Single-Atom Anchored g-C3N4 Monolayer as Efficient Catalysts for Nitrogen Reduction Reaction

Electrochemical N2 reduction reaction (NRR) is a promising approach for NH3 production under mild conditions. Herein, the catalytic performance of 3d transition metal (TM) atoms anchored on s-triazine-based g-C3N4 (TM@g-C3N4) in NRR is systematically investigated by density functional theory (DFT) calculations. Among these TM@g-C3N4 systems, the V@g-C3N4, Cr@g-C3N4, Mn@g-C3N4, Fe@g-C3N4, and Co@g-C3N4 monolayers have lower ΔG(*NNH) values, especially the V@g-C3N4 monolayer has the lowest limiting potential of −0.60 V and the corresponding limiting-potential steps are *N2+H++e−=*NNH for both alternating and distal mechanisms. For V@g-C3N4, the transferred charge and spin moment contributed by the anchored V atom activate N2 molecule. The metal conductivity of V@g-C3N4 provides an effective guarantee for charge transfer between adsorbates and V atom during N2 reduction reaction. After N2 adsorption, the p-d orbital hybridization of *N2 and V atoms can provide or receive electrons for the intermediate products, which makes the reduction process follow acceptance-donation mechanism. The results provide an important reference to design high efficiency single atom catalysts (SACs) for N2 reduction.


Introduction
Ammonia (NH 3 ) is not only an essential substance to produce fertilizers, explosives, dyes, and pharmaceuticals, but also an important clean energy carrier [1][2][3][4]. The growing demand of NH 3 has spurred researchers to seek more efficient artificial nitrogen (N 2 ) fixation. Presently, large-scale production of NH 3 mainly depends on the Haber-Bosch process. However, the process not only needs to operate under harsh conditions (300-500 • C and 200-300 atm) [5], but also requires huge energy input, and simultaneously generates a large number of greenhouse gases carbon dioxide (CO 2 ) [6][7][8]. Considering energy consumption and environmental protection, researchers hope to find a method for converting N 2 to NH 3 using renewable energy, no polluting emissions, and mild operating conditions. Electrochemical N 2 reduction reaction (NRR), a promising approach for sustainable NH 3 production under ambient conditions, has received extensive and increasing attention both in experiment and theoretic studies [9][10][11]. However, the strong bonding energy of N≡N triple bond (945 KJ/mol) and the weak adsorption of nonpolarized N 2 are two major challenges in the NRR process [12][13][14]. Therefore, an efficient catalyst for N 2 activation and reduction is urgent to improve NRR activity. On the other hand, the transition metal can accept the lone pair electrons of the N 2 molecule and weak the N≡N triple bond [15][16][17]. Dispersing metal atoms on suitable supporting materials can not only provide more active sites, but also regulate the electronic properties of substrate and enhance the catalytic efficiency. Two-dimensional materials (2D) are widely used as catalyst substrates for N 2

Computational Methods
DFT methods are performed using the Vienna Ab Initio Simulation Package (5.4.4, 2017, VASP Software GmbH, Vienna, Austria) [36,37]. The exchange-correlation potentials are described through the Perdew-Burke-Ernerhof (PBE) parametrization within the generalized gradient approximation (GGA) [38,39]. The DFT-D3 method is utilized to describe the weak van der Waals (vdW)-like interaction [40]. The kinetic energy cutoff of plane-wave expansion is set to 520 eV and a vacuum space of 20 Å is inserted in the z-direction to eliminate the interaction between adjacent periodic units [41]. A 3 × 3 × 1 supercell is constructed. For geometrical optimizations, a 1 × 1 × 1 Monkhorst-Pack mesh of k-points is employed to sample the first Brillouin zero. A finer 5 × 5 × 1 k-points grid is chosen for the density of states (DOS). The convergence tolerances for the force and energy are set to 0.02 eV/Å and 10 −5 eV, respectively. Bader charge is used analyze electron transfer [42].
The adsorption energies (E ads ) of single TM atom and N 2 molecule on substrate are calculated by the following formula: (1) where E adsorbate@substrate , E substrate , and E adsorbate are the total energies of adsorbate @substrate, substrate, and an isolate adsorbate, respectively. Negative E ads indicates that the adsorbate can be stably bond to the substrate. The cohesive energy (E coh ) is calculated by the following formula: where E TM and E TM,bulk are the energies of an isolated TM atom and bulk metal, respectively. Computational hydrogen electrode (CHE) mode [43] is used to calculate the reaction free energy (∆G). A solvation model is used to simulate the solution environment [44], and a relative permittivity of 80 is set for water [45]. The ∆G is evaluated as follows: where ∆E is the reaction energy directly obtained by DFT calculations, ∆ZPE and ∆S are the changes of zero-point energy (ZPE) and entropy computed from the vibrational frequencies at 298.15 K. ∆G U is the free energy contribution with applied potential U, ∆G U = eU, where e and U are the number of electrons transferred and the applied electrode potential. ∆G pH = k B × pH × ln10, where pH = 0 in this work for simplicity. The limiting potential (U L ) is defined as U L = −∆G max /e. Using the CHE model, the lowest ∆G(pH, U) was evaluated. The ∆G(pH, U) values can be projected onto the (pH, U) plane of lowest Gibbs free energy, resulting in a 2D plot with pH and U descriptors to build the Pourbaix diagram [46]. The Pourbaix diagrams can provide the reliability electrocatalytic reaction results under certain conditions of pH and potential U.

Structural and Electronic Properties of TM@g-C 3 N 4
Optimized structure of s-triazine-based g-C 3 N 4 monolayer is displayed in Figure 1a. It contains two C-N bonds, the bond lengths d C-N are 1.33 Å and 1.46 Å, respectively. The optimized lattice constant is g-C 3 N 4 is 4.78 Å, which is consistent with previous theoretical results [47,48]. The symmetrical distribution of spin-up and spin-down electrons in Figure 1b indicates that the g-C 3 N 4 is a non-magnetic semiconductor. TM (Sc-Cu) atom is anchored at five high-symmetry sites (Figure 1a) to investigate the stable configurations, namely, the top sites of C (T 1 ), pyridine N (T 2 ), and graphite N (T 3 ), hollow sites above the three pyridine N (H 1 ) and the triazine ring (H 2 ). Taking Mn atom as an example, the calculated adsorption energies (E ads ) are −5.16 eV, −6.01 eV, −4.49 eV, −6.20 eV, −3.70 eV, respectively, and the H 1 site is the most stable adsorption site with the lowest E ads −6.20 eV. All the TM atoms have a same stable adsorption site H 1 . The stable configuration of Mn atom adsorbed at H 1 site is shown in Figure 1c. The adsorption of Mn atom causes the distortion of g-C 3 N 4 monolayer, but no bond breaks. The other configurations of TM@g-C 3 N 4 are displayed in Figure S1.
The calculated adsorption energies (E ads ), energy difference between adsorption energy and cohesive energy (∆E), total magnetic moments (M tot ), the charge (∆Q) transferred from TM atoms to g-C 3 N 4 monolayer, and the electronic structures (ES) of TM@g-C 3 N 4 are shown in Table 1. The premise of good catalytic performance of SAC is that single metal atom has suitable E ads on the substrate. The calculated E ads values of TM@g-C 3 N 4 are in the range of −9.55~−4.76 eV, which indicates that TM atoms can bind stably to the substrate. At the same time, the energy difference (∆E) between the adsorption energy (E ads ) and the cohesive energy (E coh ) is calculated to investigate the aggregation of TM atoms on g-C 3 N 4 . The value of ∆E is in the range of −5.36~−0.92 eV, and a negative value of ∆E means that the adsorption of metal atoms on g-C 3 N 4 is stronger than the cohesive of atoms. Hence, the aggregation of TM atoms on g-C 3 N 4 can be suppressed efficiently. TM (Sc-Cu) atoms anchored to g-C 3 N 4 holes have good stability. The calculated adsorption energies (Eads), energy difference between adsorption energy and cohesive energy (ΔE), total magnetic moments (Mtot), the charge (ΔQ) transferred from TM atoms to g-C3N4 monolayer, and the electronic structures (ES) of TM@g-C3N4 are shown in Table 1. The premise of good catalytic performance of SAC is that single metal atom has suitable Eads on the substrate. The calculated Eads values of TM@g-C3N4 are in the range of −9.55~−4.76 eV, which indicates that TM atoms can bind stably to the substrate. At the same time, the energy difference (ΔE) between the adsorption energy (Eads) and the cohesive energy (Ecoh) is calculated to investigate the aggregation of TM atoms on g-C3N4.
The value of ΔE is in the range of −5.36~−0.92 eV, and a negative value of ΔE means that the adsorption of metal atoms on g-C3N4 is stronger than the cohesive of atoms. Hence, the aggregation of TM atoms on g-C3N4 can be suppressed efficiently. TM (Sc-Cu) atoms anchored to g-C3N4 holes have good stability. Table 1. Adsorption energies (Eads, eV), energy difference (ΔE, eV) between adsorption energy and cohesive energy, total magnetic moments (Mtot, μB), the charge (ΔQ, e) transferred from the TM atoms to g-C3N4 monolayer, and the electronic structures (ES) (SM: semi-metallic, M: metallic and SC: semiconductor).  Figure S2. It can be seen that spin electrons are located at TM (TM= Sc, Ti, V, Cr, Mn, Fe, Co, Cu) and surrounding atoms. Compared with Figure 1b, it is obvious that the anchored TM atoms (TM= Sc, Ti, V, Cr, Mn, Fe, Co, Cu) induce the magnetic moment of TM@g-C3N4. V@g-C3N4 has the most spin densities, which is consistent with the largest moment (4.88 μB) in Table 1.

Eads
As is well known, the anchored TM atoms can regulate the electronic structure of substrate, and the catalytic performance of the catalyst is closely related to the electronic  The spin densities distribution of TM@g-C 3 N 4 are shown in Figure S2. It can be seen that spin electrons are located at TM (TM= Sc, Ti, V, Cr, Mn, Fe, Co, Cu) and surrounding atoms. Compared with Figure 1b, it is obvious that the anchored TM atoms (TM= Sc, Ti, V, Cr, Mn, Fe, Co, Cu) induce the magnetic moment of TM@g-C 3 N 4 . V@g-C 3 N 4 has the most spin densities, which is consistent with the largest moment (4.88 µ B ) in Table 1.
As is well known, the anchored TM atoms can regulate the electronic structure of substrate, and the catalytic performance of the catalyst is closely related to the electronic structure. Hence, the total density of states (TDOS) of TM@g-C 3 N 4 and partial DOS (PDOS) are calculated and shown in Figure S3. It is seen that the anchored TM (TM=Sc, Ti, V, Cr, Mn, Fe, Co, Cu) atoms induce the asymmetry of spin-up and spin-down electrons of TM@g-C 3 N 4 , which is consistent with the spin electrons distribution in Figure S2. Second, the adsorption of Cr, Mn, Co, and Ni atoms retain the semiconductor property, but the band gap decreases from 1.57 eV to 0.38 eV, 0.46 eV, 0.58 eV, and 1.23 eV, respectively. The TDOSs of Ti@g-C 3 N 4 , V@g-C 3 N 4 , and Fe@g-C 3 N 4 exhibit metallic properties and many electronic states are located near the Fermi level, which ensures a rapid transfer of charge in the reaction.

Adsorption of N 2 Molecule
It is well known that the N 2 adsorption on the catalytic is the first step to investigate the NRR performance, which reflects the sensitivity of the catalyst and determines the most favorable adsorption manner. Stable adsorption of N 2 is a prerequisite for the subsequent reaction to generate NH 3 .
The adsorption energies [E ads (*N 2 )] of TM@g-C 3 N 4 (TM = Sc-Cu) are illustrated in Figure 2a. For TM@g-C 3 N 4 (TM = Ti, V, Cr, Mn, Fe, Co, Ni, Cu), both end-on and sideon configurations can exist, while E ads (*N 2 ) (−1.76~−0.58 eV) of end-on adsorption N 2 molecule are lower than those (−1.45~−0.26 eV) of side on patterns, which means that end-on patterns are more energetically favorable than the side-on patterns. Therefore, in the following studies of this paper, only the favorable end-on adsorption manner is discussed. These results indicate that the TM atoms (TM = Ti, V, Cr, Mn, Fe, Co, Ni, Cu) anchored at g-C 3 N 4 are active sites and have strong ability to capture N 2 molecules. This result is consistent with the result calculated by Hu et al. [35] except for Ti atom.
trons of TM@g-C3N4, which is consistent with the spin electrons distribution in Figure S2. Second, the adsorption of Cr, Mn, Co, and Ni atoms retain the semiconductor property, but the band gap decreases from 1.57 eV to 0.38 eV, 0.46 eV, 0.58 eV, and 1.23 eV, respectively. The TDOSs of Ti@g-C3N4, V@g-C3N4, and Fe@g-C3N4 exhibit metallic properties and many electronic states are located near the Fermi level, which ensures a rapid transfer of charge in the reaction.

Adsorption of N2 Molecule
It is well known that the N2 adsorption on the catalytic is the first step to investigate the NRR performance, which reflects the sensitivity of the catalyst and determines the most favorable adsorption manner. Stable adsorption of N2 is a prerequisite for the subsequent reaction to generate NH3.
The adsorption energies [Eads(*N2)] of TM@g-C3N4 (TM = Sc-Cu) are illustrated in Figure 2a. For TM@g-C3N4 (TM = Ti, V, Cr, Mn, Fe, Co, Ni, Cu), both end-on and side-on configurations can exist, while Eads(*N2) (−1.76~−0.58 eV) of end-on adsorption N2 molecule are lower than those (−1.45~−0.26 eV) of side on patterns, which means that end-on patterns are more energetically favorable than the side-on patterns. Therefore, in the following studies of this paper, only the favorable end-on adsorption manner is discussed. These results indicate that the TM atoms (TM = Ti, V, Cr, Mn, Fe, Co, Ni, Cu) anchored at g-C3N4 are active sites and have strong ability to capture N2 molecules. This result is consistent with the result calculated by Hu et al. [35] except for Ti atom. In order to further study the adsorption of N2 on TM@g-C3N4, the distance between TM atom and the N2 molecule (DTM-N), and N-N bond length (LN-N) are displayed in Figure  2b. The DTM-N (TM = Ti, V, Cr, Mn, Fe, Co, Ni, Cu) are in the range of 1.76~2.09 Å and the LN-N are stretched to 1.13~1.14 Å, which indicates that the N≡N triple bonds are weakened and good for reduction reaction. For Sc@g-C3N4, the adsorption energies for end-on and side-on patterns are −0.01 eV and 0.52 eV, the DSc-N is 4.14 Å and the LN-N is 1.11 Å, which means that Sc@g-C3N4 has poor NRR activity. TM@g-C3N4 (TM = Ti, V, Cr, Mn, Fe, Co, Ni, Cu) will be considered as SAC candidates in the following studies.

N2 electrocatalytic Reduction Reaction
The second criterion for good NRR catalyst is that it has a lower limiting potential in the N2 reduction reaction. To evaluate the catalytic activities of TM@g-C3N4 (TM = Ti, V, In order to further study the adsorption of N 2 on TM@g-C 3 N 4 , the distance between TM atom and the N 2 molecule (D TM-N ), and N-N bond length (L N-N ) are displayed in Figure 2b. The D TM-N (TM = Ti, V, Cr, Mn, Fe, Co, Ni, Cu) are in the range of 1.76~2.09 Å and the L N-N are stretched to 1.13~1.14 Å, which indicates that the N≡N triple bonds are weakened and good for reduction reaction. For Sc@g-C 3 N 4 , the adsorption energies for end-on and side-on patterns are −0.01 eV and 0.52 eV, the D Sc-N is 4.14 Å and the L N-N is 1.11 Å, which means that Sc@g-C 3 N 4 has poor NRR activity. TM@g-C 3 N 4 (TM = Ti, V, Cr, Mn, Fe, Co, Ni, Cu) will be considered as SAC candidates in the following studies.

N 2 Electrocatalytic Reduction Reaction
The second criterion for good NRR catalyst is that it has a lower limiting potential in the N 2 reduction reaction. To evaluate the catalytic activities of TM@g-C 3 N 4 (TM = Ti, V, Cr, Mn, Fe, Co, Ni, Cu), the NRR mechanisms are investigated to obtain the potential-limiting step (PLS) and limiting potential.
Theoretical and experimental studies [49] on NRR show that the end-on adsorbed N 2 molecule on TM@g-C 3 N 4 (TM = Ti, V, Cr, Mn, Fe, Co, Ni, Cu) can be reduced to NH 3 via a distal or alternating mechanism. In the distal mechanism, the proton (H + ) continuously attacks the terminal N atom until the first NH 3 molecule is released and then the other. While for the alternating mechanism, the proton (H + ) alternating bonds the two N atoms to release NH 3 molecule.
Previous studies suggest that the first H + reaction * N 2 + H + + e − = * NNH is usually the potential-limiting step (PLS) for NRR [16,50]. The free energy barrier ∆G(*NNH) is calculated and displayed in Figure 3a. It can be seen that the ∆G(*NNH) values of V@g-C 3 N 4 (0.60 eV), Cr@g-C 3 N 4 (0.76 eV), Mn@g-C 3 N 4 (0.78 eV), Fe@g-C 3 N 4 (0.63 eV), and Co@g-C 3 N 4 (0.71 eV) are lower than that on Ru (0001) surface. On the other hand, since the hydrogen evolution reaction (HER) is a major competitor for NRR [17,51], the calculated free energies change of H [∆G(*H)] are shown in Figure 3b. It is seen that ∆G(*H) are positive for V@g-C 3 N 4 , Cr@g-C 3 N 4 , and Mn@g-C 3 N 4 , which means that H atom cannot be adsorbed on V, Cr and Mn atoms. For Fe@g-C 3 N 4 and Co@g-C 3 N 4 , the ∆G(*H) is negative, while the values of ∆G(*N 2 ) are lower than ∆G(*H) and they prefer to adsorb N 2 molecule. In conclusion, V, Cr, Mn, Fe, and Co atoms anchored by g-C 3 N 4 are selected as potential SACs for NRR by the second criterion.
limiting step (PLS) and limiting potential.
Theoretical and experimental studies [49] on NRR show that the end-on adsorbed N2 molecule on TM@g-C3N4 (TM = Ti, V, Cr, Mn, Fe, Co, Ni, Cu) can be reduced to NH3 via a distal or alternating mechanism. In the distal mechanism, the proton (H + ) continuously attacks the terminal N atom until the first NH3 molecule is released and then the other. While for the alternating mechanism, the proton (H + ) alternating bonds the two N atoms to release NH3 molecule.
Previous studies suggest that the first H + reaction N 2 + H + * + e − = NNH * is usually the potential-limiting step (PLS) for NRR [16,50]. The free energy barrier ΔG(*NNH) is calculated and displayed in Figure 3a. It can be seen that the ΔG(*NNH) values of V@g-C3N4 (0.60 eV), Cr@g-C3N4 (0.76 eV), Mn@g-C3N4 (0.78 eV), Fe@g-C3N4 (0.63 eV), and Co@g-C3N4 (0.71 eV) are lower than that on Ru (0001) surface. On the other hand, since the hydrogen evolution reaction (HER) is a major competitor for NRR [17,51], the calculated free energies change of H [ΔG(*H)] are shown in Figure 3b. It is seen that ΔG(*H) are positive for V@g-C3N4, Cr@g-C3N4, and Mn@g-C3N4, which means that H atom cannot be adsorbed on V, Cr and Mn atoms. For Fe@g-C3N4 and Co@g-C3N4, the ΔG(*H) is negative, while the values of ΔG(*N2) are lower than ΔG(*H) and they prefer to adsorb N2 molecule. In conclusion, V, Cr, Mn, Fe, and Co atoms anchored by g-C3N4 are selected as potential SACs for NRR by the second criterion. In order to explore the whole N2 reduction process, it is necessary to explore the PLS of the process. The PLS dominates the catalytic efficiency of the catalyst for NRR. The results show that V@g-C3N4 is the best catalyst with the lowest ΔG(*NNH) value 0.60 eV compared with Cr@g-C3N4, Mn@g-C3N4, Fe@g-C3N4, and Co@g-C3N4. The whole free energy diagrams of V@g-C3N4 during NRR process is further investigated with the favorable end-on adsorption manner.
The NRR free energy diagrams of V@g-C3N4 are displayed in Figure 4. Optimized adsorption structures of intermediate via the alternating and distal pathways are depicted in Figure S4. The free energy diagrams via alternating and distal mechanisms are very similar. During N2 reduction process, the N2 adsorption and protonation into *NNH along the two mechanisms are same. The adsorption process of * +N 2 = N 2 * is downhill with ΔG value −0.89 eV, which shows that V@g-C3N4 can spontaneously adsorb the N2 In order to explore the whole N 2 reduction process, it is necessary to explore the PLS of the process. The PLS dominates the catalytic efficiency of the catalyst for NRR. The results show that V@g-C 3 N 4 is the best catalyst with the lowest ∆G(*NNH) value 0.60 eV compared with Cr@g-C 3 N 4 , Mn@g-C 3 N 4 , Fe@g-C 3 N 4 , and Co@g-C 3 N 4 . The whole free energy diagrams of V@g-C 3 N 4 during NRR process is further investigated with the favorable end-on adsorption manner.
The NRR free energy diagrams of V@g-C 3 N 4 are displayed in Figure 4. Optimized adsorption structures of intermediate via the alternating and distal pathways are depicted in Figure S4. The free energy diagrams via alternating and distal mechanisms are very similar. During N 2 reduction process, the N 2 adsorption and protonation into *NNH along the two mechanisms are same. The adsorption process of * + N 2 = * N 2 is downhill with ∆G value −0.89 eV, which shows that V@g-C 3 N 4 can spontaneously adsorb the N 2 molecule. The step of * N 2 + H + + e − = * NNH is uphill with ∆G value 0.60 eV, which means that the protonation process is endothermic. The second proton-electron may bond with another N atom of *N 2 molecule to generate *NHNH (alternating mechanism) or the same N atom (distal mechanism) to generate *NNH 2 , and the hydrogenation in both ways are both uphill with ∆G value 0.48 eV and 0.13 eV, respectively. In the following steps of two mechanisms, ∆G values are all negative. For alternating mechanism in Figure 4a, the calculated ∆G values are −0.89, 0.60, 0.48, −0.44, −0.12, −1.81 eV, and −0.27 eV, respectively. The first protonation step has the highest ∆G values and is confirmed to be the PLS with an uphill value 0.60 eV. When the U L is applied to −0.60 V, all the reaction steps are exothermic, therefore, the limiting potential is −0.60 V. For distal mechanism in Figure 4b, the ∆G values are −0.89, 0.60, 0.13, −0.77, −0.55, −0.70, and −0.27 eV, respectively, and it is obvious that it has the same PLS and U L as the alternating mechanism. The calculated limiting potentials of V@g-C 3 N 4 are lower than those calculated by Hu et al., where the limiting potential of V@g-C 3 N 4 with lying-on adsorbed N 2 is −0.79 V, while that of standing-on pattern is −0.93 V [35]. The final desorption of the second *NH 3 molecule requires the absorption energy of 1.16 eV, and it can also be further protonated to form NH 4 + [52].
are both uphill with ΔG value 0.48 eV and 0.13 eV, respectively. In the following steps of two mechanisms, ΔG values are all negative. For alternating mechanism in Figure 4a, the calculated ΔG values are −0.89, 0.60, 0.48, −0.44, −0.12, −1.81 eV, and −0.27 eV, respectively. The first protonation step has the highest ΔG values and is confirmed to be the PLS with an uphill value 0.60 eV. When the UL is applied to −0.60 V, all the reaction steps are exothermic, therefore, the limiting potential is −0.60 V. For distal mechanism in Figure 4b, the ΔG values are −0.89, 0.60, 0.13, −0.77, −0.55, −0.70, and −0.27 eV, respectively, and it is obvious that it has the same PLS and UL as the alternating mechanism. The calculated limiting potentials of V@g-C3N4 are lower than those calculated by Hu et al., where the limiting potential of V@g-C3N4 with lying-on adsorbed N2 is −0.79 V, while that of standing-on pattern is −0.93 V [35]. The final desorption of the second *NH3 molecule requires the absorption energy of 1.16 eV, and it can also be further protonated to form NH4 + [52]. Pourbaix diagrams can provide an effective guidance for the electrocatalytic reactions, which will not be discussed in this article.

Origin of Catalytic Activity
To explore the excellent catalytic efficiency of V@g-C3N4, the electronic properties, charge density difference, and spin electrons distribution are investigated. The charge density difference is shown in Figure 5a. It can be seen that the electron densities of the three N atoms anchoring the V atom increases, and Bader charge analysis finds that the V atom loses 1.92 e, which makes the V atom be a good active site for trapping N2 molecule. Figure 5b shows that the spin density is mainly located at the V atom, which makes V@g-C3N4 possess a total spin moment of 4.88 μB. These results prove the conclusion that the magnetism of the catalyst can increase its activity. The density of states (DOS) is presented in Figure 5c. Different from the intrinsic s-triazine-based g-C3N4 monolayer of direct bandgap semiconductor in Figure 1b [47,48], V@g-C3N4 exhibits a metallic property owing to the V atom. The spin moment is vital to activate N2 molecule, and the excellent electrical conductivity is essential to ensure good charge transfer for efficient electroreduction reaction [48,53]. Both of them play an important role in the outstanding NRR catalytic performance of V@g-C3N4. Pourbaix diagrams can provide an effective guidance for the electrocatalytic reactions, which will not be discussed in this article.

Origin of Catalytic Activity
To explore the excellent catalytic efficiency of V@g-C 3 N 4 , the electronic properties, charge density difference, and spin electrons distribution are investigated. The charge density difference is shown in Figure 5a. It can be seen that the electron densities of the three N atoms anchoring the V atom increases, and Bader charge analysis finds that the V atom loses 1.92 e, which makes the V atom be a good active site for trapping N 2 molecule. Figure 5b shows that the spin density is mainly located at the V atom, which makes V@g-C 3 N 4 possess a total spin moment of 4.88 µ B . These results prove the conclusion that the magnetism of the catalyst can increase its activity. The density of states (DOS) is presented in Figure 5c. Different from the intrinsic s-triazine-based g-C 3 N 4 monolayer of direct bandgap semiconductor in Figure 1b [47,48], V@g-C 3 N 4 exhibits a metallic property owing to the V atom. The spin moment is vital to activate N 2 molecule, and the excellent electrical conductivity is essential to ensure good charge transfer for efficient electroreduction reaction [48,53]. Both of them play an important role in the outstanding NRR catalytic performance of V@g-C 3 N 4 .
In order to further explore the understanding NRR catalytic performance of V@g-C 3 N 4 , the charge density difference and spin electrons distribution after N 2 adsorption are calculated. As shown in Figure 6a, the electrons accumulate near the *N 2 molecule, and it is confirmed by Bader charge analysis that 0.42 e transferred from V atom to the N 2 molecule. The spin electrons distribution in Figure 6b reveals that after N 2 adsorption, some spin moment on V atom can be transferred to the N 2 molecule, so that the N 2 /V@g-C 3 N 4 only possesses a magnetic moment 1.00 µ B . The change in charge and spin magnetic moment drives its further activation and reactions. The DOS of N 2 /V@g-C 3 N 4 are displayed in Figure 6c. Obvious overlap between the *N 2 -2p and V-3d orbitals around the Fermi energy happens, in which the empty d orbitals of V can accept the lone-pair electrons in N 2 , at the same time, the occupied d orbitals can donate electrons to the antibonding orbitals of N 2 , and this process follows acceptance-donation mechanism. In order to further explore the understanding NRR catalytic performance of V@g-C3N4, the charge density difference and spin electrons distribution after N2 adsorption are calculated. As shown in Figure 6a, the electrons accumulate near the *N2 molecule, and it is confirmed by Bader charge analysis that 0.42 e transferred from V atom to the N2 mole- in Figure 6c. Obvious overlap between the *N2-2p and V-3d orbitals around the Fermi energy happens, in which the empty d orbitals of V can accept the lone-pair electrons in N2, at the same time, the occupied d orbitals can donate electrons to the antibonding orbitals of N2, and this process follows acceptance-donation mechanism. To further explore the charge transferred among intermediate adsorbent, the variations of the charge transferred between them are studied. Based on the previous studies [54,55], NxHy/V@g-C3N4 is divided into three parts ( Figure S5): g-C3N4 substrate without V and the three N atoms bonded with it (moiety 1), V-3N (moiety 2), and the adsorbed To further explore the charge transferred among intermediate adsorbent, the variations of the charge transferred between them are studied. Based on the previous studies [54,55], N x H y /V@g-C 3 N 4 is divided into three parts ( Figure S5): g-C 3 N 4 substrate without V and the three N atoms bonded with it (moiety 1), V-3N (moiety 2), and the adsorbed intermediate N x H y (moiety 3). The charge variations along distal and alternating mechanisms are shown in Figure 7 (the charge variation is the charge difference of each moiety between the present step and the previous step). For N 2 adsorption on V@g-C 3 N 4 monolayer in Figure 7a, V-3N offers 0.42 e and 0.45 e to *N 2 and g-C 3 N 4 , respectively. During the protonation process * N 2 + H + + e − = * NNH, V-3N and g-C 3 N 4 substrate provide 0.15 e and 0.19 e, respectively. However, during the second protonation process * NNH + H + + e − = * NNH 2 , V-3N gains 0.52 e from the substrate and *NNH 2 . In the following hydrogenation and reduction steps, obvious charge fluctuation occurs for the three moieties, which means that V-3N is the charge transferred medium between N x H y and g-C 3 N 4 substrate, providing or receiving electrons to the adsorbed intermediate. The above results once again prove that the excellent catalytic performance of V@g-C 3 N 4 in NRR process follows the acceptance-donation mechanism. The charge variation in the alternating mechanism shown in Figure 7b is similar to those the distal mechanism.
intermediate NxHy (moiety 3). The charge variations along distal and alternating mechanisms are shown in Figure 7 (the charge variation is the charge difference of each moiety between the present step and the previous step). For N2 adsorption on V@g-C3N4 monolayer in Figure 7a, V-3N offers 0.42 e and 0.45 e to *N2 and g-C3N4, respectively. During the protonation process N 2 + H + * + e − = NNH * , V-3N and g-C3N4 substrate provide 0.15 e and 0.19 e, respectively. However, during the second protonation process NNH + H + * + e − = NNH 2 * , V-3N gains 0.52 e from the substrate and *NNH2. In the following hydrogenation and reduction steps, obvious charge fluctuation occurs for the three moieties, which means that V-3N is the charge transferred medium between NxHy and g-C3N4 substrate, providing or receiving electrons to the adsorbed intermediate. The above results once again prove that the excellent catalytic performance of V@g-C3N4 in NRR process follows the acceptance-donation mechanism. The charge variation in the alternating mechanism shown in Figure 7b is similar to those the distal mechanism. In order to explore the change of adsorbates configuration and magnetic properties during the reduction of N2, the V-N distance (DV-N), N-N bond length (LN-N), and the total magnetic moments (Mtot) are examined and displayed in Figure 8. In Figure 8a, DV-N along distal pathway decreases during protonation and reaches a minimum when the first NH3 molecule is released, and then continuously increases. While in alternating mechanism, when the proton (H + ) attacks the distal N atom, DV-N decreases, and when it attacks the N atom bonded to V, DV-N increases. DV-N increases in a zigzagging fluctuation way during the whole reduction process. For LN-N, as shown in Figure 8b, it is stretched continuously until the first NH3 is released for both distal and alternating mechanisms. Figure 8c shows the magnetic moments of the NxHy, and it can be seen that the protonation process makes the Mtot show zigzag fluctuations for both mechanisms. These results provide strong evidence for the excellent catalytic performance of V@g-C3N4. In order to explore the change of adsorbates configuration and magnetic properties during the reduction of N 2 , the V-N distance (D V-N ), N-N bond length (L N-N ), and the total magnetic moments (M tot ) are examined and displayed in Figure 8. In Figure 8a, D V-N along distal pathway decreases during protonation and reaches a minimum when the first NH 3 molecule is released, and then continuously increases. While in alternating mechanism, when the proton (H + ) attacks the distal N atom, D V-N decreases, and when it attacks the N atom bonded to V, D V-N increases. D V-N increases in a zigzagging fluctuation way during the whole reduction process. For L N-N , as shown in Figure 8b, it is stretched continuously until the first NH 3 is released for both distal and alternating mechanisms. Figure 8c shows the magnetic moments of the N x H y , and it can be seen that the protonation process makes the M tot show zigzag fluctuations for both mechanisms. These results provide strong evidence for the excellent catalytic performance of V@g-C 3 N 4 .

Conclusions
In this paper, the catalytic performance of 3d TM atoms anchored on s-triazine-based g-C3N4 (TM@g-C3N4) for N2 reduction is systematically investigated. The results show that N2 molecules can be stably adsorbed on TM@g-C3N4 (TM=Ti, V, Cr, Mn, Fe, Co, Ni, Cu), and the end-on modes are more energetically favorable than the side-on modes. The ΔG(*NNH) values indicate that TM@g-C3N4 (TM = V, Cr, Mn, Fe, Co) are potential NRR candidates with lower ΔG(*NNH) value than that on Ru (0001) surface. The free energy diagrams show that V@g-C3N4 exhibits the highly catalytic performance for NRR with a limiting potential −0.60 V, and the PLS is the first protonation of *N2 for both alternating and distal mechanisms. In order to explore the origin of excellent NRR performance of V@g-C3N4, the charge density difference, spin electrons distribution and DOS plots before and after N2 adsorption, and their corresponding charge variation of adsorbates during N2 reduction reaction are discussed. First, for V@g-C3N4, V atom as an active site provide 1.92 e and has the ability to trap N2 molecule. A spin moment of 4.88 μB of V@g-C3N4 contributed by V atom can activate N2 molecule. Second, after N2 adsorption, the strong interaction between *N2-2p and V-3d orbitals ensures the charge transferred during N2 reduction reaction. Third, the charge variation of moieties indicate that V-3N acts as a medium between adsorbates NxHy and g-C3N4 substrate, which ensures the efficient reduction of N2. The DV-N, LN-N and Mtot of NxHy/V@g-C3N4 are calculated to illustrate the change of adsorbate configurations and magnetic properties during the N2 reduction reactions, which provides an important evidence for the excellent catalytic performance of SACs TM@g-C3N4.

Conclusions
In this paper, the catalytic performance of 3d TM atoms anchored on s-triazine-based g-C 3 N 4 (TM@g-C 3 N 4 ) for N 2 reduction is systematically investigated. The results show that N 2 molecules can be stably adsorbed on TM@g-C 3 N 4 (TM=Ti, V, Cr, Mn, Fe, Co, Ni, Cu), and the end-on modes are more energetically favorable than the side-on modes. The ∆G(*NNH) values indicate that TM@g-C 3 N 4 (TM = V, Cr, Mn, Fe, Co) are potential NRR candidates with lower ∆G(*NNH) value than that on Ru (0001) surface. The free energy diagrams show that V@g-C 3 N 4 exhibits the highly catalytic performance for NRR with a limiting potential −0.60 V, and the PLS is the first protonation of *N 2 for both alternating and distal mechanisms. In order to explore the origin of excellent NRR performance of V@g-C 3 N 4 , the charge density difference, spin electrons distribution and DOS plots before and after N 2 adsorption, and their corresponding charge variation of adsorbates during N 2 reduction reaction are discussed. First, for V@g-C 3 N 4 , V atom as an active site provide 1.92 e and has the ability to trap N 2 molecule. A spin moment of 4.88 µ B of V@g-C 3 N 4 contributed by V atom can activate N 2 molecule. Second, after N 2 adsorption, the strong interaction between *N 2 -2p and V-3d orbitals ensures the charge transferred during N 2 reduction reaction. Third, the charge variation of moieties indicate that V-3N acts as a medium between adsorbates N x H y and g-C 3 N 4 substrate, which ensures the efficient reduction of N 2 . The D V-N , L N-N and M tot of N x H y /V@g-C 3 N 4 are calculated to illustrate the change of adsorbate configurations and magnetic properties during the N 2 reduction reactions, which provides an important evidence for the excellent catalytic performance of SACs TM@g-C 3 N 4 .
Author Contributions: H.C. and W.C.: investigation, writing-original draft, project administration; Z.F. and Y.L.: data curation, formal analysis; M.Z. and J.S.: software, validation, resources, methodology; Y.T. and X.D.: conceptualization, supervision, writing-review and editing and funding acquisition. All authors have read and agreed to the published version of the manuscript.

Data Availability Statement:
The data presented in this study are available on a reasonable request from the corresponding author.