Engineering unsymmetrically coordinated Cu-S1N3 single atom sites with enhanced oxygen reduction activity

Atomic interface regulation is thought to be an efficient method to adjust the performance of single atom catalysts. Herein, a practical strategy was reported to rationally design single copper atoms coordinated with both sulfur and nitrogen atoms in metal-organic framework derived hierarchically porous carbon (S-Cu-ISA/SNC). The atomic interface configuration of the copper site in S-Cu-ISA/SNC is detected to be an unsymmetrically arranged Cu-S1N3 moiety. The catalyst exhibits excellent oxygen reduction reaction activity with a half-wave potential of 0.918 V vs. RHE. Additionally, through in situ X-ray absorption fine structure tests, we discover that the low-valent Cuprous-S1N3 moiety acts as an active center during the oxygen reduction process. Our discovery provides a universal scheme for the controllable synthesis and performance regulation of single metal atom catalysts toward energy applications.


Responses to the Referees' Comments
We thank the referees for their valuable comments to our manuscript. We have carefully considered the referees' comments and revised the manuscript accordingly. Our responses and corresponding revisions are as follows:

Response to Reviewer 1：
In this work the authors present a CuN x C y catalyst containing sulphur atoms and showing an enhanced oxygen reduction reaction activity than the sulphur-free compounds. The morphology and active site structure are mainly investigated by electron microscopy techniques, X-ray absorption spectroscopy and DFT calculations. This is an interesting system, but there are a number of puzzling issues: Response: Thank you for your positive comments on our manuscript. We have revised our manuscript according to your suggestions. 1) We know that X-ray absorption spectroscopy is a bulk technique, so how the authors can exclude the formation of copper sulphide species coexisting with CuN x sites, rather than a new compound with formula CuN x S y ? This doubt is also fostered by the fact that the Cu-S bond length does not change under working conditions, and if CuS clusters are very small, they cannot be detected by TEM.
Response: Thanks for your insightful question. We agree the viewpoint that X-ray absorption spectroscopy is a bulk technique and if CuS clusters are very small they cannot be detected by conventional TEM. The demonstration of the existence for Cu-N x S y moieties in S-Cu-ISA/SNC is a key work in our project, with the applications of macro and micro characterization techniques.
Firstly, it is necessary to mention that the ZIF-8 derived synthetic strategy we applied is a typical and effective method to prepare various metal single atom catalysts, which has been identified by a series of excellent works. The precursor ZIF-8 cages can prevent the aggregation of the encapsulated metal precursor effectively, even at high pyrolytic temperature up to 800-1000 ℃. 1 To elucidate the feature of Cu atoms in S-Cu-ISA/SNC, we carried out aberration corrected HAADF-STEM measurements (Fig. 1d, e in the manuscript) with sub-angstrom resolution.
Besides, careful examinations at different areas in the sample also confirmed the absence of small copper sulphide clusters (Fig. R1). Then X-ray absorption spectroscopy (Fig. 2 in the manuscript) provides accurate evidence for the formation of Cu-S and Cu-N bonding on average view. Besides, we also try linear combination fitting (The function is embedded in Athena software) of the experimental XANES spectrum of S-Cu-ISA/SNC with the calculated spectrum for CuN 4 and experimental spectra for CuS and/or Cu 2 S, as shown in Fig. R2, Fig, R3 and Fig.   R4, respectively. The percentage of the three standard components was auto adapted by Athena software. We can find that although the fitted XANES curves near the edge seem coincide with the experimental spectrum of S-Cu-ISA/SNC in some way, the curves after the white line are quite different. In addition, powder X-ray diffraction (PXRD) of the S-Cu-ISA/SNC ( Supplementary Fig. 4b in the manuscript) reveal only two broad peaks typically for nano-sized graphitic platelets in amorphous carbon materials, suggesting the absence of copper sulphide nanoparticles in the sample. Taking these results into account, it is reliable to demonstrate the formation of CuN x S y moieties in S-Cu-ISA/SNC and exclude the existence of copper sulphide species.   Moreover, in order to further exclude the formation of copper sulphide species in S-Cu-ISA/SNC, the sample was immersed in dilute nitric acid (HNO 3 ) solution (1 mol/L) at 60 ℃ for 24 h. Due to the fairly high specific surface area and the hierarchically porous characteristics of carbon based frameworks, the dilute nitric acid solution can thoroughly permeate in the whole structure of the S-Cu-ISA/SNC polyhedron, so that the copper sulphide species can be removed if they exist in the sample. 6 After drying, the acid-treated sample was characterized by XAFS again, which we displayed in Fig. R5. It is found that both the XANES (Fig. R5a) and EXAFS  Finally, in the manuscript, we describe that "the mean bond length of Cu-S coordination is detected to be nearly unchanged and keeps at about 2.32 Å." Herein, error bar has to be considered. As we describe in the Supplementary Material, "Error bounds that characterize the structural parameters obtained by EXAFS spectroscopy were estimated as N ± 20%; R ± 1%; σ 2 ± 20%; ΔE 0 ± 20%". For the bond length which is abstracted from EXAFS, the error bar is ±0.02 Å. The change of the Cu-S bond length may be slight and it is hard to be monitored by EXAFS (Fig. R6). On the other hand, the nearly unchanged bond length for Cu-S during ORR process just demonstrates the well stability of the Cu-S 1 N 3 atomic interface structure at working conditions.   2) For the same reason of the previous point, authors cannot claim that a Cu+1 is formed and it is the active center. We know that under applied potential CuNC forms very small Cu nanoparticles is the extreme case of very low potential applied). This means that if a small fraction of Cu+2 is reduced to metallic copper, the average oxidation state will be lower than 2+. It is important to stress that EXAFS is not able to highlight a Cu-Cu contribution if the metallic fraction is small.
Still, a peak around 2.5 Å appears in the FT-EXAFS spectrum at 0.75 V, and it is not necessary to find a peak exactly at same position of metallic copper to have a metallic-based cluster with maybe a different nature than that of crystalline copper.
Both the two previous works are about electroreduction of CO 2 by atomic dispersed Cu catalysts, and they discovered the same phenomenon that small metallic Cu nanoparticles or clusters formed during CO 2 electrolysis, which was monitored by XAFS measurements at working conditions. Herein, what we have to claim is that oxygen reduction reaction (ORR) is quite a different process from CO 2 reduction reaction (CO 2 RR). In our work, CuS 1 N 3 species did not forms analogical small Cu nanoparticles or clusters under applied potential window during ORR, which had been carefully confirmed by our ex-situ XAFS and HAADF-STEM measurements ( Supplementary Fig. 42-43). Besides, some previous researches also support our viewpoint that the isolated Cu (+1) sites in S-Cu-ISA/SNC may work as the active centers for ORR. [1][2][3] Due to electron transfer between N/S atoms and Cu centers, the aggregation and further reduction of Cu atoms can be prevented effectively.
Moreover, Fig. R7 shows additional HAADF-STEM images at different areas of the S-Cu-ISA/SNC catalyst after ORR test, which also suggests the absence of Cu clusters or small copper sulphide species. Additionally, our extended in-situ XAS measurements we show in  In addition, as the reviewer has mentioned, a peak around 2.5 Å appeared in the FT-EXAFS spectrum at 0.75 V vs RHE. In our opinion, it is not the signal originated from Cu-Cu.
When we changed the FT range from "2.0-11.0 Å -1 " to "2.0-10.0 Å -1 ", the peak around 2.5 Å was erased, while the main peak around 1.5 Å almost kept the same (Fig. R8.). If the peak at 2.5 Å is attributed to Cu-Cu back scattering, there is no reason that it is such sensitive to the slight change of FT range. The weak signal at the range 2.2 to 4.2 Å may be derived from Cu-C/N back scattering at the high atomic shells and may also mix with some noise signals. It is hard for us to abstract useful structure information from this range.  Our revision: We have revised our manuscript according to the reviewer's suggestion (Line 359-361 in the revised manuscript, Supplementary Fig. 59).
3) Therefore, on the basis of the issue 1) and 2) the DFT model proposed by the authors may not be valid.

Response:
We agree the reviewer's viewpoint that the oxidation state (related to the XAS edge position) is not absolute, and it depends on the specific compound. As we known, metal foil and metal oxides (such as Cu foil and CuO) are commonly used standards for XANES and EXAFS analysis in various research works, mainly due to their fine stability and easy achievement. [1][2][3][4] Based on the viewpoint that the edge position in the XANES spectra reflects the oxidation state of Cu in different samples, 5     Our revision: We have revised our manuscript according to the reviewer's suggestion (Fig. 4d in the revised manuscript, Supplementary Fig. 20 and Supplementary Fig. 51). In addition, we carried out extended in-situ XAS measurements at lower potentials at BL14W1 in the Shanghai Synchrotron Radiation Facility (SSRF). The XANES spectra at 0.6 V and 0.4 V vs RHE were recorded (Fig. R12). We can find that the spectra at 0.6 V and 0.4 V vs RHE show little change compared to that at 0.75 V (particularly the absorption edge and white line peak as shown in Fig. R12b and Fig. R12c, respectively), which indicates that the local atomic structure and oxidation state of Cu in S-Cu-ISA/SNC at 0.6 V and 0.4 V during ORR is just the same as that at 0.75 V, mainly due to electron transfer between N/S atoms and Cu centers.

5) Why
This also means that the average oxidation state of Cu species (Cu-S 1 N 3 ) keeps stable (around +1) when the potential is down to 0.75 V, rather than further reduce tends to zero (metallic copper).
The extended in-situ XAFS measurements give strong evidences that no Cu particles or clusters form at the applied potentials and also demonstrate that that Cu (+1) sites in S-Cu-ISA/SNC may work as the active centers for ORR. Our revision: We have revised our manuscript according to the reviewer's suggestion (Line 333

References
in the revised manuscript, Supplementary Fig. 52).
6) Is the in-situ XAS spectrum reversible when you come back to the OCV potential?
Response: We appreciate the reviewer for the insightful point, which has helped us much to improve the quality of our manuscript. We performed additional in-situ XAS measurements to evaluate the reversibility of Cu K-edge XANES spectra of S-Cu-ISA/SNC at various potentials ( Fig. R13). As what we have described in Fig. 4b and Supplementary Fig. 50, from 1.05 V to 0.75 V, the absorption edge of Cu is gradually shifted to the lower energy, along with reduction of the white line peak, meaning a decrease of the average oxidation state of Cu in S-Cu-ISA/SNC during ORR. Interestingly, when the applied potential returned to OCV (from 0.75 V to 1.05 V), Cu XANES edge shifted back to higher energy along with increase of the white line peak (Fig. R13). This provides unequivocal evidence that the XAS spectra as a function of applied potential are reversible, which might be due to the strong anchor effect of N and S atom to the Cu sites. The reversible change of Cu valence state indicated the active contribution of isolated Cu atoms in the catalytic reaction for ORR.  Our revision: We have revised our manuscript according to the reviewer's suggestion (Line 335-340 in the revised manuscript, Supplementary Fig. 53-54). Figure 1e with red circles around hypothetical sites.

7) I personally do not appreciate images like
In many parts of these pictures the spots overlap and the hypothetical single-atoms site is not clear.

Response:
Thank you for your kind suggestions. We have erased the red circles around the bright dots in Fig R15. Moreover, we also modified other STEM images with red circles in the manuscript and supplementary materials. Our revision: We have modified the manuscript according to the reviewer's suggestions (Fig.   1e, Fig. 6a, Fig. 6c, Fig. 6e and Fig. 6g in the revised manuscript, Supplementary Fig. 14d,   Supplementary Fig. 15d and Supplementary Fig. 42d).
8) The authors completely forget to mention a lot of literature from other worldwide research groups, and not only from the restricted geographical area of most of the authors in the bibliography of this manuscript.
Response: Thanks to the reviewer's valuable suggestion. We have added some suitable references in the revised manuscript.
The added references:

Response to Reviewer 2：
The authors report on the synthesis, structural and electrochemical characterisation of Cu-N-C and Cu-S-N-C materials for oxygen reduction reaction in alkaline electrolyte. The main novel claim is the coordination of Cu by both S and N atoms, not only by N atoms. This seems to improve significantly the ORR activity in alkaline medium of single-atom Copper by tuning the binding of oxygen adsorbates. While demonstrating that the single Cu atoms are coordinated with a mix of N and S atoms versus only N atoms is not an easy task, the authors give convincing evidence for this. The ORR activity in alkaline is at par or higher than state of art Fe-N-C catalysts.
Response: Thank you for your time in reviewing this manuscript. We have carefully considered your comments and revised our manuscript according to your suggestions.

Major comments:
a) According to authors, the S L-edge XANES spectrum shows only S-C bonds apparently and no S-Cu signal (suppl. fig 15). Should there not be a specific signal for direct S-Cu binding in this spectrum, if the claim of CuS 1 N 3 coordination is correct?
Response: Thanks for your insightful question. In the manuscript, we claim that "The S L-edge XANES spectrum of S-Cu-ISA/SNC shows obvious peaks (peak h-j)) in the region of 163∼167 eV corresponding to C-S-C coordination species, suggesting the anchor of S in the carbon skeleton." Herein, we did not deny the existence of S-Cu signal. On the contrary, there are some other peaks in the 165~185 eV region, which are probably attributable to the S-Cu bonds (Fig.   R16.). [1][2][3] The local structure of S species can be reflected more clearly by S K-edge EXAFS spectrum as shown in Fig. R25 for Comment f. Our revision: We have revised our manuscript according to the reviewer's suggestion ( Supplementary Fig. 17).
b) The key data that directly supports the Cu-S binding is the EXAFS spectrum at Cu K edge, figure S17. However, it might also be interpreted as superimposed signal from single atom Cu  In addition, it is necessary to mention that the ZIF-8 derived synthetic strategy we applied is a typical and effective method to prepare various metal single atom catalysts, which has been identified by a series of excellent works. The precursor ZIF-8 cages can prevent the aggregation of the encapsulated metal ions efficiently, even at high pyrolytic temperature of 800-1000 ℃. [1][2][3][4] In order to further exclude the formation of copper sulphide species in S-Cu-ISA/SNC, the sample was immersed in dilute nitric acid (HNO 3 ) solution (1 mol/L) at 60 ℃ for 24 h. Due to the fairly high specific surface area and the hierarchically porous characteristics of carbon based frameworks, the dilute nitric acid solution can thoroughly permeate in the whole structure of the S-Cu-ISA/SNC polyhedron, so that the copper sulphide species can be removed if they exist in the sample. 6 After drying, the acid-treated sample was characterized by XAFS again, which we displayed in Fig. R18. It is found that both the XANES (Fig. R18a) and EXAFS (    Fig. 10 and Supplementary Fig. 22).

References
c) A puzzling issue is that the calculated XANES signal for CuS 1 N 3 is still quite different from the experimental XANES spectrum (Suppl. Figure 20b). The experimental spectrum looks in first approximation maybe more similar to that for crystalline CuS. Could the authors compare the experimental XANES spectrum to that of CuS and other copper sulfide structures in one graph, and try if linear combination fitting of the experimental spectrum with the calculated spectrum for CuN 4 and experimental spectra for CuS and/or Cu 2 S gives a possible fit as well, or impossible fit?
Response: Thanks for the reviewer's insightful suggestion. As we known, for XANES calculations, due to the variable electronic potential and broadening, it is really hard or even impossible to obtain a simulated XANES spectrum which is 100% identical to the experimental one with no difference. [1][2][3] In the Supplementary Fig. 25b  Furthermore, we compare the experimental XANES spectrum of S-Cu-ISA/SNC to that of CuS and Cu 2 S in one graph (Fig. R19). We can see that the XANES curve of S-Cu-ISA/SNC is quite different from those of CuS and Cu 2 S, which further provide evidences for the absence of copper sulphide species in the sample. Moreover, we also try linear combination fitting (The function is embedded in Athena software) of the experimental XANES spectrum of S-Cu-ISA/SNC with the calculated spectrum for CuN 4 and experimental spectra for CuS and/or Cu 2 S, as shown in Fig. R20, Fig, R21 and Fig.   R22, respectively. The percentage of the three standard components was auto adapted by Athena software. We can find that although the fitted XANES curves near the edge seem coincide with the experimental spectrum of S-Cu-ISA/SNC in some way, the curves after the white line are quite different. Based on the above analysis, we can exclude the possible existence of copper sulfide species in S-Cu-ISA/SNC.    Supplementary Fig. 27-29).

229-234 in the revised manuscript,
d) On page 6, line 136-138, the authors write that "in-situ environmental microscopic studies suggest that the permeation of sulfur plays an important role for etching the carbon frameworks".
It is not clear why and how sulfur would do that. Without addition of sulfur, ZIF-8 (with or without addition of copper salt) leads to high BET area N-C material (e.g. figure S5 shows this, with Cu-ISA/NC having BET of 1377 m 2 /g, only slightly lower than with sulfur addition). From figure S11, a hole inside the nano zif-8 pyrolyzed particles is seen for S-N-C, but barely seen for Cu/SNC in figure S13.

Response: Thanks for insightful comments. To investigate the role of sulfur, the in-situ
environmental microscopic studies were carried out to simulate the variation of carbon frameworks with temperature. From Supplementary Fig. 7 and Supplementary Video 1, it can be seen that carbon frameworks became shrunken following the increasing of temperature. Especially, the smaller particle disappeared when the temperature was up to 980℃. Interestingly, carbon frameworks without sulfur retained their shape, but just became smaller with the increasing temperature mainly due to the evaporation of Zn (Supplementary Fig. 8,   Supplementary Video 2). The results confirm that sulfur can etch the carbon frameworks effectively. Moreover, the entire carbon frameworks can be completely eroded if the sulfur amount is enough. In this work, we optimize the amount of sulfur to ensure that the carbon frameworks are moderately etched to form more porous space. The adsorption-desorption curve was further performed to evaluate the effect of sulfur to the carbon frameworks ( Supplementary   Fig. 6). The results show that S-Cu-ISA/SNC (with addition of sulfur) possesses ~20% higher BET area than that of Cu-ISA/NC (without addition of sulfur). Particularly, S-Cu-ISA/SNC obtain much larger total pore volume between 0.5 nm to 30 nm than that of Cu-ISA/NC (2.13 cm 3 g -1 : 0.67 cm 3 g-1), indicating the prominent etching effect of sulfur to construct the hierarchically porous structure for the carbon frameworks. In addition, Fig. R23 shows additional TEM images for the Cu-ISA/SNC. A hole inside the carbon matrix is obviously observed for Cu-ISA/SNC, just similar to the SNC sample in revised Supplementary Fig. 13.  The valance state of sulfur was investigated by S K-edge XANES, applied CuS (-2), S powder (0), Na 2 SO 3 (+4) and Na 2 SO 4 (+6) as references (Fig. R25). Generally, the oxidation states of S are linear correlated to the position of K-edge. [1][2][3] We found that the sulfur in S-Cu-     Our revision: We have modified the manuscript according to the reviewer's suggestions (Fig. 1 in the revised manuscript, Supplementary Fig. 14d, Supplementary Fig. 15d, Supplementary   Fig. 42d). analysis, which is much lower than that of Cu (0.73 at%) in the sample. In addition, we have carried out energy-dispersive spectroscopy (EDS) mappings and XPS measurements to confirm the existence of the elements in our S-Cu-ISA/SNC catalyst. As shown in Fig. 1c and  Table R2.

The
To investigate the influence of Zn species with ultralow content to the ORR performance, we have carefully examined the LSV analysis for the Zn-contained samples in 0.1M KOH. In Fig. 3a and Fig. 3b (Fig. R29), particularly for NC and SNC (in Table R3, the Zn content in the two samples are 0.030 and 0.025 according to ICP measurements, which is almost the same to that of S-Cu-ISA/SNC), the E 1/2 value was 0.66 V and 0.79 V respectively, significantly lower than that of S-Cu-ISA/SNC (0.918 V), Cu-ISA/SNC (0.87 V) and Cu-ISA/NC (0.86 V).
Furthermore, the catalytic activity of NC catalyst in this work is comparable to that of Zn absence NC materials in previous works. 4,5 The above results confirm that the residual Zn has little influence for ORR performance in the S-Cu-ISA/SNC system.

References:
considered the type of S and N species, as shown in Supplementary Fig. 16. The N 1s spectra ( Supplementary Fig. 16e) was classified to four types of N species, namely, 398.4 (pyridinic N), 399.1 (Cu-N), 400.2 (pyrrolic N) and 401.1 (graphitic N). The S 2p XPS spectra ( Supplementary Fig. 16f) show three types of S species. The peak at 168.3 eV assigned to the sulfate species (C-SO x ), while the two peaks at 164.9 eV and 163.6 eV were associated to C=S-C and C-S-C bond, respectively. Importantly, the characteristic peaks at 163.9 eV, corresponding to the Cu-S bond, were observed, which could stem from the partial replacement of N atoms with S to form the Cu-S coordinations. All these results indicated that the atomically dispersed Cu possessed typical Cu-N and Cu-S dual coordinating environments. So we propose the unsymmetrical isolated Cu-S 1 N 3 atomic interface structure for S-Cu-ISA/SNC.    Response: Thank you for the valuable suggestion. We read this paper carefully and strongly agree the viewpoint that the formation of hetero-coordinated moieties at atomic scale with asymmetrical electronic distribution should be conducive to the oxygen-involving catalysis, as it can benefit the adsorption and desorption of the adsorbates.
Our revision: We added the paper (Adv. Mater. 2019, 1805581) as reference in the revised version (Reference 44).