3.2.1 Development of the ORR Mechanism

Despite decades of remarkable effort has been devoted to revealing the ORR mechanism, the nature of the catalytic active sites in metal‐free ORR catalysts is still controversial and there is an ongoing debate. Regarding the origin of the ORR activity in heteroatom‐doped carbon materials, two different mechanisms have been proposed. Of these mechanisms, some researchers claim that the introduced foreign elements in the carbon structures are the ORR active sites [14, 29, 31], as they may influence the charge density of the carbon atoms, which eventually facilitates the formation of active sites for oxygen absorption and reduction. The theoretical work carried out by Jung et al. suggests that graphitic‐N is the main ORR active sites in N‐doped graphene as it could lower the reaction barrier [45]. In addition, their studies also show that the interconversion between the graphitic‐N and the pyridinic‐N during the catalytic cycle may explain why it is hard to identify which state is the active site [45]. For example, Knights and coworkers claim that graphitic‐N is the most active species in the N‐doped graphene for the ORR [14], whereas Kim et al. found that the prepared N‐doped graphene with high content of pyridinic‐N exhibits excellent ORR performance [46]. Meanwhile, it is generally regarded that the electronegativity of the heteroatoms will influence the charge density of the carbon atoms in the sp² lattice, which eventually favors the formation of adsorption sites for oxygen and thus it is beneficial for the ORR, no matter the heteroatoms are electron‐rich nitrogen and oxygen, or electron‐deficient boron (the electronegativities of C, N, O, and B atoms are 2.55, 3.04, 3.44, and 2.04, respectively) [20, 38, 47]. However, sulfur is an exception, as the electronegativities of carbon and sulfur are very close (2.55 versus 2.58), but the S‐doped graphene also shows clearly improved ORR performance [48], indicating that complementary reaction mechanisms should be established to explain this phenomenon. Moreover, other investigations revealed that the incorporated heteroatom may not be the active center but assists to create special carbon structures, such as defects, which could be the actual ORR active sites [49]. Recent studies on defect‐promoted ORR carried out in our group is a support to the second mechanism [21–24]. It is possible that both reaction mechanisms are reasonable, as the mechanism may differ from catalyst to catalyst, and the specific catalytic sites are yet to be uncovered after more systematic and in‐depth studies.

3.2.2 A Newly Proposed Defective Catalytic Mechanism for the ORR

From the above discussions, it is obvious that heteroatom doping should not be the only approach to improve the ORR activity of metal‐free electrocatalysts, and many issues regarding heteroatom doping, especially N doping, are yet to be solved. For example, the influence of nitrogen concentration on the ORR performance of the N‐doped carbon materials is still an ongoing debate, as no unified or clear relationship can be established between the N content and the ORR activity [14, 44, 50–56]. In addition, the ORR reactive species in N‐doped carbon materials are still controversial, as some researchers state that graphitic‐N is the most active species, while whereas others suggest that pyridinic‐N is more effective for the catalytic reactions [14, 26, 45, 46, 57, 58]. More importantly, it is still elusive that whether the doped foreign atoms are the ORR active sites or the carbon atoms adjacent to the incorporated heteroatoms play more significant role, as the carbon structure is modified due to the introduction of heteroatoms [20, 49, 59]. If the carbon atoms themselves are the ORR catalytic active sites, it means that any other methods that could change the electronic environment of the carbon atoms may also be capable of enhancing the ORR performance of the carbon materials, possibly more efficient than that of the foreign atom doping.

Recently, both theoretical calculations and experimental studies carried out in our group proved that non‐doped carbon materials with unique defects could effectively catalyze the ORR, and even more efficient than the N‐doped carbons [21–24]. Theoretically, a kind of divacancy was selected to investigate the influence of defects on the ORR, which is a kind of stable defect and has been received extensive research attention during the past decade [60–66]. A representative of the divacancy in graphene is the G585 defect (denoted as G585 because this topological defect contains two pentagons and one octagonal), which is illustrated in Figure 3.1a [21]. In this work, a kind of graphene with G585 defect (G585), the perfect single‐layer graphene (G), N‐doped graphene (N‐G), and an ideal catalyst (Ideal) were chosen for the investigation in terms of energy profile. From the density functional theory (DFT) calculation results in Figure 3.1b, it can be seen that N doping could reduce the adsorption energy of oxygen significantly, but compared with the ideal catalyst, it is not favorable for the subsequent reactions, such as the reduction of the chemisorbed oxygen atoms. On the contrary, G585 defects could not only promote the adsorption of oxygen on the graphene but also make the following reactions easy to proceed, which is approaching the ideal ORR catalyst. This means that pure carbons with topological defects are promising ORR catalysts, which should be more effective than the heteroatom‐doped carbon materials, according to our theoretical studies [21]. In this study, we clearly presented the concept of defect mechanism for the ORR.

Apart from the work conducted in our group, subsequently, related investigations have also been carried out to probe the influence of defects on the ORR. For instance, Xia et al. studied the electronic structure and catalytic properties of graphene clusters with different point and line defects by DFT simulations [67]. According to their calculations, both the point and the line defects in graphene with particular electronic configurations and locations are possible to catalyze the ORR, as they could tailor the local electronic structures and charge distributions of carbon materials with sp² hybridization. For example, among the investigated point defects in graphene, only the one with pentagon rings at the zigzag edge is effective for the ORR (PZ, as shown in Figure 3.2a). In addition, defective graphene (DG) with odd number of octagon rings and fused pentagon rings line defect (GLD‐558‐01, as indicated in Figure 3.2c) also exhibits ORR activity, and the proposed active sites are situated near the graphene edges as well, indicating that the high ORR performance of the defective graphene is originated from the joint effect of the specific types of defects and the corresponding edge structures. In addition, it is revealed that the two‐electron and four‐electron transfer reactions occur at the same time on the graphene with PZ or GLD‐558‐01 defect. Figure 3.2b,d show the free energy diagrams of the four‐electron transfer process on the graphene with PZ and GLD‐558‐01 defects at different potentials, respectively. As can be seen, graphene with the GLD‐558‐01 defect could catalyze the ORR more efficiently than the PZ at a potential of 1.23 V, as the GLD‐558‐01 shows a lower overpotential for the O–O rupture. Meanwhile, the calculated free energy changes of the reactions are consistent with the previous theoretical investigations for other ORR catalysts [3, 68, 69], which also illustrates that the GLD‐558‐01 is better than the PZ to catalyze the ORR. Furthermore, it is suggested that the energy barriers of the defective graphene for the ORR are comparable with that of the Pt(111) surface [70] and other N‐doped carbon materials [49].

3.2.3 Experimental Studies on Defect‐Promoted ORR

Experimentally, we used a N‐containing porous aromatic framework (PAF‐40) as the raw material to synthesize defective carbons for the ORR. By calcining the PAF‐40 sample at high temperatures, the uniformly distributed nitrogen in the PAF‐40 can be homogeneously removed. It is natural to hypothesize that porous carbons with different nitrogen contents would be obtained by treating the PAF‐40 at different temperatures. The removal of nitrogen atoms from the carbon network is also very possible to form various defects, such as the stable G585 defect. Meanwhile, the removal of more nitrogen atoms will result in a higher density of G585 defect; thus, higher ORR activity will be achieved. As expected, the PAF‐40‐derived carbons with a range of nitrogen contents (0.56–2.10 at.%) have been synthesized. It is clearly shown that the obtained carbons with lower nitrogen contents exhibit higher ORR activity. To exclude the influence of the carbonization temperature on the ORR activity of the prepared catalysts, we performed a controlled experiment at the identical temperature to fabricate a carbon material with an even lower content of nitrogen. The resulting carbon with only 0.21 at.% nitrogen shows the best ORR activity to all the carbons with higher nitrogen contents, which should be ascribed to the newly produced defects via the removal of nitrogen. In addition, this result also supports the above theoretical predications [21]. Accordingly, we can conclude from this paper [21] that (i) the concept of defect mechanism for ORR; (ii) the confirmation of the defect type, G585 for ORR, and (iii) the method to create such defects was firstly contributed to the electrochemistry field.

However, the direct observation of the proposed defects in the PAF‐40‐derived carbon sample is very difficult due to its three‐dimensional structure, which is not suitable for characterizing by the transmission electron microscopy (TEM). In order to gain direct evidences to verify the influence of defects on the electrochemical reactions, we used a single‐layer graphene as the raw material for fabricating a kind of DG. As shown in Figure 3.3a, a simple N doping and removal method was applied for the synthesis of DG. There should be some interruptions to the pristine graphene after the introduction of the heteroatom nitrogen. Theoretically, the typical carbon defects, such as pentagons, heptagons, and octagons, are very likely to be formed after the reconstructions of the carbon atoms due to the removal of the doped nitrogen atoms from the carbon network via heat treatment [21], as schemed in Figure 3.3a. The defective nature and irregularity of the DG can also be reflected from the corresponding Raman analysis, as indicated by the highest D band to G band ratio among the resulting samples [24]. The low‐magnification TEM image shows the existence of in‐plane holes in the DG sample, possibly caused by the removal of the nitrogen dopant. Impressively, various defects can be observed directly from the DG sample via the aberration‐corrected, high‐resolution TEM (Figure 3.3b). For example, the aforementioned defects, including the pentagons, heptagons, and octagons with particular arrangements, can be clearly seen. Apart from the aforementioned 585 defect, other defects, such as 75585 and 5775, are existing in the defective graphene as well, which is a strong experimental support to our previously proposed defect mechanism [21]. According to this defective catalysis mechanism, the DG should exhibit remarkable ORR activity. As shown in Figure 3.3c, compared with the pristine graphene and the N‐doped graphene, the defective graphene shows significantly improved ORR activity, in terms of onset and half‐wave potential as well as limiting current density. To further confirm the highly efficient ORR catalytic performance of the DG, the DFT simulations were employed. It can be observed from Figure 3.3d that among all the typical defects (as schemed in Figure 3.3e–g), the most active ORR site under alkaline conditions is the edge 5‐1 defect, with the smallest activation barrier of 0.470 eV. In addition, all of the edge carbon atoms (5‐1, 585‐1, and 7557‐1) are ideal for the ORR catalysis because of their low‐activation barriers [24]. It can be concluded, both experimentally and theoretically, that particular types of defects are suitable for high‐efficient electrocatalysis, which may open a new window to design catalysts using defective materials for practical fuel cell applications. Additionally, for the first time, we extended the defect catalysis mechanism to other electrochemical reactions, such as the hydrogen evolution reaction (HER) and oxygen evolution reaction (OER) [24].

In the field of catalysis, it is well known that trace impurities may activate the reactions; for example, it is revealed that metal impurities in some of the reported highly active “metal‐free” ORR catalysts are the main contributor to the catalytic reactions [71–73]. It is, therefore, natural to suspect that defects may be the actual ORR active sites in the defective carbons, but the very tiny amount of residual nitrogen that was incorporated into the graphitic carbon network may also play an important role. In this regard, we used a kind of Zn‐enriched metal organic framework (MOF), which is completely free of nitrogen, as the starting material successfully synthesized defective carbons by removing the incorporated Zn atoms at a high temperature [22]. Interestingly, the resulting carbon (without any of the so‐called dopants or Zn, verified by the X‐ray photoelectron spectroscopy (XPS) analysis) also shows significantly enhanced ORR performance, proving that it is not the impurities (N residual) but the produced defects via the removal of the Zn atoms that promoted the ORR [22]. This study not only shows strong support to the defective mechanism proposed in our group but also confirms that in addition to the nitrogen atom, the removal of other heteroatoms in the carbon network could create the valuable defects for the ORR as well.

For the mass production of fuel cells, the cathode catalyst cost is a deciding factor. It is, therefore, of paramount importance to explore inexpensive yet efficient ORR catalysts for practical applications. To this end, we used a low‐cost and readily available activated carbon (AC) as the starting material and created the unique defects in the AC by a simple N doping and removal approach [23]. The Raman spectra of the synthesized samples in Figure 3.4a provide the convincing evidences regarding the introduction of defects into the carbon structures, as indicated by the increase in the I_D to I_G ratios. As expected, the defective activated carbon (D‐AC) shows remarkable ORR activity with a four‐electron pathway (Figure 3.4b), in terms of onset potential, half‐wave potential, as well as current density. Impressively, the D‐AC sample shows comparable ORR activity as the commercial Pt/C. To probe the influence of temperature on the ORR, we treated the high surface area AC (H‐AC) at the temperature of 1050 °C for 2 h. The results show that the obtained sample H‐AC‐1050 did not exhibit obvious ORR performance improvement (Figure 3.4b), which proves that the excellent ORR activity of the D‐AC is mainly derived from the produced defects by removing the doped nitrogen from the carbon networks. Undoubtedly, the normally regarded non‐active AC becomes very active for the ORR after this simple treatment, which is an important breakthrough for practical fuel cell applications. In addition, the D‐AC is much more stable than the commercial Pt/C and free from methanol poisoning (Figure 3.4c,d), indicating its broad application prospects and the diversity of the fuel supply [23].

The subsequent studies from other groups also supported our defect mechanism for the ORR. For example, by treating a selected single‐walled carbon nanotubes (SWCNTs) under argon and ammonia plasma atmospheres, Chen et al. investigated the ORR behavior of the SWCNTs treated at different conditions [74]. It is revealed that the newly generated carbon vacancies irradiated by the argon plasma in the resulting sample are most likely contributed to the enhanced ORR performance, which was evidenced by the increase in the carbon disorderness from the corresponding Raman spectra. In comparison, the plasma ammonia is not effective to create such kind of defects for the ORR [74]. Apparently, more in‐depth studies should be carried out to probe the catalytic mechanism of the defect‐promoted ORR, as it is difficult to distinguish the structure defects of the SWCNTs before and after the plasma treatment because of the complex configurations of the SWCNTs. Similarly, Dai and coworkers employed a facile argon plasma etch approach synthesized edge/defect‐rich graphene, which shows greatly improved ORR activity compared with the pristine graphene [75]. According to the high‐resolution TEM characterizations, the argon‐plasma‐treated graphene (P‐G) retained the original structure of the pristine graphene (G), but many nano‐sized holes can be clearly observed. The Raman spectra of the pristine graphene and the P‐G further confirmed the defective nature of the P‐G, as indicated by the increased I_D to I_G ratio. Remarkably, the electrochemical test results show that the P‐G demonstrates considerably enhanced ORR performance compared with the pristine graphene, in terms of onset potential and limit current density. Meanwhile, they also used CNTs and graphite as the raw materials to produce defective ORR catalysts under argon plasma irradiations. The resulting defective samples exhibit increased ORR activity as well, revealing the universality of the plasma treatment method for producing effective defects toward the ORR [75]. This investigation fully proves the crucial role of edge/defect in carbon materials for electrocatalysis.

In the meantime, Hu and coworkers synthesized a kind of carbon nanocage full of intrinsic carbon defects but without any dopants [76]. As schemed in Figure 3.5a, the carbon cage is enriched with different types of defects, typical examples include the pentagon defects created by the corner disclinations, the edge defects produced by the surficial broken fringes, and the holes in the shell. The Raman spectra in Figure 3.5b further reveal the defective characters of the samples prepared under different temperatures, especially the one synthesized at 700 °C (CNC700) exhibits the highest I_D:I_G ratio, implying that this sample has the highest defect concentration among them. It is found that the CNC700 sample also shows the best ORR performance (Figure 3.5c), which is even superior to some of the reported N‐doped carbon catalysts. This fully demonstrates the great contribution of the carbon defects on the ORR, agreeing well with the Raman results. The corresponding DFT calculations reveal that, compared with other types of defects, the pentagon and zigzag edge defects are more crucial to the ORR in the carbon nanocages (Figure 3.5d), further revealing that certain types of defects, such as the pentagon and zigzag edge defects, are the main contributors for the ORR in defective carbon materials [76].

3.2.4 Edge Defects and Defects/Dopants Copromoted ORR

The investigations show that the defective edge carbon atoms may be crucial to the electrocatalytic reactions, as the edges could provide abundant absorption and reaction sites [77–81]. For instance, by treating two kinds of CNTs with different proportions of edge carbon atoms (fishbone CNTs (F‐CNTs) and parallel CNTs (P‐CNTs)) under various conditions, Peng et al. found that the F‐CNTs with as high as 31.7% edge carbons exhibited much higher ORR activity than the P‐CNTs that only have 1.8% edge carbons [82]. The corresponding DFT simulations indicate that the edge carbons more easily adsorb oxygen than their basal counterparts, as the edge defective sites could make the adjacent carbon atoms carry more positive charge due to the electron transfer between them [82]. The positive charge is able to promote the adsorption of oxygen, which is similar to the N doping for the ORR [20, 83]. To verify the effect of edge defective sites on the ORR, we synthesized a 3D web‐like, horizontally aligned carbon nanotube‐graphene (CNT‐G) hybrid by a facile plasma‐assisted method, which shows abundant active edge sites. The sample CNT‐G demonstrates more defects than that of the N‐doped graphene, as evidenced by the Raman test. The defective CNT‐G is proved to be a more efficient material to anchor and disperse the Pt to boost the ORR [84].

In addition to the pioneering work conducted in our group to probe the influence of the unique carbon defects on the ORR, Zhang and coworkers also studied how the topological defects correlated with the enhanced electrochemical activity of metal‐free catalysts [85, 86]. For example, by using the in‐situ‐formed Mg(OH)₂ precipitation as a template, sticky rice and melamine as the carbon and nitrogen sources, respectively, they synthesized a kind of N‐doped graphene mesh (denoted as NGM, nitrogen content is identified to be 7.60 at.% by the XPS test) enriched with abundant edges and topological defects [85]. A reference sample graphene mesh (GM) without adding the nitrogen source melamine was prepared as well. It is revealed that the graphene flakes are featured with nano‐sized holes over the plane, which were inherited from the porous structures of the template MgO during the carbonization process. The defective nature of the NGM can be gleaned from the corresponding Raman spectra, as indicated by the higher I_D:I_G ratio [85]. For further comparison, they fabricated a N‐doped graphene with fewer in‐plane holes and edges but similar nitrogen content (7.48 at.%) compared with that of the NGM. As expected, the defective NGM sample shows the highest ORR activity among the resulting catalysts, which is comparable with that of the commercial Pt/C. The experimental results demonstrate that both the nitrogen dopant and the defective structure are essential to the remarkable ORR performance of the as‐synthesized NGM sample. Furthermore, the DFT calculations were applied to verify and confirm the origin of the ORR activity of the NGM. It is revealed that the nitrogen‐doping‐induced sites that are close to the edge show much lower overpotentials, indicating the crucial roles of edge defects on the ORR. In addition, it is suggested from their DFT simulations that the carbon rings with different configurations could exhibit significantly different overpotentials toward the ORR. For example, a pure carbon arrangement with the five‐carbon ring near the seven‐carbon ring (C5 + 7) shows the optimal adsorption for the oxygen intermediates; thus, the overpotential of the C5 + 7 ORR sites can be decreased to as low as 0.14 V. This work is beneficial for further understanding and uncovering the complicated reaction mechanism of the ORR, especially the great contributions from particular types of defects. Meanwhile, in order to unveil the active sites of the ORR, Terakura and coworkers applied the first‐principles‐calculations probed the ORR mechanisms [87]. It is found that the Stone–Wales defects are pivotal to form the active site to boost the ORR of the N‐doped carbon catalysts, which provides a feasible approach to fabricate highly efficient electrocatalysts, by combining the unique carbon defects and selected dopants.

3.2.5 The Effect of Defect Density on Electrocatalysis

As discussed above, certain types of topological carbon defects are the actual active sites for the electrochemical reactions. Therefore, extended investigations should be conducted to probe the relationship between defects and the electrocatalytic properties of carbon materials. For example, to quantitatively study the density of defect on the electrocatalysis, Ren and coworkers used a single‐layer graphene as the precursor and created defects on the graphene sheet by Ar plasma irradiation [88]. By utilizing scanning electrochemical microscopy (SECM) and Raman spectroscopy in conjunction with ab initio simulations, they investigated the relationship between the electrochemical activity and the defect density of the graphene. It is found that the optimized electrochemical performance of the single‐layer graphene can be reached at a moderate defect density, by balancing the heterogeneous electron transfer (HET) rate and the conductivity. In order to achieve precise control of the experimental conditions, they created patterns with different densities of vacancy defects on the same single‐layer graphene sheet. It can be seen from the Raman mapping images of the D band in Figure 3.6a that different densities of defects were successfully created on the 100 × 100 μm² squares by adjusting the Ar⁺ plasma irradiation time and ion dose, as indicated by the brightness of the corresponding patterns. Of this, region A is the pristine single‐layer graphene showing a dark color because of the lack of the D band. From Figure 3.6a, it can also be observed that pattern E exhibits the highest defect density. Figure 3.6b shows the SECM images of the relevant defective graphene regions presented in Figure 3.6a. Apparently, all of the normalized feedback currents of the defective patterns are higher than those of the pristine graphene, suggesting the crucial role of defects on the electrochemical reactions. The differences of the feedback currents of the defective patterns also reveal that the concentration of defects affects the electrocatalytic activity of the defective graphene significantly. The two‐stage SECM results of the defective graphene agree well with the corresponding Raman D‐band‐mapping images, implying that the density of defect of the graphene has strong relationship with the electrochemical performance. Meanwhile, they used the approaching curve measurement to quantify the HET rate of the prepared graphene samples. The testing results are presented in Figure 3.6c, which is in good accordance with the corresponding SECM images shown in Figure 3.6b. This investigation shows that graphene with different defect densities can be successfully produced by tuning the Ar⁺ irradiation time and ion dose. Figure 3.6d–f display the microscopic models with different concentrations of defects. As can be seen, when the mean distance between defects L_D is greater than 6 nm, between 6 and 2 nm, less than 2 nm, the corresponding defective graphene is with low, moderate, and high defect density, respectively. In particular, for the graphene with moderate defect density, it is fully activated, but at the same time maintained the structure integrity; thus, the standard HET rate constant k⁰ reaches the peak value at such defect concentration (Figure 3.6g), which is in good accordance with the results in Figure 3.6h, showing that the optimal distance between defects is around 2 nm [88]. For example, at a defect density of (7.39 ± 0.58) × 10¹² cm⁻² (L_D = 2.10 ± 0.10 nm), the k⁰ of the corresponding defective graphene is 50‐fold faster than that of the pristine graphene. It is, therefore, concluded that graphene with a moderate concentration of defects should show the highest electrochemical activity, as it is not only fully activated but also retained the excellent conductivity of the graphene. However, it is still not easy to precisely control the defect density of graphene, and the homogeneous creation of defects on the graphene is another critical issue to be solved.