Boosting the oxygen reduction reaction activity of dual-atom catalysts on N-doped graphene by regulating the N coordination environment

Abstract

Development of low-cost and high-efficiency oxygen reduction reaction (ORR) catalysts is of significance for fuel cells and metal–air batteries. Here, by regulating the N environment, a series of dual-atom embedded N₅-coordinated graphene catalysts, namely M₁M₂N₅ (M₁, M₂ = Fe, Co, and Ni), were constructed and systematically investigated by DFT calculations. The results reveal that all M₁M₂N₅ configurations are structurally and thermodynamically stable. The strong adsorption of *OH hinders the proceeding of ORR on the surface of M₁M₂N₅, but M₁M₂N₅(OH₂) complexes are formed to improve their catalytic activity. In particular, FeNiN₅(OH₂) and CoNiN₅(OH₂) with the overpotentials of 0.33 and 0.41 V, respectively, possess superior ORR catalytic activity. This superiority should be attributed to the reduced occupation of d-orbitals of Fe and Co atoms in the Fermi level and the apparent shift of d_yz and d_z² orbitals of Ni atoms towards the Fermi level after adsorbing *OH, thus regulating the active sites and exhibiting appropriate adsorption strength for reaction intermediates. This work provides significant insight into the ORR mechanism and theoretical guidance for the discovery and design of low-cost and high-efficiency graphene-based dual-atom ORR catalysts.

---

I. Introduction

The development of renewable energy conversion and storage devices, such as fuel cells and metal–air batteries, is of considerable importance for reducing the reliance on fossil fuels and resolving the environmental issues.^1,2 However, the sluggish cathodic oxygen reduction reaction (ORR) heavily limits the overall performance of these devices and requires catalysts to accelerate the reaction process.³ At present, platinum (Pt)-based materials are still the most commonly used ORR catalysts, but the high cost and natural scarcity of Pt have severely restricted their widespread applications.^4,5 Therefore, there is an urgent need to explore low-cost and high-performance alternatives to substitute Pt-based catalysts.

Metal-embedded nitrogen-doped graphene (M–N–G) single-atom catalysts (SACs) have attracted increasing attention due to their abundant active sites, outstanding mechanical flexibility, and unique electronic properties.^6–8 Unfortunately, their long-term stability and electrocatalytic performances are restrained by their simple geometric structure and single active site.⁹ Introducing a second metal atom into the single metal active site to form double-atom catalysts (DACs) not only inherits the advantages of SACs but also is beneficial to overcoming the restrictions above.^10–13 Firstly, the chemical interactions between two metal atoms in DACs can prevent the agglomeration of individual species, which leads to higher metal loading and more stable active sites than SACs.¹⁰ Furthermore, DACs can provide adjacent coordinated reaction sites, modulate the adsorption strength and mode of reaction intermediates, and optimize the reaction pathway, thus leading to the remarkable improvement of the ORR catalytic activity.^11,12 For example, Fe–Co and Co–Zn embedded N-doped graphene catalysts exhibit excellent ORR catalytic activity, operation durability, and structural stability, outperforming the commercial Pt/C.^14,15 The incorporation of an Ir atom into the single Co active site can induce the re-arrangement of the d-electron of the Co atom toward intensified spin polarization and thus propel faster reaction kinetics.¹⁶ Recently, Fe–Se and Fe–Cu DACs were experimentally identified to possess better structural stability and ORR catalytic activity than Fe-based SACs.^17,18

For N-doped graphene-based dual-metal–atom catalysts, the local coordination environment of dual metal atoms plays important roles in the modulation of their electronic structures and thus affecting their ORR catalytic activity. In particular, the N numbers and species have a significant effect on the ORR performance. FeCoN₆ was identified as the best ORR catalyst among the FeCoN_x (x = 1–6) DACs and the Fe atom was the main active center.¹⁹ Subsequently, the DFT calculations showed that FeCoN₇ and FeCoN₈ possess better catalytic performance as compared with FeCoN₆, and the main active center was transformed from the Fe to Co atom.²⁰ In addition, M₁M₂N₈ (M₁, M₂ = Co, Pt, Fe, Ni) was confirmed to have better ORR catalytic activity than M₁M₂N₆, and CoPtN₈ has the smallest overpotential of 0.30 V.²¹ However, FeFeN₆, FeCoN₆, and CoZnN₆ were found to exhibit the highest ORR catalytic activity by screening 63 M₁–M₂ (M₁, M₂ = Mn, Fe, Co, Ni, Cu, and Zn) embedded N₈ and two kinds of N₆ coordinated graphene.²² In our previous work, Ni–Ni dual-atom active sites displayed the highest ORR catalytic activity among all M₁M₂N₆ (M₁, M₂ = Mn, Fe, Co, Ni, Cu, and Zn) catalysts through constructing a new N₆ coordinated graphene support.²³ Therefore, regulating the N coordination environment in DACs can be expected to further improve the ORR catalytic activity and enrich the structural possibilities of N-doped graphene-based dual-metal-atom catalysts.

In this work, by regulating the N numbers and sites, a N₅-coordinated graphene support was designed to anchor the dual metal atoms. Considering that Fe, Co, and Ni are currently common participating metal atoms in ORR DACs with good catalytic performance, they were selected to construct M₁M₂N₅ catalysts by embedding different dual metal combinations into N₅-coordinated graphene. The structural stability, the electronic properties, the adsorption behaviors of intermediates, the ORR catalytic activity, and the reaction mechanism of these M₁M₂N₅ catalysts were thoroughly investigated by DFT calculations. This work not only schemes out FeNiN₅ and CoNiN₅ DACs with excellent ORR catalytic activity, but provides a significant insight into the fundamental understanding of the ORR reaction mechanism at the atomic level.

II. Computational method

All the spin-polarized DFT calculations were performed using the Vienna ab initio simulation package (VASP).^24,25 Electronic exchange–correlation interactions were treated with the Perdew–Burke–Ernzerhof (PBE) functional method within the generalized gradient approximation (GGA).²⁶ The kinetic energy cutoff was chosen as 500 eV to sufficiently yield good energy convergence. The convergence criteria for the atomic force and the total energy were respectively set to 0.02 eV Å⁻¹ and 10⁻⁵ eV. A sufficient vacuum of 15 Å in the c direction was used to eliminate the interactions between mirror images. The Brillouin zone was sampled with a 3 × 3 × 1 Monkhorst–Pack k-point grid for structural optimizations and a 11 × 11 × 1 grid for the electronic-structure calculations. The default setting of the ferromagnetic state for metal atoms in VASP was adopted and the initial magnetic moment for each metal atom was set to 3 μ_B. The DFT-D2 dispersion correction was adopted to describe the van der Waals interactions between catalysts and reaction intermediates.^27,28 Ab initio molecular dynamics (AIMD) simulations were performed at the temperature of 300 K in the canonical ensemble with a time step of 2 fs and a total time of 10 ps.

To evaluate the structural stability of M₁M₂N₅, the formation energy (E_f) and the binding energy (E_b) between metal atoms and N₅-coordinated graphene are calculated using the following equations

E_f = E_M1M2N5 − E_M1 − E_M2 − 5μ_N − 42μ_C, (1)

where E_M1M2N5, E_N5-G, E_M1, and E_M2 are the energies of M₁M₂N₅, N₅-coordinated graphene, and isolated metal atoms in a vacuum, respectively. μ_N and μ_C respectively denote the chemical potentials of N and C atoms, which are obtained from nitrogen gas and pristine graphene.

The adsorption energy (E_ads) of the reaction intermediates on the catalysts is defined as

E_ads = E_catalyst+X − E_catalyst − E_X, (3)

where E_catalyst+X, E_catalyst, and E_X represent the energies of the entire adsorption systems, the catalysts, and the isolated reaction intermediates, respectively.

The free energy change (ΔG) of each reaction step is determined according to the following equation:

ΔG = ΔE + ΔZPE − TΔS + ΔG_U + ΔG_pH, (4)

where ΔE, ΔZPE, and ΔS are the reaction energy, the changes of the zero-point energy, and the entropy of each reaction step, respectively. T is the room temperature of 298.15 K. ΔG_U = −neU, where n is the number of transferred electrons and U is the standard electrode potential. ΔG_pH = k_BT × ln 10 × pH, in which k_B is the Boltzmann constant, and pH is set as 0.

III. Results and discussion

The schematic diagram for constructing the M₁M₂N₅ structure is illustrated in Fig. 1. Firstly, three connected C atoms were removed from a 5 × 5 pristine graphene supercell to form the tri-vacancy defect. Subsequently, five C atoms around the vacancy site were replaced by N atoms to form the N₅-coordinated graphene. Finally, nine M₁M₂N₅ catalysts (i.e., Fe₂N₅, Co₂N₅, Ni₂N₅, FeCoN₅, CoFeN₅, FeNiN₅, NiFeN₅, CoNiN₅, and NiCoN₅) were constructed by putting different diatomic combinations of Fe, Co, and Ni atoms into the vacancy site of N₅-coordinated graphene. Due to the symmetry of N atoms in N₅-coordinated graphene, CoFeN₅, NiFeN₅, and NiCoN₅ respectively exhibited the same structure as FeCoN₅, FeNiN₅, and CoNiN₅. Therefore, there are six possible structures in total, namely Fe₂N₅, Co₂N₅, Ni₂N₅, FeCoN₅, FeNiN₅, and CoNiN₅. Among all the structures, the M₁ atom was deviated from the plane of graphene, that is, these structures were divided into the raised side (RS) and the dented side (DS) (see step 3 in Fig. 1). To evaluate the structural stability of these catalysts, the formation energy (E_f) for M₁M₂N₅, the binding energy (E_b) between metal atoms and N₅-coordinated graphene, and the cohesive energy (E_coh) for the bulk metal were calculated (see Table S1 for detailed results, ESI†). Evidently, the E_f values for all structures are negative. Moreover, the results of AIMD simulations show that the energy of all structures smoothly oscillates around a certain value, and there is no bond breakage, as indicated by Fig. S1 (ESI†). Furthermore, the larger E_b compared with E_coh for all structures suggests that metal atoms tend to strongly combine with the N₅-coordinated graphene rather than to aggregate together. Therefore, all the M₁M₂N₅ structures mentioned above are thermodynamically stable and are considered in the following discussions.

Fig. 1 Schematic diagram for the formation of the M₁M₂N₅ structure. The orange, olive, blue, green, and yellow balls denote C in perfect graphene, removed C, N, M₁, and M₂ atoms, respectively.

In an acidic medium, the ORR process includes multiple reaction steps, and the adsorption of reaction intermediates plays a key role in the whole reaction process. Hence, the most stable adsorption structures of reaction intermediates on M₁M₂N₅ were firstly explored by considering various possible adsorption configurations, and the calculated results are presented in Fig. 2(a). One can find that there is only one type of stable adsorption configuration for *O₂ and *OH, while two possible stable configurations exist for *OOH (*OOH-I for Fe₂N₅, Co₂N₅, FeCoN₅, and CoNiN₅ as well as *OOH-II for Ni₂N₅ and FeNiN₅) and *O (*O-I for FeNiN₅ and CoNiN₅ as well as *O-II for Fe₂N₅, Co₂N₅, Ni₂N₅, and FeCoN₅). Among all stable adsorption structures, the M₁ atom is the main active site to adsorb reaction intermediates. As is expected, the raised metal atom is more likely to capture the reaction intermediates. Meanwhile, the corresponding E_ads and the M–O bond length (d_M–O) are listed in Table 1 and Tables S2, S3 (ESI†).

Fig. 2 (a) Schematic illustration of the most stable adsorption structures of reaction intermediates on M₁M₂N₅. (b) Free-energy diagrams of the ORR on M₁M₂N₅. The orange, blue, green, yellow, red, and white balls denote C, N, M₁, M₂, O, and H atoms, respectively.

Table 1 Adsorption energy (E_ads) of O₂ on different M₁M₂N₅ catalysts. RS and DS represent the raised and dented sides of all structures, respectively

Model	E ads (eV)	Model	E ads (eV)		Model	E ads (eV)
			RS	DS		RS	DS
Fe2N5	−3.06	Fe2N5(OH)	−1.97	−1.06	Fe2N5(OH2)	−0.93	−1.12
Co2N5	−2.84	Co2N5(OH)	−2.11	−1.85	Co2N5(OH2)	−0.43	−0.97
Ni2N5	−2.44	Ni2N5(OH)	−1.02	−0.40	Ni2N5(OH2)	−0.23	−0.34
FeCoN5	−2.89	FeCoN5(OH)	−1.56	−1.00	FeCoN5(OH2)	−0.61	−0.87
FeNiN5	−3.16	FeNiN5(OH)	−1.22	−0.38	FeNiN5(OH2)	−0.28	−0.45
CoNiN5	−2.91	CoNiN5(OH)	−1.06	−0.33	CoNiN5(OH2)	−0.24	−0.33

---

To further unveil the ORR process, the free-energy diagrams of the four-electron reaction pathway (*O₂ → *OOH → *O + H₂O → *OH + H₂O → 2H₂O) and the theoretical overpotential for all structures are presented in Fig. 2(b). It can be seen that only the free energy for the hydrogenation of *OH to H₂O is uphill at zero electrode potential (U = 0 V), implying that this step is endothermic and thermodynamically unfavorable, and thus the ORR process is hindered by the formation of the second H₂O molecule. Therefore, the rate-determining step (RDS) for the ORR on all M₁M₂N₅ structures is the hydrogenation of *OH to H₂O. This RDS arises from the strong adsorption of *OH on M₁M₂N₅ (see Table S2, ESI†) and leads to the large overpotentials of 2.19, 1.89, 1.52, 2.12, 2.35, and 2.07 V, respectively. Meanwhile, the *OH can be regarded as a ligand to modify M₁M₂N₅, thus forming M₁M₂N₅(OH) complexes. Hence, the catalytic performances of M₁M₂N₅(OH) complexes need to be further investigated.

Since the chemisorption of O₂ is usually regarded as a prerequisite for triggering the ORR, the adsorption characteristics of O₂ on the RS and DS of M₁M₂N₅(OH) were firstly investigated. Through considering all the adsorption possibilities, the energetically favorable O₂ adsorption configurations on the RS and DS of M₁M₂N₅(OH) were obtained (see Fig. 3(a) and Fig. S2 for details, ESI†). It can be seen from Fig. 3(a) that O₂ tends to be adsorbed on the RS of Fe₂N₅(OH), Co₂N₅(OH), and FeCoN₅(OH) by bonding with dual metal atoms, but prefers to bond with the M₁ atom of Ni₂N₅(OH), FeNiN₅(OH), and CoNiN₅(OH). This result is understandable since the spin charges are distributed on both M₁ and M₂ atoms in the former but only on the M₁ atom in the latter, as indicated by Fig. 4. Different from the above results, O₂ only bonded with the M₂ atom on the DS of M₁M₂N₅(OH) (see Fig. S2, ESI†). Furthermore, E_ads of O₂ on the RS and DS of M₁M₂N₅(OH) were calculated to determine the most stable adsorption configurations. As shown in Table 1, the adsorption of O₂ on the RS has higher E_ads than that on the DS, suggesting that it is more likely to adsorb O₂ on the RS of M₁M₂N₅(OH). Therefore, the adsorptions of other reaction intermediates and the subsequent reaction steps of the ORR should occur in the RS of M₁M₂N₅(OH). In addition, after forming the M₁M₂N₅(OH) complexes, the E_ads of O₂ was remarkably decreased compared with M₁M₂N₅ (see Table 1), and d_M–O was stretched (see Table S3, ESI†). This is because the *OH obtained some charges from dual metal atoms (see Table S4, ESI†), thus reducing their adsorption capability. Besides O₂, the most stable adsorption structures of other reaction intermediates (*OOH, *OH, and *O) on the RS of M₁M₂N₅(OH), the corresponding E_ads, and the d_M–O are illustrated in Fig. 3(a) and Tables S2, S3 (ESI†). It can be seen that these intermediates are favorably bonding with a single M₁ atom or dual M₁ and M₂ atoms. Moreover, the introduction of *OH also lowers the interactions between these reaction intermediates with M₁M₂N₅, which may lead to the reduced overpotential.

Fig. 3 (a) Schematic illustration of the most stable adsorption structures of reaction intermediates on M₁M₂N₅(OH). (b) Free-energy diagrams of the ORR on M₁M₂N₅(OH). The orange, blue, green, yellow, red, and white balls denote C, N, M₁, M₂, O, and H atoms, respectively.

Fig. 4 Spin-resolved charge density of M₁M₂N₅(OH) with an isosurface value of 0.02 e Å⁻³. Purple and cyan bubbles represent positive and negative charges, respectively. The orange, blue, green, gray, yellow, red, and white balls denote C, N, Fe, Co, Ni, O, and H atoms, respectively.

Based on the most stable adsorption structures of the reaction intermediates above, the free energy change and the overpotential for the ORR on M₁M₂N₅(OH) were further investigated to unveil the catalytic activity. Different from M₁M₂N₅, the free energy for all reaction steps of the ORR on M₁M₂N₅(OH) is downhill at U = 0 V except for Fe₂N₅(OH) [see Fig. 3(b)], while the RDS is still the hydrogenation of *OH to H₂O. Meanwhile, the overpotentials of ORR on Fe₂N₅(OH), Co₂N₅(OH), Ni₂N₅(OH), FeCoN₅(OH), FeNiN₅(OH), and CoNiN₅(OH) are respectively reduced to 1.31, 1.04, 0.85, 1.13, 1.17, and 0.98 V, which could be attributed to the reduced binding strength between the reaction intermediates and these catalysts (see Table S2 for details, ESI†). However, these values are still significantly higher than those for Pt (111) (0.44 V)²⁹ and Pt (100) (0.47 V).³⁰ Fortunately, the *OH can be adsorbed on M₁M₂N₅(OH) to form M₁M₂N₅(OH₂) complexes for further enhancing the ORR catalytic activity.

After incorporating *OH into M₁M₂N₅(OH), its adsorption characteristics were significantly changed. The most stable adsorption configuration of O₂ on M₁M₂N₅(OH₂), the corresponding E_ads, and d_M–O are presented in Table 1 and Fig. S2, Table S3 (ESI†). It can be found that the E_ads of O₂ on the RS of Ni₂N₅(OH₂), FeNiN₅(OH₂), and CoNiN₅(OH₂), respectively, are only −0.23, −0.28, and −0.24 eV, implying the weak interactions between O₂ and these catalysts. Moreover, the O–O bond lengths of O₂ on these catalysts, respectively, are 1.250, 1.258, and 1.253 Å, which are very close to the O–O bond length of 1.232 Å in gas phase O₂. These results indicate that an O₂ molecule can be only physically adsorbed on the RS of Ni₂N₅(OH₂), FeNiN₅(OH₂), and CoNiN₅(OH₂), and hence the ORR cannot proceed on the RS of these complexes. However, Fe₂N₅(OH₂), Co₂N₅(OH₂), and FeCoN₅(OH₂) can chemically adsorb O₂ on their RS with E_ads of −0.93, −0.43, and −0.61 eV, respectively. As compared with the RS of Fe₂N₅(OH₂), Co₂N₅(OH₂), Ni₂N₅(OH₂), FeCoN₅(OH₂), FeNiN₅(OH₂), and CoNiN₅(OH₂), the DS of these structures show a stronger adsorption capacity for O₂, and the E_ads rises to −1.12, −0.97, −0.34, −0.87, −0.45, and −0.33 eV, respectively, indicating that the O₂ prefers to be adsorbed on the DS of M₁M₂N₅(OH₂) to proceed the subsequent reactions of the ORR. Moreover, these values are smaller than that on M₁M₂N₅(OH) because dual metal atoms lose some charges to *OH (see Table S4, ESI†). It is noteworthy that the values of −0.34, −0.45, and −0.33 eV fall in the range of −0.24 and −0.80 eV for some excellent dual-metal-atom embedded N-doped graphene ORR catalysts.^19,31–33 Thus, Ni₂N₅(OH₂), FeNiN₅(OH₂), and CoNiN₅(OH₂) may possess outstanding ORR catalytic activities.

Following the aforementioned O₂ adsorption results, the free-energy diagrams of the whole reaction process on M₁M₂N₅(OH₂) and the corresponding overpotential are presented in Fig. 5(a) to identify their ORR catalytic activities. Compared with M₁M₂N₅(OH), the RDS of the ORR on Ni₂N₅(OH₂), FeNiN₅(OH₂), and CoNiN₅(OH₂) was transferred into the hydrogenation of O₂ to *OOH. The overpotentials of the ORR on Fe₂N₅(OH₂), Co₂N₅(OH₂), Ni₂N₅(OH₂), FeCoN₅(OH₂), FeNiN₅(OH₂), and CoNiN₅(OH₂) were reduced to 0.90, 0.73, 0.51, 0.81, 0.33, and 0.41 V, respectively. In particular, the overpotentials of 0.33 V for FeNiN₅(OH₂) and 0.41 V for CoNiN₅(OH₂) are smaller than or comparable to those for FeNiN_6/8 (0.35–0.90 V),^{21–23,31,34–36} CoNiN_6/8 (0.35–0.79 V),^{21–23,31,37} Pt (111) (0.44 V),²⁹ and Pt (100) (0.47 V),³⁰ indicating that the FeNiN₅(OH₂) and CoNiN₅(OH₂) complexes possess excellent ORR catalytic activity. The corresponding schematic illustration of the ORR process on FeNiN₅(OH₂) and CoNiN₅(OH₂) is given in Fig. 5(b). Considering that the FeNiN₅(OH₂) and CoNiN₅(OH₂) complexes are formed during the ORR process on FeNiN₅ and CoNiN₅ due to the strong adsorption of *OH, FeNiN₅ and CoNiN₅ should be promising alternatives to substitute for Pt catalysts.

Fig. 5 (a) Free-energy diagrams of the ORR on M₁M₂N₅(OH₂). (b) Schematic illustration of the ORR process on FeNiN₅(OH₂) and CoNiN₅(OH₂). The orange, blue, green, yellow, red, and white balls denote C, N, M₁, M₂, O, and H atoms, respectively.

To concretely explore the origin of the outstanding catalytic activities of the FeNiN₅(OH₂) and CoNiN₅(OH₂), the projected density of states (PDOS) of FeNiN₅, FeNiN₅(OH), FeNiN₅(OH₂), CoNiN₅, CoNiN₅(OH), and CoNiN₅(OH₂) are presented in Fig. 6. It can be seen from this figure that d_xy, d_z², and d_x²−y² orbitals of the Fe atom in FeNiN₅ and d_xy, d_yz, and d_z² orbitals of Co atom in CoNiN₅ are mainly distributed near the Fermi-level, but there are no d orbital electronic states of Ni atoms, indicating that Fe and Co atoms are the main active sites for the ORR. These results are consistent with the intermediate adsorption structures on FeNiN₅ and CoNiN₅. When incorporating *OH into FeNiN₅ and CoNiN₅, the *OH mainly obtained some electrons from Fe and Co atoms (see Table S4, ESI†), resulting in decreased d-orbitals of Fe and Co atoms near the Fermi-level. However, d_yz and d_z² orbitals of Ni atoms in FeNiN₅ and CoNiN₅ only slightly moved towards the Fermi-level. Therefore, Fe and Co atoms are still the main active sites to adsorb reaction intermediates (as shown in Fig. 3), but their adsorption capabilities were weakened to reduce the overpotential. After forming FeNiN₅(OH₂) and CoNiN₅(OH₂) complexes, d_yz and d_z² orbitals of Ni atoms continued to move towards the Fermi-level compared with those in FeNiN₅(OH) and CoNiN₅(OH). Meanwhile, the Fe and Co sites were occupied by the adsorbed *OH, and thus the active sites were transformed from Fe and Co atoms to Ni atoms, corresponding to moderate adsorption strength for reaction intermediates and improved ORR catalytic activity. For Fe₂N₅, Co₂N₅, Ni₂N₅, and FeCoN₅, and the adsorption of *OH also moved the d-orbitals of M₂ atoms towards the Fermi-level, as shown in Fig. S3 (ESI†). However, more d-orbital electronic states of these atoms distributed near the Fermi level lead to the stronger adsorption capability for intermediates, thus presenting lower ORR catalytic activity compared with FeNiN₅ and CoNiN₅. In addition, the above results also indicate that *OH ligand modification can be an effective strategy to modify the electronic structures and improve the ORR catalytic activity of graphene-based dual metal atom catalysts.

Fig. 6 The PDOS of (a) Fe and (b) Ni atoms in FeNiN₅, FeNiN₅(OH) and FeNiN₅(OH₂) as well as (c) Co and (d) Ni atoms in CoNiN₅, CoNiN₅(OH) and CoNiN₅(OH₂). The dashed line indicates the Fermi level.

IV. Conclusions

In summary, a series of dual-atom embedded N₅-coordinated graphene catalysts M₁M₂N₅ (M₁, M₂ = Fe, Co, and Ni) were constructed by regulating the N coordination environment. The geometric structure, electronic structure, ORR catalytic activity, and reaction mechanism of M₁M₂N₅ catalysts were systematically investigated by DFT calculations. The results reveal that all M₁M₂N₅ configurations are structurally and thermodynamically stable. The strong adsorption of *OH hinders the proceeding of ORR on the surface of M₁M₂N₅ and M₁M₂N₅(OH), but M₁M₂N₅(OH₂) complexes are formed in the reaction process. Among all M₁M₂N₅(OH₂), FeNiN₅(OH₂) and CoNiN₅(OH₂) respectively with the overpotentials of 0.33 and 0.41 V exhibit the best catalytic activity superior to Pt catalysts. The reason for this lies in the fact that the absorbed *OH lowers the occupation of the d-orbitals of Fe and Co atoms in the Fermi level, and shifts the d_yz and d_z² orbitals of the Ni atom toward the Fermi level, thus transforming the active sites from Fe and Co atoms to Ni atoms and regulating the adsorption strength for reaction intermediates. The *OH ligand modification was identified to be an effective strategy for improving the ORR catalytic activity. This work provides a significant insight into the fundamental understanding of the ORR mechanism and greatly facilitates the computational screening and design of low-cost and high-efficiency graphene-based dual-atom ORR catalysts.

Conflicts of interest

The authors declare no competing financial interest.

Acknowledgements

This work was financially supported by the Natural Science Foundation of Inner Mongolia Autonomous Region (Grant No. 2023QN01010), the Research Program of Science and Technology at Universities of Inner Mongolia Autonomous Region (Grant No. NJZZ23019, NJZZ23025), and the Fundamental Research Funds for the Inner Mongolia Normal University (Grant No. 2023JBYJ016).

References

1. Z. P. Cano, D. Banham, S. Y. Ye, A. Hintennach, J. Lu, M. Fowler and Z. W. Chen, Nat. Energy, 2018, 3, 279–289.
2. H. F. Wang and Q. Xu, Matter, 2019, 1, 565–595.
3. X. L. Tian, X. F. Lu, B. Y. Xia and X. W. Lou, Joule, 2020, 9, 45–68.
4. M. K. Debe, Nature, 2012, 486, 43–51.
5. L. Huang, S. Zaman, X. L. Tian, Z. T. Wang, W. S. Fang and B. Y. Xia, Acc. Chem. Res., 2021, 54, 311–322.
6. L. Li, R. Huang, X. R. Cao and Y. H. Wen, J. Mater. Chem. A, 2020, 8, 19319–19327.
7. M. B. Gawande, P. Fornasiero and R. Zbořil, ACS Catal., 2020, 10, 2231–2259.
8. C. W. Ye and L. Xu, J. Mater. Chem. A, 2021, 9, 22218–22247.
9. H. Y. Zhuo, X. Zhang, J. X. Liang, Q. Yu, H. Xiao and J. Li, Chem. Rev., 2020, 120, 12315–12341.
10. W. Y. Zhang, Y. G. Chao, W. S. Zhang, J. H. Zhou, F. Lv, K. Wang, F. X. Lin, H. Luo, J. Li, M. P. Tong, E. Wang and S. J. Guo, Adv. Mater., 2021, 33, 2102576.
11. R. Z. Li and D. S. Wang, Adv. Energy Mater., 2022, 12, 2103564.
12. H. S. Shang and D. Liu, Nano Res., 2023, 16, 6477–6506.
13. D. C. Zhong, Y. N. Gong, C. Zhang and T. B. Lu, Chem. Soc. Rev., 2023, 52, 3170–3214.
14. J. Wang, Z. Q. Huang, W. Liu, C. R. Chang, H. L. Tang, Z. J. Li, W. X. Chen, C. J. Jia, T. Yao, S. Q. Wei, Y. E. Wu and Y. D. Lie, J. Am. Chem. Soc., 2017, 139, 17281–17284.
15. Z. Y. Lu, B. Wang, Y. F. Hu, W. Liu, Y. F. Zhao, R. O. Yang, Z. P. Li, J. Luo, B. Chi, Z. Jiang, M. S. Li, S. C. Mu, S. J. Liao, J. J. Zhang and X. L. Sun, Angew. Chem., Int. Ed., 2019, 58, 2622–2626.
16. M. L. Xiao, J. B. Zhu, S. Li, G. R. Li, W. W. Liu, Y. P. Deng, Z. Y. Bai, L. Ma, M. Feng, T. P. Wu, D. Su, J. Lu, A. P. Yu and Z. W. Chen, ACS Catal., 2021, 11, 8837–8846.
17. Y. Wang, J. Wu, S. Tang, J. Yang, C. Ye, J. Chen, Y. Lei and D. Wang, Angew. Chem., Int. Ed., 2023, 62, e202219191.
18. Y. X. Zhang, S. Zhang, H. Huang, X. Liu, B. Li, Y. Lee, X. Wang, Y. Bai, M. Sun, Y. Wu, S. Gong, X. Liu, Z. Zhuang, T. Tan and Z. Niu, J. Am. Chem. Soc., 2023, 145, 4819–4827.
19. Y. L. Sun, J. Wang, Q. Liu, M. R. Xia, Y. F. Tang, F. M. Gao, Y. L. Hou, J. Tse and Y. F. Zhao, J. Mater. Chem. A, 2019, 7, 27175–27185.
20. R. M. Hu, Y. C. Li, Q. W. Zeng and J. X. Shang, Appl. Surf. Sci., 2020, 525, 146588.
21. M. A. Hunter, J. M. T. A. Fischer, Q. H. Yuan, M. Hankel and D. J. Searles, ACS Catal., 2019, 9, 7660–7667.
22. F. T. Wang, W. B. Xie, L. J. Yang, D. Q. Xie and S. Lin, J. Catal., 2021, 396, 215–223.
23. L. Li, Y. M. Li, T. Zhao, R. Huang, X. R. Cao, T. E. Fan and Y. H. Wen, J. Electrochem. Soc., 2022, 169, 026524.
24. G. Kresse and J. Furthmuller, Phys. Rev. B: Condens. Matter Mater. Phys., 1996, 54, 11169–11186.
25. G. Kresse and J. Furthmüller, Comput. Mater. Sci., 1996, 6, 15–50.
26. J. P. Perdew, K. Burke and M. Ernzerhof, Phys. Rev. Lett., 1996, 77, 3865–3868.
27. S. Grimme, J. Comput. Chem., 2006, 27, 1787–1799.
28. S. Grimme, J. Antony, S. Ehrlich and H. Krieg, J. Chem. Phys., 2010, 132, 154104.
29. V. Tripkovic, E. Skúlason, S. Siahrostami, J. K. Norskov and J. Rossmeisl, Electrochim. Acta, 2010, 55, 7975–7981.
30. Z. Y. Duan and G. F. Wang, J. Phys. Chem. C, 2013, 117, 6284–6292.
31. Z. Liang, M. M. Luo, M. W. Chen, C. Liu, S. G. Peera, X. P. Qi, J. Liu, U. P. Kumar and T. X. Liang, J. Colloid Interface Sci., 2020, 568, 54–62.
32. L. Li, Y. M. Li, R. Huang, X. R. Cao and Y. H. Wen, ChemCatChem, 2021, 13, 4645–4651.
33. K. W. Huang, W. H. Yang, L. Li, Y. M. Li, R. Huang and Y. H. Wen, Phys. Chem. Chem. Phys., 2023, 25, 18266–18274.
34. Y. N. Meng, C. Yin, K. Li, H. Tang, Y. Wang and Z. J. Wu, ACS Sustainable Chem. Eng., 2019, 7, 17273–17281.
35. Y. D. Zhou, W. Yang, W. Utetiwabo, Y. M. Lian, X. Yin, L. Zhou, P. W. Yu, R. J. Chen and S. R. Sun, J. Phys. Chem. Lett., 2020, 11, 1404–1410.
36. X. F. Zhu, D. T. Zhang, C. J. Chen, Q. R. Zhang, R. S. Liu, Z. H. Xia, L. M. Dai, R. Amal and X. Y. Lu, Nano Energy, 2020, 71, 104597.
37. J. Y. Xu, A. Elangovan, J. Li and B. Liu, J. Phys. Chem. C, 2021, 125, 2334–2344.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d3cp04831d

---