Computational design of spatially confined triatomic catalysts for nitrogen reduction reaction

Abstract


INTRODUCTION
Ammonia (NH3) is a pivotal chemical compound, serving as a primary precursor for the production of chemical fertilizers inducing nitric acid, biofuel energy, plastic, synthetic fiber, and other chemical products.Its importance cannot be overstated, as it plays an indispensable role in industrial production and the daily lives of people globally [1][2][3][4].Currently, the Haber-Bosch process remains the primary method for the industrial synthesis of NH3, achieved by the reaction between N2 and H2 under high temperature (T > 700 K) and pressure conditions (P > 200 atm) [5][6][7][8], which requires an excessive amount of global energy consumption, leading to the generation of a substantial amount of greenhouse gases [9,10].
Therefore, it is an urgent need to develop sustainable and clean methods for yielding NH3 products.
Moreover, theoretical research and calculation models play a key role in predicting highly active and selective catalytic materials [22], providing important reference values for catalyst preparation.According to density functional theory (DFT) calculations, Sun and co-workers reported that V3C2 possesses the best NRR activity with 0.64 eV activation barrier among d 2 -d 4 M3C2 Mxenes [23].Some theoretical work has studied the NRR catalyzed by TM@N4-G, in which the central TM atom is coordinated by four pyridinic nitrogen atoms.The results show that the limiting potential of Ti@N4 (0.69 eV) and V@N4 (0.87 eV) [24], are shown to exhibit lower free energy for NRR than that of the Ru(0001) stepped surface (0.98 eV) [25].
Despite significant progress in this field, many obstacles remain, including high overpotential (> 0.6 V) and low FE (9% ~ 29.6%) for NRR.Hence, the development of highly efficient and selective NRR electrocatalysts to facilitate mild condition synthesis of ammonia is crucial.
In recent years, the emergence of atomically dispersed catalysts has brought about a revolution in the design and synthesis of catalysts [26].Single-atom catalysts (SACs) have gained significant attention due to their high-specific activity and maximum metal utilization efficiency [27][28][29][30][31].For example, He et al.
found that 11 transition metal atoms supported on a graphdiyne monolayer (TM@GDY) are highly stable electrocatalysts for hydrogen evolution reaction (HER) and oxygen evolution reaction (OER) involving overpotential of 0.01 ~ 0.46 V [32].Moreover, Liu et al. conducted research on the electrocatalytic generation of NH3 from N2 at room temperature and atmospheric pressure, using nitrogen-doped porous carbon embedded in cobalt, achieving a high ammonia generation rate of 0.86 μmol•cm -2 h -1 [33].
Alternatively, other single metal atoms anchored on N-modified carbon-based materials such as graphitic carbon nitride (g-C3N4) and defective graphene, promising electrocatalysts for NRR, with an onsetpotential of 0.34 V [34].Despite the potential benefits of SACs, a significant challenge persists in balancing the reaction rate and FE for NH3 synthesis, mainly due to the involvement of multiple reactive species in the NRR.This challenge remains a major obstacle in the practical application of SACs for the controlled and efficient electrocatalytic generation of NH3.
Atomic clusters, on the other hand, possess unique geometric and electronic properties due to their strong quantum size effects, rendering them highly promising for catalytic applications.The catalytic activities and product selectivity of supported metal nanoclusters (NCs) was atomically-precise governed by tuning its size, composition, and interfacial electron coupling between metal NCs and underlying substrates.In experiment, well-dispersed Pt2 dimers on graphene catalyzed the hydrolytic dehydrogenation of ammonia borane at a specific rate nearly 17-fold higher than the isolated single Pt atoms [35].Furthermore, Mo3 trimer on graphdiyne nanosheets was found to be most active toward NRR with high selectivity and stability by the designed screening criteria, involving the calculated onset potential of -0.32 V [36].More importantly, the presence of multiple active sites in the catalytic center of a catalyst can significantly enhance its versatility in adsorbate binding and enable the catalysis of a broader spectrum of complex reactions, yielding diverse products.Li et al. fabricated diatomic Fe2 NCs anchored on mesoporous carbon nitride, which exhibits superior catalytic performance for the epoxidation of trans-stilbene to trans-stilbene oxide, showing outstanding selectivity of 93% at high conversion of 91% [37].Ru3 stabilized on nitrogendoped carbon nanosheets was reported to efficiently catalyze the selective oxidation of alcohols [38].Furthermore, supported metal NCs have been theoretically proposed to have outstanding capability for catalyzing the reactions that require the activation of inert reactant molecules or demand multiple reaction centers [39][40][41][42][43].For instance, a tetramer immobilized in nitrogen-doped graphene monolayer shows optimal surface activities for N2 activation and next reaction to yielding NH3 products, involving the calculational on-set potential of -0.45 V, which much low than that of supported single atoms and dimers [43].Li et al. found that the catalytic performance of Rh3-C2N is more efficient than Rh1-C2N and Rh2-C2N, in the enzymatic pathway, limiting potential is only -0.45 V [44].Furthermore, a series of triple-atom catalysts (TACs) as electrocatalysts are designed by Zheng et al. for NRR, and the theoretical results reveal that Mn3-N4, Fe3-N4, Co3-N4, and Mo3-N4 can be promising non-noble electrocatalysts for NRR with high activity, selectivity, stability, and feasibility.Especially, the Co3-N4 system exhibits the highest activity with a limiting potential of -0.41 V through the enzymatic mechanism [42].
Inspired by the significant experimental and theoretical advancements, we have systematically investigated using DFT calculations the stability of 3d-5d TM trimers embedded on C3N3 nanosheets (marked as TM3@C3N3), and evaluate their electrocatalytic performances for NRR.The theoretical results indicate that TM3@C3N3 (TM = Re, Ru, Pt) hold significant promise electrocatalysts for the NRR, showcasing exceptional levels of activity, selectivity, stability, and feasibility.Notably, the Re3@C3N3 system exhibits the highest activity with a limiting potential of -0.11 V through the consecutive mechanism.Furthermore, we established the relationship between the intrinsic electronic properties and the catalytic activity of TM3@C3N3 and proposed key physical parameters.Furthermore, we advocate the concept of designing triple-atom catalysts as a strategic approach to drive the development of advanced electrocatalysts within the framework of green hydrogen economics.
The cut-off energy for the plane-wave basis was set to 500 eV.The convergence criteria for energy and force were set to 10 -4 eV and 0.02 eV•Å -1 , respectively.A (2 × 2) supercell of C3N3 monolayer with a vacuum space of 16 Å along the z-direction was applied to avoid the interaction between two periodic units.Moreover, the k-point in the Brillouin zone was sampled with a 2 × 2 × 1 grid [52].To evaluate the thermodynamic stability of TM3@C3N3 systems, ab initio molecular dynamics simulations (AIMD) simulations were performed at 300 K in an NVT ensemble for 10 ps, with a time step of 3 fs.
To assess the thermodynamic stabilities of the designed TM3@C3N3 systems, we calculated the binding energies (Eb), as follows: where ETM3@C3N3, EC3N3 and ETM3 denote the total energy of C3N3 substrate anchored with transition-metal trimeric clusters, the total energy of the optimized pristine C3N3, and the total energy of the isolated transition-metal trimeric clusters, respectively.The electronic adsorption energy (Eads) of reaction intermediates on TM3@C3N3 substrate can be calculated by the following formula: where Etot is the total energy of TM3@C3N3 substrate adsorbed by the intermediate,    1.Furthermore, according to the above calculation, it's worth noting that Pd3@C3N3 possesses the weakest binding strength among the 21 selected systems.In light of this, we conducted ab initio molecular dynamics (AIMD) at 300 K for 10 ps to evaluate the stability of TM3@C3N3, with Pd3@C3N3 considered as a representative.Figures 2B and 2C, illustrate the oscillations of the DFT total energies relative to the initial states (E) and temperature (T), as well as the bond lengths of Pd-N (dPd-N) and Pd-Pd (dPd-Pd), indicating dynamic fluctuations near the initial condition.According to snapshots of atomic structures of Pd3@C3N3 with different time (Figure 2D), it is preserved well and maintain structural stability at the imposed conditions throughout the entire molecular dynamics, in which the vertical buckling exhibits minimal fluctuation, measuring less than 0.10 Å.Therefore, by amalgamating the findings from the binding energy calculations and molecular dynamics calculations, we can confidently assert that these TM3@C3N3 systems have robust structural stabilities.

Activation of N2 on TM3@C3N3
In the overall electrochemical NRR process, the primary and critical step involves the adsorption and activation of N2 molecules, which is of utmost importance as it is responsible for activating the inert N≡N triple bond, laying the foundation for the subsequent smooth occurrence of protonation.

The electronic properties of N2 on TM3@C3N3
Revealing the source of NRR activity of electrocatalysts will provide guidance for the design and development of highly active catalysts.To understand the underlying mechanism of the N2 activation, we analyzed the interactions between the metal trimers and N2 using partial density of states (PDOS).
Compared to the molecular orbitals of free N2, the strong ability for Pt3 To further study the interaction between N2 and TM3@C3N3 catalysts, we calculated the ΔGN2* value.The corresponding adsorption free energy for N2 adsorption on TM3@C3N3 is calculated by Equation (3) with the detailed data presented in Figure 4B and Supplementary Table 1, respectively.It is not difficult to see that for most TM3@C3N3 Catalysts, which the ΔGN2* value is close to the zero, in the range of 0.35 eV to -5.12 eV, which means that N2 molecule is beneficial to adsorption on these catalysts.Moreover, as shown in Table 1, we also calculated the bond length of N2 molecular during adsorption.Compared to the original N≡N triple bond length in N2 molecules, there is a slight stretch in the N≡N triple bond of N2 adsorbed on the catalyst, regardless of whether it is end-on or side-on adsorption configuration, which indicates that N2 adsorption on the selected 21 catalysts can weaken the N≡N triple bond, which is conducive to activating the subsequent NRR process.
According to Bader charge analysis, as presented in Figures 4C and 4D, show that there is a significant charge transfer between the transition metal atoms and the N atoms with the number of electrons lost per metal trimer cluster ranges from 0.24 to 1.97 e with the detailed data presented in Table 1.As shown in Figure 4E, the adsorption energy of N2 is linearly related to the d-orbital center (εd) of the loaded metal cluster.higher d orbital center of the TM3@C3N3 system, the more favorable its interaction with the π* orbital of N2 and provides more charge transfer to the N2 molecule, thus showing a stronger binding capacity for N2 adsorption [59].Guided by these key physical parameters, the catalytic behavior of transition metal trimer clusters can be precisely controlled by designing moderate cluster-carrier interactions through the selection of suitable metal elements.As shown in Figure 5A, there is one reaction mechanism for electrocatalytic synthesis of ammonia, namely, enzymatic mechanism and another consecutive mechanism is shown in Supplementary Figure 1A.The distal mechanism and the alternative mechanism are also presented in Supplementary Figure 1B.Their common feature is that the N atom at one end of N2 is adsorbed on the catalyst, while the N element at the other end is not adsorbed, while the N element at the far end preferentially reacts with the H proton.In the distal mechanism, N atoms leaving the surface of the catalyst preferentially react with H protons and release as ammonia, leaving *N adsorbed on the catalyst, which will start hydrogenation and form ammonia.In the alternative mechanism, two N atoms are alternately hydrogenated by six proton-electron pairs, resulting in the formation of two NH3 molecules.In the enzymatic mechanism, N2 molecule is first decomposed into adsorbed *N atom, and then *N is gradually hydrogenated into ammonia.The hydrogenation process is the same as the alternative mechanism.In the consecutive mechanism, one of the two N atoms is hydrogenated, and then continues to react with the remaining N atoms to form a second NH3.Due to the different structure and performance of catalysts, the NRR reaction mechanism is often different.However, many studies have shown that the first step of *N2 + (H + + e − ) → *NNH or the last step of *NH2 + (H + + e − ) → *NH3 is likely to be the potential determination step, independent of the NRR mechanism.We define the protonation step with the maximum positive free energy changes (∆Gmax) the PDS.It is well known that the ideal catalyst for an electrochemical NRR should meet the following criteria: (1) The ∆Gmax of the two key steps is less than 0.55 eV, that is, ∆G*N2→*NNH ≤ 0.55 eV and ∆G*NH2→*NH3 ≤ 0.55 eV, so the catalysts considered may have better performance than the best pure metal and nano-metal cluster catalysts.( 2) Eads(*N2) − Eads(*H) < 0, which proves that N2 has good selectivity in catalyst.
Table 2. DFT-calculated the optimal reaction mechanism, potential-determining step, and corresponding limiting potentials (UL) for TM3@C3N3 (TM = Re, Pt, Ru, Rh, Ta, Ir ).  2 and Supplementary Figure 2. A total of four classifications are considered to evaluate the critical steps, with a critical point set at 0.55 eV.As a result, six systems met the above criteria, demonstrating decent catalytic activity in electrochemical NRR.In the case of Pt3@C3N3, the associated free energies and intermediate geometries for each step are presented in Figures 5B and 5D.
The Pt3@C3N3 exhibits a propensity for NRR via enzymatic mechanism, wherein N2 adopts a side-on configuration.In the enzymatic mechanism, the first hydrogenation step involves the bonding of a hydrogen atom to one of the N atoms, followed by alternating bonding to the two N atoms until a second NH3 molecule is produced.The ∆Gmax change is 0.24 eV, establishing the final hydrogenation step as the PDS of the entire electrochemical NRR.Similarly, we investigated the NRR pathway on Ru3@C3N3, as depicted in Figure 5C, reveling a ∆Gmax of 0.35 eV.It is evident that throughout the entire NRR process, the first hydrogenation step (*N2 → *NNH) is identified as the PDS.Briefly speaking, the corresponding UL values for Pt3@C3N3 and Ru3@C3N3 are -0.24V and -0.35 V, respectively.Therefore, after applying UL to Pt3@C3N3 and Ru3@C3N3 surfaces, all electron transfer steps can be downhill, which is beneficial to the production of NH3, where the reaction process of Ru3@C3N3 also follows an enzymatic mechanism.
It is worth emphasizing that the development of efficient NRR catalysts remains challenging due to the competition with HER.The ideal NRR catalyst would exhibit significantly higher NRR activity while displaying considerably lower HER activity.To assess selectivity, we calculated the N2 adsorption energy and hydrogen adsorption energy on the designed TM3@C3N3 catalysts using Equation (2).A more negative difference between N2 adsorption energy and hydrogen adsorption energy indicates higher selectivity for NRR.The results are presented in Supplementary Figure 3.It is evident that both Pt3@C3N3 and Ru3@C3N3, which were previously identified as having the highest NRR activity, exhibit markedly higher selectivity for NRR over HER.These findings imply that these catalysts can ensure a high Faraday efficiency in catalytic electrochemical NRR.

CONCLUSIONS
In summary, this study provides a systematic investigation into the potential of C3N3-loaded triple-atom catalysts (TACs) in electrocatalytic NRR using DFT calculations.Employing a stringent "five-step" filtering strategy, we have identified TM3@C3N3 (TM = Pt, Ru, Re) as highly promising candidates with the attributes of low energy cost, high selectivity, remarkable stability (both thermodynamic and kinetic), and remarkably low limiting potentials ( −0.35 ~ −0.11 V).Our analysis of electronic properties highlights that the exceptional NRR activity can be ascribed to the electron acceptance and donation mechanism involving d-π* interactions.This mechanism, combined with charge density differences and PDOS, underscores the qualifications of TM3@C3N3 as electrocatalysts.Furthermore, we have explored the linkage between chemical activity and the electronic structure of TM3@C3N3 surfaces, revealing a pivotal physical parameter that allows precise control over the catalytic performance of transition metal trimer clusters.In conclusion, this work not only enhances our comprehensive understanding of the stability, activity, and selectivity of TM3@C3N3 electrocatalysts but also offers an effective strategy for the screening and design of novel TACs for NRR.We anticipate that this study will inspire further experimental and theoretical endeavors to unlock the potential of TACs in NRR and other related electrochemical reactions.Based on the insightful understandings, herein we computationally demonstrate that the spatially confined trimetric transition metal clusters are the smallest catalysts that enable the fixation of plural N2 molecules.They have outstanding stability and feasibility to synthesize in laboratory, and provide an exclusive reaction pathway for N2 reduction to NH3 products.Remarkably, hydrogen evolution reaction is effectively suppressed on these trimeric metal centers, which is competitive side reactions that strongly affect the generation of senior products from N2 reduction.
These theoretical results provide keystone knowledge that is highly demanded for impelling the synthesis and application of atomically precise catalysts for direct conversion of N2 to useable energies, and thus will be appealing to a wide readership.We believe that the prestigious Journal of Materials Informatics is the best journal to publish it in a timely manner.We really appreciate if our manuscript can be reviewed and considered for publication in your esteemed journal soon.

Figure 1 .
Figure 1.The Schematic diagram for screening NRR candidate catalysts and electrocatalytic N2 to produce NH3 on supported metal trimers.

Figure 2 .
Figure 2. (A) Computed The binding energies (ΔEbind) of triple-atoms anchored on C3N3.The insets display the optimized structures of TM3@C3N3 for two types of transition metal elements.The C, N and metal atoms are shown by gray, blue and light green colors, respectively.(B) Variations of the temperature (T) and total energies (E) and (C) the bond length of Pd-N (dPd-N) and Pd-Pd (dPd-Pd) vs. time for AIMD simulations of Pd3@C3N3.(D) The snapshots of atomic structures of Pd3@C3N3 within 10 ps,the Pd is shown by green color.

Figure 3
shows the N2 activation is facilitated through the electron acceptance and donation mechanism, where the partially occupied d orbitals of TM atoms accept lone pair electrons from *N2 molecule, while concurrently donating d electrons to the anti-bonding orbitals (π*) of *N2 in the reverse direction.Therefore, this interaction serves enhance the TM-N bond, while weakening the N≡N bond.Notably, the metal double or triple-atom centers play a crucial role in maximize the activation effect through donating a greater number of electrons to N2 in comparison to monatomic active sites[56][57][58].Therefore, the dual-metal or tripleatom centers provide a more efficient strategy for promoting N2 activation, which is beneficial for overcoming the barrier of the first hydrogenation step of NRR.

Figure 3 .
Figure 3. Simplified schematic illustration of N2 binding to single and double-atom sites.
/g-C3N3 to adsorb/activate N2 is primarily associated with their availability of unoccupied and occupied d orbitals.As displayed in Figure 4A, the unoccupied 5d orbitals of Pt3 accept electrons from the 2π and 3σ molecular orbitals of N2 to form bonding states.Meanwhile, strong d−2π* coupling leads to the 2π* orbital of N2 partially occupied near the Fermi level, which is a result of electron back-donation from the occupied d orbitals of Pt3 to the 2π* orbital of N2.The result further confirms the existence of strong d−π* orbital coupling around the Fermi level, as well as the significant overlap between the d orbitals of trimeric clusters and the occupied orbitals of adsorbed N2, which is also very consistent with the above analysis of N2 activation mechanism.

Figure 4 .Figure 5 .
Figure 4. (A) Calculated the molecular orbitals of free N2, absorbed N2 on Pt3/g-C3N3.(B) Adsorption TM3@C3N3 catalysts for their NRR performance, we computed the ∆G of the initial and final stages of NRR under open-circuit conditions (U = 0).The resulting values, along with corresponding diagrams, are presented in Table our manuscript entitled "Computational design of spatially confined triatomic catalysts for nitrogen reduction reaction" for publication as a research paper on Journal of Materials Informatics.The renewable-electricity-driven N2 reduction to generating NH3 products has long been pursued but faces many challenges.However, conventional metal-based catalysts usually lack effective reaction activities and channels for the N2 activation and next protonation to forming NH3 in mild conditions.There remains a big gap of knowledge in the N2 activation mechanism and the principles for manipulating this key process.Recently, our group has designed a series of nano and subnano catalysts for small molecules conversion to high-valued chemical products, including N2 to NH3 [Chem.Sci., 2020, 11, 2440; J. Mater.Chem.A, 2020, 8, 20570], and CO2 to C1 and C2 products via C-C coupling [Angew.Chem.Int.Ed. 2020, 59, 1919; Nano Energy 2020, 76, 105049], and illuminated how to utilize the quantum confinement effect for atomically precise control of the product.

Table 1 .
DFT -calculated the average distances of TM-N (dTM-N) and TM-TM (dTM-TM), vertical buckling between TM3 and C3N3 monolayer (dTM-C), bind energy (Ebind), d orbital center (d), magnetic monument (Mag) and the number of charge transfer between trimer clusters and underlying C3N3 substrate.

Wei Pei www.jmijournal.com Journal of Materials Informatics Computational design of triple-atom catalysts for electrocatalytic nitrogen reduction on g-C3N3 MAIN TEXT SupplementaryTable 1 .
DFT-calculated the adsorption Gibbs free energy of *N2 on TM3@C3N3 via the end-on and side-on mode.