Experimental and Theoretical Elucidation of Metal‐Free Sulfur and Nitrogen Co‐Doped Porous Carbon Materials with an Efficient Synergistic Effect on the Oxygen Reduction Reaction

Metal‐free carbon‐based catalysts have attracted significant attention owing to their unique electronic structure and excellent catalytic activity for the oxygen reduction reaction (ORR). Recently, several heteroatom‐doped carbon catalysts with ORR performances comparable with those of state‐of‐the‐art Pt‐based catalysts have been reported. However, the intrinsic influence of heteroatoms on the catalytic activity has not been thoroughly investigated to date. This paper reports porous carbons co‐doped with N and S, prepared using an ion‐exchange resin and tetramethylammonium cations, with an onset potential of 0.93 V versus reversible hydrogen electrode (RHE) and a half‐wave potential of 0.79 V versus RHE. First‐principles calculations reveal that the change in the atomic charge caused by the co‐doping of N and S is important for inducing a high ORR activity. This paper presents a rational synthetic strategy for metal‐free catalysts and provides fundamental insights into their electrocatalysis.

heteroatom doping induces charge delocalization and spinstate variation at the sp 2 carbon surface. [31] Several studies have reported that charge delocalization and changes in spin state due to the introduction of heteroatoms, such as N and S, facilitate the adsorption of oxygen and its intermediates on the catalyst and promote the ORR. [32] It is recognized that the catalytic active site is composed of heteroatoms and adjacent carbon atoms with negative electronegativity located on the surface of the carbon matrix.
Carbon materials containing two or more types of heteroatoms are expected to be the more facile and efficient catalysts than current precious-metal and transition-metal catalysts because the coexistence of hetero-atoms can give carbon atoms specific bonding configuration which suitable for ORR. [26,30] However, the active sites of these multi-heteroatom-doped metal-free carbon catalysts have not been identified thus far because their structure and bonding states are easily disordered by pyrolysis during catalyst synthesis. [31,32] In short, the relationship between experimentally observed catalytic activity and catalyst synthesis almost remains empirically understood. Therefore, elucidating the bonding states of heteroatoms in metal-free catalysts is essential for the replacement of commercial Pt-based catalysts.
Synthetic methods for incorporating heteroatoms into carbon materials have been reported extensively, such as chemical vapor deposition and thermal annealing. [17,33,34] However, controlling the bonding state of the active sites is extremely challenging. Recently, a synthetic method was developed to precisely control the effective ORR active sites using ionexchange resins as starting materials. [35] However, this method has only been used to synthesize N-doped catalysts with Co nanoparticles.
This study focuses on the use of a quaternary ammonium salt (tetramethylammonium; TMA) with a cationic charge and an ion-exchange resin (Amberlyst 15; Amb) to synthesize metal-free dual-heteroatom-doped carbon materials. The N/Sco-doped porous carbon materials with precisely controlled active sites are prepared by sequential treatments of primary carbonization, CO 2 activation, and secondary carbonization on the ion-exchanged samples. To demonstrate the validity of the approach, the structures and active sites of various heteroatom-containing catalysts derived from Amb and TMA are analyzed. This study provides a design policy for next-generation metal-free carbon catalysts by elucidating the synergistic effect between N and S.

Results and Discussion
As shown in Scheme 1, the N/S-co-doped carbon catalysts were synthesized using an ion-exchange method. Firstly, Amb was mixed with an aqueous TMA-containing solution, and protons from the sulfone groups in Amb were exchanged with TMA cations. Subsequently, the ion-exchanged precursor was carbonized at 1100 °C (denoted as NSC). After sequential CO 2 activation for 4 h and secondary carbonization, NSC-act4R was obtained. For comparison, different N,S-doped carbon materials were synthesized at various carbonization temperatures (NSC-900 and NSC-1000), activation times (NSC-act2 and NSC-act4), and number of carbonization treatments (NSC-act4R). Transmission electron microscopy results confirmed the morphologies of all the carbonized catalysts ( Figure S1, Supporting Information). The N/S-co-doped catalyst after the second carbonization process (NSC-act4R) ( Figure S1d, Supporting Information) had a rough surface with several defects in relation to the N/S-co-doped catalyst (NSC) ( Figure S1a, Supporting Information), suggesting that the carbon matrix was gasified during the CO 2 activation process, leading to the formation of nanopores. Furthermore, N 2 adsorption isotherms of the synthesized catalysts were obtained to elucidate the effect of CO 2 activation. Table 1 lists the Brunauer-Emmett-Teller (BET) surface areas of the samples after primary and secondary carbonization. The BET surface area increased with increasing CO 2 activation time and after the secondary carbonization treatment; the BET surface area of NSC-act4R prepared by CO 2 activation for 4 h and secondary carbonization for 3 h was 2065 m 2 g −1 . Figure 1a shows the N 2 adsorption-desorption isotherms of all the NSC samples. All samples except NSC-act4R exhibited type Scheme 1. Schematic of the synthesis of the NSC samples. The gray, blue, and yellow balls represent C, N, and S atoms, respectively. www.advmatinterfaces.de I isotherms, indicating the presence of micropores, whereas NSC-act4R had a type IV adsorption isotherm, indicating the presence of mesopores. Moreover, the corresponding pore size distributions of each sample confirmed the formation of mesopores, with a peak observed at approximately 6.8 nm for NSC-act4R ( Figure 1b). In this case, intrinsic carbon defects, such as edge and topological defect groups, which formed during the CO 2 activation process, were removed by graphitization of the carbon matrix during the secondary carbonization process, resulting in mesopore formation.
The elemental analysis results show the relative C, H, N, and S contents of the NSC samples (Table 1). During the primary carbonization process, a decrease in the N content and an increase in the S content were observed with increasing carbonization temperature. Moreover, the N content decreased from 1.30 wt% (NSC-act4) to 0.50 wt% (NSC-act4R) after the secondary carbonization process. The relative S content of NSC-act4R was 0.57% higher than that of NSC-act4. The atomic ratio of S to N in NSC-act4R was 1.22. Based on the elemental analysis results of NSC-act4R, it is concluded that there is approximately one N atom in the vicinity of the S species in the carbon matrix.
Additionally, the chemical properties of the NSC samples were characterized by X-ray photoelectron spectroscopy (XPS) to identify the specific configuration states of each element, as shown in Figure 2 and Figures S2-S4 (Supporting Information). As shown in Figure 2a, the high-resolution XPS spectra in the N 1s region was fitted with two peaks corresponding to pyridinic N (398.6 eV) [36,37] and graphitic N (401.2 eV). [38,39] Notably, the NSC samples exhibited a single configuration state of graphitic N, indicating that the ion-exchange method using TMA is suitable for controlling the bonding state of heteroatoms. The origin of pyridinic N may be explained by the activation process described by the following equation: [40,41] C CO C 2CO n 2 n 1 The gasification of the carbon matrix by CO 2 activation to CO caused graphitic N to be exposed on the outer surface, leading to pyridinic N generation. However, the graphitic N content in NSC-act4R was significantly higher than the pyridinic N content in NSC-act4 and NSC-act4R. Although it is still unclear whether pyridinic N or graphitic N affects catalytic activity, it is generally recognized that the presence of a high amount of graphitic-N enhances the ORR catalytic activity. [19,42] The C 1s spectra of the synthesized catalysts were divided into four species at 283.8 eV (C=C), 284.6 eV (C−C), ≈285.8 eV (C−N, C−S, and C−O), and ≈287.4 eV (C=O) (Figures S2-S4, Supporting Information). [43][44][45][46][47] The presence of covalent bonds between carbon atoms and heteroatoms was confirmed in all the synthesized samples, indicating the successful doping of heteroatoms into the carbon matrix using the ion-exchange method. The S 2p spectra (Figure 2b) of the NSC samples exhibited one distinct peak at 164.4 eV, corresponding to thiophene-derived C−S−C bonds. [48,49] The absence of a peak corresponding to S(0) suggests that the S species are incorporated into the carbon matrix. Particularly, the N−S peak, which indicates a covalent bond between the S and N atoms, was not detected because of the high energy barrier for bond formation. [50,51] The evaluation of the catalytic activity of the S-doped catalysts (SC), N-doped catalysts (NC), NSCs, and NSC-act4R for the ORR was performed in 0.1 m KOH; the performance of these catalysts was compared with that of a commercial catalyst, 20 wt% Pt/C (Figure 3a). Linear sweep voltammetry (LSV) results confirmed that the co-doping of N and S increased the current density and the onset and half-wavelength potentials, suggesting that the synergistic effect of the heteroatoms enhances the catalytic activity for the ORR (Table S1, Supporting Information). The LSV curves indicated that NSC-act4R exhibits a higher electrocatalytic performance than other hetero-atom-doped samples, with a positive onset potential of 0.93 V versus the reversible hydrogen electrode (RHE) and half-wavelength potential of 0.79 V versus RHE, which is For the sample name, the roman numerals after "NSC", "act", and "R" represent the carbonization temperature, activation time, and the secondary carbonation process, respectively; b) Calculated using the Brunauer-Emmett-Teller (BET) equation.  www.advmatinterfaces.de comparable with that of the commercial Pt/C catalyst (1.01 V and 0.86 V versus RHE, respectively) ( Figure S5, Supporting Information). Furthermore, the Tafel gradient of NSC-act4R in the half-wavelength potential region (90.6 mV dec −1 ) was lower than that of the other metal-free and Pt catalysts (98.2 mV dec −1 ), indicating its higher ORR rate (Figure 3b); the lower Tafel slope value indicates that the charge transfer from the active site to the intermediate species is rapid owing to optimal adsorption and facile desorption of the intermediate. Thus, these results indicate that NSC-act4R exhibits the highest electrocatalytic activity among the previously reported catalysts (Table S2, Supporting Information).
To further investigate the ORR pathway of the NSC electrode, the amount of H 2 O 2 produced during the ORR process was calculated using a rotating ring disk electrode (RRDE). N/S/C-act-R exhibited the lowest ring current, generating 8% of the maximum H 2 O 2 produced (Figure 3c). Additionally, the electron transfer number per oxygen molecule (n) for the ORR was determined by RRDE experiments, and the results are shown in Figure 3d. The n of NSC-act4R in the diffusion-controlled potential region was 3.8-4, which is higher than those of the other NSC samples and previously reported catalysts. [52,53] These results confirm that oxygen molecules are efficiently reduced to OH − by an approximately four-electron pathway (n > 3.8), and the amount of peroxide species produced by NSC-act4R is low (<8%).
In addition to catalytic activity and H 2 O 2 production, longterm stability and methanol tolerance are crucial properties of ORR catalysts. NSC-act4R exhibited good stability, the relative current density decreased slightly by 8.3% after 7200 s, as shown in Figure S6 (Supporting Information). After the longterm stability test, NSC-act4R retained a stable relative current when exposed to methanol, whereas the current rapidly decreased for the commercial Pt/C catalyst. Thus, the NSC samples are expected to be promising alternatives to current Pt catalysts because of their high catalytic activity, excellent stability, and methanol tolerance.
To gain further insights into the ORR catalytic mechanism of the N/S-co-doped metal-free carbon materials, first-principles calculations using the density functional theory (DFT) method were conducted along with electrochemical experiments to identify the catalytic reactions and electronic structures of the heteroatom-doped carbon structures (Figure 4). The incorporation of N and S into the carbon matrix can adjust the geometric distribution and electronic state of catalytically active sites. An active site model was constructed with various N and S

www.advmatinterfaces.de
functionalities predicted based on the XPS and elemental analysis results; this model included a nondoped graphene matrix (GC), graphitic-N-doped carbon matrix (N-1 and N-2), thiophenic S-doped carbon matrix (S-1), and graphitic N-thiophenic S co-doped carbon matrices (NS-1, NS-2, NS-3, and NS-4). The effect of oxygen functional groups was not considered in the predicted model because previous studies using carbon-based materials with different oxygen contents did not exhibit major differences in the catalytic activity for the ORR. [42] The active sites of all the heteroatom-containing model structures are shown in Figure 4. Generally, carbon atoms with low charge densities are undesirable active sites for reactant adsorption. The geometric optimization of the adsorption sites of GC and oxygen molecules indicates that oxygen molecules are repulsed by the carbon matrix, thus hindering reactant adsorption. This was clearly explained by population analysis, which indicated that the basal carbon atoms in the GC have delocalized electrons, causing molecular repulsion with oxygen molecules ( Figure S7, Supporting Information). Similarly, previous studies reported that the incorporation of heteroatoms with different atomic charges and spin densities significantly influenced the ORR. [54,55] To clarify the effects of charge and electron density on oxygen adsorption, the atomic charge was further evaluated. The charge population was calculated by Mulliken population analysis for all heteroatom-doped models, and the results are shown in Figure 5 and Figures S8-S10 (Supporting Information). Positive charges favorable to the ORR were observed for carbon atoms adjacent to the N and S atoms. Furthermore, N doping increased the atomic charge of the edge-type and quaternary carbon atoms adjacent to N. In contrast, S doping decreased the atomic charge of adjacent carbon atoms. Further computational experiments were performed on the reaction mechanism using each heteroatom-incorporated model ( Figures S11-S14, Supporting Information). Typically, the elementary steps of oxygen reduction suggested by Nørskov consist of five oxygen intermediates (details are provided in the Experimental section). [56] The free energy calculations of S-1, NS-3, and NS-4 were excluded because repulsion was observed between the oxygen molecules and intermediates during the geometric optimization process in each model. The potential energy diagrams at 0 V versus RHE for the four models (NS-1, NS-2, N-1, and N-2) shown in Figures 6  and 7 indicate that the free energy change for all the elementary steps was negative, that is, the ORR occurs as a spontaneous reaction for each model. The free energy diagram at the equilibrium potential of U = 1.23 V versus RHE revealed the effect of N/S-co-doped carbon materials on the ORR. Notably, except edge-type NS-2, the adsorption energy of O* was positive in all S-doped models, whereas it was negative in all N-only models (N-1 and N-2). These results suggest that the ORR activity of N-doped catalysts is primarily limited by the desorption of *O, which is strongly bound to the carbon atom that serves as the active site, and S addition destabilizes the adsorption energy of the O* intermediate. These results were also supported by  the atomic charge values calculated using Mulliken population analysis, which showed a decrease in the atomic charge of carbon atoms adjacent to S and an increase in that adjacent to N atoms ( Figure 5). Overall, these results indicate that thiophenic S promotes charge delocalization and enhances the ORR performance of the N/S-co-doped carbon materials.
Additionally, the rate-determining step (RDS) and overpotential (η ORR ) are essential for reducing the differences between experimental and theoretical results ( Figures S15 and 8, Supporting Information). With the exception of edge-type NS-2, the RDS of the ORR was different for the models with and without S ("*OOH + e − ↔ *O + OH − " and "*O + e − ↔ *OH", respectively). These results correspond to the free energy calculation results, indicating the high adsorption energy of *O for the N-doped models, as shown in Figures 6 and 7. In this study, the free energy of the RDS (ΔG RDS ) had a negative value for each model. Based on the results, both S-doped models exhibited lower ΔG RDS values than the models without S doping. This indicates that the addition of S atoms to N-containing carbons enhances the catalytic activity for the ORR. Furthermore, NS-1 exhibited a lower ΔG RDS than N-1 without S doping. In contrast, ΔG RDS was reduced in NS-2, where S is located closer to N. Thus, the SN model with S and N coordination provides a reasonable ΔG for the N-only doped model, and the catalytic performance varies significantly with the distance between S and N atoms. Moreover, the validity of the DFT calculations was evaluated by combining the overvoltage calculations and electrochemical experimental results. As shown in Figure 8, the overpotentials of all basal-type models were lower than those of the corresponding edge-type models. The basal-type model showed an overpotential similar to the experimentally obtained overpotentials of NC and NSC (Table S3, Supporting Information), confirming the validity of the calculation results obtained in this study. Furthermore, the N/S-co-doped model (basal-type NS-2) exhibited a lower overvoltage (by 0.03 V) than the N-only model (basal-type N-1), which corroborates with the difference between the NC and NSC materials ( Figure S16, Supporting Information). Based on these results, it is concluded that the ORR of carbon materials with thiophenic S and graphitic N is caused by basal-type carbons adjacent to graphitic N, and the positions of S and N significantly enhance the catalytic performance.

Conclusion
Carbon catalysts with configuration-controlled heteroatom doping were synthesized using the interaction between TMA and an ion-exchange resin. The obtained N/S-co-doped porous carbon materials had larger specific surface areas and exhibited  www.advmatinterfaces.de higher ORR performance than the N/S-co-doped carbon catalysts reported in the literature, with an onset potential of 0.93 V versus RHE and a high electron transfer number of >3.8. The experimental results demonstrated that the synergistic effect of N and S improves the catalytic performance for the ORR. First-principles simulations revealed that the change in the charge population of carbon atoms owing to the co-doping of N and S is essential for inducing high ORR activity. Moreover, by combining experimental and DFT calculation results, key issues regarding the function of S and N in carbon materials with different predicted structures were identified. However, the type of heteroatoms that can be incorporated into the carbon material is limited. Thus, in this study, only the synergistic effects of N and S were elucidated. Further research on carbon materials with other types of heteroatoms will be conducted to determine the optimal solution for metal-free carbon-based catalysts.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.