V atoms are coordinated with N atoms to form atomically dispersed active sites in the carbon skeleton while Zn is evaporated substantially owing to its relatively lower boiling point[41]. The Fe1-NC and V1-NC were synthesized by using the same method without respective V and Fe salts in the precursors. Note that Fe1V1-NC-1:3 and Fe1V1-NC-3:1 indicate the corresponding molar ratio of Fe to V salts is 1:3 or 3:1 during the preparation process respectively.
Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) are employed to analyze the morphology of the catalysts. Fe1-NC, V1-NC, and Fe1V1-NC consist of uniform polyhedral structure with carbon nanotubes anchored to its surface, which is inherited from ZIF-8. However, Fe1V1-NC-1:3 and Fe1V1-NC-3:1 with unmatched metal ratios demonstrate nanotubes as the major morphology with much fewer polyhedral structures (Figure 3b-c and S4). The TEM and corresponding energy dispersive spectroscopy (EDS) element mapping indicate even dispersions of metal elements in the carbon skeleton without obvious clusters or metal particles (Figure 3f and Figures S5-6). The aberration-corrected high-angle annular dark-field scanning transmission electron microscopic (AC-HAADF-STEM) measurements reveal that metal elements are dispersed at atomic levels as featured by the highlighted bright spots appearing in pairs for Fe1V1-NC in Figure 3d. Additionally, atomically dispersed Fe atoms are observed on Fe1-NC, with isolated bright spots (Figure 3e). An inductively coupled plasma optical emission spectrometry (ICP-OES) analysis determines the content of Fe and V in Fe1V1-NC to be 0.84 wt.% and 0.73 wt.% respectively, which is a near the 1 : 1 atomic ratio (1.04 : 1) for Fe : V as summarized in Table S1.
 
 
Figure 3. (a) Schematic illustration of the synthesis of Fe1V1-NC. (b) TEM image and (c) HR-TEM image of Fe1V1-NC. AC-HAADF-STEM images of (d) Fe1V1-NC (The red squares represent Fe-V diatomic pairs) and (e) Fe1-NC (The red circles represent Fe single atoms.). (f) Elemental mapping for Fe1V1-NC. 
Powder X-ray diffraction (XRD) patterns of Fe1V1-NC, Fe1-NC, V1-NC, Fe1V1-NC-1:3, and Fe1V1-NC-3:1 are shown in Figure S7. All the catalysts show diffraction peaks at 26.2° and 43.2° belonging to the (002) and (101) crystal planes of graphitic carbon, respectively[42]. No diffraction peaks associated with metallic Fe or V species were detected in Fe1-NC, V1-NC and Fe1V1-NC, corroborating the highly dispersed nature of Fe and V species in the carbon substrate. However, additional diffraction peaks of metal and/or their oxide nanoparticles were apparent for Fe1V1-NC-1:3 and Fe1V1-NC-3:1, indicating that precise control of the metal content is crucial for the synthesis of atomically dispersed catalysts. Furthermore, the defects and graphitization of the catalysts were analyzed by Raman spectroscopy (Figure S8). Two prominent peaks are observed at 1333 and 1569 cm-1, corresponding to D and G bands, respectively. The peak intensity ratios of G and D bands (IG/ID) for Fe1-NC, V1-NC, and Fe1V1-NC are calculated to be 1.32, 1.53, and 1.42, respectively, implying that the introduction of V species improves the degree of graphitization. Such prominence of G band relative to D band in all the catalysts due to the high graphitization degree is advantageous for electrochemical conductivity. The specific surface area and structural porosity were characterized by Brunauer-Emmett-Teller (BET) N2 adsorption-desorption isotherms. As shown in Figure S9, the three catalysts demonstrate type-IV isotherm with abundant micropores and mesopores. Fe1V1-NC features a large surface area of 648.0 m2·g-1 compared to Fe1-NC (365.5 m2·g-1) and V1-NC (438.8 m2·g-1), suggesting exposure of more active sites with Fe1V1-NC. The total pore volume of Fe1V1-NC is measured to be 0.351 cm3·g-1 with an average 2.169 nm pore size (Table S2). The abundant micropores and mesopores are beneficial to increasing accessibility of active sites and enhancing mass transfer kinetics.[43] These characteristics highlight a promising potential of Fe1V1-NC for catalyzing ORR process. O2 temperature-programmed desorption (O2-TPD) measurements indicate that the V1-NC exhibits a stronger adsorption capability for O2 and O2-containing intermediates compared to Fe1-NC as evidenced by the higher O2 desorption temperature in Figure S10, demonstrating the importance of V sites in Fe1V1-NC for ORR. [44] 
The chemical composition and electron interaction of the catalysts were investigated by X-ray photoelectron spectroscopy (XPS). The XPS survey spectrum confirms the presence of C, N, O, Fe, and V in Fe1V1-NC (Figure S11). The atomic contents of Fe and V in Fe1V1-NC obtained from XPS analysis are 1.72% and 1.64% (close to 1:1), respectively, as summarized in Table S3, and consistent with the result obtained by ICP-OES in Table S1. The high-resolution C 1s spectrum (Figure S12) was deconvoluted into three peaks, which are assigned to C-C (284.8 eV), C-N (285.4 eV), and C=O (286.9 eV). The C-N bond indicates the successful incorporation of N into carbon matrix, which provides anchoring sites for metal atoms[45]. The high-resolution N 1s spectra of Fe1-NC, V1-NC, and Fe1V1-NC are deconvoluted into four peaks respectively, including pyridinic N (398.4 eV), metal-N (398.9 eV), pyrrolic N (399.7 eV), and graphitic N (401.2 eV) (Figure S13a). The distinct metal-N peak implies the formation of coordination between N and Fe/V. Importantly, the higher contents of metal-N and graphitic N are induced by incorporating V element for Fe1V1-NC with respect to Fe1-NC and V1-NC (Figure S13b and Table S4), in good agreement with Raman results. The rich graphitic N is beneficial for improving limiting diffusion current density and corrosion resistance of the carbon substrate. The high-resolution Fe 2p spectrum of Fe1V1-NC in Figure 4a can be fitted to the sum of five peaks. The peaks at 709.5 (Fe 2p3/2) and 723.2 eV (Fe 2p1/2) correspond to Fe2+ species, while the peaks located at 712.5 (Fe 2p3/2) and 726.2 (Fe 2p1/2), belong to Fe3+ species. The peak at 717.0 eV is the satellite peak. Such an analysis indicates that Fe mainly coordinates with N to form Fe2+/3+-Nx moieties. Notably, compared with Fe1-NC, the Fe 2p peaks of Fe1V1-NC exhibit a slight shift by 0.5 eV toward the lower binding energy, suggesting an electron enrichment effect at Fe sites. The high-resolution XPS spectra of V 2p (Figure S14) are all fitted to individual peaks at ~532.5 eV (V 2p1/2). Meanwhile, the binding energy of V 2p for Fe1V1-NC exhibits a significant positive shift by 0.6 eV compared to that of V1-NC. These results suggest the existence of a strong inter-site electron interaction between Fe and V sites, which can modulate the reaction electron state for excellent ORR activity.[46]
X-ray absorption spectroscopy (XAS) was used to further elucidate the coordination environment and electronic structure of Fe and V sites in the catalysts. As shown in X-ray absorption near-edge spectroscopy (XANES) (Figure 4b), the Fe K-edge absorption edges of both Fe1V1-NC and Fe1-NC are located at higher energy position relative to Fe foil, proving an oxidized state of the metal centers. Importantly, the Fe K-edge of Fe1V1-NC exhibits a slightly lower adsorption edge than that of Fe1-NC, implying a comparatively electron-enriched state of Fe site for Fe1V1-NC. Similarly, the V K-edge XANES spectrum of Fe1V1-NC exhibits a positive energy shift relative to V foil (Figure S15). The XAS analysis agrees well with XPS results, jointly confirming the site-to-site electron interaction for Fe1V1-NC. In addition, the Fourier transforms (FT) of the K-edge extended X-ray absorption fine structures (EXAFS) for Fe1V1-NC and Fe1-NC show a dominant peak at ~1.4 Å, corresponding to Fe-N scattering path (Figure 4c). Fe-Fe coordination is absent in both Fe1V1-NC and Fe1-NC, confirming the atomically dispersed nature of Fe. Similarly, the dominant peak at ~1.3 Å in the V-edge EXAFS spectrum is assigned to V-N scattering path with the absence of V-V bond in Fe1V1-NC. Furthermore, the quantitative least-squares fitting of EXAFS was performed (Figure 4d), with the corresponding fitting results summarized in Table S5. The coordination numbers (CN) of Fe-N and V-N in Fe1V1-NC are determined to be 3.90 and 3.83 respectively. This finding confirms that Fe and V are coordinated with four N atoms, forming a N2Fe-N2-VN2 configuration in Fe1V1-NC, where Fe and V are bridged by N atoms with no direct bonding between Fe and V. This is consistent with AC-HADDF-STEM analysis that Fe single atom is adjacent to V atom site within the Fe-V atomic pair. Additionally, Fe1-NC fits well with typical Fe-N4 model, in which the Fe is coordinated with four pyridinic N atoms (Figure S16). Both Fe and V foils fit well with obvious Fe-Fe and V-V configuration (Figure S17). The wavelet transform (WT) for EXAFS oscillations of Fe1V1-NC and Fe1-NC is performed to provide high-resolution visualization of the metal atomic sites with both radial distance and k-space (Figure 4e). Notably, the WT signals of Fe-Fe (~9.0 Å-1) and V-V (~9.1 Å-1) for Fe foil and V foil, respectively, are clearly observed. The WT contour plots of Fe1-NC and Fe1V1-NC exhibit maximum intensities at ~5.2 Å-1, corresponding to the Fe-N scattering path. Meanwhile, the maximum intensity of WT contour appears at ~2.4 Å-1 for Fe1V1-NC, which is attributed to the V-N scattering path. Compared to Fe foil and V foil, the Fe-Fe or V-V signals are absent in Fe1-NC, V1-NC, and Fe1V1-NC, confirming the absence of metal nanoparticles. These findings demonstrate that Fe and V atoms exist in atomically dispersed form in Fe1V1-NC and feature N2Fe-N2-VN2 configurations.
 
 
Figure 4. (a) Fe 2p XPS spectra for Fe1-NC and Fe1V1-NC. (b) Fe K-edge XANES spectra (the inset shows a partial enlargement of the area near the onset of the near edge). (c) Fourier transform of k3-weighted EXAFS spectra of Fe1V1-NC, Fe1-NC, and reference Fe and V foils. (d) FT-EXAFS fitting curves of Fe1V1-NC in R space (the inset is optimized configuration of Fe1V1-NC). (e) Wavelet transforms of k3-weighted EXAFS spectra of Fe1-NC, Fe1V1-NC (Fe), Fe foil, Fe1V1-NC (V), and V foil.
 
Electrocatalytic ORR performance
The ORR performances of the catalysts were evaluated via cyclic voltametric (CV) and linear sweep voltametric (LSV) measurements in 0.1 M KOH electrolyte using a three-electrode system equipped with a rotating ring disk electrode (RRDE). The CV curves of all the catalysts (Fe1V1-NC, Fe1-NC, V1-NC, Fe1V1-NC-1:3, and Fe1V1-NC-3:1) and commercial Pt/C exhibit rather significant reduction peaks (prominent ORR responses) in an O2 saturated electrolyte than an Ar-saturated environment (Figure 5a and S18), with Fe1V1-NC exhibiting the most positive reduction peak position and, therefore, the best ORR activity. Figure 5b and Figure S19 present LSV curves of all the catalysts and commercial Pt/C with a scan rate at 10 mV·s-1 with a rotation speed at 1600 rpm. As expected, Fe1V1-NC demonstrates superior ORR activity in terms of both onset potential (Eonset) and half-wave potential (E1/2), reaching an ultrahigh Eonset at 1.02 V and E1/2 at 0.89 V. In comparison, the values for other catalysts are as following: Fe1-NC (Eonset = 0.98 V, E1/2 = 0.85 V), V1-NC (Eonset = 0.94 V, E1/2 = 0.75 V), and commercial Pt/C (Eonset = 0.99 V, E1/2 = 0.85 V), respectively. The ORR kinetics is verified by the Tafel plot in Figure 5c and S20. The lower Tafel slope of Fe1V1-NC (103.8 mV·dec-1) indicates its enhanced reaction kinetics than that of commercial Pt/C (111.7 mV·dec-1). Moreover, the kinetic current density (Jk) at 0.85 V for Fe1V1-NC (50.1 mA·cm-2) is also 2.5, 5.8 and 1.8 times higher than that of Fe1-NC (20.2 mA·cm-2), V1-NC (8.6 mA·cm-2), and commercial Pt/C (28.6 mA·cm-2) (Figure 5d), suggesting a fast ORR reaction. The electrochemical double-layer capacitance from CV curves is calculated to evaluate the electrochemically active surface area (ECSA). Shown in Figure S21, the ECSA of Fe1V1-NC is comparable to that of Fe1-NC and V1-NC, which excludes the interference of site density when comparing ORR activities. The electron transfer number (n) is determined based on rotating disk electrode (RDE) measurements at different rotating speeds varying from 400 to 2500 rpm. The corresponding Koutecky-Levich (K-L) curves (Figures S22-26) show that the calculated n values for Fe1V1-NC in the potentials ranging from 0.2 to 0.6 V are between 3.93 to 3.97, which are very close to the standard 4e- transfer process. The electron transfer number and H2O2 yield are further investigated using a RRDE. Figure S27 confirms once again the high-efficiency 4e- ORR process catalyzed by Fe1V1-NC with an approximate electron transfer number of 3.98. Importantly, the H2O2 yield by Fe1V1-NC is maintained below 4.7% throughout the potentials range of 0.1-0.8 V vs. RHE, which is much more consistent and lower than that of Fe1-NC (9.8%) and V1-NC (14.9%) (Figure 5e). This reveals that the dissociation path followed by Fe1V1-NC efficiently inhibits H2O2 generation and promotes 4e- ORR, as well as mitigates the attack by the corrosive peroxide and hydroxyl radials (•OOH/•OH) all the while maintaining the superior ORR activity. The electrochemical impedance spectra (EIS) in Figure S28 demonstrate the similar shape of semicircles for all the catalysts. The EIS for Fe1V1-NC exhibits the smallest diameter, which unveils the lowest charge-transfer resistance and the enhanced ORR reaction kinetics. The anti-methanol ability was evaluated by measuring chronoamperometric responses (i-t) of ORRs by Fe1V1-NC and Pt/C to an injection of 3 mL methanol into the electrolyte at 500 s as shown in Figure S29. Fe1V1-NC exhibits a stable relative current with little to no disturbance upon the injection, whereas the commercial Pt/C demonstrates a sharp decline in the relative current upon the injection with a gradual and partial current recovery afterward under the same condition.
With the excellent ORR performance achieved above, we now evaluate the electrocatalytic ORR stability. Firstly, the long-term stabilities of Fe1V1-NC, Fe1-NC, and commercial Pt/C were evaluated by measuring chronoamperometry (I-t) at 0.5 V vs. RHE (Figure 5f). Fe1V1-NC demonstrates superior stability with the relative current being maintained over 95% throughout the continuous ORR process for 24 h. In sharp contrast, Fe1-NC displays a rapid attenuation of the current density to ~60%. The commercial Pt/C shows a better stability than Fe1-NC but far inferior than Fe1V1-NC. In addition, we performed accelerated durability tests (ADT) to further probe the stability of the catalysts (Figure 5g-h). Fe1V1-NC observes a loss in the E1/2 by only 16 mV after 50,000 CVs, showcasing its remarkable stability. For Fe1-NC, a test of 15,000 CVs already leads to a loss in E1/2 by 31 mV. The stability of Fe1V1-NC in this work is also at the record-high level when compared to the catalysts reported previously (Figure 5i and Table S6), with the loss in E1/2 as low as 3.2×10-4 mV per CV cycle for 50,000 cycles.
 
 
Figure 5. (a) CV curves in O2-saturated (solid line) and Ar-saturated (dashed line) electrolyte solutions respectively. The scan rate is 50 mV·s-1.  (b) LSV polarization curves in O2-saturated 0.1 M KOH solution. (c) Transferred Tafel slope from the LSV curves. (d) Comparison of the Jk at 0.85 V and E1/2. (e) H2O2 yield and (f) I-t curves of Fe1-NC, Fe1V1-NC and Pt/C. LSV curves of (g) Fe1V1-NC and (h) Fe1-NC before and after ADT tests. (i) E1/2 decays compared with those reported in the literature. These data are also summarized in Table S6 with references.  
Investigation of ROS radical generation
As reported previously[47], the generation of corrosive ROS during ORR causes issues such as carbon oxidation and demetallation, which is the major reason for the deactivation of Fe-N-C catalysts. To investigate the mechanism of stability enhancement, we focus on the ORR pathway and its effect on ROS generation process for Fe1-NC and Fe1V1-NC. The key reaction intermediates of Fe1-NC and Fe1V1-NC during ORR process were monitored in situ using attenuated total reflection surface-enhanced infrared absorption spectroscopy (ATR-SEIRAS). Figure 6a shows the spectra for the Fe1-NC catalyzed ORR, in which the characteristic absorption peaks all three intermediate species (OOH* at ~1027 cm-1, OH* at ~1200 cm-1, and *O at ~950 cm-1) are present[48]. However, only signals attributed to OH* (~1116 cm-1) and *O (964 cm-1) are observed for the Fe1V1-NC catalyzed ORR (Figure 6b). These results confirm that Fe1V1-NC follows a dissociation pathway without the formation of OOH* intermediate, thereby preventing formation of ·OOH, H2O2, and associated ROS radicals. Furthermore, the OH* signal for Fe1V1-NC is more distinct than for Fe1-NC, which implies a stronger adsorption of OH* onto Fe1V1-NC. The strong adsorption facilitates rapid intermediate conversion while suppressing desorption-induced formation of ·OH radical. The ability of Fe1V1-NC to circumvent ROS radical generation is consistent with the DFT analysis in Figure 2. We further performed in-situ electron paramagnetic resonance (EPR) spectroscopy to detect characteristic signals of the intermediates to corroborate the ATR-SEIRAS analysis above. 5,5-dimethyl-1-pyrroline N-oxide (DMPO) is used to capture the active species, for which an aliquot is then transferred from the quartz reactor cell to the EPR sample chamber during the corresponding ORR.[49] As shown in Figures 6c and 6d, Fe1-NC exhibits significantly higher ·OH signals in the range of 3415 – 3455 Gauss, whereas these characteristic peaks are barely visible in the spectra for Fe1V1-NC, hence corroborating the unique pathway on Fe1V1-NC with no radical generation. Interestingly, the characteristic peaks of ·OH remain undiminished after 18 min for Fe1-NC, indicative of a slow conversion of the ·OH radicals with a longer lifetime. We also design radical trapping experiments to monitor the ROS formation during ORR. We chose 2,2-azinobis-(3-rthylbenzoline-6-sulfonate) (ABTS) to trap ROS. ABTS is easily oxidized by ROS to form ox-ABTS and the color of the system changes from green to colorless. This can be quantified by monitoring the absorbance changes at 417 nm. The measurement is initiated by the addition of a controlled amount of H2O2 to the reaction system. In Figure 6e-6g, the absorbance for Fe1V1-NC is notably lower at each concentration point when compared to that of Fe1-NC, confirming again that Fe1V1-NC generates fewer ROS radicals during ORR. These investigations collectively demonstrate that, unlike Fe1-NC, Fe1V1-NC follows a rather unique pathway that effectively reduces the generation of ROS radicals, which suppresses the associated catalyst corrosions and promotes robust long-term ORR activity. 
 
 
 
  

Figure 6. In situ ATR-SEIRAS spectroscopies of (a) Fe1-NC and (b) Fe1V1-NC, respectively. In situ EPR spectra for monitoring ·OH radical of (c) Fe1-NC and (d) Fe1V1-NC, respectively. UV-vis spectra of (e) Fe1-NC and (f) Fe1V1-NC obtained by ABTS radical test. (g) Comparison of the absorbance at 417 nm for Fe1-NC and Fe1V1-NC. 
 
 
 
Figure 7. (a) Configuration diagram of the ZABs. (b) Open-circuit potential (OCP) curves (the inset is the picture of Fe1V1-NC-based ZAB connected to a multimeter). (c) Discharge polarization curves and corresponding power density curves of Fe1V1-NC and Pt/C. (d) Galvanostatic discharge curves of Fe1V1-NC-based and Pt/C-based ZABs at various current densities from 10 to 50 mA·cm-2. (e) Galvanostatic discharge curves (Zn-mass-normalized specific capacities) of Fe1V1-NC and Pt/C at 10 mA·cm-2. (f) Comparison of Zn-air batteries performance with that of the PGM-free electrocatalysts reported previously. These data are also summarized in Table S7 with references. (g) Galvanostatic discharge/charge cycling performance of Fe1V1-NC-based and Pt/C-based ZABs at 10 mA·cm-2, and (h) corresponding discharge voltage stability.
 
Zn-Air Batteries Performance 
To evaluate the performance of the catalysts at the device level, we assembled a rechargeable zinc-air battery (ZAB) using Fe1V1-NC as the cathode catalyst and a zinc plate as the anode in an electrolyte solution containing 6 M KOH and 0.2 M Zn(Ac)2 (Figure 7a and S30). For comparison, another ZAB was assembled with a mixture of commercial Pt/C (20 wt.%) and RuO2 (weight ratio at 1:1) as the cathode catalyst while the rest of the components kept unchanged. The Fe1V1-NC-based ZAB registers a slightly higher open-circuit potential (OCP, up to 1.50 V) than the commercial Pt/C-based ZAB (1.47 V) as shown in Figure 7b, indicative of a lower overpotential of Fe1V1-NC than that of commercial Pt/C. The discharge polarization and corresponding power density curves are shown in Figure 7c, where the Fe1V1-NC-based ZAB presents a higher peak power density (237 mW·cm-2) than the Pt/C-based ZAB (211 mW·cm-2). The galvanostatic discharge curves in Figure 7d depict the similar trend where the operating potential of the Fe1V1-NC-based ZAB outperforms that of the Pt/C-based ZAB across the range of discharge current densities from 10 to 50 mA·cm-2. Both batteries’ operating potentials show a quick and reasonable recovery with returning the current density from 50 to 10 mA·cm-2. The specific capacity (795 mAh·g-1, normalized to the mass of zinc) of the Fe1V1-NC-based ZAB at j = 10 mA·cm-2 also surpasses that of the Pt/C-based ZAB (766 mAh·g-1) as shown in Figure 7e. The cycling performances of both ZABs are evaluated via galvanostatic discharge-charge cycling tests. The cycling stability of the Fe1V1-NC-based ZAB for both the discharge and charging processes at 10 mA·cm-2 exceeds 400 h (600 cycles) without notable discharge voltage changes, showing its impressive cycling durability (Figure 7g). In contrast, the benchmark Pt/C-based ZAB already starts experiencing an obvious drop in the discharge voltage and, therefore, a loss energy efficiency after 200 h. Figure 7h shows the voltage efficiency of the Fe1V1-NC-based ZAB stays at ~60% with negligible changes (less than 1%) throughout the 400 hours of discharge and charge cycles. The performance of the Fe1V1-NC-based ZAB presented here is leading amongst the ZABs that employ PGM-free catalysts in the latest reports in the literature as summarized in Figure 7f and Table S7. Particularly, the exceptional cycling durability achieved by the Fe1V1-NC-based ZAB is a strong testimony to the strategy of circumventing ROS generation to improve the stability of catalysts during an ORR by introducing Fe-V atomic pair in Fe1V1-NC.
Conclusion
In summary, we demonstrated an effective strategy to switch the mechanistic ORR pathway and, therefore, improve the intrinsic ORR stability via designing catalysts guided by DFT calculations. The as-prepared Fe1V1-NC with Fe-V atomic pairs exhibit a unique active site configuration (N2Fe-N2-VN2), allowing an O=O bond direct-cleavage pathway. This unique ORR pathway not only generates a lower reaction barrier that enhances its ORR activity, but also circumvents the ROS radical generations, which significantly improve the stability of the catalyst. Fe1V1-NC achieved an outstanding ORR activity with Eonset and E1/2 at 1.02 and 0.89 V vs. RHE, respectively. The remarkable ORR stability of Fe1V1-NC is characterized by losses in E1/2 at the low rate of 3.2×10-4 mV per cycle over 50,000 CV cycles. In situ characterizations and analyses revealed that the transformation of ORR pathway between Fe1V1-NC and Fe1-NC led to distinct differences in ROS radical generation, fundamentally enhancing ORR stability of Fe1V1-NC. The Fe1V1-NC-based ZAB exhibited excellent cycling durability with no detectable voltage decay over 400 h charge-discharge operation, outperforming most of the PGM-free catalysts reported lately. These findings elucidate the effect of different ORR pathways on the catalytic stability of ORR catalysts, providing fresh insights for developing stable and efficient catalytic systems towards sustainable energy conversion technologies.
"