Flexibility Tuning of Dual‐Metal S─Fe─Co─N5 Catalysts with O‐Axial Ligand Structure for Electrocatalytic Water Splitting

The electrocatalytic performance of metal–nitrogen–carbon (M─N─C) single‐atom catalysts remains a significant challenge due to their rigid active center. Controllable tuning of the local microenvironment and electronic structure is critical for M─N─C single‐metal site catalysts in improving the electrochemical performance and exploring the reaction mechanism. Herein, Co─N4 is selected as a benchmark among various M─N─C catalysts based on theoretical prediction and experimental studies. A dual‐metal S─Fe─Co─N5 catalyst is constructed by embedding Fe and S into the structure of Co─N4 motifs. Theoretical analysis and in situ characterizations illustrate that the active sites will in situ combine an O‐axial ligand to form S─Fe─Co─N5─O structure during the oxygen evolution reaction (OER), which can reduce the reaction energy of O*→OOH*. The Ab Initio Molecular Dynamics simulations and deformation energy for H*/O* adsorption reveal that the Fe─Co and S─Fe bonds exhibit flexible characteristics compared to the Co/Fe─N bonds. This flexibility of S─Fe─Co─N5─O structure facilitates the OER performance by reducing the OOH*→O2, which is the OER rate‐determining step, resulting in superior performance. The optimized S─Fe─Co─N5─O catalyst exhibits excellent OER (260 mV@50 mA cm−2) and hydrogen evolution reaction (138 mV@10 mA cm−2) performance in alkaline electrolytes. The reported regulation strategy ameliorates the micro‐environment of Co─N4 with tunable flexibility, which helps allow a basic comprehension of the electrochemical reaction mechanism.


Introduction
Owing to low economic cost, high atomic utilization, and isolated active site features, non-noble metal atoms have attracted much attention in electrocatalytic systems. [1]However, the migration and aggregation problems in the electrocatalytic process DOI: 10.1002/aenm.202301547cannot be ignored. [2]Among them, materials with non-noble metal structures stabilized by nitrogen functional groups on the carbonaceous surface are usually coordinated in the stable form of M─N 4 , effectively avoiding the formation of M─M bonds. [3]Thus, there remains essential challenges to improve the intrinsic M─N structural properties of single-metal site M─N─C materials by tuning their coordination environment. [4]nchoring a second metal atom is considered to be a rational method to regulate the coordination environment of M─N─C. [5]For instance, Li et al. reported the Fe─Co─N 6 partial structure formed by Fe─Co diatom. [6]By decorating N-doped carbon material with hostguest strategy, Fe─Co─N 6 structure illustrated superior oxygen reduction reaction performance in the acidic electrolyte.Introducing nonmetals to modify the specific coordination conditions around M can also enhance the electrochemical activity of M─N materials.DFT calculations performed by Cui et al. demonstrate that Fe─Co─N 5 ─A (where A represents B, N, O, P, or S) coordination can effectively regulate the adsorption strength of ΔG H* and reduce the reaction energy barrier of intermediates. [7]hese findings suggest that changing the properties of coordination environment in single-metal site M─N─C materials can significantly influence their electrochemical performance.However, these materials have not been widely applied or considered in the field of electrochemical overall water splitting.
Motivated by the above discussion, this study aims to develop an efficient electrocatalyst for overall water splitting by integrating appropriate elements into the dual-metal M─N─C with a controllable coordination environment.The optimal intermediates adsorption strength of multiple M─N─C materials was obtained based on density functional theory (DFT) calculations, among which the most promising Co─N 4 was selected as a benchmark.Inspired by the excellent electrochemical performance in ORR after introducing single-atom Fe into Co─N 4 , [8] a host-guest strategy was employed to obtain S─Fe─Co─N 5 structure by simultaneously anchoring Fe and non-metallic S element on this basis.In situ characterization and DFT calculations showed that the constructed S─Fe─Co─N 5 catalyst would be in situ transformed into S─Fe─Co─N 5 ─O with O axial ligand during OER/HER process, which acts as the real active center.At the same time, compared with the original Co─N 4 (rigid structure), the variation range of bond length in the introduced S─Fe bond was expanded by 0.8 Å, which facilitates to increase the electronic activity near the active center to achieve superior electrochemical performance.This work not only explains the structural reorganization and the real active center of S─Fe─Co─N 5 catalysts during the water splitting process, but also reveals the role of introduction of S element in enhancing the electrochemical activity of M─N─C materials.Therefore, this study provides new insight for construction of dual-metal M─N─C site with controllable coordination environment and tunable flexibility to enhance the electrocatalytic water splitting performance of M─N─C materials.

Theoretical Prediction
DFT calculation was first adopted to perform a preliminary screening of M─N─C catalysts to find the most appropriate candidate material for study.First, according to the Sabatier principle, the optimal catalyst should bind reaction intermediates with moderate strength.In other words, it should conform to the principle of "just right", and the adsorption of reactants should be neither too strong (unfavorable for the escape of products) nor too weak (unfavorable for the activation of reactants). [9]The electrochemical performance of OER/HER is closely related to the adsorption-free energy of intermediates (ΔG OH* and ΔG H* ). [10]he intrinsic characteristics of the active center can partially reflect the binding strength of the adsorbent. [11]Therefore, the values of ΔG OH* and ΔG H* can be employed as a guidance for preliminary screening of appropriate catalysts.According to other relevant reports, [12] the catalyst achieves optimal adsorption/desorption performance when ΔG OH* and ΔG H* is between 1-1.5 and −0.5-0.5 eV, respectively. [11]Herein, the ΔG H* and ΔG OH* values of M (M = Mn, Fe, Co, Ni, and Cu)-N 4 were calculated, and the calculation results illustrate that Co─N 4 is the best candidate.Then M─Co─N 6 was utilized as the basic model to obtain the corresponding ΔG H* -ΔG OH* relationship diagram (Figure 1a; see detailed models in Figures S1 and S2, Supporting Information).Since no effective information can be obtained from theoretical efforts, some pre-experiments were carried out (Figure 1d).Fe─Co─N 6 , Ni─Co─N 6 , Cu─Co─N 6 , and Mn─Co─N 6 catalysts were synthesized, and Fe─Co─N 6 exhibits the lowest overpotential in the electrocatalytic OER experiment.Therefore, Fe─Co─N 6 was finally chosen as the target eletrocatalyst for further performance optimization.However, the Fe─Co─N 6 (OER: 380 mV@50 mA cm −2 ) still shows unsatisfactory overpotential compared to the reported state-of-art materials. [1]n order to break this inherent property, our group embed metalloid S element to modulate the M─N coordination microenvironment in an attempt to further enhance its catalytic performance.Modeling of Co/Fe sites in Co─N 4 , Fe─Co─N 6 , S─Fe─Co─N 5 , and Fe─Co─S─N 5 were performed, and the density of states (DOS) of these materials were obtained.The hybridization condition of the surface metal atoms (including Co and Fe sites) in d orbitals with H * and OOH * was calculated.Generally, the larger the orbitals are overlapped, the stronger the hybridization degree is, which is not conducive to the desorption of H * and OOH * . [13]As shown in Figure 1b,e, the hybridization degree between Fe/Co reactive centers of different materials and the OOH * and H * follows the sequence of Fe─Co─N 6 > Fe─Co─S─N 5 > Co─N 4 > S─Fe─Co─N 5 (for H * ), Fe─Co─S─N 5 > Fe─Co─N 6 > Co─N 4 > S─Fe─Co─N 5 (for OOH * ), respectively.Meanwhile, the binding strength during the reaction was also evaluated by the COHP-(Figure 1c,f).
It is worth noting that the integration area of the negative part represents the anti-binding strength, while the integration area of the positive part represents the binding strength. [14]COHP analysis of H * and OOH * adsorbates showed that the construction of Fe─Co sites would modulate the adsorption of H * (−1.22 → −1.37 eV) and OOH * (−1.15 → −1.45 eV).While embedding of S also properly optimized the bonding between the catalyst and H * and OOH * .Therefore, it can be tentatively suggested that incorporation of S to form S─Fe─Co─N 5 moieties would strengthen the adsorption of the two key intermediates (H * and OOH * ), which might give rise to highly boosted HER and OER activities.

Experimental Verification: Catalyst Synthesis and Characterization
As shown in Figure 2a, the dual-metallic Zn/Co MOF was first synthesized at room temperature, and metal Fe was then embedded into the bimetallic metal-organic frameworks (BMOFs) using the double solvent method. [15]The formed BMOFs and sulfur powder were continuously annealed in Ar atmosphere at 820 °C for 2 h.This "molecular encapsulation" method was chosen mainly due to the porous nature of BMOFs and the stable anchoring of metal atoms with N and C after high-temperature carbonization.The initially formed BMOFs in shape of rhomboidal dodecahedrons were converted into hollow ultrathin nanosheets due to the evaporation of Zn combined with the participation of S element into coordination moieties of N and C. It should be emphasized that embedding of Fe atoms has insignificant effect on the morphology of BMOFs.However, the collapse of the BMOFs skeleton may be attributed to the fact that the S element participates into of the M─N coordination structure and results in the formation of hollow nanosheets, which can be tracked through the transmission electron microscope images shown in Figures S3-S6 (Supporting Information) and the corresponding EDS mapping results.The structural characteristics of all the samples were investigated by Raman spectroscopy (Figure 2b), which revealed the change in the disorder-related D band at ≈1330 cm −1 and the graphitization-related G band at ≈1575 cm −1 .The D/G ratio of S─Fe─Co─N 5 (1.12) was significantly higher than that of Fe─Co─N 6 (1.05) and Co─N 4 (1.01).A Higher D to G ratio implies an increased structure disorder, which should be mainly caused by S doping.This confirms the formation of S─Fe─Co─N 5 moieties.Finally, high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) was employed to verify the existence of singly dispersed metal atoms.As shown in Figure 2e (see HAADF-STEM image of Fe─Co─N 6 in Figure S7, Supporting Information), Co singleatom and Fe─Co dual-atomic materials were obtained.The line scan profile in Figure 2f shows the distance between two bright spots is ≈2.3 Å, which is consistent with the Fe─Co distances based on Extend X-ray absorption fine structure (EXAFS) spectra and DFT-optimized S─Fe─Co─N 5 structure.Electron energyloss spectroscopy (EELS) spectrum confirmed the presence of Fe, Co, and S elements (Figure 2d).Moreover, Fe, Co, and S elements were identified (Figure 2g), denoting that the three elements were evenly distributed, which preliminarily confirmed the existence of coordination between S, Fe, and Co.
X-ray photoelectron spectroscopy (XPS) was conducted to investigate the electronic states and surface chemical composition of the synthesized samples.High-resolution spectra were obtained to clearly compare the offset in binding energies.In the high-resolution Fe 2p spectrum (Figure 3a), Compared with Fe─Co─N 6 , the binding energy of Fe 2p in S─Fe─Co─N 5 was positively shifted to a higher value, indicates a decreased electron density on the site of Fe in S─Fe─Co─N 5 .Meanwhile, a distinct Fe─S bond shown up at 707.8 eV in S─Fe─Co─N 5 also suggests a decreased electron density surrounding Fe. [16] Compared to Fe─Co─N 6 , a positively shifted BE of Fe as well as a deconvoluted Fe─S bond located at 707.8 eV can be found in S─Fe─Co─N 5 , suggesting that electrons are transferred from Fe to S. Considering the higher electronegativity possessed by S (2.58) as compared to Fe (1.83), it is reasonable to have a partial electron transfer between Fe and S, which leading to a charge redistribution and results in positively charged S and electronegatively charged Fe.This charge redistribution result can be adopted to optimize the absorption energy of intermediate species, such as OH * , and improve the corresponding electrocatalytic OER performance.Interestingly, the presented high-resolution Co 2p spectrum (Figure 3d) indicates that introduction of Fe elements also changes the electronic structure around Co atoms. [17]oreover, X-ray absorption near-edge structure (XANES) spectra were carried out to detect the coordination environment and electronic structure of Fe─Co dual-metal sites.As shown in Figure 3b,e, combined with the corresponding atom information in Fe 2 O 3 , Fe foil, FeO, Co 2 O 3 , Co foil, and CoO, etc.The absorption edge energies of Fe in Fe─Co─N 6 and S─Fe─Co─N 5 are positioned between the values of Fe and Fe 2 O 3 as shown in Figure 3b, indicating that the oxidation state of Fe species is between 0 and +3. [8,18]Furthermore, from the insets of Figure 3b,e, the involvement of S element causes a peak shift in the edge structure of Fe─Co─N 6 (Fe) toward higher binding energy.The result confirms that the Fe monatomic local electronic state of S─Fe─Co─N 5 transfers to S element, which is also more inclined to affect the coordination environment around Fe element compared to Co monatomic element.The Fourier transform extended -EXAFS-spectra are illustrated in Figure 3c,f, while the corresponding R and K space spectra were given in Figures S8-S11 (Supporting Information).For Fe elements, Fe─Fe bonding was not observed in Fe─Co─N 6 and S─Fe─Co─N 5 , Fe─Co─N 6 have the main peak at 1.58 Å, attributing to the scattering of Fe─N bonding, verifying the presence of single dispersed Fe atoms.A sharp peak with R = 2.4 Å was observed, indicating that the coordination around the Fe center in the first shell of Fe─Co─N 6 and S─Fe─Co─N 5 includes a metal-metal path in addition to the Fe─N/S coordination path.The EXAFS fitting results further confirmed the presence of Fe─Co coordination (Tables S1 and  S2, Supporting Information).Compared with the main peak of Fe─Co─N 6 (1.58 Å), the main peak of S─Fe─Co─N 5 was shifted to the right at 1.71 Å, [18,19] proving the existence of asymmetric coordination at the S─Fe─Co─N 5 site.The wavelet transforms EXAFS analysis of Fe─Co─N 6 and S─Fe─Co─N 5 (Figure 3g; see the wavelet transform diagrams of standard FeO and CoO in Figures S12 and S13, Supporting Information) showed a maximum intensity at 5 and 5.5 Å, further demonstrating the contribution of Fe─N bonding.The observed shift may be attributed to the embedding of S element. [20]For Co element, it has been confirmed by XANES and EXAFS that there is no Co─Co bond, however the embedding of S element has not significantly changed its surrounding coordination environment.The maximum intensities of the Fe─Co─N 6 , S─Fe─Co─N 5 wavelet transform EXAFS are both at 7 Å, and there is no obvious shift, which again confirms that the embedding of S element mainly affects the Fe coordination structure instead of Co element.

Electrocatalytic Performance
In order to investigate the effect of Fe and S heteroatoms on the OER performance of Co─N 4 under alkaline conditions, the electrochemical performance of the as-prepared catalysts was evaluated via a three-electrode system in N 2 -saturated 1 m KOH electrolyte.The OER performance of bare Nickel Foam (NF) was first tested to avoid interference to the final activity evaluation.All the obtained polarization curves were pre-sented with 85% IR compensation.As shown by the linear sweep voltammetry (LSV) polarization curve (Figure 4a), S─Fe─Co─N 5 (260 mV@50) exhibited excellent catalytic activity, much lower than that of Fe─Co─N 6 (370 mV@50), Co-N 4 (420 mV@50) and NF (460 mV@50), indicating that the introduction of Fe and S ameliorated the local coordination environment of the rigid framework Co─N 4 .Moreover, it is surprisingly to find that S─Fe─Co─N 5 a completed conversion of Fe 2+ /Co 2+ →Fe 3+ /Co 3+ was achieved in a very short voltage range (1.4-1.47V vs RHE) in S─Fe─Co─N 5 , which suggested that the introduction of S element enhances the electronic activity near the active center Fe/Co. [21]The electrocatalytic OER performance of as-prepared samples with different S contents was also explored (Figure S14, Supporting Information).Furthermore, to understand the catalytic reaction kinetics comprehensively, the Tafel slope of the corresponding logarithmic current density (logJ) to  was plotted in Figure 4b.The Tafel slopes of S─Fe─Co─N 5 , Fe─Co─N 6 , Co─N 4, and NF catalysts are 59.2, 77.4,86.6, and 123.9 mV dec −1 , respectively, revealing that S─Fe─Co─N 5 has the fastest catalytic kinetics.The electrochemical active surface area (ECSA) of the catalyst was obtained through double-layer capacitance (C dl ) to further reveal its intrinsic activity in Figure 4c (see the corresponding CV curves in Figure S15, Supporting Information).The C dl value of S ─Fe─Co─N 5 (39.15 mF cm −2 ) was higher than that of Fe─Co─N 6 (24.63 mF cm −2 ), Co─N 4 (10.69 mF cm −2 ), NF (4.52 mF cm −2 ), indicating that the new coordination environment formed by S─Fe─Co─N 5 was conducive to more the exposure of the active site.Combined with the HAADF-STEM images, it is not difficult to find that the high catalytic performance is mainly results from the formation of dual-metal (Fe─Co).To compare intrinsic activity, LSV curves are normalized by ECSA (Figure 4a), which still suggests the superior performance of S─Fe─Co─N 5 over other samples.
The charge-transfer resistance of the catalyst was evaluated by electrochemical impedance test (EIS).S─Fe─Co─N 5 exhibited the smallest charge-transfer resistance, confirming the electron transfer rate near the active center was enhanced due to the formation of new coordination environment (Figure 4d).Excellent stability of the catalyst is a prerequisite for commercial application.Then the electrocatalytic stability test was conducted for 20 h by amperometry in S─Fe─Co─N 5 (Figure 4e).The results showed that the attenuation in performance at a current density of 10 mA cm −2 can be ignored.The overpotentials at different current densities were also presented (Figure 4f).Taking the current density of 50 → 100 mA cm −2 as an example, the change in overpotential is smallest in S─Fe─Co─N 5 (40 mA cm −2 ), indicating that S─Fe─Co─N 5 always maintains superior electrochemical activity during the process of increasing current density.
As shown in Figure 5a, a small overpotential of 138 mV was needed in S─Fe─Co─N 5 to reach a current density of 10 mA cm −2 , which is significantly lower than that of Fe─Co─N 6 (225 mV), Co─N 4 (255 mV) and NF (297 mV).In fact, both anchoring of Fe and introduction of S elements can significantly improve the electrochemical activity of Co─N 4 .The HER performance of S─Fe─Co─N 5 with different S contents was also obtained (Figure S16, Supporting Information).The reaction paths and rates of these samples can be confirmed through the Tafel slope, as shown in Figure 5b.Compared with original Co─N 4 (119.8mV dec −1 ), S─Fe─Co─N 5 (52.1 mV dec −1 ), and Fe─Co─N 6 (65.6 mV dec −1 ) exhibited much lower Tafel slopes.Obviously, anchoring of Fe atoms accelerated the electron transfer rate.Similarly, the introduction of S elements also reformed the inherent coordination of Co─N 4 structure and ultimately improve HER performance.To investigate the intrinsic activity of the catalyst, the TOF was calculated.The TOF value for S─Fe─Co─N 5 is obtained to be 1.64 s −1 ( = 100 mV), which is higher than those of Fe─Co─N 6 (0.63 s −1 ) and Co─N 4 (0.84 s −1 ).
In general, there are many reasons for the difference in catalyst HER activity, including local electron transfer rate, active site exposure, H * adsorption/desorption, etc.Therefore, it is necessary to analyze the ECSA value of the catalyst to reveal its intrinsic activity (Figure 5c; see the corresponding CV curves in Figure S17, Supporting Information), S─Fe─Co─N 5 (75.3 mF cm −2 ) showed the largest ECSA value, confirming that the HER performance would enhance with the increase of active sites.Then, it should be the main contribution that Fe─Co dual-metal formed by anchoring Fe atoms and the new coordination environment formed by the introduction of S element.Furthermore, S─Fe─Co─N 5 exhibits an ultra-small charge transfer resistance (Figure 5d), confirming a favorable kinetic mechanism during the reaction.The stability test was carried out at a current density of 10 mA cm −2 , and the S─Fe─Co─N 5 material still has considerable stability (Figure 5e).Based on the above results, it is apparent that several factors contribute to the excellent HER performance of S─Fe─Co─N 5 .Therefore, the catalytic mechanism was further investigated using DFT calculations, AIMD, and in situ characterization techniques to gather further information.

Electrocatalytic Mechanism Investigation
Inspired by the excellent OER and HER performance of S─Fe─Co─N 5 , the sample was applied as bifunctional electrode for overall water splitting in a 1 m KOH electrolyte.As shown in Figure S18a (Supporting Information), the S─Fe─Co─N 5 //S─Fe─Co─N 5 bifunctional electrode delivers a current density of 10 mA cm −2 at a cell voltage of only 1.55 V, which is superior to the benchmark sample Pt/C(−)//RuO 2 (+) (1.58 V) commonly used in industry. [22]The overall water splitting performance did not show significant decline during 20 h stability test (Figure S18 , Supporting Information).There is no crystalline metal sulfides in spent S─Fe─Co─N 5 catalyst, and meantime, the dual Fe─Co site still exists after electrochemical reactions as evidenced by the HAADF-STEM image, which suggesting the good stability of unique S─Fe─Co─N 5 site (Figure S19b, Supporting Information).It is widely acknowledged that the M─N─C materials would undergo surface reconstruction during the OER process, in situ forming M─O/M─OOH species on the catalyst surface, which have been proven to be the real active species for OER.Therefore, the possible active center phase diagrams of these sam-ples during the reaction were calculated to further determine the possible active species.Generally speaking, the smaller the value at the same voltage, the more favor to the water splitting reaction.As shown in Figure S20 (Supporting Information), Fe─Co─N 6 , S─Fe─Co─N 5 , and Fe─Co─S─N 5 materials are more inclined to play the role of M─O in the OER (U (V) > 1.23V) process.The Co─N 4 active material realizes the conversion of Co→Co─O after the voltage of 1.6 V, which may be attributed to the slow reconstruction process (due to the inherent rigid structure of Co─N 4 ), which can also be found during LSV testing process.S─Fe─Co─N 5 always adopts M─O as the active center during HER reaction (−0.5V < U (V) < 0V).It should be noted that the main active species of Co─N 4 , Fe─Co─S─N 5 , and Fe─Co─N 6 are M and M─O, respectively.The voltage-based phase diagrams of different models were shown in Figure 6a, and the structures with more negative chemical potential values were more favorable in real reaction media.Then, the possible changes in the metal surface during the OER process was investigated through quasi in situ XPS technique, and the XPS spectra of Co, Fe, and O at different voltages were discussed.As the voltage increased, the peak of M─O gradually intensified, confirming the formation of the M─O axial ligand during the OER reaction (Figure 6b).At the same time, the binding energy of Co/Fe shifted to a higher position with increasing voltage, indicating that Co and Fe elements jointly participated in the reconstruction process (Figure 6d,e).Moreover, a small amount of sulfate also precipitated along with the process (Figure S21, Supporting Information).In addition, the structural evolution of the electrochemical catalytic reaction process was also monitored using in situ Raman spectroscopy.The in situ Raman spectra of Fe─Co─N 6 and S─Fe─Co─N 5 during the anodic water oxidation process are shown in Figure 6c,f.For the Fe─Co─N 6 material, two new vibrational peaks appeared when the applied potential increased from 1.47 to 1.6 V.These two peaks can be assigned to the Co─O vibration (446 cm −1 ) [23] of the surface CoOOH intermediate and the Fe─O vibration (627 cm −1 ) [24] of the FeOOH intermediate, indicating the in situ formation of axial oxygen-containing ligands of the prepared Fe─Co─N 6 during the OER reaction process.After the introduction of S element (Figure 6f), new peaks (400 and 510 cm −1 ) [25] appeared when the voltage was higher than 1.4 V, which can be assigned to Co/FeOOH and FeOOH, respectively.Notably, the peak intensity of the Co/Fe─O intermediate increased, which may be attributed to the increased local electronic disorder after introducing the S element.This phenomenon indicates that S element reduces the kinetic barrier for O─O bond formation and accelerates the reconstruction process of M─O.The conclusion was further confirmed by subsequent AIMD and Barder charge analysis.
DFT calculation and the AIMD techniques were employed to reveal the role of S and Fe on the electronic structure and catalytic behavior of Co─N 4 moiety.The differential charge density and related planar average charge along b axis direction were calculated as given in Figure 7a.The gray and yellow regions represent electron depletion and accumulation with an iso-surface level of 0.002 e Å −3 , respectively.The results clearly indicate the incorporation of S led to localized electronic rearrangement, resulting in a deficient electron state of Fe and Co, well matches the XPS results. [26]It is well known that the overall water splitting efficiency is limited by the OER semi-reaction which requires more transferred electrons. [27]Therefore, the OER reaction energy diagrams on Co─N 4 , Fe─Co─N 6 , S─Fe─Co─N 5 , and Fe─Co─S─N 5 models were further explored (Figure S22, Supporting Information and see corresponding structural models in Figures S23-S25, Supporting Information).The results display that Co─N 4 exhibits the lowest reaction energy of rate-determining step (RDS) of 0.46 eV, which is inconsistent with our experimental observations.Together with the results in Figure 6, it can be speculated that these ideal models undergo structural reorganization to in situ produce newly reactive centers, which were also reported in previous studies. [28]In this sense, the OER reaction diagrams based on these in situ formed O-terminated reactive centers were re-calculated (Figure 7b; see corresponding structural models in Figure S26, Supporting Information).The RDS of Co─N 4 is OH * →O * , and the highest reaction energy is 0.46 eV.After the introduction of Fe and S elements, M recombines with the generated O axial ligand-during the reaction process and the in situ formed Fe─Co─N 6 ─O and S─Fe─Co─N 5 ─O were deemed as the real reactive center.Accordingly, the OOH * →O 2 step becomes RDS which require reaction energy barriers of 0.33and 0.22 eV.The results prove that Fe and S effectively regulated the coordination environment around Co─N 4 toward decreased OER barrier, which is consistent with the improved activity in our experiments.To determine the location of the S element (adjacent to Co or Fe), the OER reaction energy of Fe─Co─S-N 5 ─O model was also calculated, which shows higher barrier than the S─Fe─Co─N 5 ─O site (0.33 vs 0.22 eV, Figure S27, Supporting Information).The results were consistent with our XANES discussions that indicate the S elements mainly regulate the electronic structure of Fe site for improved performance in S─Fe─Co─N 5 catalyst.The unique role of S element in reducing OER reaction barrier was further evidenced by the inferior barriers over the P─Fe─Co─N 5 and O─Fe─Co─N 5 models, which involve other heteroatoms (Figures S28 and S29, Supporting Information).
Next, the corresponding HER reaction energy diagrams were also calculated.28b] The results show that asprepared materials favor the V-H pathway (Figures S30-S37, Supporting Information), and S─Fe─Co─N 5 ─O has the lowest H * binding energy (−0.12 eV) in the reaction (Figure 7c).Above theoretical calculations indicate that introducing Fe and S can effectively decrease the reaction barrier of OER and HER compared to parent Co─N 4 site.To further clarify the role of Fe and S, the variations of Co─N, Fe─Co, and Fe─S bonds using the explicit solvent model were also simulated and analyzed. [29]The calculation details were described in the experimental section.As shown in Figure 7d (see the cases for CoN 4 , Fe─Co─N 6 and Fe─Co─S─N 5 in Figure S38, Supporting Information), Fe─S bonds (yellow lines) and Fe─Co bonds (blue lines) show severe distortion (Accelerated the adaptive changes of active centers during catalytic processes), and the remaining Co─N and Fe─N bonds have little change, which is not conducive to the reconstruction of the active center and the rapid transfer of electrons in the catalytic reaction process.The changes in Co─N and Fe─Co bond lengths of other samples were also calculated (Figure S38d,e, Supporting Information), which all exhibited inconspicuous fluctuation.Therefore, it can be considered that the modification by the S element in particular caused changes in the coordination microenvironment around Fe, which is also consistent with the results of XANES.In Figure 7e, the mean absolute error values of bonding lengths for these samples were further explored, and the results showed that Fe─S bonding (0.065) was significantly better than Co─N, Fe─N, Fe─Co, and Co─S.The newly formed Fe─S bond has a unique flexible skeleton to ameliorate the rigid structure of Co─N, thereby exhibiting excellent electrochemical performance.As expected, the smaller deformation energy for H * and O * (key intermediates in HER and OER) adsorption on S─Fe─Co─N 5 structure again indicates the better flexibility of S─Fe─Co─N 5 moiety can better adapt to H * and O * adsorption (Figure 7f), and therefore contributes to the water splitting performance.

Conclusion
A typical bifunctional electrocatalyst consisting of S and Fe─Co dual-metal sites with flexible structures and axial oxygen ligands was proposed for enhancing the electrochemical performance in

Experimental Section
For experimental details please see Supporting Information.

Figure 2 .
Figure 2. a) Schematic illustration of the preparation process of the S─Fe─Co─N 5 .b) Raman spectra of Co─N 4 , Fe─Co─N 6 , and S─Fe─Co─N 5 .The inset shows the enlarged view of the G band. c) TEM and d) EELS spectra of S─Fe─Co─N 5 without background removal, showing the presence of Fe, Co, and S elements, indicating the presence of S elements involved in the coordination process and g) corresponding EELS mapping images of Co, Fe, and S. e) HAADF-STEM image of S─Fe─Co─N 5 wherein the frames and cycles marked the diatomic and atomic site, respectively f) line scan profile of the red box along red arrow.

Figure 3 .
Figure 3. Analysis of chemical states and coordination environments.a,d) Peak fitting of Fe 2p and Co 2p core-level XPS spectra of corresponding catalysts.b,e) Fe K-edge XANES spectra and Co K-edge XANES spectra with a zoomed-in view in the inset.c,f) Corresponding Fe, Co K-edge EXAFS fittings of Fe─Co─N 6 and S─Fe─Co─N 5 .g) Wavelet transform (WT) images of Fe─Co─N 6 and S─Fe─Co─N 5 samples.

Figure 5 .
Figure 5. Electrocatalytic HER performance in 1 m KOH.a) HER polarization curves of as-prepared catalysts at a scan rate of 5 mV s −1 .Illustrations: polarization curves normalized by ECSA for S─Fe─Co─N 5 , Fe─Co─N 6 , Co─N 4. b) Corresponding Tafel plots of different catalysts.c) Compared C dl data of S─Fe─Co─N 5 , Fe─Co─N 6 , Co─N 4, and NF.d) Electrochemical impedance spectra of S─Fe─Co─N 5 , Fe─Co─N 6 , Co─N 4, and NF.e) Stability test of S─Fe─Co─N 5 during HER process.f) Corresponding HER overpotentials at 10, 50, and 100 mA cm −2 .

Figure 6 .
Figure 6.a) Active center phase diagrams for Co─N 4 , Fe─Co─N 6 , S─Fe─Co─N 5 , and Fe─Co─S─N 5 .b,d,e) The quasi-in situ XPS spectra of O, Co, Fe for S─Fe─Co─N 5 , respectively.c,f) In situ Raman spectroscopy of Fe─Co─N 6 and S─Fe─Co─N 5 in a potential range between 1.10 and 1.60 V in 1 m KOH (vs Ag/AgCl).

Figure 7 .
Figure 7. a) Differential charge density diagrams and related planar average along b axis direction of S─Fe─Co─N 5 configuration.b) Calculated free energy diagram of OER intermediates at zero potential (U = 0), c) Calculated free energy diagram of HER intermediates.d) AIMD simulation Co─N, Co─S, Fe─Co, and Fe─S distances in S─Fe─Co─N 5 catalyst under reduction potential.e) Mean absolute errors plots for Co─N 4 , Fe─Co─N 6 , S─Fe─Co─N 5 , and Fe─Co─S─N 5 .f) Deformation energy for H * and O * adsorption for Co─N 4 , Fe─Co─N 6 , S─Fe─Co─N 5 , and Fe─Co─S─N 5 .
alkaline.It is demonstrated that recombination of axial O ligand and M behaves as the real active site by forming S─Fe and Fe─Co bonds, which accelerates the electron transfer rate around the active center and reduces the OH * →O * reaction energy barrier.After successfully regulating the Co─N 4 coordination environment, the formed S─Fe─Co─N 5 exhibited excellent overall water splitting performance and negligible attenuation in performance during the 20 h stability test.This study is the first to anchor Fe in Co─N 4 materials to form dual-metal sites and introduce S elements to regulate the local coordination environment of M─N─C materials, confirming that introducing S elements allows the possibility to form new flexible structures while reducing the H * adsorption/desorption barrier in HER.This work provides new insights into a comprehensive understanding of the unique role of sulfides and the structural reorganization of OER catalysts for overall water splitting.