Iron Dual-atom Catalyst combine with ″vicinal nonmetallic sites″ for efficient Ammonia Synthesis


 Ammonia synthesis from N2 under mild conditions is a long-term pursuit and goal, which theoretically limited by the Brønsted–Evans–Polanyi (BEP) relation in industrial transformation via the N2 dissociation. Here we show that the Fe2 catalyst combined with the “vicinal nonmetallic sites” may break the BEP limitation to fulfill the efficient ammonia synthesis. The catalyst supported on boron doped graphitic carbon nitride (Fe2/B/mpg-C3N4) strongly favors hydrogenation of *N2 to form a *NHNH2 species, which leads to low energy barriers for N-H formation (0.57eV) and N-N dissociation (0.51eV). Constructed B-N “Lewis pairs” on the mpg-C3N4 serve as nonmetallic sites can activate and transfer hydrogen, which reduce the competitive adsorption of N2 and H2. Through co-activated H2 on the vicinal site, synergistic Fe2 catalyst shows a significant advantage among Fen/mpg-C3N4 (n=2, 3, 4) catalysts and thus can avoid harsh reaction condition for the thermal conversion of N2 to NH3.


Introduction
Ammonia as a key precursor for nitrogenous fertilizers is crucial to ensure human survival and sustain the world's population. 1 Currently, ammonia synthesis is dominated by the industrial Haber-Bosch (HB) process using heterogeneous iron-based catalysts. 2 Although this important catalytic reaction has been developed for a hundred years, it still requires harsh reaction conditions which are intensely energy dependent. 3,4 Previous theoretical studies indicate that ammonia synthesis via the N 2 dissociative mechanism (N 2 +3H 2 →2N+3H 2 →→2NH 3 ) 5 abides by the Brønsted-Evans-Polanyi (BEP) relationship. [6][7][8] This implies the requirement for a moderate nitrogen adsorption strength on an applicable metal catalyst, enabling a compromise between N 2 dissociation and NH x desorption. In 2017, the Chen group proposed that a LiH-mediated Fe surface breaks the BEP limitation for ammonia synthesis 9 (Scheme 1a). In biological nitrogen xation processes, nitrogenase enzymes, containing FeMo, FeV, or FeFe cofactor, as catalytically active sites, 10 which are able to reduce N 2 at low temperatures and pressures probably proceeding by the associative mechanism of N 2 11, 12 rather than the dissociative mechanism as in the HB process. 5 A similar catalytic reduction of N 2 under mild conditions was also proposed for electrochemical ammonia synthesis, based on theoretical studies. [13][14][15] However, there are still rare examples of ammonia synthesis via the associative mechanism in thermal catalysis (N 2 +3H 2 →N 2 H+5/2H 2 →→2NH 3 ), and the recent theoretical progress seems to indicate this associative mechanism in the catalytic reactions is underestimated. [16][17][18][19] Skúlason et al 17 suggested that a certain proportion of associative mechanism of N 2 on Ru(001) surfaces may reduce the N-N activation energy and lead to more e cient production of NH 3 . Moreover, Li and coworkers revealed, through theoretical studies, that the associative hydrogenation mechanism has the potential to facilitate N 2 reduction by using small metal clusters as heterogeneous catalysts (Rh 1 Co 3 /CoO(011) or Fe 3 /θ-Al 2 O 3 (010)) (Scheme 1a). 18,19 Even so, because the hydrogenation of nitrogen in the associative mechanism may not be facile, the competition of the hydrogenation and N-N dissociation in NH x NH y (x=0-3, y=0-3) will become a key issue, which need to be solved for idea catalysts of ammonia synthesis. (Scheme 1b) Binuclear iron is very important in biological enzymes, [20][21][22][23][24][25][26] such as [FeFe]-hydrogenase 20,25, and diiron monooxygenase, 21,22 etc. In view of their superior performance in biology, binuclear iron complexes as mimic enzymes continue to intrigue interests, 27-33 especially for compounds used in N 2 activation. [29][30][31][32][33] Under the development of single-cluster catalysts, diatomic catalysts comprising the smallest type of metal clusters may provide a unique platform for bridging heterogeneous and homogeneous catalysis. 34,35 Although diatomic catalysts remain a relatively obscure area, they have emerged with great potential for increasing catalytic performance. 34 Recently, a series of highly dispersed dimer clusters supported on mesoporous carbon nitride (mpg-C 3 N 4 ) have been successfully prepared and it has been demonstrated that the obtained Fe 2 /mpg-C 3 N 4 sample has superior catalytic performance for alkene epoxidation. 48 Because the "Fe 2 N 4 " structural center is similar to the low-coordinated iron complexes for the N 2 reduction, 44-46 diatomic iron models Fe 2 /mpg-C 3 N 4 were initially established to explore the high e cient catalyst for ammonia synthesis. In this work, we predict that Fe 2 diatomic catalyst strongly favors hydrogenation of *N 2 to form a *NHNH 2 species with the assistance of co-activated H 2 on the ''vicinal nonmetallic sites'', which leads to a low energy barrier for the N-N dissociation. In addition, the partial doping of boron on the carbon sites of the support was designed in Fe 2 /B/mpg-C 3 N 4 to overcome the relatively high energy barrier of the hydrogen transfer and promote the hydrogenation of nitrogen (Scheme 1b). We also further reveal that with increasing coordination number of Fe n /mpg-C 3 N 4 (n=2,3,4), the associative mechanism tended to be less competitive, and the dissociative mechanism became dominant in ammonia synthesis, which is controlled by the spin polarization of Fe n . The Fe 2 cluster catalyst with its support strongly involved may provide an alternative perspective to bridge heterogeneous and homogeneous catalysis that bestows the potential for e cient ammonia synthesis.

Results
The "vicinal nonmetallic sites" -promoted Hydrogenation Mechanism Inspired by the low-coordinated iron complexes of the "Fe 2 N 4 " for the N 2 reduction, 44-46 a diatomic catalyst, Fe 2 /mpg-C 3 N 4 , was developed for the thermal conversion of N 2 to NH 3 . The favorable con guration of Fe 2 /mpg-C 3 N 4 is shown in Figure 1a, where each Fe atom binds with two adjacent nitrogen atoms forming a at-lying Fe 2 right above mesoporous graphitic carbon nitride. The binding energy of Fe 2 with mpg-C 3 N 4 is thermodynamically stable at -7.53eV due to the back donation interactions of iron d orbitals to π* orbitals of the mpg-C 3 N 4 support, substantiated by +1.08|e| Bader charges. The Fe 2 moiety is calculated with magnetic moment of 6μB. Such low oxidation state and high spin polarization characteristics on Fe 2 facilitate to activate N 2 . 47 Our investigation started with the adsorption of N 2 to understand the in uence of the activation of nitrogen for the reaction mechanism. The adsorptions of N 2 on Fe 2 /mpg-C 3 N 4 were calculated in three models that were further veri ed by Ab Initio molecular dynamics (AIMD) simulations (Figure 1b and Figure S1-2). The most stable con guration is the side-on/side-on (μ-ŋ2:ŋ2) mode with E ads = -1.52eV, which is different from the known "Fe-NN-Fe" complexes, 28, 29, 46, probably due to the change in geometry impacting the Fe−NN-Fe bonding interaction. The calculated energies for side-on/end-on (μ-ŋ2:ŋ1) and terminal end-on (ŋ1:ŋ1) modes were 0.02 and 0.23eV less stable, respectively, see Table S1 for details. On close inspection of the μ-ŋ2:ŋ2 mode, the bond length of four Fe-N bonds varied from 1.92 to 1.98Å. The remarkably elongated N-N bond for chemisorbed N 2 (from 1.11 Å of free N 2 to 1.25 Å) can be ascribed to the back donation interactions of iron d orbitals to two π* orbitals of N 2 , supported by the increase of negative charge (-0.86|e|) on the *N 2 . Therefore, the μ-ŋ2:ŋ2 mode was used for the following mechanistic investigations of N 2 hydrogenation and N-N dissociation. 48 Since the calculated N-N dissociative mechanism in Fe 2 /mpg-C 3 N 4 is particularly unfavorable via a 2.75eV energy barrier of N-N cleavage ( Figure S3), the associative hydrogenation of N 2 will be an alternative choice. In contrast to the electro-catalytic reaction where protons and electrons are transferred directly to the adsorbed nitrogen, H 2 activation and transfer are important in the thermal catalytic hydrogenation of nitrogen. Due to lack of binding site on the Fe 2 cluster, H 2 molecules prefer to approach and absorb dissociatively on "C-N" site in the support rather than the Fe site with reaction energy -2.30eV versus -1.26eV, respectively ( Figure 1c and Figure S4). 49,50 The dissociation of H 2 tends to be hemolytic,  Figure 2a). Subsequently, the Fe1 site favors accepting the second *H on the C1 site by overcoming an activation energy barrier (E a ) of 1.38eV rather than the *H on the N2 site (a4-a5, E a =1.91eV), see Figure S5 for more details. The synergistic activation of H 2 at "vicinal nonmetallic sites" not only reduces the adsorption competition between N 2 and H 2 , but also promotes the subsequent hydrogenation process of N 2 (A further discussion of hydrogen activation and transfer are in the last Section.).
After forming the *NNH intermediate (a5 in Figure 2), four different hydrogenation pathways are proposed to generate NH 3 , dependent on different combinations of alternative/distal hydrogenation on *N 2 H and the N-N bond dissociation of *NH x NH y (x=0-3, y=0-3) intermediates ( Figure 2 and Schematic depiction in Figure S6). It is favorable for *NNH to continue hydrogenation rather than to break the N-N bond due to the relatively high dissociation energy barrier (E a =1.10eV in a5→d6, AH3). The following hydrogenation favors the formation of *HN-NH via an alternate hydrogenation (AH) mode with a 0.36eV energy barrier (a5→b6 in AH1), which is 0.30eV lower than the formation of *N-NH 2 via distal hydrogenation (DH) mode (a5→a6 in DH1). Then, the Fe1 site can accept and activate additional H 2 by breaking the Fe-N bond of *HN-NH to form co-adsorption con guration c7, which is ready to further transfer H* to nitrogen bonded to Fe2 site (c7→c8 via 0.36eV energy barrier in AH2). The resulting *NHNH 2 (c8) is unusual example in the thermal catalytic reaction, although it is very common in electro-and enzyme catalysis via the with 0.50eV energy barrier is facile and not a rate-limiting step, which is distinguished from direct cleavage of nitrogen in the Haber-Bosch process. 6,7 Subsequently, the remaining H* on the Fe1 center is further transferred to the μ-*NH forming the μ-*NH 2 species. The last H 2 would still be activated by the Fe1 center, and delivers the active H* species to generate two NH 3 . Since the hydrogenation mechanism of nitrogen in Fe 2 /mpg-C 3 N 4 relies on synergistic catalysis of Fe 2 active sites and "vicinal nonmetallic sites", we call it "vicinal nonmetallic sites" promoted hydrogenation mechanism. The key intermediates of the preferred reaction pathway are illustrated in Figure S7. More details for four reaction pathways such as optimized intermediates, transition states and energy diagrams are also given in Figure S8-11.
In fact, the reaction pathways calculated above mainly include two types of reactions: N-N bond dissociation and N-H bond formation. Energy barriers for N-N bond dissociations will signi cantly decrease depending on the gradual hydrogenations to form different *NH x NH y (N-NH, NH-NH, N-NH 2 and NH-NH 2 ) intermediates in Figure 2b. This can be rationalized by electron accumulations on related *N 2 and elongated N-N bond lengths during the hydrogenation (Table S2). Therefore, promoting the hydrogenation of *N 2 would be an applicable way to lower the energy barrier of N-N dissociation. As shown in Figure 3, the complicated N-H bonds formation reactions in different pathways can be resolved to two kinds according to whether the N-N bond is dissociated or not. The nitrogen hydrogenation energy barriers on Fe 2 clusters are low and vary from 0.22 to 0.66eV, probably due to the electronic structures of *N 2 and the proper Fe-H bond strength. In addition, it is more feasible for the Fe 2 cluster with low positive Bader charges to transfer their hydrogen (e.g. a11-a12/c9-c10 or a8-a9/b11-a12, etc. in Figure 3). Overall, Fe 2 /mpg-C 3 N 4 could be a potential catalyst for N 2 -to-NH 3 conversion with a lower hydrogenation barrier (0.36eV in TS-a5-b6 and 0.36eV in TS-c7-c8) and a lower N-N bond dissociation barrier (0.50eV in TS-c8-c9) in the AH2 reaction pathway.

The Mechanisms of Nitrogen Reduction on various Iron clusters
As diverse multinuclear systems in active sites can signi cantly change the adsorption mechanism and catalytic performance, we preformed more mechanistic studies of nitrogen reduction by using the newly designed Fe 3 /mpg-C 3 N 4 , 51 Fe 4 /mpg-C 3 N 4 and the Fe (211) surface 19 for the comparisons with the Fe 2 /mpg-C 3 N 4 system. The reaction pathways and corresponding structures can be obtained in Figure   S12-S13. Catalytic N-H bond formation and N-N bond dissociation as key reaction steps are illustrated in Figure 4 to probe the relationship between mechanisms and iron coordination numbers (Fe n ) of the adsorption sites. 52  As shown in Figure 5 and Figure S14, key reaction pathways of related iron clusters indicate the Fe 3 /mpg-C 3 N 4 and Fe 4 /g-C 3 N 4 systems tend to form *NNH via *N 2 hydrogenation (0.95eV in Fe 3 /mpg-C 3 N 4 and 1.32eV in Fe 4 /mpg-C 3 N 4 ), followed by N-NH bond cleavage with energy barriers of 0.69eV and 0.31eV respectively. The continued hydrogenations to form *NH-NH or *N-NH 2 were unfavorable due to high energy barrier ( Figure S12-13). These mechanistic results in Fe n /mpg-C 3 N 4 (n=3,4) are consistent with results in Fe 3 /Al 2 O 3 clusters from Li's group. 19 There might be an energy crossover point in the Fe 4 /mpg-C 3 N 4 , because the energy barriers of N-N bond cleavage (1.34eV) and N-H bond formation (1.32eV) are quite close, indicating that Fe coordination number over four will cause the dissociative mechanism of nitrogen to dominate. For example, the energy barrier for the direct dissociation of the N 2 molecule is only 0.50eV at the Fe (211) C7 site. 19 On the contrary, the Fe 2 /mpg-C 3 N 4 catalyzed system favors continuous hydrogenation of *NNH until the formation of *NHNH 2 , which can promote the breaking of the N-N bond.
Such a synergistic associative mechanism can break the traditional BEP relationship with low N-N dissociation energy barrier and low NH x adsorption energy compared with metal surfaces 53, 54 such as Figure 5b. We can rationalize the results through charges of *N 2 as shown in their relationship with energy barriers of N-N dissociation and N-H formation (Figure 5c). With more coordination of Fe n , the negative charges of *N 2 will increase that enable N-N bond cleavage to be facile, but inhibit *N 2 hydrogenation via Fe-hydride intermediates. Therefore, the fewer electrons on nitrogen, the more favorable it is for the hydride to transfer hydrogen to adsorbed *NH x NH y (x=0-1, y=0-2) on the iron sites.

Electronic structure analysis
Besides the Bader charge analysis, projected density of states (PDOS) of adsorbed N 2 on different clusters can provide further explanation. As shown in Figure 6a, the α-spin orbitals of the Fe 2 clusters are much lower energy than the 2π* orbitals of N 2 , which cannot ful ll the orbital interaction. The well matched β-spin d orbital of Fe 2 can partially donate its electron to the 2π*orbital of N 2 forming a β-spin dπ* interaction, which leads to the strong spin polarization of the *N 2 . This result was further supported by differential charge density and spin density (Figure 6b). Through fragment orbital analysis between an isolated Fe 2 cluster and N 2 in Figure 6c and Figure S15, the occupied β-d xy-xy /β-d xz-xz orbitals of the Fe 2 species can interact with empty nitrogen 2π* orbitals to form two bonding orbitals (β-d xy-xy +π* and β-d xzxz +π*), which is consistent with molecular orbital interaction between nitrogen and Fe 2 /mpg-C 3 N 4 ( Figure  S16) and indicate the electron transfer from Fe 2 to nitrogen. Therefore the strong spin polarization of the activated *N 2 will facilitate to accept the electron and therefore make hydrogen transfer process on the metal-hydride accessible. In terms of the Fe coordination number increase (Figure 6a), the α-spin orbitals of Fe n clusters have shifted to relatively high energy levels and match with the energy level of the 2π* orbitals of N 2 . The additional electron transfer from α-spin orbitals of Fe clusters will lead to an increase in the electron and a decrease in the spin polarization for *N 2 , further weakening of the bond strength of *N 2 , which promotes the N-N bond dissociation. When the coordination numbers of Fe is over 4, such as C7 site of Fe (211), the energy barrier for direct cleavage of N-N bond is favorable. For Fe n /g-C 3 N 4 (n=2, 3) catalysts, the N 2 associative mechanism becomes dominant in nitrogen reduction to ammonia, supported by previous discussion in Figures 4 and 5.
To understand the unusual hydrogenation mechanism of N 2 on the smallest cluster-Fe 2 /mpg-C 3 N 4 , the projected density of states (PDOS) and spin densities for key intermediates were shown in Figure 7. With continuous hydrogenation, both Fe 2 clusters and the absorbed hydrogens transfer their electrons to the N 2 2π* orbitals, which decreased the energy of the N-N anti-bond orbital and activates the N-N bond in Figure 7a. The formed *NNH has relatively high spin polarization to make the next hydrogenation favorable, forming *NHNH via an alternate pathway. Because the α-spin electrons of the N 2 2π*orbital keep increasing from *NN to *NHNH 2 , the spin density for each intermediate is gradually eroded until *NHNH 2 with no obvious spin density. After that, the following transformation would favor the dissociation of the N-N bond rather than additional hydrogenation. In contrast, the spin density of *NNH in Fe 3 /g-C 3 N 4 is relatively low, hindering further hydrogenation and making N-N bond cleavage accessible.
The e cient catalyst by B doping mpg-C3N4 Although the iron diatomic cluster shows a catalytic advantage in the synergistic associative mechanism for ammonia synthesis, the hydrogen transfer from the C1 site of the mpg-C 3 N 4 support to Fe active sites seems unfavorable (E a (C-H) = 1.38eV in a4 a5 of Figure 8a and Figure 2), which might severely hinder the following transformation. To accelerate this step, we tried to modulate hydrogen adsorption by doping other heteroatoms (including B, O, and N) at the C1 site as shown in Figure S17 and Figure 8. Theoretical simulation indicates that doping with B atoms would not only maintain the stability of the structure, but also promote the activation and conversion of H 2 .
In B-doped Fe 2 /mpg-C 3 N 4 (Fe 2 /B/mpg-C 3 N 4 ), two H 2 molecules were absorbed dissociatively on the C-N and B-N sites with adsorption energy of -2.33eV (a2 in Figure 8b). Different from the C-N site on the support, the dissociation of H 2 in the B-N site tends to be heterolysis forming B-Hand N-H + that are supported by the -0.57|e| and +0.45|e| charges on hydrogens, respectively. The *H stabilized by the electron-de cient B atom (as Lewis acid) will change the transfer order of *H on supports acting as a noninnocent ligand in homogenous catalysis. 55,56 In the Fe 2 /B/mpg-C 3 N 4 , *H on the B site would preferentially transfer to the Fe1 site with an energy barrier of 0.75eV and continue to attack the activated *N 2 to form the *N 2 H intermediate. After that, the *H on the N1 site will shift to the Fe1 site via a 0.86eV barrier that is slightly higher energy compared to that in the Fe 2 /mpg-C 3 N 4 system (0.75eV). However, the overall hydrogen transfer process by doping the support with B would reduce the energy barrier from 1.38eV to 0.86eV. This designed B-N system can be regarded as a typical "Lewis pair" that enables heterolysis of the H-H bond and facilitates the hydrogen transfer. 57 Stimulated by the excellent theoretical performance of Fe 2 /B/mpg-C 3 N 4 for H 2 activation and transformation, we recalculated the nitrogen reduction mechanism to further investigate whether doping with B will affect the conversion of N 2 . As shown in Figure S18 and Figure 9a, the doping of B atoms does not change the reaction mechanism, and the variation of energy barriers for the subsequent N 2 reduction reaction can be ignored (±0.02eV). Therefore, B doping of the support was selectively targeted to optimize of the H transfer process without hampering other catalytic processes, it would be a good candidate to improve the catalytic performance of N 2 reduction. Furthermore, Microkinetic analysis of Fe 2 /mpg-C 3 N 4 and Fe 2 /B/mpg-C 3 N 4 by the rate determining states method, proposed by Kozuch et al, 58 are conducted to further probe the reaction rate for the in uence of B doping and the catalytic performance. The calculated TOF of ammonia synthesis on Fe 2 /B/mpg-C 3 N 4 is 6.34 × 10 −1 s −1 site −1 at 100 bar and 700 K, which is 2 orders of magnitude faster than the reaction rate of Fe 2 /mpg-C 3 N 4 (3.07×10 −3 s −1 site −1 ). The TOF of ammonia production on Fe 2 /B/mpg-C 3 N 4 is less than 10 -10 s −1 site −1 below 400 K, which is due to the stable NH x adsorption species at low temperature. Due to entropy effects for free gases, the bare site will increase accordingly after desorption of NH x upon the temperature rise.
Under the condition of constant pressure, the temperature increased from 300K to 700K, and the reaction rate increased by 13-14 orders of magnitude. At constant 700K temperature, the pressure has a relatively gentle effect on the reaction rate, so it is expected to realize the nitrogen conversion reaction at low pressure ( Figure 9d in Fe 2 /B/g-C 3 N 4 ). In order to achieve maximum TOF, our calculated partial pressure of N 2 is 0.3 (P N2 /P (N2+H2) ), close to ideal ratio of 1/4 (Figure 9c), indicating that the Fe 2 /B/g-C 3 N 4 catalyst can reduce the competitive adsorption of N 2 and H 2 . That is distinct from Fe 3 /Al 2 O 3 forming the coadsorption of N 2 and H 2 on Fe 3 clusters, which leads to a 0.44 partial pressure of N 2 for maximum TOF.
In contrast, it is necessary to remarkably change the partial pressure of nitrogen in classical metal catalysts, for example 0.06 (P N2 /P (N2+H2) ) in the Fe (211) C7 site and 0.78 in the Ru (001) B5 site. 19

Discussion
In conclusion, we have revealed that, with the decrease of coordination number in Fe n /mpg-C 3 N 4 (n=4, 3, 2), the hydrogenation mechanism tends to be dominant, and the dissociative mechanism is no longer favored for ammonia synthesis, which is controlled by the spin polarization of Fe n . The mechanism is conducive to the continuous hydrogenation of *N 2 on the Fe 2 cluster owing to the highest energy gap of spin d-orbitals in Fe n /mpg-C 3 N 4 (n=4, 3, 2), mimicking the nitrogen xation process in biological nitrogenase. The N-N bond can be activated by the continuous hydrogenation of *N 2 because of the increasing electron occupancy of the N 2 's 2π* orbitals and the decreasing energy of the N-N anti-bond orbital.
As shown in Figure 10, we propose that iron diatomic clusters anchored on boron doped mpg-C 3 N 4 (Fe 2 /B/mpg-C 3 N 4 ) can achieve e cient thermal conversion of N 2 to NH 3 by a "vicinal nonmetallic sites"promoted hydrogenation mechanism. This reaction pathway mainly includes the activation and transfer of H 2 , the formation of NHNH 2 , the N-N bond break and the formation of ammonia. The *N 2 can be hydrogenated to form *NHNH 2 , leading to a low energy barrier (0.51eV) in the N-N bond dissociation. By partially doping boron on the carbon site of the support, the mpg-C 3 N 4 support can act as a mimic non-innocent ligand of homogenous catalysts to activate and transfer hydrogen (the rate determining step of the hydrogen transfer energy barrier was reduced form 1.38eV in mpg-C 3 N 4 to 0.86eV in B/mpg-C 3 N 4 ), which reduces the competitive adsorption of N 2 and H 2 . And then, the synergistic effect of Fe 2 and "vicinal nonmetallic sites" of mpg-C 3 N 4 achieves the most favorable hydrogenation pathway for thermal conversion of N 2 to NH 3 .
Notably, the Fe 2 /B/mpg-C 3 N 4 catalyst with synergistic hydrogenation mechanism in the combined advantages of traditional thermal catalysis and electro-catalysis in ammonia synthesis, which not only breaks the BEP limitation of metal surface catalysts, but also avoids competitive reactions in electrocatalysis. Through theoretical calculations, we propose that the synergistic effect of Fe 2 and "vicinal nonmetallic sites" of mpg-C 3 N 4 may lead to better catalytic performance for the idea thermal conversion of N 2 to NH 3 . The current Fe 2 catalyst model may still be improved, however, diatomic catalysts with high spin polarization are expected to be used to activate and transform more inert molecules. Modifying the support based on "Lewis pair" is an effective method and provides a distinct strategy for the development of more e cient catalysts.

Methods
The computations were performed with the spin-polarized density functional theory (DFT) method, as implemented in the Vienna ab initio Simulation Package (VASP version 5.4.4). 59 The exchange correlation energy was modeled with the Perdew−Burke−Ernzerhof (PBE) functional 60 within the generalized gradient approximation (GGA). For models of Fe n /mpg-C 3 N 4 (n=2,3,4) and Fe 2 /B/mpg-C 3 N 4 , we chose 2×2 supercell of C 3 N 4 and the Brillouin zone was sampled by 2 × 2 × 1 k-points using the Monkhorst−Pack scheme in structural optimizations. An energy cutoff of 500eV was adopted for the plane-wave basis. To avoid interactions between repeated images, a vacuum gap of 15 Å was used in the direction perpendicular to the 2D layer. The energy and force convergence thresholds for the iterations in the self-consistent eld (SCF) were set to 10 −5 eV and 0.02 eV/Å, respectively. The transition states were obtained by climbing image nudged elastic-band method (CI-NEB) 61 combined with the Dimer 62 method and further con rmed by vibrational frequency analysis.
Fragment Molecular orbitals (FMOs) analyses were performed using spin-unrestricted DFT with the PBE exchange-correlation functional and DZP Slater basis sets as implemented in the Amsterdam Density Functional (ADF 2019. 304) program. 63 The scalar relativistic (SR) effect was included by the zero-orderregular approximation (ZORA). 64 At the same time, spin multiplicity calculations for Fe 2 /mpg-C 3 N 4 and N 2 adsorption in Table S3-4 and partial molecular orbitals (MOs) of N 2 adsorption con guration in Figure   S16 were performed with the ω-B97XD functional 65           (a) Energy diagrams of optimal associative mechanism for ammonia synthesis on Fe2/mpg-C3N4 and

Supplementary Files
This is a list of supplementary les associated with this preprint. Click to download.