Highly Active and Durable Metal‐Free Carbon Catalysts for Anion‐Exchange Membrane Fuel Cells

The development of highly active and durable platinum‐free oxygen reduction reaction (ORR) catalysts is of vital importance for the practical application of anion‐exchange membrane fuel cells (AEMFCs). Herein, a metal‐free carbon catalyst (marked as NDPC‐1000) with a graphitic N‐regulating defect structure is specifically designed and developed for AEMFCs by integrating theoretical calculations and experiments. Density functional theory calculations first reveal that the graphitic N can tailor the charge density of pentagon and armchair defects to reach the top of the adsorption energy‐activity volcano plot, while the enhanced durability is attributed to the high dissociation energy of the CN covalent bond. Under this guidance, the synthesized NDPC‐1000 demonstrates its high ORR activity and durability in alkaline media. With H2/O2 reacting gases, the AEMFC with this catalyst as the cathode delivers a peak power density of 913 mW cm−2. Unprecedented fuel cell durability is verified via continuous operation over 100 h at 0.25 A cm−2 with only a voltage decay of ≈25%, which is the greatest among all reported metal‐free‐based AEMFCs. Here a theory‐guided experiment strategy is provided for the development of high‐performance and durable ORR catalysts for AEMFCs.


Introduction
Hydrogen fuel cells, with a key advantage of zero emission, have been widely regarded as the most ideal candidate for their unsatisfied stability. For instance, the most encouraging PEMFCs equipped with Fe-SACs as cathodes generally cannot last 100 h with smooth and consistent operation accompanied by more than 50% current density attenuation. [11] As is well known, the bonding strength of bridged components is the momentous determinant in studying material stability. [12] To a large extent, the weak bond strength between the metal center and surrounding N induces the demetallation of metal centers in SACs because their metal active centers are anchored on the carbon support via coordination bonds. [13] In the field of porous polymer design, covalent bonds (such as C-N or C-C) are stronger than coordination bonds, which empowers the derived covalent organic frameworks (COFs) to inherit more rigid structure and better stability than metal-organic frameworks (MOFs). [14] Triggered by this, constructing metal-free catalysts only containing covalent bonds is expected to reinforce the intrinsic durability of electrocatalysts.
The uneven distribution of charge triggered by N-doping, edges, holes, or topological defects endows the metal-free N-doped carbon catalysts with comparable or even superior ORR activity than that of benchmarking Pt/C in alkaline medium. [15] However, the reports concerning metal-free catalyst-based AEMFCs is still rare. For instance, nitrogen-doped carbide-derived carbon/carbon nanotube composites were utilized as the AEMFC cathode and the resulting device exhibited a peak power density (PPD) of 310 mW cm −2 under H 2 -O 2 . [4c] Moreover, the AEMFC based on multiheteroatom-doped defect-enriched carbon nanotubes as the cathode displayed a PPD of 250 mW cm −2 under H 2 -O 2 . [16] Unfortunately, the PPDs of these reported AEMFCs still has a large gap compared to the Fe-N-C-based catalysts, not to mention the need to explore the stability of these catalysts. Indeed, little or no previous literature has explored the stability of metal-free-based AEMFCs. [10,16,17] In this situation, exploiting the mechanisms that affect the activity and stability of metal-free-based AEMFCs is urgently needed. Increasing the degree of graphitization and coupling multiple active sites, e.g., graphitic nitrogen (G-N) with defects, is a possibility in addressing the current bottleneck to achieve long-term AEMFC operation.
Inspired by the above, we first adopted density functional theory (DFT) calculations to forecast the synergetic catalytic mechanism of G-N and defects in carbon substrate. The results reveal that the graphitic nitrogen-connected pentagon and armchair (AC) structures optimize the intermediate adsorption thus enhancing the ORR; meanwhile, the C-N bond with high dissociation energy empowers its excellent stability. Based on theoretical predictions, the assembled AEMFC with a prepared NDPC-1000 as the cathode achieves a remarkable PPD of 913 mW cm −2 with H 2 /O 2 fuel gas. More importantly, this AEMFC can operate at 0.25 A cm −2 for 100 h with a voltage decay of only 25% as well as 60 h with a slight voltage decay of 22% at 0.5 A cm −2 in a H 2 /O 2 system, ranking top among all reported metal-free-based AEMFCs.

Theoretical Forecast
DFT calculations were adopted to investigate the synergistic mechanism of the G-N and defect in the carbon substrate. First, the possible defect types generated by removing the unstable pyrrolic N (Pr-N) and pyridinic N (Pd-N) were explored. The results indicate that three defect types of the Armchair, Pentagon 1#, and Pentagon 2# can be formed by removing the Pr-N, Pd-N1, and Pd-N2, respectively (Figure 1). Compared to Pentagon 2#, the Armchair and Pentagon 1# defects are more easily generated due to their smaller formation energy, which is consistent with the later X-ray photoelectron spectroscopy (XPS) results (a certain amount of Pd-N still be remained after pyrolysis) and reported literature from Yao and Dai et al. [18] Therefore, two main defect types of AC and Pentagon 1# (C5) were selected. Since the adjacent local structure provides an important effect on the electronic structure tailoring of materials, [19] we only consider the structures of G-N connected to the defects.
Afterward, we constructed six models with nine active sites to investigate the synergistic effects of the G-N with AC and C5 defects, respectively (Figure 2a). The free energy diagrams of different active sites ( Figure 2b) clearly reveal that the introduction of G-N can significantly improve the ORR performance. According to the Sabatier principle, [20] a value of free energy approaching 0 eV is more suitable for the reaction, where positive free energy represents that it is difficult for intermediate adsorption, while negative free energy means there will be a challenge of intermediate desorption. Obviously, the AC-1# and AC-2# present inferior ORR activity due to the difficult adsorption (+1.287 and +1.062 eV, Figure 2b), but the free energy of the identical sites significantly decreases (+0.145 and +0.520 eV, Table S1, Supporting Information) after G-N regulation. In particular, the pattern of OOH* adsorption has altered toward AC-2# ( Figure S1, Supporting Information), that is from physical adsorption (AC-2#) to chemical adsorption (AC-N2-2#) on account of G-N regulation, and the bond length shortens from 3.069 to 1.490 Å (Table S2, Supporting Information). This phenomenon is also proven by the crystal orbital Hamiltonian population theory, [21] as shown in Figure S2 in the Supporting Information, where AC-N2-2#-OOH* possesses obvious bonding and antibonding states (i.e., chemical adsorption), but the AC-2#-OOH* does not (i.e., physical adsorption). On the contrary, the C5 defect bespeaks poor ORR performance due to difficult desorption (C5, −0.910 eV), but it transfers into easier desorption (C5-N1, +0.510 eV; C5-N2, −0.446 eV) upon the G-N regulation. As a result, the onset potential of the active sites is improved obviously through G-N regulation (Table S1, Supporting Information). It is 0.710 V for AC-N2-2#, 0.168 V for AC-2#, 0.784 V for C5-N2, and 0.320 V for C5.
The above results indicate that G-N regulation can efficiently tailor the activity of different defects in ORR process. After the G-N regulation, the activity of both AC and C5 defects climb to the volcano top in Figure 2c, but they have opposite optimization mechanisms. With respect to AC defect, the G-N regulation increases the adsorption strength, but decreases the adsorption strength for C5 defect. This phenomenon can be explained by the charge attraction ability of G-N and defect sites. The Bader charge data (Table S3, Supporting Information) show that the number of electrons on the active site of the AC defect is reduced (AC-1#, −0.421 e; AC-2#, −0.406 e) upon G-N regulation, but it is increased on C5 defect (C5, +0.097 e). That is to say, G-N regulation enhances the activity of AC configuration by withdrawing electrons. In parallel, it promotes the intermediate desorption of a strong electron-withdrawing C5 species via electron donation.
To explore the inherent stability for G-N-regulating defect configuration, the thermodynamic stabilities for two catalysts (plus Fe-N 4 model) were evaluated by the dissociation energy (E dis ). [22] The computational method is detailed in the Supporting Information. Figures S3 and S4 in the Supporting Information illustrate the dissociation energies of multiple sites for C5-N2 and AC-N2, in which the edge carbon atom with pseudo-hydrogen (C edge ) is more easily dissociated than other sites due to its lower E dis ( Figure S4b,c, Supporting Information). Furthermore, the E dis of C edge on C5-N2 is higher than that on AC-N2, indicating that it is more stable for C5-N2. Thus, the C5-N2 is presumably an active center. In contrast, the E dis of C edge on C5-N2 is significantly higher than that of Fe in FeN 4 catalysts (Figure 2d), which may be attributed to the fact that the covalent bond in the C5-N2 configuration is apparently more stable than the coordination bond in the FeN 4 structure.

Synthesis and Characterizations of NDPC-X
In light of the theoretical prediction, we adopted MgCl 2 6H 2 O as a hard template and adenine as carbon and nitrogen sources to synthesize a N-doped porous carbon (mark as NDPC-X, X: carbonization temperature) catalyst with a hierarchical pore structure. The schematic diagram of the NDPC synthesis process is depicted in Figure 3a. First, the uniformly mixed MgCl 2 6H 2 O and adenine were pyrolyzed under Ar atmosphere at 900 °C. Subsequently, the pre-catalyst (Pre-NDPC-900) was obtained via acid etching and drying process. Ultimately, Pre-NDPC-900 was further annealed at X °C to obtain the target product of nitrogen-doped porous carbon NDPC-X (X: 700-1100).
Transmission electron microscopy (TEM) images of a series of NDPC-X samples all exhibit 3D staggered nanosheet structures with rich holes and wrinkles that provide more edge defects (Figure 3b,c and Figure S5, Supporting Information). The uniform distribution of elemental N on the carbon support was verified by energy dispersive X-ray spectroscopy (EDS) elemental mappings (Figure 3d-g), which complies with the XPS result clarifying the presence of C, N, and O elements in NDPC-1000 ( Figure S6 and Table S4, Supporting Information). The oxygen of NDPC-1000 mainly comes from the surface oxidation upon exposure to air ( Figure S7, Supporting Information). Furthermore, the aberration-corrected scanning TEM (ac-STEM) images also present AC and Pentagon 1# (C5) defect types for NDPC-1000 (Figure 3h,i), echoing the above DFT configuration.
The power X-ray diffraction patterns of representative NDPC-X samples display two broad peaks in the range of 25-30° and 40-45°, corresponding to the graphitized carbon, where no peaks from metal-based species were detected ( Figure S8, Supporting Information). Inductively coupled plasma optical emission spectrometry data demonstrate that there is no metal residual for the NDPC-1000 sample accompanied by negligible residual Mg element content (≈0.014 wt%, Table S5, Supporting Information) that has been verified to be insufficient to promote the ORR reaction. [23] The NDPC-1000 possesses a large Brunauer-Emmett-Teller (BET) surface area of 2446 m 2 g −1 and pore volume of 2.62 cm 3 g −1 , along with a rich hierarchical pore structure including micropores, mesopores, and macropores, which would play a key role in strengthening the mass transfer rate in AEMFCs ( Figures S9 and S10, Supporting Information). In addition, other NDPC-1000 samples have similar BET surface area, indicating that the hard template in the primary pyrolysis mainly brings rich defect porous morphology, while the transformation of N species occurs in the second carbonization.

ORR Activity of NDPC-X
The ORR electrocatalytic performance of the NDPC-X was evaluated by cycle voltammetry (CV) and linear sweep voltammetry (LSV) in 0.1 m KOH electrolyte. As depicted in Figure S11 in the Supporting Information, only featureless CV curves are discovered for NDPC-1000 in N 2 -saturated KOH solution, while the reduction peak is observed in O 2 -saturated electrolyte, implying its high activity toward ORR. As expected, high onset potential (high onset potential (E on ) of 0.97 V versus reversible hydrogen electrode (vs RHE)) and half-wave potential (E 1/2 of 0.87 V vs RHE) are achieved for NDPC-1000 (Figure 4a,b), which is better than that of benchmark 20% Pt/C (E 1/2 of 0.86 V vs RHE). Besides, as shown in Figure 4b, the kinetic current density (J k , 47.1 mA cm −2 at 0.8 V) is also higher than 20% Pt/C (28.1 mA cm −2 at 0.8 V). Meanwhile, NDPC-1000 possesses the best ORR activity among a series of NDPC-X (700-1100) samples, illustrating that carbonization temperature is also a linchpin optimization factor used to regulate the type of active sites. The ORR activity of NDPC-1000 exceeds most metal-free electrocatalysts reported to date (e.g., E 1/2 : 0.76 V of N, S@ C M -1000, [24] 0.83 V of P-CD/G(900), [25] 0.84 V of B,N-carbon, [26] 0.86 V of N-doped carbon defects (NDC), [7a] 0.843 V of pentagon defect-rich carbon (PD/N-C), [27] 0.83 V of nanotube-like porphyrin-based conjugated microporous polymers derived porous carbon (TPP-CMP), [28] etc.). The detail comparison is summarized in Table S6 in the Supporting Information.
In addition, NDPC-1000 displays the smallest Tafel slope of 80.2 mV dec −1 among all reference samples (Figure 4c), illustrating the faster ORR kinetics. The ORR selectivity in terms of peroxide yield and electron transfer number (n) of NDPC-1000 were evaluated using a rotating ring disk electrode. The sample has a low HO 2 − yield of less than 2.5% and an n value between 3.8 and 4 in the potential scope of 0.2-0.7 V, vs RHE, similar to the n (approximately average 3.97) calculated by the K-L equation (Figure 4d and Figure S12, Supporting Information). This observation reveals a four-electron-dominated ORR pathway for NDPC-1000. The electrochemical active surface area was evaluated by double-layer capacitance (C dl ). As shown in Figure S13 in the Supporting Information, NDPC-1000 owns the largest C dl value of 21.6 mF cm −2 compared to other samples, proving the full exposure and high utilization of active sites. We also performed chronoamperometry measurements to monitor the stability for both NDPC-1000 and 20% Pt/C at a fixed 0.6 V vs RHE. As displayed in Figure 4e, a 95% current retention was obtained for NDPC-1000 after 12 consecutive hours, significantly higher than the 83% current retention of Pt/C in the same manner. Moreover, the stability testing of NDPC-1000 was performed based on the accelerated stress test (AST) protocol. After 5000 potential cycles within a potential range of 0.6-1.0 V in the O 2 -saturated 0.1 m KOH electrolyte, the almost imperceptible attenuation (≈7 mV) toward E 1/2 (Figure 4f) was observed. This outcome further shows the preeminent stability of NDPC-1000.

Active Centers Identified by Experiment
To further identify active sites by experiment, Raman spectroscopy, XPS, and ac-STEM characterizations were carried out. The D peak is attributed to the disordered carbon with defect while the G peak is correlated with the sp 2 -hybridized carbon in Raman spectroscopy. As depicted in Figure 5a, the high values of the D band to G band, I D /I G , of 1.43-1.49 with minor fluctuation are obtained for a series of NDPC-X samples, indicating all samples are rich in defects, which is line with the TEM results (Figure 3c). The abundant defects mainly originate from the key role of the hard template and the significant reduction of N content in the first carbonization process. Several previous literature studies report that the defects are the active centers for ORR. [27,29] In this work, however, we did not discover the close correlation of ORR performance (represented by the E 1/2 ) and the defects (represented by the ratio of I D /I G , Figure S14, Supporting Information). As shown in Figure 5b, the high-resolution N 1s XPS spectra are deconvoluted into four N-bonding species, including Pd-N (398.34 eV), Pr-N, 400.1 eV, G-N (401.2 eV), and oxidized N (O-N, 404 eV), respectively. As the temperature increased, the thermochemically unstable Pr-N and Pd-N are reduced gradually along with a partial conversion to relatively stable G-N (Table S7, Supporting Information). In general, Pd-N is deemed as an ORR active site; [30] however, no direct correlation is identified between ORR activity and Pd-N content (Figure 5c). Instead, the ORR reactivity is closely related to G-N content (Figure 5d). A previous study demonstrates that the G-N doping in the basal plane carbon matrix cannot enhance ORR activity. [31] Therefore, we believe that the synergistic effect of the G-N and defects is the origin of the high ORR performance of as-synthesized metal-free catalysts, appropriately validating the DFT outcome.

AEMFC Performance with NDPC-1000 as the Cathode
The performance of the NDPC-1000 catalyst-based AEMFC was thoroughly evaluated (Figure 6a), where the NDPC-1000 (0.5 mg cm −2 ) was used as a cathode and PtRu/C (0.4 mg PtRu cm −2 ) as an anode. The AEMFC with 0.8 mg cm −2 single atom Fe on N-doped carbon (SA-Fe/NG) sample (it exhibits the best AEMFC performance via optimizing different loadings of catalysts, Figure S15, Supporting Information) as a comparison, single-atom Fe catalyst (SA-Fe/NG) in the cathode was synthesized by the same method as in our previous publication [8b] and the characterization is shown in Figures S16 and S17 in the Supporting Information. As depicted in Figure 5b, the NDPC-1000 based AEMFC delivers a current density of 261 mA cm −2 at 0.8 V higher than that of SA-Fe/ NG (180 mA cm −2 at 0.8 V) under H 2 /O 2 system, indicating its rapid kinetics. Impressively, its PPD was 913 mW cm −2 , comparable to SA-Fe/NG-based AEMFC (986 mW cm −2 ), which is the highest performer among all the metal-free-based AEMFCs until now (Figure 6b), even if still does not outperform the Pt/C-based one (1.52 W cm −2 in Figure S18, Supporting Information). The performance of the NDPC-1000 based AEMFC is even superior to most NPM catalysts as cathodes, e.g., PPD: Cobalt-and N-doped carbide-derived carbon/carbon nanotube (Co-N-CDC/CNT) (577 mW cm −2 ), [32] Fe-N-C (450 mW cm −2 ), [33] Fe salt-containing polyacrylonitrile fiber with ionic liquid (IL) additive derived pyrolysis sample (Fe/IL-PAN-A1000) (289 mW cm −2 ), [34] organic molecules derived Fe-N-C (Fe-NMG) (218 mW cm −2 ), [35] Co@G/C_600 (412 mW cm −2 ), [36] Ce and Fe on polypyrrole nanowire derived pyrolysis sample (Ce/Fe-NCNW) (496 mW cm −2 ), [37] and Co-NC (271 mW cm −2 ). [38] The detailed information is listed in Figure 6c and Table S8 in the Supporting Information. Meanwhile, the PPD of the NDPC-1000-based AEMFC in H 2 /air gas flow reaches 597 mW cm −2 as well as PPDs at different backpressures, as shown in Figure S19 in the Supporting Information. Besides its high PPD, the excellent stability of the NDPC-1000-based AEMFC was demonstrated in H 2 /O 2 flow, with a voltage loss of 25% at 250 mA cm −2 after a 100 h continuous operation, while the 60 h continuous operation at 500 mA cm −2 was accompanied with a voltage loss of 22% (Figure 6d and Figure S20, Supporting Information). The initial faster attenuation with voltage loss of 9% through a short time about 3 h at 500 mA cm −2 ( Figure S20, Supporting Information), where an analogous situation also occurs in the stability measurement at 250 mA cm −2 (Figure 5d). This phenomenon is possibly caused by the remaining OH − in the catalyst layer from the activation of the membrane electrode in KOH solution that was confirmed by the residual OH − test, i.e., if the activated membrane electrode was not washed, the NDPC-1000 based AEMFC would reach a surprising PPD of 1.340 W cm −2 ( Figure S21, Supporting Information). As a comparison, the AEMFC with SA-Fe/NG as the cathode exhibits a voltage decay of ≈41% after a 10 h stability test in an identical way ( Figure S22, Supporting Information), indicating that the N-doped carbon ORR catalyst is better than SA-Fe/NG with respect to long-term AEMFC operation.
To unveil the degradation mechanism, cathode catalysts were collected from the membrane electrode assemblies (MEAs) after AEMFC stability testing for further characterization. XPS results show element Fe was generally not detected from the SA-Fe/NG catalyst after the AEMFC stability test (Table S9, Supporting Information), verifying that Fe demetallation causes the loss of active sites of the catalyst, therefore giving insufficient stability of the SA-Fe/NG-based AEMFC, agreeing with the conclusion of previous reports. [11b,c] Moreover, this is also confirmed by EDS elemental mapping, where the Fe elements are significantly reduced from 0.49 to 0.08 wt% (Figures S17 and S23 and Table S10, Supporting Information). Impressively, the morphology of NDPC-1000 maintains a flake structure and the N element is uniformly distributed on the carbon support with the almost constant N content (before: 1.54 wt%; after: 1.63 wt%, tested by TEM-EDS) after AEMFC stability testing, as displayed in Figures S24 and S25 and Table S11 in the Supporting Information. This phenomenon was also confirmed by XPS analysis and the result demonstrates the total N content of 2.31 at% and proportion of G-N of ≈63.4% after the fuel cell stability testing ( Figure S26, Supporting Information), which is similar to the initial NDPC-1000 sample (total N content of 2.24 at% and G-N of 65.2% in Table S7, Supporting Information). For Fe-SA/NG, the weak coordination bond of Fe connected with N may be the principal reason for the easy dissolution of the Fe active center, while the covalent bonds produced by N and C in N-doped carbon materials are more stable than the coordination bonds, which may result in the excellent stability of NDPC-1000 based AEMFCs. Therefore, we believe that developing metal-free catalysts with highly active centers connected by covalent bonds is a sensible strategy to improve the performance and stability of AEMFCs simultaneously.

Conclusion
In summary, a metal-free carbon (NDPC-1000) catalyst with graphitic N and defects was designed via DFT calculation and experiments. Theoretical results first predicted the synergistic mechanism, in which the graphitic N flexibly tailors the charge density of pentagon (C5-N2) and armchair defects (AC-N2) can optimize the intermediate adsorption hence boost ORR activity, where the stronger covalent bond brings about excellent stability. As expected, the E 1/2 of the obtained NDPC-1000 reached 0.87 V vs RHE in alkaline media and the NDPC-1000 based AEMFC also exhibited a high peak power density of 913 mW cm −2 and long-term durability with a voltage loss of 25% at 250 mA cm −2 after a 100 h at H 2 /O 2 reacting gas. This study provides a useful strategy of using covalent bondcontaining catalysts to improve the stability of AEMFCs.

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