DFT Study on the CO2 Reduction to C2 Chemicals Catalyzed by Fe and Co Clusters Supported on N-Doped Carbon.

The catalytic conversion of CO2 to C2 products through the CO2 reduction reaction (CO2RR) offers the possibility of preparing carbon-based fuels and valuable chemicals in a sustainable way. Herein, various Fen and Co5 clusters are designed to screen out the good catalysts with reasonable stability, as well as high activity and selectivity for either C2H4 or CH3CH2OH generation through density functional theory (DFT) calculations. The binding energy and cohesive energy calculations show that both Fe5 and Co5 clusters can adsorb stably on the N-doped carbon (NC) with one metal atom anchored at the center of the defected hole via a classical MN4 structure. The proposed reaction pathway demonstrates that the Fe5-NC cluster has better activity than Co5-NC, since the carbon-carbon coupling reaction is the potential determining step (PDS), and the free energy change is 0.22 eV lower in the Fe5-NC cluster than that in Co5-NC. However, Co5-NC shows a better selectivity towards C2H4 since the hydrogenation of CH2CHO to CH3CHO becomes the PDS, and the free energy change is 1.08 eV, which is 0.07 eV higher than that in the C-C coupling step. The larger discrepancy of d band center and density of states (DOS) between the topmost Fe and sub-layer Fe may account for the lower free energy change in the C-C coupling reaction. Our theoretical insights propose an explicit indication for designing new catalysts based on the transition metal (TM) clusters supported on N-doped carbon for multi-hydrocarbon synthesis through systematically analyzing the stability of the metal clusters, the electronic structure of the critical intermediates and the energy profiles during the CO2RR.

1. Introduction

The electrochemical CO2 reduction reaction (CO2RR), as a useful method to convert CO2 into value-added chemical products, which not only helps to solve the energy and environmental problems caused by fossil fuel combustion but also achieves sustainable development [1,2,3,4]. The main products of CO2RR are generally divided into C1 products (e.g., CO, CH4, CH3OH, HCOOH, etc.) [5] and C2 products (e.g., C2H4, C2H5OH, CH3COOH, etc.) [6]. Cu and Cu-derived materials have been considered the most common electrocatalysts for the CO2RR in the early stages [7,8]. Furthermore, Ag-based [9,10] and Au-based [11,12,13] catalysts can selectively reduce CO2 to CO at low overpotentials. However, they suffer from low utilization of metal atoms and a low C2+ selectivity.

Recently, the single-atom catalysts (SACs) of metal loaded on carbon substrates (metal nitrogen-doped carbon-based catalysts) have become a rather hot frontier for the maximized atom utilization efficiency and defined active centers. Rossmeisl et al. [14,15] found that the transition metal nitrogen-doped carbon-based catalysts (M-N-C, M = Mn, Fe, Co, Ni or Cu) performed a high CO selectivity for CO2RR. Furthermore, both Mn-N-C and Fe-N-C also possessed CO selectivity as well as trace amounts of CH4, which was assigned to the stronger CO binding of the Fe and Mn porphyrine-like structures. Zu et al. [16] successfully synthesized atomically dispersed Sn sites on nitrogen-doped carbon, which performs excellent activity and stability for formate generation at a kilogram scale with a quick freeze-vacuum drying-calcination method. Many Ni-based, Fe-based and Co-based SACs have exhibited high electrocatalytic activity and Faradaic effectivity (FE) for the CO2RR with CO as the primary product due to the moderate adsorption energies of *COOH and *CO intermediates, as well as the high activation barrier for the hydrogen evolution reaction (HER) [17,18].

Though the widespread study on the single-atom catalysts enhanced the utilization efficiency of metal atoms, most of the current studies are limited to the reduction of CO2 to C1 products. Compared to C1 products, C2+ products have a higher economic and chemical utilization value [19,20,21]. Cu-based SACs, up to now, have still performed good electrochemical reduction of CO2 to C2+ chemicals [22,23]. However, Karapinar [24] revealed that the atomically dispersed CuNx sites could reversibly convert into Cu clusters during CO2RR, which are suggested as the real multiple active sites for CH3CH2OH production. Considering the fact that a single metal atom can accommodate only a single CO, it is difficult to activate two CO2 molecules simultaneously to trigger the C-C coupling reaction based on an isolated metal center. Thus C-C coupling proceeding on the single metal atoms is quite difficult. Therefore, the catalysts with multiple active sites need to be considered to achieve the conversion from CO2 to C2 products [25].

Transition metal (TM) clusters with precise atomic numbers can offer multiple active sites, tune the size-dependent catalytic activity [26], and allow them to find the highest reactivity for the activation and dissociation of strong chemical bonds from CO2. Xu et al. [27] reported a facile underpotential deposition technique to fabricate Cu clusters on carbonaceous substrates via rationally introducing S dopants in graphite foam. The obtained free-standing electrode exhibited high activity and excellent long-term stability toward oxygen reduction reaction. Pei et al. [28] found the trimeric metal clusters anchored on N-doped porous graphitic sheets possess a good selectivity and superiority towards CO2RR to multi-carbon products due to the multiple active sites.

Considering the loading of metal clusters with a precise number of atoms on the active substrate can not only avoid the problem of low stability of bare metal cluster catalysts at room temperature but also further enhance their stability and catalytic efficiency. Graphite-based materials are currently widely used as substrates for electrocatalysts. To access C2 products more efficiently, herein, we employed density functional theory (DFT) calculations to explore the CO2RR catalyzed by Fen (n = 1, 3–5) anchoring on N-doped carbon (Fen-NC) to C2H4 and C2H5OH in this work. Furthermore, the Co5 cluster supported on N-doped carbon (Co5-NC) was explored comparatively. We found that Fe5 loaded on NC exhibit significant activity for promoting the reduction of CO2 to C2 products, while the Co5 cluster has higher priority for the selective synthesis of C2H4. Our findings provide insights into the design of highly active catalysts for CO2RR and create a platform for developing metal cluster-NC electrocatalysts.

2. Theoretical Method

First principle calculations were performed using DFT with spin polarization utilizing the Vienna Ab initio Simulation Package (VASP). The projected augmented wave (PAW) [29,30,31] was used, and the generalized gradient approximation (GGA) realized by the Perdew–Burke–Ernzerhof function (PBE) was adopted to incorporate the exchange-correction functional [32]. A 2 × 2 × 1 Monkhorst-Pack K-point was sampled in the Brillouin zone, and a cut-off energy of 500 eV was set for geometric optimization. The convergence criteria are of 10−5 eV in energy between two electronic iteration steps and 0.02 eV/Å in force for every atom [33]. Our calculations of catalytic performance are based on the computational hydrogen electrode (CHE) proposed by Nørskov et al. [34]:

The change of the free energy for the step can be equal to the reaction: at 0 V versus the reversible hydrogen electrode (RHE) at all pH values.

We employ five types of small iron clusters Fen (n = 1, 3–5) supported on nitrogen-doped carbon sheets as the calculation models. The defects of NC provide ideal anchor sites for the iron cluster. To estimate the stability of supported TM clusters, the binding energy (E) of TM cluster on NC is calculated by Equation (2)

Here is the total energy of the optimized TM cluster supported on NC. The terms and refer to the energies of isolated TM cluster and support. We calculated the cohesive energy (E) of each TM cluster to further evaluate the stability of TM catalysts, with E defined as:

Here the and represent the energy of the total energy of the TM cluster and the energy of single TM atom; n is the number of TM atoms in the cluster. The more negative cohesive energy (E) indicates a more stable structure.

The adsorption energy (E) of every intermediate species is defined by Equation (4) where refers to the total energy of the adsorbed species on the supported TMn cluster, and is the energy of supported TMn cluster. refers to the energies of in gas phase, respectively. The more negative adsorption energy indicates a stronger binding between TM cluster and NC support. Gibbs free energy change (ΔG) [35,36] is defined as: where , , and are the total energy difference, the zero-point energy difference, the difference in enthalpic correction and the entropy change between the products and reactants obtained from DFT calculations, respectively. The zero-point energies (ZPE) and total entropies of the gas phase were computed from the vibrational frequencies, and the vibrational frequencies of the adsorbed species were also computed to obtain the ZPE contribution to the free energy expression. Only vibrational modes of the adsorbates were computed explicitly, while the catalyst sheet was fixed (assuming that vibration contribution to the free energy from the substrate is negligible) [37,38]. T is the temperature (298.15 K). The influence of applied potential is: , where U is the external potential versus RHE, e is the electron transfer, and n is the number of proton–electron pairs. is the free energy correction due to the concentration of H+. , where kB is the Boltzmann constant, and the value of pH was assumed to be zero for acidic conditions.

3. Results and Discussion

3.1. The STABILITY Analysis of Fen-NC

For the Fen clusters supported on the NC substrate (Fen-NC), n = 1, 3, 4 and 5 were chosen to be studied here since the Fe2 cluster is unstable [39]. As shown in Figure 1, for the adsorption of a single Fe atom on the NC substrate (Figure 1a), the mono Fe atom coordinated with the four nitrogen atoms and Fe-NC maintains a perfect monolayer structure, which is in agreement with previous results [40]. For the adsorption of Fen clusters with n ranging from 3 to 5 (Figure 1b–d), one Fe atom is anchored at the same position with that in a single Fe atom. Two Fe atoms bound to the doped nitrogen atoms with distances of about 2.1~2.3 Å, respectively, while the other Fe atoms bound together through Fe-Fe metal bonds.

Figure 1

Top-view and side-view of optimized structures of Fe1 (a), Fe3 (b), Fe4 (c) and Fe5 (d) clusters supported on four nitrogen-doped carbon. The gold, blue and gray spheres represent Fe, N and C atoms, respectively. The different Fe sites are marked with white numbers.

As listed in Table 1, all the binding energies of Fen clusters on NC supports are thermodynamically favorable (Eb < 0). With the increase in Fe atoms, the binding energy decreases except for the magic Fe5 cluster, which means that the small Fe clusters may tend to aggregate from small clusters to bigger clusters on NC support. The reason for the decreased binding energy of the magic Fe5 cluster lies in that the Fe5site located at the top site, as shown in Figure 1d. Hence there is no interaction with the NC support. What is more, the cohesive energy of various Fen clusters was also explored according to Equation (3), as shown in Table 1. It can be found that with the increase in Fe atoms in the cluster, the cohesive energy becomes thermodynamically favorable.

Table 1

The Bader charge of Fe atom in various Fen-NC, binding energies (E) between the Fen cluster and NC, and the cohesive energy (E) of Fe atoms in various Fen clusters.

Catalyst	* Bader Charge (e)					Eb (eV)	Ec (eV)
	Fe 1site	Fe 2site	Fe 3site	Fe 4site	Fe 5site
Fe1-NC	−1.07	-	-	-	-	−9.01	0
Fe3-NC	−0.80	−0.32	−0.31	-	-	−9.73	−0.35
Fe4-NC	−0.81	−0.34	−0.33	−0.19	-	−9.79	−0.62
Fe5-NC	−0.81	−0.36	−0.35	−0.28	−0.01	−8.34	−1.01

* negative Bader charge means electron loss.

3.2. Electrocatalytic CO2RR

The charge difference between two active transition atoms plays a key role during CO2RR, and the mixed oxidation state of the catalytic centers can boost the C-C coupling [41,42]. Thus the Bader charges of various Fe clusters adsorbed on NC were investigated. As listed in Table 1, one can find that the electrons can transfer from the clusters to the support, making the whole Fe cluster positively charged, and each Fe atom in the cluster is positively charged as well. Herein, the Fen clusters are favorable to the CO2RR; thus the electron-accepting properties of the positively charged Fe and Co sites are favorable for stabilizing the CO2RR intermediates [43]. The largest charge transfer can be determined for the single Fe atom configuration since the monatomic Fe interacts with four coordinated N atoms. Furthermore, significant discrepancies for each charged Fe atom in the Fe3, Fe4 and Fe5 clusters can be determined, which is beneficial for the C-C coupling reaction.

The conversion of CO2 to CO catalyzed by various Fen clusters was calculated, as shown in Figure 2; the PDS is the *CO to CO for all the Fen-NC with a maximum Gibbs free energy (). The ΔG3 are 0.91, 1.14 and 1.58 eV for the Fe-NC, Fe3-NC and Fe4-NC, respectively. This value increases to 1.94 and 2.29 eV at two different sites of Fe5-NC. Thus it can be deemed that the Fe-NC, Fe3-NC and Fe4-NC require a lower overpotential to drive the desorption of CO, indicating that these structures favor the conversion of CO2 to CO. The step with strong *CO binding leads to a positive ΔG of CO desorption. The relatively strong binding of *CO on Fe-Nx is fully supported by the experimentally confirmed exclusive ability of the Fe-Nx catalyst to produce the hydrocarbon CH4 [44].

Figure 2

(a) The optimized structures of each intermediate during the CO2 reduction reaction from CO2 to CO, and (b) the Gibbs free energy profiles of CO2 reduction to CO on Fen-NC catalysts during CO2RR.

In simple terms, one could say that to produce subsequent reaction products from CO during the CO2RR; the CO molecule must be bound strong and long enough to undergo subsequent dissociation and hydrogenation steps to arrive at CH4 or other small organic molecules. Herein, our work focuses on the reduction of CO2 to C2 products, which requires improved selectivity and activity by inhibiting the unwanted, i.e., C1 hydrocarbon reaction pathway, which favors both the stabilization of *CO on the catalyst surface and the formation of C-C bonds. The strongest interaction for CO on the two Fe5-NC sites means that CO does not leave the iron cluster surface easily, which, in turn, favors the subsequent C2 product conversion. Therefore, the following study focuses on the performance of electrocatalytic CO2 reduction to C2H4 and CH3CH2OH over the Fe5-NC cluster. In order to further extend this result to other systems, the Co5 cluster supported on NC was chosen to be studied comparatively (Figure S1). As shown in Table 2, the Eb value between the Co5 cluster and the NC is −10.00 eV, and the Ec value of the Co atom is −1.39 eV, which means the Co5 cluster can adsorb stably on the NC.

Table 2

The Bader charge of Co atoms in Co5-NC, binding energy (E) between the Co5 cluster and NC and the cohesive energy (E) of Co atoms in the Co5 cluster.

Catalyst	Bader Charge (e)					Eb (eV)	Ec (eV)
	Co 1site	Co 2site	Co 3site	Co 4site	Co 5site
Co5-NC	−0.71	−0.34	−0.31	−0.24	0.06	−10.00	−1.39

To further study the mechanisms of CO2RR catalyzed by Fe5-NC and Co5-NC, the optimized structures and the energy profiles along the reaction coordinate for CO2RR to both C2H4 and CH3CH2OH on Fe5-NC and Co5-NC are calculated as shown in Figure 3 and Figure 4, respectively. It can be found that two strongly adsorbed CO molecules adsorbed on the two adjacent metal atoms through carbon atoms either on the Fe5 or the Co5 clusters before the C-C bond formation. The two CO molecules will couple with each other via the top Fe or Co atom in the following steps of CO2RR. For the CO2RR on Fe5-NC, the Gibbs free energy for the hydrogenation of the *CO dimer is uphill with an energy value of 0.79 eV, which is the highest energy during the formation of both C2H4 and CH3CH2OH. Thus, it can be deemed that the C-C coupling reaction for the CO2RR from CO2 to C2 chemicals is the PDS. Furthermore, it can be speculated that both C2H4 and CH3CH2OH products can be achieved with Fe5-NC catalyst, and the amount of C2H4 should be much more than the CH3CH2OH. Because the Gibbs free energy for the hydrogenation of *CH2CHO to C2H4 is energy thermodynamically favorable, while the hydrogenation of *CH2CHO to *CH3CHO is an uphill reaction with a free energy of 0.26 eV.

Figure 3

Optimized structures of intermediate Fe5-NC (a) and Co5-NC catalysts (b) during the CO2RR process.

Figure 4

Gibbs free energy profiles for CO2RR on Fe5-NC and Co5-NC.

For the CO2RR on Co5-NC, the Gibbs free energy for the hydrogenation of the *CO dimer is uphill with an energy of 1.01 eV, which is 0.22 eV higher than that of the Fe5-NC. However, the Gibbs free energy for the hydrogenation of *CH2CHO to *CH3CHO is (0.07 eV higher) comparable with that in the C-C coupling reaction, which means that the hydrogenation of *CH2CHO to *CH3CHO becomes the PDS. Thus, most of the final C2 products should be C2H4. In general, the Fe5-NC has good catalytic activity towards the C2 chemicals with relatively lower free energy change (0.53 eV), while the selectivity is not as good as Co5-NC. However, Co5-NC possesses better selectivity while the activity is lower than Fe5-NC.

The d-band center of the TM and its electronic occupancy can affect the bonding strength between the intermediate and the catalytic surface. As shown in Figure 5, the PDOS of Fe d orbit from the top- and sub-layer structures show much more differences than that of the Co d orbit on Co5-NC. Furthermore, the d band center of the top Fe atom in the Fe5 cluster is −3.48 eV, while it becomes −1.59 eV for the sub-layer atoms. However, the d band center of the top Co atom is −1.34 eV, and it only changes to −1.47 eV for the sub-layer atoms. A much bigger discrepancy of the d band center between the top Fe atom and sub-layer atoms than that in the Co5 cluster may boost the C-C coupling reaction, which could be called the synergy effect between the top- and sub-layer metal atoms. Our findings are consistent with the synergy effect between Cu+ and Cu0, and the surface can significantly improve the kinetics and thermodynamics of both CO2 activation and CO dimerization. Cu metal embedded in an oxidized matrix catalyst can promote CO2 activation and CO dimerization for electrochemical reduction of CO2 [41].

Figure 5

The PDOS of the d orbitals of the single top Fe atom (red) and average three middle Fe atoms (blue) in Fe5-NC (a), and the PDOS of the d orbitals of the single top Co atom (red) and average three middle Co atoms (blue) in Co5-NC (b).

Our theoretical calculations found that the multiple active sites in both the Fe5 and Co5 cluster-based catalysts facilitate the stabilization of *CO on the catalyst surface and the formation of C-C bonds. Both geometrical effects and electronic effects are the key factors leading to the Fe5 and Co5 clusters exhibiting better activity and/or selectivity over the single metal component. Furthermore, the tunable synthesis of Fe and Co alloys supported on NC may promote both their activity and selectivity toward CO2RR. Therefore, Fe5, Co5 and the related tunable alloy clusters show great potential applications in electrocatalytic CO2RR, and our methods provide a concept for designing the improved CO2RR electrocatalysts.

4. Conclusions

The stability of the Fen (n = 1, 3, 4, and 5) clusters was studied first, and it can be found that the Fen anchors stably on the nitrogen-doped carbon via a basic Fe-NC structure. With the increasing of Fe atoms in the cluster, both the binding energy and cohesive energy become thermodynamically favorable, which means a small cluster tends to aggregate to be a bigger one. While for the Fe5 cluster, the binding energy decreases because there is no interaction between the topmost Fe atom with the NC support anymore. In addition, the CO desorption is the most difficult on the Fe5 cluster, which is beneficial to the subsequent reaction products from CO. Hence, the Fe5-NC cluster was chosen to be studied as our C2 catalyst, and Co5-NC was comparatively studied as well. The results show that Fe5-NC has better activity towards CO2RR, and the products should be the mixed C2H4 and CH3CH2OH, since the PDS is the C-C coupling reaction with a free energy change of only 0.79 eV. The free energy change is only 0.53 eV for the reduction of CH2CHO to CH3CHO, and the reduction of CH2CHO to C2H4 is a spontaneous step without any free energy change. Considering the fact that C2H4 is a gas, Fe5-NC should be a good catalyst for CO2RR to liquid ethanol with a relatively lower yield since part of the C2H4 gas will also be produced. Furthermore, Co5-NC possesses a relatively good selectivity, but bad activity since the reduction of CH2CHO to CH3CHO is the PDS, and the free energy change is 1.09 eV. The PDOS and d band center analysis demonstrates that the relative energy favorable C-C coupling reaction on the Fe5 cluster could be attributed to the larger discrepancy of d electrons of the two CO-adsorbing Fe atoms. This paper predicts a good application prospect of TM clusters supported on nitrogen-doped graphene for CO2RR, and the new insight into the relationship between selectivity and activity sheds light on a new route for understanding and designing highly efficient non-precious catalysts for CO2RR.