Enabling the conversion of chemical energy of fuels directly into electricity without combustion, fuel cells are arousing great interest in both academia and industry. A typical case is the proton exchange membrane fuel cell (PEMFC), already commercialized by automobile giants. For mass popularization, however, three major criteria must be balanced: performance, durability and cost. The electrocatalysts used in both the anode and cathode are the kernel of PEMFCs, being essential for efficient operation. First in the firing-line is the oxygen reduction reaction (ORR) at the cathode, which is normally very sluggish: over six orders of magnitude slower than the anode hydrogen oxidation reaction (HOR) [1]. Thus, considerable efforts have been made to improve the cathode ORR. Identifying the main active sites is key to the design of optimum materials for enhanced ORR. Considering the complex balance of preparation, performance and cost, the active sites of metal-nitrogen-carbon (M-N-C) catalysts are particularly promising. Coupled with the single metal atom (SMA) catalysts [2-5], two excellent M-N-C catalysts were recently reported [6, 7]. New insights were thereby gained into the delicate architecture of carbon-based SMA catalysts for ORR.

One of the best ORR catalysts is known to be Pt. Intensive efforts have focused on downsizing and alloying of Pt to maximize the exposure of active sites. A recent work reported that Pt single atoms supported on nitrogen (N)-doped carbon black exhibit excellent ORR activity and fuel cell performance [6]. Pt supported on N-doped carbon (Pt₁-N/C) was found to consist entirely of highly dispersed individual Pt atoms, which differs from Pt supported on pure carbon (Pt₁/C), where some Pt nanoparticles (NPs) and clusters were also observed. This indicates that doped N atoms can anchor single Pt atoms. Such anchoring prevents their aggregation into NPs. Additionally, the strong interaction between N and Pt prevents the oxidation of Pt atoms in air. Thus, the ORR activity of Pt₁-N/C was demonstrated to be superior to Pt₁/C and N-doped carbon, and close to commercial Pt/C. Particularly, Pt₁-N/C exhibits high tolerance to methanol or carbon monoxide, attributed to its stronger adsorption of O₂ than CO according to theoretical calculations. Meanwhile, a density functional theory (DFT) calculation confirmed that Pt₁ atoms can be strongly trapped by pyridinic N sites, which favors the dispersion of Pt single atoms on carbon and the durability of Pt₁-N/C for ORR. The mechanism of ORR catalyzed by this material was analyzed (Fig. 1(a)). The adsorption mode of O₂ was found to be a side-on configuration on the graphene-supported single Pt atom (g-Pt₁-N/C). The O–O bond on g-Pt₁-N/C is longer than in O₂ adsorbed on bulk Pt NP-based active sites (g-Pt₁/C), indicating the relative ease of breaking the O–O bond on Pt single atoms compared with the bulk. Free energy calculations reveal that the ORR activity of g-Pt₁-N/C is close to conventional Pt/C catalysts, attributed to the synergistic effect between Pt single atom centers and pyridinic N. The strong oxidation resistance of g-Pt₁-N/C was also demonstrated by calculating the formation energies and Bader charges of oxidized g-Pt₁ sites, g-Pt₁-N sites and the entire system. Moreover, the small adsorption energy and short O–O bond length indicate that O₂ molecules are difficult to activate on oxidized g-Pt₁ sites, which explains the poor ORR activity of Pt₁/C, inferior to pure N-doped carbon.

After comprehensive experimental characterization and theoretical calculations, it was ultimately concluded that single Pt atom anchored on single pyridinic N was the main active site for ORR [6]. Specifically, the presence of single Pt atoms as the major components was also experimentally identified in pure carbon supports (Pt₁/C) [6]. Note that pure carbon black contains intrinsic defects, and N-doping probably enables the creation of more defects on carbon. Another scenario, which we propose here, is that defects on carbon may be true anchoring sites for Pt single atoms, in addition to the previously proposed N-doped sites.

Another recent report [7] reinforces the viewpoint that defects in carbon can anchor SMAs, thus promoting ORR. An atomic-level insight into M-C-N catalysts, with direct visualization of the proposed ORR active sites (N-coordinated Fe atoms (FeN₄) embedded in carbon), was obtained by low-voltage aberration-corrected scanning transmission electron microscopy (AC-STEM) [7]. Individual Fe atoms embedded in a few-layer graphene sheet were observed by high-angle annular dark-field (HAADF)-STEM (Fig. 1(b)). Electron energy-loss spectroscopy (EELS) demonstrated the presence of N around single Fe atoms in the bulk regions, as shown in Fig. 1(c), where the N K-edge only appears around the Fe atom sites and is absent from the Fe-free pure graphene regions. The quantitative Fe/N ratio acquired from several sites implied an average composition of FeN₄ according to the EELS data. Because of the hierarchical pore structure, an abundance of graphite (002) basal-plane edges and steps was found in the catalyst. The single Fe atoms observed by AC-STEM were preferentially positioned at exposed edges and steps, where N coordination can hardly be verified with STEM-EELS. Nevertheless, the edge-hosted FeN₄ structures satisfy the indirect evidence obtained by M ssbauer, X-ray absorption fine structure and X-ray photoelectron spectroscopies. The concentration of edge-hosted structures is higher than the bulk-hosted sites. Quantum chemistry calculations demonstrated that the FeN₄ sites follow different ORR pathways, in terms of their hosting sites, at the edge or in the bulk of graphene. Since the edge-hosted FeN₄ sites in a fuel cell can be spontaneously ligated by OH^-, the highly ORR-active structure gives credit to the key involvement of edge-hosted FeN₄ sites [7].

In fact, since most Fe atoms are positioned at the edges and steps, whether they are only associated with N-doped sites, remains an open question. Undoubtedly, the edges and steps in carbon-based materials are usually cataloged as defects. Since the Fe-N ligands cannot be detected directly in the edges and steps [7], another likely possibility cannot be excluded: single Fe atoms may be anchored by various other defects, such as vacancies, dislocations and grain boundaries (Fig. 2), rather than the proposed N sites in carbon.

Taking all these inferences into account, it is completely possible that various defects, rather than the only candidate of N-doped sites, are true anchoring sites for SMAs. That is, grain boundaries, dislocations and other point defects, such as B, P, S, and F atoms doped in carbon, can also serve as anchoring sites for SMAs, as shown in Fig. 2. The defects may induce variations of bond length, valence state, and modulation of the surface electronic structure of neighboring carbon atoms, and consequently enable trapping of SMAs. Careful synthesis and the ability to discern various defects are undoubtedly crucial for preparing SMA-anchoring materials. The suggested detection methods for each defect are shown at the bottom of Fig. 2 accordingly. The delicate design of defects on carbon supports probably presents a general route to SMA anchoring. If this assumption holds, it should provide a new approach to the preparation of SMA catalysts for improved ORR performance. In addition, it remains challenging to identify various crucial factors in the mechanism of carbon-based SMA catalysts for 4e^- ORR (Fig. 2), as other possible active sites have also been proposed [8]. Typically, carbon atoms with Lewis basicity adjacent to pyridinic N, instead of the N atoms themselves, turn out to be ORR active sites in metal-free N-doped carbon materials [9]. Defects in nanocarbons have also been discerned as playing an active role in ORR [10-12]. Therefore, the authentic mechanism of active sites in carbon-based ORR catalysts are not yet clear. We have outlined a new viewpoint, that is, defect-anchored SMAs can serve as active centers in carbon for ORR. Their occurrence probably accounts for the excellent ORR activity of such materials. The microporous nature of zeolites may contribute to the generality of SMA catalysts [13]. In fact, most such catalysts are somewhat versatile, catalyzing not just one reaction but several. Thus, in our view, the use of defects in carbon to trap SMAs for enhanced ORR activity may become a universal strategy in electrocatalysis.

In summary, two excellent recent works by Liu et al. [6] and Chung et al. [7] demonstrated that single Pt or Fe atoms can be effectively anchored on N-doped carbon. These carbon-based single atom catalysts exhibit excellent ORR performance, which is attributed to the active centers: SMAs, anchored by N dopants. Unfortunately, two facts were not underscored: some SMAs were also found on carbon free of N-doping and these catalysts showed ORR activity as well; most single atoms tend to be sited on the defect-rich edges or steps of the carbon support. This indicates that N-doped sites are not the only candidates for anchoring SMAs. Inspired by the advanced characterization in these two reports [6, 7], we put forward that various defects, not just N-doped sites, may be the anchoring sites for SMAs. Thus, cost-effective single atom catalysts for excellent ORR can be probably obtained by delicate design of carbon defects.

Acknowledgments

Support by the Jilin Province/Jilin University co-Construction Project-Funds for New Materials (SXGJSF2017-3, Branch-2/440050316A36), the National Key R&D Program of China (2016YFA0200400), the NSFC (51372095), the Program for JLU Science and Technology Innovative Research Team (JLUSTIRT), “Double-First Class” Discipline for Materials Science & Engineering, and the Special Funding for Academic Leaders are greatly acknowledged.