The reduction of carbon dioxide (CO 2 ) to other carbon-based feedstocks and fuels has been considered as a promising solution to cope with the growing greenhouse effect and energy crisis. To date, numerous catalysts for CO 2 reduction, such as metals, metal alloys and oxides (e.g., Au, Ag, and Pd nanoparticles, Cu-Pd and Au-Cu alloys, Cu 2 O, and SnO 2 ), [1][2][3][4][5][6][7][8][9][10][11][12] 2D metal dichalcogenides (e.g., MoS 2 and WSe 2 ), 13,14 metalorganic frameworks (MOFs), 15,16 covalent organic frameworks (COF), 17 and N-doped carbon materials, [18][19][20][21] have been developed. Remarkably, by CO 2 reduction, C 1 products such as carbon monoxide (CO) and formate (HCOO -) can be produced with high efficiency and high selectivity (i.e., faradaic efficiency) at a low energy cost. However, high value-added multi-carbon chemicals, namely, C n (n ≥2), have been rarely harvested due to the large overpotential and low selectivity for C-C coupling. [22][23][24][25][26] Thus, to achieve an effective and sustainable carbon-cycle utilization, further reducing C 1 products to high value-added C n chemicals with high efficiency is an alternative strategy for direct reduction of CO 2 to C n chemicals. 23,[27][28][29] Note that CO is one of the key intermediates for C-C coupling during the reduction of CO 2 to C n products. 27,28 Directly using high pressure CO as the initial reactant has great advantages to promote the C-C coupling. In addition, CO reduction can be carried out in an alkaline electrolyte without producing undesirable carbonate, which can also inhibit the competitive hydrogen evolution reaction, thus enhancing the productivity of C n chemicals. 23 Therefore, it is of great economic and scientific importance to develop new catalysts with high efficiency and selectivity but low energy consumption to reduce CO to C n chemicals based on C-C coupling.
Recently, the metal-free frustrated Lewis pairs (FLPs) comprised of a sterically encumbered donor and acceptor, for examples, intra-or inter-molecular B-P, B-N and B-B atomic pairs, have been demonstrated to be effective for the activation of many small molecules, such as H 2 , CO, CO 2 , NO, SO 2 and so on. [30][31][32][33][34][35] In particular, the B species in the FLPs shows strong potential to capture and activate CO, in which B species accepts electrons from CO by σ donation and simultaneously donates electrons from its filled p orbital to the π* orbital of CO, thus weakening the CuO triple bond. This resembles the widely reported transition metal-CO and B-N 2 interactions. [36][37][38][39][40]53 Therefore, we speculate that the FLPs with B and N dual active sites might be effective at facilitating the C-C coupling during the CO reduction to C n chemicals, in which the B site acts as a Lewis acid to strongly capture one CO molecule, 39,40 while the N site serves as a Lewis base to push another CO molecule to couple with the first CO. From this point of view, we have constructed a metal-free catalyst for CO reduction, namely, single boron (B) atom decorated on monolayer C 2 N (i.e., B/C 2 N), which has been widely adopted as the substrate for single atom deposition due to its two-dimensional (2D) structure with uniformly distributed cavities, as well as pyridine-like host N atoms as coordination sites. [41][42][43] The decorating B atom and host N atoms of monolayer B/C 2 N may form intra-molecular B-N FLPs with the pull-push function, thus promoting the CO coupling and facilitating the CO reduction to C n chemicals.
In this work, by means of density functional theory (DFT) computations, we firstly evaluated the structural stability of the B/C 2 N complex based on ab initio molecular dynamics (AIMD) simulations and phonon spectrum calculations, confirmed the existence of B-N FLPs, and then explored its catalytic role in promoting multi-CO coupling and reduction. Our DFT results show that CO can be efficiently reduced to ethylene (C 2 H 4 ) and propylene (C 3 H 6 ) on B/C 2 N both with a rather low free energy barrier (0.22 eV) for the potential-determining step under visible light irradiation due to the presence of B-N FLPs. This study not only has identified a highly efficient metalfree photocatalyst for CO reduction to high value-added C n chemicals but also may guide the future development of photo/ electrochemical catalysts with high efficiency and selectivity.
Stability of the B/C 2 N monolayer with B-N FLPs Before exploring its possible catalytic role in CO reduction, we examined the structural stability of our designed B/C 2 N in order to evaluate its experimental feasibility, which is the prerequisite for practical applications of newly designed catalysts with long-term activity. In the case of monolayer C 2 N, there are five possible sites for the decoration of an isolated B atom, and thus we investigated the stabilities of all these five B/C 2 N complexes (Fig. 1 and Table S1 in the ESI †).
Among these five possible decoration sites, we found that the corner (site 2 in Fig. 1) of the six-fold cavity is energetically the most favorable site for anchoring an isolated B atom (with an adsorption energy of -5.46 eV), and the resultant B/C 2 N complex is featured by a N 2 C 2 B 1 five-membered ring. Such a strong binding (-5.46 eV) implies a high possibility to experimentally achieve this B/C 2 N complex containing N 2 C 2 B 1 fivemembered rings.
As shown in Fig. 1, except for the two N atoms which form the N 2 C 2 B 1 five-membered ring after B atom decoration, the remaining four N atoms in the cavity of C 2 N have large distances (3.1 and 4.2 Å) from the decorating B atom. Such large B-N distances can effectively avoid the quenching of electron donors and acceptors, thus making it possible to form the B-N FLPs in this case.
We further evaluated the thermal stability of this B/C 2 N complex by the AIMD simulations at 500 K and 1000 K for 10 ps with a time step of 2 fs (Fig. S1 in the ESI †). For both cases, the energy of the B/C 2 N complex oscillates near the equilibrium state, and the atomic structure remains intact without obvious structure distortion during the 10 ps AIMD simulation period, indicating its high thermodynamic stability. We also confirmed that this B/C 2 N monolayer is dynamically stable, as no imaginary modes were observed in the phonon dispersion spectrum (Fig. S2 in the ESI †).
All the above results and analyses have confirmed that this lowest-energy structure of B/C 2 N monolayer (see Fig. 1, the right panel) has a high stability featuring possible B-N FLPs. Thus, in the following sections, this model was adopted to examine the catalytic performance of the B/C 2 N monolayer for CO reduction.
As is well known, the stable adsorption of multiple CO molecules at the catalytic active site is critical for multi-CO coupling and subsequent CO reduction to C n products. 25,[44][45][46] Fig. 2 illustrates the fully optimized structures for the adsorption of one to five CO molecules on B/C 2 N, and Table 1 summarizes the corresponding adsorption energies and adsorption free energies. The adsorption energy of a single CO molecule on B/C 2 N is -2.41 eV, due to a strong capture ability of the decorated B atom for the CO molecule. As the number of adsorbed CO molecules increases, each newly introduced CO molecule is linked with the former one by C-C coupling, giving rise to a gradually prolonged carbon skeleton (see Fig. 2). During this process, the carbon skeleton is well supported together by the decorated B atom and host N atoms of the B/C 2 N complex. Meanwhile, the calculated adsorption energy (ranging from -2.85 to -3.02 eV) of multi-CO on B/C 2 N does not change too much upon increasing the number of CO molecules from two to five (see Table 1). After considering the contribution of zeropoint energy and entropy, the calculated adsorption free energies of CO on B/C 2 N gradually increase (i.e., less favorable) upon increasing the number of adsorbed CO molecules.
Note that all the calculated adsorption free energies of two, three and four CO molecules on B/C 2 N are negative, indicating that the multi-CO coupling is thermodynamically spontaneous for these three cases. As for the 5 th CO adsorption, a positive but still moderate adsorption free energy (0.58 eV) is obtained. In this case, we believe that the multi-CO coupling for the 5 th CO adsorption may also occur by increasing the concentration of gaseous CO at somewhat energy cost. Therefore, we speculate that different C n chemicals including C 2 , C 3 , C 4 , and even C 5 can be produced by CO reduction on B/C 2 N. Considering the extremely complex and overlong reaction path of CO reduction to C 4 and C 5 chemicals, we only take C 2 and C 3 as examples of possible C n products from CO reduction to evaluate the catalytic performance of B/C 2 N in this work.
For the C-C coupling process, in addition to the thermodynamics as we discussed above, kinetics is the other impor-tant issue. Thus, we studied the detailed reaction pathway for C-C coupling by the climbing image nudged elastic band (CI-NEB) method. Starting from the barrier-free chemisorption of the 1 st CO molecule on B/C 2 N, we sequentially calculated the energy barriers from physisorption to chemisorption (C-C coupling) for the 2 nd and the 3 rd CO molecule. As shown in Fig. 3, when the 2 nd CO approaches and bonds with the 1 st prechemisorbed *CO on B/C 2 N to form *COCO, a moderate energy barrier of 0.73 eV is observed with an energy decrease of 0.31 eV. When introducing the 3 rd CO, the formation of *COCOCO leads to a negligible energy release (0.01 eV), and the calculated energy barrier from *COCO + CO to *COCOCO is 0.83 eV. The moderate energy barriers (0.73 and 0.83 eV) for these two C-C coupling processes (*CO + CO to *COCO and *COCO + CO to *COCOCO) on B/C 2 N are comparable with those (about 0.48 and 0.90 eV) at the interface of between Cu (111) and Cu (100); both have been demonstrated to be able to reduce CO to n-propanol with a 20% faradaic efficiency at a low potential of -0.45 V versus the reversible hydrogen electrode. 28 The above results clearly show that the C-C coupling is energetically favorable and occurs spontaneously for the adsorption of up to three CO molecules on B/C 2 N, probably due to the existence of the B-N FLPs in our designed B/C 2 N complex and its catalytic role in C-C coupling. To verify this speculation, we further calculated the difference charge density of B/C 2 N with two adsorbed CO molecules. As shown in Fig. S3a and 3b in the ESI, † the charge accumulation occurs between the decorating B atom and the C atom of the first CO, while charge depletion appears between the host N atom and the C atom of the second CO. Furthermore, when considering the multi-CO (one to five CO molecules) adsorption, the dec- orating B atom always accepts electrons and acts as the Lewis acid, while the host N atom binding with the subsequent approaching CO molecules always donates electrons and serves as the Lewis base, based on Hirshfeld population analysis (see Table S2 †). As confirmed by the adsorption energy and free energy calculations (see Table 1), the decorating B atom (electron acceptor) strongly captures the first CO molecule; meanwhile the host N atoms (electron donors) push the approaching CO molecules toward the foregoing adsorbed CO molecule or the already formed carbon skeleton, thus promoting the C-C coupling. This highly resembles the pull-push effect observed in borane, boron-metal cation, and B doped BP for the NuN bond activation. [47][48][49] Therefore, we believe that the decorating B atom and the host N atoms of the B/C 2 N complex do act as the Lewis acid and base, respectively, forming the B-N FLPs with the so-called pull-push function and thus facilitating the multi-CO coupling.
Considering the spontaneous occurrence for C-C coupling on B/C 2 N, it is reasonable to believe that the hydrogenation of CO on the B/C 2 N complex prefers to start from the spontaneously formed CO dimer or trimer rather than from an isolated CO, which triggers the CO reduction reaction to produce C 2 and C 3 chemicals with the help of B-N FLPs. Fig. 4 summarizes the most energetically favorable reaction pathway (with the lowest positive free energy change between any two elementary steps) towards the formation of ethylene/propylene (C 2 H 4 /C 3 H 6 ), together with the atomic structures of the corresponding reaction intermediates. As shown in Fig. 4a, reducing the CO dimer to C 2 H 4 follows the following reaction path:
Along this path, the formation of *COHCOH species with the largest positive free energy change (0.22 eV) is the potential-determining step during this process. As for the reduction of the CO trimer, the reduction process proceeds with three O atoms sequentially removed (see Fig. 4b). After removing the 2 nd O atom, the center C atom of the CO trimer attaches its nearest N atom, forming a C-N bond. After the last hydrogenation step, the distal C atom bonding with three H atoms lifts up from the B/C 2 N basal plane. In this case, the formation of *CH 2 CHCHO species is the potential-determining step with a free energy increase of 0.22 eV. Note that the total decreases (4.15 and 5.09 eV) of the Gibbs free energy from the beginning to the end of the hydrogenation for CO dimer and trimer reduction are much larger than the energy costs (2.49 and 2.12 eV) to form gaseous C 2 H 4 and C 3 H 6 molecules, which indicates an energetically feasible process to release the final products (C 2 H 4 and C 3 H 6 ) from the B/C 2 N substrate. Energetically, the reduction of the CO dimer and trimer prefers to produce ethylene and propylene, respectively, based on our DFT calculations. However, we cannot exclude the production of other possible oxygenated organics.
In the above analyses, we theoretically demonstrated the high catalytic activity of the B/C 2 N complex towards the reduction of the CO dimer or trimer. However, the adsorbed single CO molecules may also be reduced to other C 1 chemicals such as CH 4 and CH 3 OH on the B/C 2 N surface, which may affect the selectivity towards the production of C n chemicals. Thus, we examined the selectivity towards multi-carbon chemicals.
Encouragingly, according to our DFT computations, producing the C 1 chemicals (CH 4 and CH 3 OH) from CO reduction has to overcome large free energy barriers (1.85 and 1.80 eV for CH 4 and CH 3 OH, respectively; see Fig. S4 in the ESI †). Such high free energy barriers are due to the strong bonding between the decorated B atom and the relevant intermediates during the CO reduction, which severely hinders the formation of saturated hydrocarbons or their oxygenated derivatives.
Significantly, extremely low energy costs (0.22 eV) are needed to be overcome to reduce the CO dimer and trimer to C 2 H 4 and C 3 H 6 , respectively, implying a high catalytic efficiency of the B/C 2 N catalyst for CO reduction to C n chemicals due to the presence of the B-N FLPs. Note that the energy cost (0.22 eV) of reducing the CO dimer to C 2 H 4 is identical to that (0.22 eV) of reducing the CO trimer to C 3 H 6 , giving rise to a mixture of gaseous C 2 H 4 and C 3 H 6 in the final products, which can be effectively separated by an industrially mature cryogenic distillation method or newly proposed zeolitic imidazolate frameworks. 50 As a major side reaction of CO reduction, the hydrogen evolution reaction (HER) on our proposed catalyst was also examined by computing its corresponding free energy profile, [51][52][53] and three active sites (labelled B, N 1 , and N 2 , respectively, see Fig. S5 in the ESI †) were considered. Our simulations demonstrated that the HER process on the three sites is hindered by the Heyrovsky step (H* + H + + e -→ H 2 ) with the free energy barriers of 2.27, 0.51 and 0.68 eV, respectively, which are much larger than that of CO reduction to C 2 H 4 and C 3 H 6 (0.22 eV, Fig. 4a andb). Thus, B/C 2 N was predicted to exhibit outstanding catalytic selectivity for CO reduction compared to the HER.
In addition, we note that Cao et al. proposed the ultra-high catalytic activity of two B atom embedded C 2 N nanosheets (B 2 /C 2 N) for the nitrogen reduction reaction, which is attributed to the synergetic effects of two B atoms. 54 Thus, we expected that the C 2+ products could also be generated on B 2 /C 2 N due to the high chemical reactivity of two B atoms. However, the corresponding catalytic pathways will be greatly complicated compared to that on B/C 2 N, in which the two B atoms and their around N atoms could be the active sites for CO reduction reactions rather than a mere frustrated Lewis pairs-based catalyst that is composed of one B and host N atoms. Some related computations are being performed and are thus not considered in the present work.
Note that all the above results and discussion on the catalytic activity of B/C 2 N to reduce CO to C n chemicals (C 2 and C 3 ) are based on the well-known proton-coupled electron transfer electrocatalytic reaction mechanism. Actually, the pristine C 2 N is also a metal-free semiconducting photocatalyst with excellent visible light adsorption due to its moderate bandgap. [55][56][57] Therefore, we further studied the electronic structure of the B/C 2 N complex to check its possibility as a metal-free photocatalyst for CO reduction.
Fig. 5a-c present the calculated band structures and light absorption spectra of the monolayer C 2 N and B/C 2 N, respectively. As shown in Fig. 5a, the pristine C 2 N shows a direct bandgap of 2.46 eV, which is consistent with previous reports. [55][56][57] After decorating the B atom, the bandgap of the B/C 2 N complex decreases to 1.20 eV (see Fig. 5b), which further extends its absorption range even into the infrared region (see Fig. 5c). Therefore, the as-designed B/C 2 N shows a wider light absorption range than the pristine C 2 N, implying its better photocatalytic performance.
More importantly, as shown in Fig. 5d, though slightly lower than that of pristine C 2 N due to B atom decoration, the conduction band minimum (CBM) level of B/C 2 N still lies about 0.61 and 0.65 eV higher than the calculated reduction potentials of CO/C 2 H 4 (0.17 V vs. NHE) and CO/C 3 H 6 (0.21 V vs. NHE), respectively. Therefore, the photoexcited electrons of B/C 2 N remain strong enough driving force to fulfill the CO reduction to C 2 and C 3 chemicals, considering the extremely low free energy barrier (0.22 eV) for the potential-determining step to produce both C 2 and C 3 chemicals from CO reduction. All these results clearly demonstrate that our designed B/C 2 N complex should act as a highly efficient photocatalyst for CO reduction to C n chemicals under visible and even infrared light irradiation.
In summary, we have designed a frustrated Lewis pairs photocatalyst for the reduction of CO to high value-added multicarbon chemicals (C 2 H 4 /C 3 H 6 ) by decorating single B atom on the monolayer C 2 N (i.e., B/C 2 N). Free energy calculations have clearly revealed that the B/C 2 N catalyst can efficiently convert CO into C 2 H 4 and C 3 H 6 both with the largest free energy increase of 0.22 eV, due to the presence of B-N FLPs in the B/C 2 N complex with the pull-push function. Electronic structure calculations indicated that the newly designed B/C 2 N complex possesses excellent visible light and even infrared adsorption with a suitable band position, rendering it an ideal photocatalyst for CO reduction to multi-carbon chemicals under visible and even infrared light irradiation. This study not only identified the B/C 2 N monolayer as a very promising metalfree photocatalyst for CO reduction to C n chemicals, which has substantial economic benefits and a significant effect on the long-term application of CO 2 reduction, but also calls for much effort in developing new concepts in modern catalyst design.
The spin-polarized DFT calculations were performed by using the Vienna Ab Initio Simulation Package (VASP). 58 The generalized gradient approximation (GGA) combined with the Perdew-Burke-Ernzerhof (PBE) functional was employed to describe the exchange-correlation term. The electron-ion interactions were described by the projector augmented wave (PAW) method. 59,60 All geometries were fully optimized with a plane wave energy cutoff of 500 eV and the convergence criterion for the residual force and energy was set to 0.03 eV Å -1 and 10 -5 eV, respectively. We used the empirical correction in Grimme's method (DFT+D3) to describe van der Waals interactions. 61 The B/C 2 N catalyst was constructed by introducing one B atom into a 2 × 2 × 1 supercell of the C 2 N monolayer. The vacuum space was set to 15 Å in the z direction, which was enough to negligible interactions between periodic units. A 3 × 3 × 1 Monkhorst-Pack k-point mesh was used to sample the Brillouin zone. The band structures and light adsorption spectra were calculated by using the hybrid functionals based on the Heyd-Scuseria-Ernzerhof (HSE06) method. 62 The climbing image nudged elastic band (CI-NEB) method was used to calculate the minimum energy path and search the reaction saddle point. 63 Ab initio molecular dynamics (AIMD) simulations were performed in the canonical ensemble (NVT) using a Nose-Hoover thermostat to evaluate the thermodynamic stability of the B/C 2 N catalyst. 64,65 The phonon band structure was computed using the density functional perturbation theory (DFPT) 66 as implemented in Phonopy package. 67 Hirshfeld population analysis 68 was performed using the PBE functional with a DNP basis set using the DMol 3 program. 69 Other computational details including calculation methods of adsorption energy, free energy and reduction potential can be found in the ESI. †