Structural stability and O₂ dissociation on nitrogen-doped graphene with transition metal atoms embedded: A first-principles study

Graphene is a two-dimensional nanomaterial with sp²-bonded carbon atoms packed in a honey comb lattice. By showing extraordinary properties like large surface area (2630 m²/g),¹ remarkably high charge carrier mobility (∼200 000 cm²V ⁻¹s⁻¹),² strong tensile modulus (∼1 TPa)³ and abnormal quantum Hall effect (QHE),⁴ it has been extensively studied and recognized as a potential candidate for various applications.^5–8 However, the absence of a band gap limits its application in devices especially in graphene-based electronics. Therefore, tuning its electronic structure is important and many methods have been developed successfully. Among them, chemical doping by different kinds of species is one of the popular methods and can effectively tune its electronic, optical and magnetic properties.^9–11 Recently, nitrogen-doped graphene (N-graphene) has attracted great attentions due to the comparable atomic size of N to C atom and its unique catalytic and electronic properties. For examples, several groups reported that the N-graphene exhibits excellent stability and catalytic activity, which can be used as an effective catalyst for oxygen reduction reaction (ORR) in fuel cells and Li-air batteries.^12–14 Besides, the N-graphene can also be used as field-effect transistors (FET),³ ultracapacitors,¹⁵ biosensors,¹⁶ photocatalysts.¹⁷ In N-graphene, there are three bonding configurations for N: graphitic N, pyridinic N and pyrrolic N.^3,18,19 Graphitic N refers to the N atoms that substitute for C atoms in the carbon hexagonal ring and makes the N-graphene behaves like n-type semiconductor. Pyridinic N is sp² hybridized with substituting carbon atom in the defective site. The sp³ hybridized pyrrolic N is formed in the defective site or at the edges and contributes two p electrons to the π system. The proportion of these three bonding configurations in N-graphene varies with different experiments and related to the experimental conditions (catalyst, precursors, etc.).^19–21 

Graphene with the impurity of transition metal (TM) is expected to have potential applications in spintronic devices and chemical catalyst, but the binding energies for TM in pristine graphene is weak and it prefers clustering. Thus, modifying graphene by inducing vacancies or doping is suggested. Previously, the TM-embedding in the nitrogen-doped graphene or carbon nanotubes have been synthesized in experiments and proved to be very stable.^22–25 For example, in Fe- and N-doped CNTs, a square-planar configuration for Fe-N4 was found to be distributed uniformly throughout the whole tube and as the catalytic sites for electrocatalytic reduction of oxygen.²⁶ Due to the complex nature of the N-doped graphene with TM-embedding, the atomic structure and the catalytic activity remain largely elusive, and the theoretical studies are very limited.^27,28

In this paper, we study the structures, energetics and O₂ dissociation on different types of N-doped graphene with and without TM-embedding. We illustrate the structural stability for different types of N-doping in graphene and the energetics for TM-embedded N-doped graphene. In recent experiments, N-doped graphene was suggested to be used as electrocatalyst for ORR,^12,29,30 and here we show which type of N dominates the catalytic reaction and how the embedded TM atoms further enhance the catalytic performance. Our results give deep insights into N-doped graphene and show the promising application of N-doped graphene for ORR in fuel cells.

Spin-polarized self-consistent field electronic structure calculations were performed using density functional theory (DFT) implemented in DMol³ program.³¹ The exchange-correlation functional was treated by the generalized gradient approximation (GGA) with the Perdew−Burke− Ernzerhof (PBE) parametrization³² as well as a DFT-based relativistic semi-core pseudopotential (DSPP),³³ and double numerical basis including d- and p-polarization function (DNP) was used. All systems were fully relaxed without any symmetry constraint. An 8×8 supercell of graphene with the lattice parameters of 19.68 Å is used with adjacent layers separated by a 15 Å thick vacuum region. During the geometry optimizations, the Γ point was used to sample the k space. For the calculations of electronic properties, the Brillouin zone was sampled by an 8×8×1 Monkhorst-Pack mesh of k-points. The relaxation of atomic positions was considered to be converged when the change in total energy is less than 1.0 × 10⁻⁵ Ha/Å, and the force on each atom is less than 0.1 eV/Å. To investigate the reaction path of O₂ dissociation and locate the transition states, we use the synchronous method with conjugated gradient (CG) refinements.³⁴ This method involves linear synchronous transit (LST) maximization, followed by repeated CG minimizations, and then quadratic synchronous transit (QST) maximizations and repeated CG minimizations until a transition state is located. Then we calculated the reaction barrier by the energy of reactant minus the energy of transition state and the formation enthalpy by the energy of reactant minus the energy of product.

We investigate the structural stability of graphene doped by different types of N including graphitic N, pyridinic N and pyrrolic N. To compare the structural stability, the formation energies are defined as following:

where E_N-graphene and E_graphene represent the total energies of the N-doped graphene and the pristine graphene, respectively, and E_N and E_C are the energies of N atom in N₂ gas and C atoms in graphene. The n and m are the numbers of N atoms for doping and C atoms for removing in the structures of N-doped graphene. The optimized structures and formation energies for different N-doped graphene are shown in Fig. 1(a) and 1(b). First, we consider the doping of graphitic N. By substituting C with N atom, the graphitic-N doped graphene was formed. The formation energy for one N doping in graphene is 1.23 eV with N-C bond length of 1.42 Å, which is not very difficult to form. Two graphitic-N atoms prefer to locate at the opposite site in a hexagonal carbon ring with the formation energy of 1.19 eV per N.

The single vacancy (SV) by missing one carbon atom and double vacancies (DV) by missing two adjacent carbon atoms offer a high possibility for pyridinic-N doping. The pyridinic-N doping was formed by replacing the C atoms at the SV or DV sites with N atoms. In Fig. 1(a), The N₁V ₁, N₂V ₁, N₃V ₁ were formed in SV defect by doping one to three N atoms, respectively. The formation energies decrease with the number of the pyridinic-N atoms increase. The formation energy of N₃V ₁ structure is 3.55 eV, which is much lower than that of the N₁V ₁ with 5.93 eV and N₂V ₁ with 5.36 eV, respectively. In the N₃V ₁ structure, the three C atoms with dangling bonds at the SV site were replaced by N atom and 3N-centered structure was formed. The energy of the 3N-centered structure is comparable to that of the graphitic N₃ structure and the energy difference is only 0.07 eV. Then we dope one more graphitic N atom to the N₃V ₁ structure and it is named N₄V ₁ structure. This structure is more energetically favorable than graphitic N₄ by 1.25 eV. Similar to N₃V ₁, we obtain the N₄V ₂ structure by substituting four C atoms at the DV site with N atoms, which is also more stable than the N₄ structure by 0.67 eV. Finally, we consider the doping of pyrrolic-N by replacing the C atom in the pentagonal ring with N atom at the DV site, named N₃V ₂. Thermodynamically, however, the pyrrolic-N doping is less stable compared to the doping of graphitic-N and pyridinic-N with the formation energy of as high as 6.22 eV due to the existence of the pentagonal ring in graphene. The vacancies especially the 3N-centered and 4N-centered sites in graphene lead to the possibility for embedding metal atoms, which avoid the metal clustering and will be discussed in the following section.

Seven different types of N-doped graphene were chosen to study the binding energies with transition metal embedded. The binding energies for TM atoms embedded into N-doped graphene are calculated according to the following formula:

where E_{TM-N-graphene}, E_N-graphene and E_TM represent the energies for the systems of TM embedded N-graphene, the doped N-graphene and the free TM atoms, respectively. Fig. 2(a) shows the optimized structures of TM atoms (Ti atom as an example) embedded into different types of N-graphene. Compared to the Ti atom adsorbed on pristine graphene, N-doping, especially the pyridinic-N, can greatly enhance the adsorption capability of the graphene. The adsorption energy for Ti atom on graphitic-N doping graphene (i.e., N₂) is about 1.37 eV, which is much stronger than that on pristine graphene (about 0.1 eV), and the adsorption distance between Ti atom and the N₂ structure is 1.90 Å. When the Ti atom adsorbed at the defective site with pyridinic-N doping, it forms strong covalent bonds with adjacent N and C atoms. Since the atomic radius of Ti is larger than that of the C atom, Ti atom is protruded from the graphene plane by about 1.9 Å. However, the Ti atom can be embedded into DV defect of N₃V ₂ and N₄V ₂ structures with forming a planar structure due to a large defective hole. The binding energy for Ti atom on N₄V ₂ structure is as large as 7.88 eV with the Ti-N bond length of 1.95 Å. Similar to Ti-4N structure, the Fe-4N configuration in carbon nanotube has been observed in previous experiments.³⁵ The binding energy of Fe atom in N₄V ₂ is as large as 7.85 eV, which indicates a greatly stable Fe-4N configuration in graphene. Besides, embedding TM atom into N₃V ₂ structure possesses a low binding energy of less than 2 eV since it breaks a C-C bond at the vacancy site.

We compare the binding energies for TM atoms embedded into different types of N-graphene and plot the results in Fig. 2(b). The binding energies for TM atoms embedded into seven types of N-graphene exhibit the similar curves. The binding energy on graphitic-N doped graphene is very low (less than 1 eV). For the pyridinic-N doped graphene, generally speaking, the binding energies for TM atoms become small as the number of N atoms increases, but the N₁V ₁ and N₂V ₁ possess the similar values. For the particular case of N₃V ₂, the binding energies of TM atoms are smallest (less than 2 eV), which are not thermodynamically stable. The reason is that the TM atoms embedded break the C-C bonds in a perfect hexagonal carbon ring and induce structural distortions. While, the TM atoms embedded into N₄V ₂ structure possess extremely large binding energies (more than 7 eV), and among them, Ti, Fe and Ni atoms possess the largest binding energies as high as 8 eV. The bond lengths of Ni-N and Fe-N are 1.88 Å and 1.90 Å, respectively, which are shorter than the others (i.e., 2.0 Å for Sc). Therefore, the TM-N₄V ₂ structure exhibits extraordinary stability. We will use it for further catalytic studies.

In the previous experimental studies, the Fe-N4 site in which Fe is embedded into N₄V ₂ structure, is proved to be active toward to oxygen reduction reaction,²⁶ which is a key electrocatalytic reaction for fuel cell applications. Here, we systematically compare the catalytic ability for N-doped graphene with and without different TM atom embedded. We consider the O₂ dissociation on the surface of N₂ structure with two graphitic N and N₄V ₂ structure with four pyridinic N configurations. O₂ dissociation is a common reaction in fuel cell and Li-air battery.^36,37 Previous studies demonstrated that the O₂ dissociation reaction can be catalyzed by graphene³⁸ and the reaction barrier can be decreased by N doping.^39,40 Our work here demonstrates that the reaction barrier could be decreased efficiently by embedding TM atoms. In addition, it will provide more active sites on graphene for O₂ dissociation.

First, we study the O₂ molecule dissociated on graphene by doping two graphitic N (N₂ structure). When the O₂ molecule physically adsorbed on N₂, the adsorption energy is very weak with the adsorption distance of 2.78 Å. After dissociation, one oxygen atom forms bond with one C atom and locates on top of it, while the other O atom forms bonds with two C atoms and it locates at bridge site as the epoxy configuration (see Fig. 3(a)). The barrier energy for this reaction is 1.15 eV, which is consistent with a recent theoretical study.³⁹ For the doping of pyridinic N in graphene, we choose the N₄V ₂ structure for O₂ dissociation. The O₂ molecule can be physically adsorbed on N₄V ₂ with a low adsorption energy of 0.1 eV. It is at the center of four N atoms and parallel to the graphene plane with the adsorption distance of 3.1 Å. After dissociation, two O atoms separately bonded to two pyridinic N atoms at the N₄V ₂ site with a reaction barrier of 2.17 eV, which is shown in Fig. 3(b). Therefore, in comparison with the pyridinic N, the graphitic N is more preferable to dissociate O₂ molecules with relatively low reaction barrier energy.

Although the pyridinic N shows poor catalytic capability for O₂ dissociation, we expect it can be improved by embedding TM atoms. When O₂ molecule adsorbed on the TM-N₄V ₂ surface, there are two configurations, top-on and side-on as proposed in the previous studies.^41,42 In the top-on configuration, there is only one oxygen atom directly bonded with the TM atoms and O₂ molecule tilts away from the TM-N₄V ₂ plane, while the side-on configuration has both oxygen atoms bonded to the TM atoms with the O₂ molecule in parallel with the TM–N₄V ₂ plane. Our results show that, O₂ molecule on Fe-N₄V ₂ forms a top-on configuration, and the other TM atoms (Ti, V, Cr and Mn) prefer to adsorb O₂ as the side-on configuration. The adsorption energy of O₂ molecule decreases as the d-shell electrons increases from Ti to Fe. The O₂ molecule adsorbed on Ti-N₄V ₂ possesses the largest adsorption energy of 4.93 eV, and the adsorption energy is 3.64 eV on V-N₄V ₂. On Fe-N₄V ₂ surface, it is only 1.02 eV, which is the same as reported in a previous literature with 1.06 eV on Fe-4N site in carbon nanotube.³⁵ After adsorption, the O₂ molecule has a longer O-O bond length and subsequently a weaker O-O bond. The adsorption energies, distance, O-O bond length and charge transfer from O₂ to TM atoms are summarized in Table I. The elongation of the O-O bond correlates with the charge transfer from 3d orbitals of TM atoms to 2p anti-bonding orbitals of O₂ molecule. More charge transfer corresponds to longer O-O bond length. For example, the O-O bond length increases from 1.23 Å to as large as 1.45 Å for Ti after adsorption and the charge transfer is 0.61e. The charge transfer decreases from 0.61e for Ti to 0.24e for Fe, since the electrons of 3d orbital increases from Ti to Fe.

The reaction path for O₂ dissociation on the different TM-N₄V ₂ structures (five TM atoms are considered here) are similar, since the O₂ can be easily adsorbed on the TM-N₄V ₂ structures. The reaction barrier and formation enthalpy for different TM atoms are listed in Table I and plotted in Fig. 4(b). After the O₂ dissociation, two O atoms are chemisorbed separately on TM atoms and TM atom moves upwards from the plane. For V-N₄V ₂ structure as an example, the O-O distance for the transition state (TS) is 1.76 Å, and after dissociation, the distance between two O atoms is 2.53 Å. Among the TM atoms considered here, only the reaction on V-N₄V ₂ is exothermic with the formation enthalpy of -0.58 eV. The reaction barrier of O₂ dissociation varies with different TM atoms embedded. The reaction barriers for Ti, V, Cr, Mn and Fe are 1.03, 0.61, 1.22, 0.9 and 1.15 eV, respectively. And among them, V-N₄V ₂ exhibits the lowest reaction barrier of only 0.61 eV, which is much lower than that of O₂ dissociation on N₄V ₂ structure without TM embedded (2.17 eV), and is only a half of barrier energy on N₂ structure (1.15 eV). V exhibits better catalytic performance for the reaction of O₂ dissociation than the other TM atoms due to its distinctive configuration after O₂ dissociated into two O atoms. As shown in Fig. 4(a), V only bonded with two N atoms after forming two V-O bonds, while the other TM atoms (Fe, Cr, Ti and Mn) form six bonds with four N atoms and two O atoms. Therefore, by embedding different TM atoms into N₄V ₂ structure, the catalytic activity for O₂ dissociation can be efficiently improved. Among the TM atoms, V atom shows the best catalytic activity for O₂ dissociation. Our results suggest an efficient method to provide more catalytic sites in N-doped graphene (not only limited to graphitic-N) by embedding TM atoms, and give useful insights into N-doped graphene for its applications in the oxygen reduction reactions.

In summary, we have investigated the structures and energetics of different types of N-doped graphene, and studied the TM-embedded N-graphene and its enhanced catalytic activity for O₂ dissociation. Our study demonstrates that: (1) at low coverage of N, the doping of graphitic-N is thermodynamically more stable than the doping of pyridinic-N and pyrrolic-N with defects, but the pyridinic 3N-centered and 4N-centered structures show extreme stability; (2) TM atoms from Sc to Ni embedded into different types of N-doped graphene have been systematically studied, and among them, N₄V ₂ possesses the largest binding energies for TM atoms (more than 7 eV), while Ti, Fe and Ni atoms possess the largest binding energies of about 8 eV on N₄V ₂ structure; (3) by comparing the reaction barrier for O₂ dissociation on N-doped graphene with and without TM-embedding, the results show that the graphitic-N is more preferable to dissociate O₂ molecules with a relatively low reaction barrier of 1.15 eV compared to the pyridinic-N with 2.17 eV. The catalytic activity on pyridinic-N doped graphene can be greatly improved by embedding different TM atoms, i.e., the barrier energy can be reduced to 0.61 eV with V atom embedded. Our results provide an approach to enhance the catalytic performance of N-doped graphene and show its applications in the oxygen reduction reactions for fuel cell.

Authors acknowledge the support from the National Natural Science Foundation of China (Grant No. 21403146, 91233115, 21273158 and 91227201), Natural Science Foundation of Jiangsu Province (Grant No. BK20140314), the National Basic Research Program of China (973 Program, Grant No. 2012CB932400), a Project Funded by the Priority Academic Program Development of Jiangsu Higher Education Institutions (PAPD). This is also a project supported by the Fund for Innovative Research Teams of Jiangsu Higher Education Institutions, Jiangsu Key Laboratory for Carbon-Based Functional Materials and Devices, Collaborative Innovation Center of Suzhou Nano Science and Technology.