Fe 3 Cluster Anchored on the C 2 N Monolayer for E ﬃ cient Electrochemical Nitrogen Fixation

: Under the current double challenge of energy and the environment, an e ﬀ ective nitrogen reduction reaction (NRR) has become a very urgent need. However, the largest production of ammonia gas today is carried out by the Haber–Bosch process, which has many disadvantages, among which energy consumption and air pollution are typical. As the best alternative procedure, electrochemistry has received extensive attention. In this paper, a catalyst loaded with Fe 3 clusters on the two-dimensional material C 2 N (Fe 3 @C 2 N) is proposed to achieve e ﬀ ective electrochemical NRR, and our ﬁrst-principles calculations reveal that the stable Fe 3 @C 2 N exhibits excellent catalytic performance for electrochemical nitrogen ﬁxation with a limiting potential of 0.57 eV, while also suppressing the major competing hydrogen evolution reaction. Our ﬁndings will open a new door for the development of non-precious single-cluster catalysts for e ﬀ ective nitrogen reduction reactions.


Introduction
Ammonia (NH 3 ) is an important chemical in agriculture and industry [1], and the direct reduction of nitrogen to ammonia is still considered to be one of the most important and challenging chemical transformations [2]. Dinitrogen (N 2 ), as the component with the largest volume fraction (78.08%) in the Earth's atmosphere, is the main source of nitrogen [3]. Therefore, ammonia can be directly synthesized from the Earth's abundant nitrogen resources. Usually, the Haber-Bosch process (N 2 + 3H 2 = 2NH 3 ) is used to synthesize ammonia on large scales. In this process, iron and ruthenium-based metal catalysts are used to convert atmospheric nitrogen (N 2 ) into NH 3 by reacting with hydrogen (H 2 ) [4,5]; the adsorbed N 2 molecule is firstly dissociated on specific active sites of the catalysts [6][7][8], and then the dissociated N species are hydrogenated, which is known as the dissociation mechanism [9]. However, extreme reaction conditions are usually required, such as high pressure (~100 bar) and high temperature (~700 K) [10]; as the reaction proceeds, a large amount of carbon dioxide is emitted, and the energy consumed each year is huge. On the other hand, the conditions of high temperature and pressure may not be necessary because the reaction is actually exothermic [10]. For this reason, it is of significance to search for a green and cost-effective method for the production of ammonia [11].
With continuous efforts, scientists have found that the electrochemical nitrogen reduction reaction (NRR, N 2 + 6H + + 6e − = 2NH 3 ) [12] under environmental conditions is extremely attractive because of its greatly reduced energy input and good environmental compatibility. Here, the electrocatalyst is the core component for reducing the reaction limit potential and increasing the reaction speed and the selectivity of NH 3 [13]. For NRR, one of the key challenges is the activation of nitrogen, which is

Results and Discussion
2.1. The Geometry, Stability and Electronic Properties of the Fe 3 @C 2 N Two-dimensional (2D) C 2 N is a porous material similar to graphene, and the hexagonal lattice parameter of the C 2 N unit cell was calculated to be 8.32 Å, which matches well with previous studies [29]. The optimized geometric structure of C 2 N is given in Figure 1a. In the C 2 N monolayer, the C 6 rings are connected by N atoms, resulting in a six-membered nitrogen pore with a para-N distance of 5.51 Å; the pores provide desired sites to anchor metal atoms or clusters. Previous theoretical studies [30,31] showed that adding the second Fe atom to Fe 1 @C 2 N is energetically favored: the binding energy of the second Fe atom is comparable to that of the first one, and Fe 2 @C 2 N maintains its original structure throughout a 10 ps molecular dynamics simulation at 800 K [30]. Based on the porous structural feature in the C 2 N monolayer and the stable Fe 2 @C 2 N complex [30,31], we conjectured that adding one more Fe atom to Fe 2 @C 2 N, i.e., Fe 3 @C 2 N, may also have good thermodynamic and thermal stability. According to our computations, the lowest-energy supported Fe 3 cluster adopts a buckled triangle structure, with each Fe bonded to two N atoms, and the buckling height of the Fe 3 structure is 1.29 Å, while the C 2 N support almost maintains its flat configuration (Figure 1b), similar to the model proposed by Pei et al. [32]. The Fe−Fe and Fe−N bond lengths are in the range of 2.28~2.32 Å (slightly shorter than the 2.48 Å in the bulk phase) and 1.97~2.01 Å, respectively. The average binding energy per Fe atom is −4.12 eV, which is slightly lower than the value of −4.84 eV per Fe atom in the bulk phase at the same level of theory, indicating the good thermodynamic stability of Fe 3 @C 2 N. Furthermore, a first-principles molecular dynamics (FPMD) simulation in an NVT ensemble with the temperature controlled by the Nosé-Hoover method [33] was performed, and the structure of Fe 3 @C 2 N was well kept through a 10 ps FPMD with a time step of 0.5 fs at 800 K ( Figure S1). Thus, Fe 3 @C 2 N is extraordinarily stable. was performed, and the structure of Fe3@C2N was well kept through a 10 ps FPMD with a time step of 0.5 fs at 800 K ( Figure S1). Thus, Fe3@C2N is extraordinarily stable. Good electrical conductivity is required for fast charge transfer during efficient electrocatalytic processes. Compared to the semiconducting C2N monolayer, whose band gap was calculated to be 1.68 eV (Figure 1c), in line with a previous report [29], Fe3@C2N is a ferromagnetic (each Fe atom carries the magnetic moment of ~3 μB, Figure S2) metal (Figure 1d), and the metallic feature originates from the states that have crossed the Fermi level, which is dominatingly attributed to the Fe-d orbitals ( Figure S3). The charge transfer between the Fe3 cluster and C2N is 2.12 |e|, indicating the high activity of the Fe3 cluster.

N2 Adsorption on Fe3@C2N
Previous investigations proposed the following criteria for an efficient electrocatalyst for NRR: (1) the catalyst can facilitate the chemisorption of N2 molecules to sufficiently activate the inert N≡N triple bonds, and (2) the catalyst can selectively stabilize N2H* and (3) destabilize NH2* species to lower the limiting potential [12]. In the electrochemical synthesis of ammonia, the adsorption of nitrogen is considered to be the first step, and the initial nitrogen adsorption configuration plays an important role in subsequent reactions [9]. On pristine C2N, the adsorption of N2 is very weak, and the adsorption energy (Eads) is only −0.01 eV, which indicates that the pristine C2N monolayer cannot effectively activate nitrogen. Zhao's theoretical group also found that N2 adsorption on C and N atoms is rather weak (Eads < −0.20 eV), or the N2 molecule is spontaneously trapped by the central TM atoms after structural optimization on TM@C2N [34]. In our work, we considered two adsorption configurations, namely, end-on and side-on structures of N2 on the Fe3 cluster anchored on C2N (Figure 2), and the adsorption energies of the two configurations are −1.08 and −1.45 eV, respectively. Good electrical conductivity is required for fast charge transfer during efficient electrocatalytic processes. Compared to the semiconducting C 2 N monolayer, whose band gap was calculated to be 1.68 eV (Figure 1c), in line with a previous report [29], Fe 3 @C 2 N is a ferromagnetic (each Fe atom carries the magnetic moment of~3 µ B , Figure S2) metal (Figure 1d), and the metallic feature originates from the states that have crossed the Fermi level, which is dominatingly attributed to the Fe-d orbitals ( Figure S3). The charge transfer between the Fe 3 cluster and C 2 N is 2.12 |e|, indicating the high activity of the Fe 3 cluster.

N 2 Adsorption on Fe 3 @C 2 N
Previous investigations proposed the following criteria for an efficient electrocatalyst for NRR: (1) the catalyst can facilitate the chemisorption of N 2 molecules to sufficiently activate the inert N≡N triple bonds, and (2) the catalyst can selectively stabilize N 2 H* and (3) destabilize NH 2 * species to lower the limiting potential [12]. In the electrochemical synthesis of ammonia, the adsorption of nitrogen is considered to be the first step, and the initial nitrogen adsorption configuration plays an important role in subsequent reactions [9]. On pristine C 2 N, the adsorption of N 2 is very weak, and the adsorption energy (E ads ) is only −0.01 eV, which indicates that the pristine C 2 N monolayer cannot effectively activate nitrogen. Zhao's theoretical group also found that N 2 adsorption on C and N atoms is rather weak (E ads < −0.20 eV), or the N 2 molecule is spontaneously trapped by the central TM atoms after structural optimization on TM@C 2 N [34]. In our work, we considered two adsorption configurations, namely, end-on and side-on structures of N 2 on the Fe 3 cluster anchored on C 2 N (Figure 2), and the adsorption energies of the two configurations are −1.08 and −1.45 eV, respectively. The E ads values are sufficiently strong to capture and activate N 2 , as indicated by the elongated N≡N lengths of 1.16 and 1.26 Å (the isolated N≡N length was calculated to be 1.11 Å) and a charge transfer of −0.57 and −1.14 |e| for the end-on and side-on configurations, respectively. Compared to the maximum vibrational frequency of free N 2 (2420 cm −1 ), the maximum vibrational mode of the adsorbed N-N in the side-on (end-on) configuration is 1380 (2011) cm −1 . The remarkably reduced frequency in the side-on structure suggests an apparent elongation/weakening in the N-N bond upon N 2 being adsorbed on Fe 3 @C 2 N.
Catalysts 2020, 10, x FOR PEER REVIEW 4 of 9 The Eads values are sufficiently strong to capture and activate N2, as indicated by the elongated N≡N lengths of 1.16 and 1.26 Å (the isolated N≡N length was calculated to be 1.11 Å) and a charge transfer of −0.57 and −1.14 |e| for the end-on and side-on configurations, respectively. Compared to the maximum vibrational frequency of free N2 (2420 cm −1 ), the maximum vibrational mode of the adsorbed N-N in the side-on (end-on) configuration is 1380 (2011) cm −1 . The remarkably reduced frequency in the side-on structure suggests an apparent elongation/weakening in the N-N bond upon N2 being adsorbed on Fe3@C2N.

N2 Reaction on Fe3@C2N
Considering the energetically preferred side-on configuration and the greater charge transfer between N2 and Fe3@C2N, as well as the much longer stretched N≡N length, we selected the N2 adsorption with the side-on structure as the starting state for NRR.
Previous studies revealed that the N2 molecule could dissociate directly on the specific iron surface, such as Fe(111) and Fe(211) surfaces [35,36]; we first examined the dissociative mechanism. However, our computations showed that the dissociated NN state on Fe3@C2N is not energetically favored, since it is 2.67 eV higher in energy than the N2 adsorbed state ( Figure S4), suggesting a very high activation energy barrier (>2.67 eV) of N2 dissociation. For comparison, the calculated barrier of N2 dissociation is as high as 1.89 eV on Fe3/θ-Al2O3(010) [25]. Thus, the dissociative mechanism of NRR over Fe3@C2N was not further studied in our work. Quite recently, Wang's theoretical group proposed a new mechanism for NRR, namely, a surface-hydrogenation mechanism [37], where the surface hydrogenation can drive the N2 reduction reaction. We also tested the surface-hydrogenation mechanism of NRR on Fe3@C2N; nevertheless, N2 adsorption on the hydrogenated Fe3 cluster is endothermic at 2.41 eV (Figure S5), indicating an unfavorable pathway for NRR. Therefore, we focused on the associative mechanism in the following experiment.

N 2 Reaction on Fe 3 @C 2 N
Considering the energetically preferred side-on configuration and the greater charge transfer between N 2 and Fe 3 @C 2 N, as well as the much longer stretched N≡N length, we selected the N 2 adsorption with the side-on structure as the starting state for NRR.
Previous studies revealed that the N 2 molecule could dissociate directly on the specific iron surface, such as Fe(111) and Fe(211) surfaces [35,36]; we first examined the dissociative mechanism. However, our computations showed that the dissociated NN state on Fe 3 @C 2 N is not energetically favored, since it is 2.67 eV higher in energy than the N 2 adsorbed state ( Figure S4), suggesting a very high activation energy barrier (>2.67 eV) of N 2 dissociation. For comparison, the calculated barrier of N 2 dissociation is as high as 1.89 eV on Fe 3 /θ-Al 2 O 3 (010) [25]. Thus, the dissociative mechanism of NRR over Fe 3 @C 2 N was not further studied in our work. Quite recently, Wang's theoretical group proposed a new mechanism for NRR, namely, a surface-hydrogenation mechanism [37], where the surface hydrogenation can drive the N 2 reduction reaction. We also tested the surface-hydrogenation mechanism of NRR on Fe 3 @C 2 N; nevertheless, N 2 adsorption on the hydrogenated Fe 3 cluster is endothermic at 2.41 eV ( Figure S5), indicating an unfavorable pathway for NRR. Therefore, we focused on the associative mechanism in the following experiment.
Catalysts 2020, 10, x FOR PEER REVIEW 5 of 9 (*NNH → *NNH2) has a maximum free energy change of 0.57 (0.83) eV via the enzymatic (consecutive) mechanism. Thus, the limiting potential (η) of NRR on Fe3@C2N is as low as −0.57 V, lower than the η values on the Ru (0001) step surface [40] and Fe2@C2N [24] (0.98 eV and 1.23 eV, respectively). Typically, the potential limiting step on most metal surfaces (such as Re, Ru, Rh and Fe) or two-dimensional MBenes is either the first hydrogenation (forming *NNH) or the last protonation (*NH3 formation or NH3 desorption) [9,41]. However, on Fe3@C2N, the free energy change of *N2 → *NNH is only 0.28 eV. We also calculated the barriers of the potential limiting steps of the two mechanisms, and the barriers are both 1.11 eV based on the CI-NEB method ( Figure S6), lower than the corresponding values (1.24 and 1.31 eV) on Fe3@θ-Al2O3(010) [25]. All of the above results suggest the high activity of Fe3@C2N for NRR. In addition to the catalytic activity, another important aspect for the catalytic performance is the selectivity of NRR [22]. Due to the high content of protons in the acidic solution, the major competitive reaction of NRR is the hydrogen evolution reaction (HER). The adsorption energy of *H on Fe3@C2N is −0.94 eV, weaker than Eads of N2 (−1.45 eV), and expectedly, the charge transfer between *H and the catalyst is smaller (0.43 |e| vs. 1.14 |e|). By comparing the energies of two *H on Fe3@C2N (−1.69 eV) and the formation of H2 (1.79 eV, illustrated in Figure S7), we found that the reaction energy of HER is as high as 3.45 eV, indicating that Fe3@C2N has a high ability to suppress the competing HER during NRR under an acidic environment. The stable and metallic Fe3@C2N not only possesses low NRR limiting potential but also exhibits high selectivity against HER, and therefore, the low-cost Fe3@C2N is a highly promising electrocatalyst for NRR.

Materials and Methods
The spin-polarized density functional theory (DFT) calculations were carried out using the Vienna Ab initio Simulation Package (VASP 5.4.4) [42]. The exchange-correlation functional was described by the Perdew, Burke, and Ernzerhof (PBE) parameterization of the generalized gradient approximation (GGA) [43]. The projector augmented-wave (PAW) potential [44] was employed, and an electron configuration of 3d 7 4s 1 was adopted for Fe. According to our tests, we found that the system is ferromagnetic, with each Fe atom carrying a magnetic moment of ~3 μB ( Figure S2), which In addition to the catalytic activity, another important aspect for the catalytic performance is the selectivity of NRR [22]. Due to the high content of protons in the acidic solution, the major competitive reaction of NRR is the hydrogen evolution reaction (HER). The adsorption energy of *H on Fe 3 @C 2 N is −0.94 eV, weaker than E ads of N 2 (−1.45 eV), and expectedly, the charge transfer between *H and the catalyst is smaller (0.43 |e| vs. 1.14 |e|). By comparing the energies of two *H on Fe 3 @C 2 N (−1.69 eV) and the formation of H 2 (1.79 eV, illustrated in Figure S7), we found that the reaction energy of HER is as high as 3.45 eV, indicating that Fe 3 @C 2 N has a high ability to suppress the competing HER during NRR under an acidic environment. The stable and metallic Fe 3 @C 2 N not only possesses low NRR limiting potential but also exhibits high selectivity against HER, and therefore, the low-cost Fe 3 @C 2 N is a highly promising electrocatalyst for NRR.

Materials and Methods
The spin-polarized density functional theory (DFT) calculations were carried out using the Vienna Ab initio Simulation Package (VASP 5.4.4) [42]. The exchange-correlation functional was described by the Perdew, Burke, and Ernzerhof (PBE) parameterization of the generalized gradient approximation (GGA) [43]. The projector augmented-wave (PAW) potential [44] was employed, and an electron configuration of 3d 7 4s 1 was adopted for Fe. According to our tests, we found that the system is ferromagnetic, with each Fe atom carrying a magnetic moment of~3 µ B (Figure S2), which is quite similar to the case of a Fe 3 cluster on the θ-Al 2 O 3 (010) surface [25]. A cutoff energy of 550 eV was adopted. The van der Waals interactions were described using the empirical correction in the Grimme scheme (DFT-D2) [45], where the dispersion energy is corrected based on the pairwise atomic -C/R 6 terms. The D2 scheme was widely used in these theoretical studies for the oxygen reduction reaction [29] and NRR [2]. During the structure relation, the maximum force and energy on each atom were less than 0.01 eV/Å and 10 −5 eV, the width of smearing was chosen as 0.2 eV, and k-points were sampled using the 5 × 5 × 1 Monkhorst−Pack mesh [46]. To avoid the interaction between two neighboring surfaces, a vacuum space over 20 Å was used [2,40]. To simulate a C 2 N monolayer, a periodic 1 × 1 unit cell (with a lateral dimension of~8.34 Å) containing 12 carbon atoms and 6 nitrogen atoms was constructed. The barriers of the potential limiting steps were identified by using the climbing image nudged elastic-band (CI-NEB) method [47].
The average binding energy (E b ) of each Fe atom on the C 2 N substrate in our work is given by where E Fe 3 @C 2 N , E C 2 N and E Fe are the energies of the Fe 3 @C 2 N system, the C 2 N monolayer, and an isolated Fe atom, respectively. The adsorption energy (E ads ) of the single atom or NRR intermediates was determined according to the following equation: where E tot , E Fe 3 @C 2 N and E adsorbate represent the total energies of the systems containing the Fe 3 @C 2 N catalyst and the adsorbate, Fe 3 @C 2 N, and the adsorbate, respectively. According to the calculated hydrogen electrode (CHE) model proposed by Nørskov and co-workers [48], the reaction free energy of each basic step ∆G is calculated by In the above equation, ∆E is the reaction energy difference between the products and reactants of the NRR occurring on the catalyst, which can be directly obtained from DFT computations, and ∆E ZPE is the change in zero-point energies, which was calculated from the vibrational frequencies [12]. The vibrational frequencies were calculated based on the finite differences method, and only the adsorbed species were included. ∆S is the change in entropy at 298.15 K. Among them, the free energy correction of pH is expressed by ∆G pH , which is equal to k B T × ln10 × pH, where k B is the Boltzmann constant. In this study, it is assumed that pH = 0.
The limiting potential (η) of the entire reduction process is determined by the potential limiting step, which has the most positive ∆G (∆G Max ), as computed by η = −∆G Max /e [49].
The reaction rates of the key reaction steps were estimated based on Nørskov's model [48], provided that there is no extra barrier: k = k 0 exp(−∆G/k B T), where ∆G is the free energy change and k 0 is the prefactor. The approximated values are given in Table S3.

Conclusions
In summary, we designed a supported catalyst, i.e., Fe 3 clusters anchored on a two-dimensional C 2 N monolayer (Fe 3 @C 2 N), to electrocatalyze the reduction of nitrogen to ammonia. Through our density functional theory computations, we found that Fe 3 @C 2 N is excellent as an electrocatalyst for NRR owing to the stability at high temperature (800 K), the metallic feature, the absence of precious metal, the low limiting potential (0.57 eV of the enzymatic mechanism) and the high selectivity against the competing side reaction, HER. Such superiority can be ascribed to the partially occupied d orbitals and largely negative charged Fe 3 cluster, which is beneficial in activating the inert N≡N triple bond. Since it was feasible to reach Cu 2 @C 2 N using CuCl 2 as a metal precursor in a previous theoretical protocol [50], Fe 3 @C 2 N could be synthesized using FeCl 2 and other proper iron precursors. We hope that our work can inspire more experimental and theoretical studies to further explore the potential of non-precious metal clusters for NRR.
Supplementary Materials: The following are available online at http://www.mdpi.com/2073-4344/10/9/974/s1: Figure S1: Top and side views of the final structure of Fe 3 @C 2 N through a 10 ps FPMD simulation at 800 K; Figure S2: The magnesium distribution of Fe 3 @C 2 N; Figure S3: Density of states (DOS) of C 2 N and Fe 3 @C 2 N; Figure S4: The energy diagram of N 2 dissociation on Fe 3 @C 2 N; Figure S5: The energy diagram of NRR via the surface-hydrogenation mechanism on Fe 3 @C 2 N; Figure S6: The reaction pathway of the potential limiting step of the enzymatic mechanism and the consecutive mechanism; Figure S7: The energy diagram of the hydrogen evolution reaction (HER0 on Fe 3 @C 2 N; Table S1: The calculated zero-point energy (ZPE) and entropy of different molecules (T = 298.15 K and P = 101.325 Pa, in eV); Table S2: The calculated E(DFT), ZPE and TS of the intermediates (T = 298.15 K and P = 101.325 Pa, in eV); Table S3: The estimated reaction rate of each step on different paths (k 0 is the prefactor of the estimated reaction rate k = k 0 exp(−∆G/kT)).