Selective C−C Coupling by Spatially Confined Dimeric Metal Centers

Summary

Direct conversion of carbon dioxide (CO₂) to high-energy fuels and high-value chemicals is a fascinating sustainable strategy. For most of the current electrocatalysts for CO₂ reduction, however, multi-carbon products are inhibited by large overpotentials and low selectivity. Herein, we exploit dispersed 3d transition metal dimers as spatially confined dual reaction centers for selective reduction of CO₂ to liquid fuels. Various nitrogenated holey carbon monolayers are shown to be promising templates to stabilize these metal dimers and dictate their electronic structures, allowing precise control of the catalytic activity and product selectivity. By comprehensive first-principles calculations, we screen the suitable transition metal dimers that universally have high activity for ethanol (C₂H₅OH). Furthermore, remarkable selectivity for C₂H₅OH against other C₁ and C₂ products is found for Fe₂ dimer anchored on C₂N monolayer. The role of electronic coupling between the metal dimer and the carbon substrates is thoroughly elucidated.

Graphical Abstract

Highlights

• •Dual metal centers provide an exclusive pathway for CO₂ reduction to C₂ products
• •Activity and selectivity are modulatable by the metal-support interaction
• •Fe₂ dimer anchored on C₂N leads to remarkable selectivity for ethanol

Catalysis; Atomic Electronic Structure; Energy Sustainability; Numerical Method in Materials Science

Introduction

Production of liquid fuels by catalytic convertion of CO₂, the main greenhouse gas and meanwhile an abundant carbon feedstock, has been regarded as an appealing strategy to solve both energy and environmental crises, albeit facing great challenges (Birdja et al., 2019, Jia et al., 2019, Amal et al., 2017). Copper-based materials have been widely adopted as catalysts for electro-reduction of CO₂ to multi-carbon (C₂ or C₂₊) products (Zheng et al., 2019). Although fairly good activity can be achieved by modification or morphology engineering of copper, such as sculpturing it into nanoparticles or nanocubes, doping or alloying, and making oxide-derived copper, the selectivity and efficiency of most copper-based electrocatalysts remain unsatisfactory for commercialization of the CO₂ conversion technique to high-energy fuels and high-value chemicals (Gao et al., 2019, Kim et al., 2017, Wang et al., 2018a, Zhou et al., 2018).

Recenlty, transition metal atoms dispersed on nitrogen-doped porous carbon nanomaterials emerge as a promising category of electrocatalysts for CO₂ reduction, which have maximum atomic efficiency, high electrical conductivity and good durability, and can be facilely synthesized in the laboratory (Bayatsarmadi et al., 2017, Chen et al., 2019a, Cheng et al., 2018, Wang et al., 2019a). The transition metal atoms are usually anchored in the pores of the carbon matrix and coordinated with the nitrogen atoms, exhibiting unique electronic states and acting as isolated reaction centers for CO₂ reduction. Remarkable activity and selectivity toward carbon monoxide (CO) has been observed for various dispersed transition metal atoms (Fe, Co, Ni, Mn, and Cu) on N-doped graphene, carbon nanosheets or nanospheres, with selectivity up to 97% and Faradaic efficiency above 80% (Jiang et al., 2018a, Ren and Zhao, 2020, Wang et al., 2019b, Yang et al., 2018, Zhang et al., 2018). First-principle calculations show that the activity highly depends on the type of metal atoms, which provide different binding strengths with the reaction intermediates (Ju et al., 2017). The single metal sites also have an advantage of suppressing the competing hydrogen evolution reaction (HER), due to the unique adsorption configuration of H^∗ species compared with those on the transition metal surfaces (Bagger et al., 2017).

Furthermore, homonuclear and heteronuclear dimers of transition metal immobilized in carbon-based nanostructures, such as Fe₂ and Fe-Co on nitrogenated graphitic carbon materials, Fe-Ni on N-doped graphene, and Pt-Ru on g-C₃N₄, have been synthesized in the laboratory (Wang et al., 2017, Wang et al., 2018b, Ye et al., 2019, Zhou et al., 2019). This opens up the windows for a broader range of chemical processes that require dual reaction centers either with enhanced activity or carrying different functionalities simultaneously. For instance, Ren et al. fabricated diatomic Fe-Ni sites embedded in nitrogenated carbon (Ren et al., 2019). By taking advantage of the strong binding capability of Fe with CO₂ molecule and the weak adsorption of CO on Ni, they achieved impressively high selectivity of 99% for CO and Faradaic efficiency above 90% over a wide potential range from −0.5 to −0.9 V, reaching 98% at −0.7 V versus reversible hydrogen electrode (RHE). On the theoretical side, a Cu₂ dimer supported on the C₂N monolayer was predicted to have high selectivity for methane (CH₄), whereas dimerization of two CO species leading to the formation of ethene (C₂H₄) is possible with an energy cost of 0.76 eV (Zhao et al., 2018). Heteronuclear dimers such as V-Mo on 2D C₂N and Cu-B on g-C₃N₄ have been shown to effectively reduce CO₂ to ethanol (C₂H₅OH) and C₂H₄, owing to the synergistic interaction and asymmetric coupling between two reaction centers yielding favorable binding strength for the formation of C₂ intermediates (Li et al., 2019, He et al., 2020).

Two adjacent metal atoms that are spatially confined in a hole of N-doped carbon materials as unique active sites not only enable the simultaneous fixation of two CO₂ molecules but also sterically limit the reaction pathways that may be beneficial for C−C coupling toward C₂ or C₂₊ products. Moreover, various combinations of metal dimers and carbon substrates give high degrees of freedom for modulating the catalytic performance. However, the atomistic mechanism and composition recipe of such heterogeneous catalysts remain largely unknown, which impede their rational design and experimental synthesis for practical uses.

Here we exploit 3d transition metal dimers immobilized on various nitrogenated holey carbon sheets for selective reduction of CO₂ to C₂ products. By systematic first-principle calculations, the detailed C−C coupling mechanism on the spatially confined dual metal centers has been elucidated for the first time. The suitable transition metal elements and carbon substrates that lead to high activity and selectivity for C₂H₅OH and C₂H₄ are screened, and the underlying electronic structure-activity relationship is unveiled. These theoretical explorations illuminate important clues for precisely engineering the dispersed metal catalysts on porous carbon nanomaterials for direct conversion of greenhouse gas to multi-carbon hydrocarbons and oxygenates.

Results and Discussion

In the laboratory, N-doped graphitic carbon materials with controllable doping contents (up to 16.7% of N content) and atomic geometries can be achieved via either direct synthesis or posttreatment (Xue et al., 2012, Xu et al., 2018, Qu et al., 2010). Here we focused on pyridine N dopants in graphene, which are the main doping species at high N contents and are usually associated with the vacancies or pores of the carbon basal plane (Sheng et al., 2011, Sarau et al., 2017). As displayed in Figure 1, we considered a series of N-doped holey graphene monolayers, comprising C vacancies of various sizes (denoted as V_n, n = 2, 3, 4, 6) with the edges coordinated with different numbers of N atoms (denoted as mN, m = 4, 5, 6). Specifically, 4N-V₂, 5N-V₃, and 6N-V₄ systems can be viewed as four, five, and six N atoms decorating the edges of di-vacancy, tri-vacancy, and tetra-vacancy in graphene, respectively, all of which have been commonly observed in experiment (He et al., 2014, Lin et al., 2015, Wang et al., 2018c). Note that the V₆ pore in graphene is a favorable defect according to transmission electron microscopy experiment (Robertson et al., 2015), and our previous calculation showed that N-doped V₆ (namely 6N-V₆) has extraordinary thermodynamic stability (Luo et al., 2013). We further created a number of randomly N-doped graphene lattices, which shows that the 6N-V₆ configuration would emerge as N doping content reaches 10% (see Figure S1 for details). Besides the N-doped graphitic sheets, we also considered the synthetic carbon nitride monolayers, including g-C₃N₄ and C₂N (Zhao et al., 2014, Mahmood et al., 2015). All these porous N-coordinated carbon sheets have formation energies (defined by Equation S1 in Supplemental Information) in the range of 0.16–0.21 eV/Å, whereas the N-free V₆ is higher in energy by over 0.46 eV/Å than the others (Table 1). These nitrogenated 2D holey carbon materials are ideal templates to stabilize and disperse metal atoms or small clusters. Indeed, isolated Fe₂, Fe-Ni, and Fe-Co dimers embedded in 6N-V₄(a), as well as Fe₂ and Pt-Ru dimers anchored on g-C₃N₄ have already been realized in experiment (Ye et al., 2019, Wang et al., 2017, Zhou et al., 2019, Ren et al., 2019, Tian et al., 2018).

Figure 1.

Atomic Structures of a Fe₂ Dimer Anchored on Various Nitrogenated Holey Carbon Monolayers (Top Panel: Top View; Bottom Panel: Side View)

The C, N, and Fe atoms are shown in gray, blue, and orange colors, respectively.

Table 1.

Structural and Energetic Properties of Supported Fe₂ Dimer

Substrate	Eform (eV/Å)	Eb (eV)	d (Å)		CT (e)	ΔECO2∗ (eV)
			Fe−Fe	N−Fe
4N-V2	0.17	−5.01	2.09	1.98	0.71	−1.02
5N-V3	0.19	−7.33	1.91	1.87	0.97	−1.15
6N-V4(a)	0.20	−9.47	2.21	1.94	0.96	−0.11
6N-V4(b)	0.21	−7.48	2.14	1.98	0.81	−1.15
6N-V6	0.16	−6.02	1.96	2.00	0.76	−1.12
C2N	–	−5.80	2.01	1.97	0.74	−0.64
g-C3N4	–	−5.09	1.98	1.99	0.72	−1.58
V6	0.67	−12.03	2.21	1.94	0.87	−0.30

Formation energy (E_form) of various nitrogenated 2D holey carbon materials, binding energy (E_b) of a Fe₂ dimer on the carbon sheet, bond length (d) of Fe−Fe and N−Fe/C−Fe bonds, Mulliken charge transfer (CT) from Fe₂ to the carbon sheet, and adsorption energy of a CO₂ molecule (ΔE_CO2∗) on the supported Fe₂ dimer.

To evaluate the capability of various supported metal dimers for CO₂ reduction toward C₂ products, we first explored the atomic structures, electronic and adsorption properties of dimeric 3d transition metal clusters on the 6N-V₆ monolayer (as will be shown later, this substrate gives metal dimers the highest activity for CO₂ reduction). As presented in Figures 1 and S2, all metal dimers are embedded in the hole of the graphitic sheet, except that Sc₂ with a larger atomic size induces a noticeable buckling of 0.94 Å in the out-of-plane direction. Four N−metal bonds are formed with bond length of 1.95–2.09 Å, and the metal−metal bond length ranges from 1.96 Å to 2.79 Å (Table 2). The binding energy (defined by Equation S2 in Supplemental Information) between the metal dimer and the graphitic sheet is −4.29 to −10.28 eV, excluding the possibility of dissociation or aggregation of the metal dimer. The thermal stability of these carbon-substrate-anchored metal dimers was further assessed by ab initio molecular dynamics (AIMD) simulations, which manifest that they can sustain at least 800 K for 10 ps with small vertical displacement of metal atoms (<0.2 Å) (see Figure S3 for details), suggesting superior thermal stability for practical uses.

Table 2.

Structural and Energetic Properties of Various Supported 3d Transition Metal Dimers

Metal Dimer	Eb (eV)	d (Å)		ΔE (eV)		εd (eV)
		M−M	N−M	CO2	2CO2
Sc2	−10.28	2.79	2.09	−3.40	−3.71	1.16
Ti2	−8.50	2.17	1.99	−2.85	−3.31	0.62
V2	−8.80	2.14	1.96	−2.26	–	0.41
Cr2	−4.99	2.16	1.97	−1.44	−0.76	0.07
Mn2	−6.52	2.04	2.01	−1.05	−0.48	−0.42
Fe2	−6.02	1.96	2.00	−1.12	−0.50	−1.00
Co2	−5.71	2.10	1.95	−1.20	–	−1.09
Ni2	−5.93	2.17	2.00	−0.82	–	−1.12
Cu2	−4.29	2.35	1.96	−0.35	–	−2.08

Binding energy (E_b) of various 3d transition metal dimers anchored on the 6N-V₆ monolayer, bond lengths (d) of metal dimer (M−M) and N−metal (N−M), adsorption energy of single and dual CO₂ molecules on the supported metal dimers (ΔE), and the d band center (ε_d) of the supported metal dimers (Hammer and Nørskov, 2000).

A CO₂ molecule can favorably chemisorb on these dispersed metal dimers except Cu₂. The molecule is bended in the bidentate configuration with O−C−O angle of 124.90–141.96°. The C atom and one of the O atoms of CO₂ form two bonds with the underlying metal atoms; the C−O bond length is elongated to 1.21–1.36 Å, compared with 1.16 Å for a free CO₂ molecule. The dynamic process of CO₂ adsorption was also examined by AIMD simulations at 100 K and 300 K, respectively, both showing that the molecule can quickly chemisorb on the dimeric metal centers within a simulation time of 1 ps (see Videos S1 and S2 for the dynamic movies). The adsorption energy (defined by Equation S3 in Supplemental Information) of CO₂ ranges from −0.82 eV to −3.40 eV. Overall speaking, stronger binding is provided by the metal element with fewer d electrons. The trend of activity can be understood by the electronic density of states (DOS) shown in Figure 2A. Taking Fe₂@6N-V₆ as an example, hybridization between the d orbitals of Fe₂ dimer and the p orbitals of 6N-V₆ monolayer substrate is evident, with prominent electronic states close to the Fermi level mainly contributed by the Fe atoms (see Figure S4 for projected DOS). Electron transfer of 0.73 e occurs from Fe₂ to 6N-V₆ monolayer, which lifts the Fermi level of the hybrid system above the 2π^∗ state of CO₂. As a result, Fe₂@6N-V₆ can favorably donate about 0.71 electrons to the antibonding orbital of CO₂, as manifested by the differential charge densities in Figure 3A, which is a general mechanism for activation of reactant molecules on metal active centers (Liu et al., 2018). As depicted in Figure 2B, CO₂ adsorption energy generally follows a linear relationship with the d band center of the supported metal dimers (relative to the Fermi level), as the metal dimer with a higher d band center would provide stronger binding with CO₂ (Hammer and Nørskov, 2000).

Figure 2.

Electronic Structure-Activity Relationship

(A) From left to right: molecular orbital levels or local density of states (DOS) of a free and a bended (with C−O−C angle of 130°) CO₂ molecule in vacuum, an adsorbed CO₂ molecule on Fe₂@6N-V₆, an individual 6N-V₆ monolayer, and a Fe₂ dimer. The insets display the HOMO and LUMO charge densities of CO₂. The energy is relative to the vacuum. The dashed line shows the Fermi level, with the occupied states shadowed. The hybridization region between d orbital of Fe atoms and 2π∗ state of CO₂ is shadowed in green. The dark blue and orange colors represent the d orbital of Fe atoms and p orbital of N atoms, respectively.

(B) The d band center (ε_d) of various supported 3d transition metal dimers as a function of the adsorption energy of single CO₂ molecule. The blue/orange/gray symbols denote that two/one/none CO₂ molecule can be chemisorbed on the metal dimer. The dashed line is a linear fit of the data points.

(C) Charge transfer (CT) from the Fe₂ dimer to various nitrogenated carbon holey monolayer as a function of the adsorption energy of dual CO molecules. The dashed line is a linear fit of the data points. The insert shows the differential charge density of Fe₂@6N-V₆. The yellow and cyan colors represent the electron accumulation and depletion regions, respectively, with an isosurface value of 0.005 e/ų.

Figure 3.

Structures and Energies of Molecular Adsorption

(A) From left to right: differential charge densities of single and dual CO₂ molecules, single and dual CO molecules adsorbed on Fe₂@6N-V₆. The yellow and cyan colors represent the electron accumulation and depletion regions, respectively, with an isosurface value of 0.005 e/ų.

(B) Adsorption energies of single and dual CO₂ and CO molecules on the Fe₂ dimer anchored on various nitrogenated holey carbon monolayers. The C, N, O, and Fe atoms are shown in gray, blue, red, and orange colors, respectively.

In addition, we examined the capability of various dispersed 3d transition metal dimers for activating two CO₂ molecules simultaneously, which is a prerequisite for C−C coupling to yield C₂ products. Several candidate systems including Sc₂, Ti₂, Cr₂, Mn₂, and Fe₂ dimers on the 6N-V₆ monolayer have adsorption energies of −3.71 to −0.48 eV for fixation of two CO₂ molecules (Figures 3 and S5), whereas the other metal dimers are only able to bind one CO₂ molecule. Considering that Fe is an earth-abundant element and dispersed Fe atoms and dimers can be readily obtained in the experiment (Ye et al., 2019, Tian et al., 2018), thereafter we explored Fe₂ dimer on various nitrogenated 2D holey carbon materials as a representative of dual metal centers.

Figure 1 presents the structures of a Fe₂ dimer immobilized on several 2D carbon substrates. The dimer forms 4–6 bonds with the neighboring N or C atoms, having bond lengths of 1.91–2.21 Å for Fe−Fe and 1.87–2.00 Å for N−Fe (C−Fe) bonds, respectively, and the binding energies are −5.01 to −12.03 eV (Table 1). The Fe₂ dimer exhibits different buckling height in the out-of-plane direction (0.01–2.06 Å) and meanwhile induces some local vertical distortions on the carbon basal plane (0.09–0.35 Å). The dimer-substrate coupling strength depends on the size of the hole as well as the saturation degree of the edge atoms. For instance, binding strength between Fe₂ and 4N-V₂, 5N-V₃, and 6N-V₄(a) increases with both N content and hole size. The bonding interaction between Fe₂ and g-C₃N₄ or C₂N is relatively weak, due to the electronic saturation of these two semiconducting carbon nitride monolayers (as manifested by their large band gaps). In sharp contrast, Fe₂ is strongly anchored on the nitrogen-free V₆ defect that has six unsaturated carbon atoms on the hole edge, thereby leading to the largest binding energy of −12.03 eV.

All the supported Fe₂ dimers are able to chemisorb two CO₂ molecules with total adsorption energies of −0.23 to −1.62 eV (compared with −0.11 to −1.58 eV for adsorption of single CO₂ molecule), as revealed by Figure 3B. Our nudged elastic band (NEB) calculations show that adsorption of the second CO₂ molecule involves kinetic barriers of 0.29–1.04 eV. Both CO₂ molecules are bended with O−C−O angle of 141.00–152.13° and elongated C−O bond lengths of 1.17–1.29 Å. The C atom in each CO₂ is bonded to the underlying Fe atom with Fe−C bond length of 1.93–2.12 Å. Furthermore, we investigated the interaction between the dispersed Fe₂ dimers and the CO molecule, which is an important reaction intermediate in the CO₂ reduction process. Our calculations indicate strong binding of CO on the anchored Fe₂ dimers, with adsorption energies of −2.94 to −4.04 eV (−1.94 to −2.70 eV) for two (one) CO molecules. Consequently, desorption of CO from dual metal centers would be rather difficult, which allows further protonation of CO and thus provides the opportunity for successive C−C coupling.

The distinct binding capability of various supported Fe₂ dimers with gas molecules can be related to the electronic coupling between Fe₂ and the carbon substrate. As displayed in Figure 2C, the amount of charge transfer from Fe₂ to the substrate varies from 0.71 e to 0.97 e. Generally speaking, less electron transfer leads to higher activity of the Fe₂ dimer for CO₂ and CO chemisorption, which is consistent with the trend of binding energies between Fe₂ and the carbon templates discussed before (Table 1). It is the N content, the degree of electronic saturation of the hole edge, and the bond configuration of Fe₂ in the hole that jointly determine the coupling strength between the metal dimer and the carbon sheets. Therefore, the nitrogenated 2D holey carbon materials with diverse morphologies and controllable N contents can not only stabilize and disperse metal dimers but also dictate the electronic structures and activity of the anchored metal dimers. By choosing proper metal elements and substrates, it is possible to delicately mediate their coupling strength and charge transfer, endowing large degree of freedom to optimize the activity and selectivity of various supported metal dimers for CO₂ reduction.

Note that graphitic N species are inevitably present in the experimentally synthesized N-doped carbon materials (Lin et al., 2014). To clarify their effect on the activity of the dispersed Fe₂ dimer, we investigated the CO₂ adsorption on Fe₂@6N-V₆ containing various numbers of graphitic N atoms at different distances from the 6N-V₆ hole (Figure S6). For all the considered systems, the CO₂ adsorption energies on the catalysts with and without substitutional N atoms on the graphene lattice differ by less than 0.16 eV, suggesting that existence of the graphitic N species has only minor impact on the catalytic properties of the Fe₂ dimer supported on pyridine holes of 2D carbon substrates.

Figure 4 shows the most efficient pathways for CO₂ reduction toward possible C₁ and C₂ products on the Fe₂ dimer immobilized on various nitrogenated carbon sheets, and the corresponding free-energy diagrams of various model systems calculated by the computational hydrogen electrode (CHE) model (Peterson et al., 2010) are given by Figures 5, S7, and S8. We used point (.) to represent the co-adsorption of two carbon intermediates on the catalyst and strigula (−) to indicate the coupling between two carbon intermediates. The maximum Gibbs free energy of formation ΔG among all the reaction steps defines the rate-determining step (RDS) and is thus denoted as ΔG_RDS. Overall speaking, formation of C₂ products first requires the activation of dual CO₂ molecules on the catalyst. By going through the carboxyl (COOH^∗) pathway, two CO^∗ intermediates can be generated; then protonation of CO^∗ leads to C₁ products such as methanol (CH₃OH) and CH₄. Alternatively, it paves a way to the coupling between two neighboring carbon intermediates, which is energetically favorable and kinetically easy, and finally yields C₂ products (C₂H₅OH and C₂H₄). A similar path for C−C coupling was also found for the other metal dimers anchored on the nitrogenated carbon sheet, as revealed by Figure S9 for Ni₂@6N-V₆ as an example.

Figure 4.

The CO₂ reduction pathways to various C₁ and C₂ products on the supported Fe₂ dimer

The H, C, N, O, and Fe atoms are shown in light blue, gray, blue, red, and orange colors, respectively.

Figure 5.

CO₂ Reduction Pathway

(A and B)(A) Free-energy diagram of CO₂ reduction to various C₁ and C₂ products (indicated by different colors) on Fe₂@C₂N. The blue numbers, from left to right, give the Gibbs free energy of formation for the rate determining step of C₂H₅OH, C₂H₄, and CH₃OH/CH₄. The local structures of selected reaction intermediates are presented in (B). The H, C, N, O, and Fe atoms are shown in light blue, gray, blue, red, and orange colors, respectively.

(C) Competing reactions of CO.HCOH^∗ to form C₁ and C₂ intermediates on Fe₂@C₂N. The insets display the structures of initial state (IS), transition state (TS), and final state (FS). The blue numbers give the kinetic barriers (middle) and heat of reaction (right).

Specifically, formation of two CO^∗ species on most of the considered Fe₂ dimers is uphill in the free-energy profile, involving energy steps of 0.24–0.87 eV. Then, reduction of CO^∗ gives rise to HCO^∗ species, which is lower in energy by up to 1.13 eV than the other possible intermediates such as COH^∗ (Figure S10). The CO^∗ → HCO^∗ conversion is endothermic with ΔG = 0.49–1.01 eV. Further protonation of HCO^∗ leads to HCOH^∗ and then produces a CH^∗ species by release of a H₂O molecule. The C−C coupling reaction is most likely to occur between a CH^∗ (or CH₂^∗) species and the neighboring CO^∗. Our NEB calculations suggest that the CO−CH^∗ coupling is exothermic and barrierless on all the considered Fe₂ dimers, except for Fe₂@C₂N and Fe₂@C₃N₄ that involve a small kinetic barrier of about 0.22 eV (Table S1). According to previous theoretical studies (Goodpaster et al., 2016, Jiang et al., 2018b), Cu(211) and (100), as the typical active surfaces for CO₂ reduction, favor dimerization of CO^∗ or CO−HCO^∗ coupling involving ΔG = −0.17–0.48 eV. For the present Fe₂ dimers on nitrogenated carbon sheets, however, CO−CO^∗ or CO−HCO^∗ coupling has higher ΔG than the values of CO−CH^∗ by 0.84–2.42 eV and thus is unlikely to occur.

Following the C−C coupling, successive reduction of CO−CH^∗ leads to CO−CH₂^∗, HCO−CH₂^∗, HCOH−CH₂^∗, HCOH−CH₃^∗, and finally yields C₂H₅OH. Alternatively, reduction of HCOH−CH₂^∗ can give rise to CH−CH₂^∗ with release of a H₂O molecule, and further protonation of CH−CH₂^∗ eventually produces C₂H₄. These elementary reactions involve relatively small steps of 0.15–0.73 eV in free-energy profile and thus would take place readily from the thermodynamic point of view. At the last step, desorption of C₂H₅OH^∗ and CH₂CH₂^∗ is endothermic by 0.11–0.59 eV and 0.23–1.69 eV, respectively. For most of the considered Fe₂ dimers, the rate-determining step for C₂H₅OH production is the CO^∗ → HCO^∗ conversion. The release of C₂H₄ mainly suffers from the strong binding of CH₂CH₂^∗ on the catalyst, which can be overcome by the reaction heat of the corresponding reduction step (0.61–2.07 eV) (Chen et al., 2019b), as well as by adopting some strategies such as the pulse electrolysis mode to accelerate desorption of the final products (Yano et al., 2007, Qiao et al., 2014).

On the other hand, formation of C₁ products is also possible on the dispersed Fe₂ dimers. As discussed earlier, HCOH^∗ can be reduced to CH^∗, followed by the CO−CH^∗ coupling. Alternatively, HCOH^∗ may be protonated to H₂COH^∗. Then, reduction of H₂COH^∗ yields CH₃OH or produces CH₂^∗ with release of a H₂O molecule followed by the generation of CH₃^∗ and CH₄. For Fe₂@4N-V₂, Fe₂@6N-V₆, and Fe₂@g-C₃N₄, the CO^∗ → HCO^∗ conversion is the rate determining step for both C₁ products. For Fe₂@6N-V₄(b) and Fe₂@5N-V₃, formation of CH₃OH from H₂COH^∗ protonation requires ΔG_RDS = 1.45 and 0.98 eV, respectively. In particular, Fe₂@C₂N encounters ΔG_RDS = 0.94 eV and a kinetic barrier of 0.77 eV during the reaction of HCOH^∗ → H₂COH^∗ for both C₁ products, whereas the competing step of CO−HCOH^∗ → CO−CH^∗ + H₂O has much reduced ΔG = −0.30 eV and a lower kinetic barrier of 0.42 eV (Figures 5A and 5C). This would lead to high selectivity for C₂ products on Fe₂@C₂N.

Figure 6A plots ΔG_RDS values for various C₁ and C₂ products from CO₂ reduction on the anchored Fe₂ dimers. Among the four products, C₂H₅OH exhibits the lowest ΔG_RDS = 0.57–1.01 eV, and the highest activity is achieved by Fe₂@6N-V₆ owing to its moderate adsorption strength with the reaction intermediates (indicated by the dashed blue line in Figure 6A). Formation of C₂H₄ is less favorable with ΔG_RDS = 0.58–1.76 eV due to the strong binding of CH₂CH₂^∗ on the Fe₂ dimers. Fe₂@6N-V₄(a), Fe₂@4N-V₂, Fe₂@6N-V₆, and Fe₂@g-C₃N₄ exhibit similar selectivity for C₂H₅OH, CH₃OH, and CH₄, whereas Fe₂@5N-V₃ favors both C₂H₅OH and CH₄ products. Remarkable selectivity for C₂H₅OH is obtained for Fe₂@C₂N and Fe₂@6N-V₄(b) with ΔG_RDS = 0.70 and 0.59 eV, respectively, notably lower than ΔG_RDS values for the other products (above 0.94 and 0.85 eV, respectively). Hence, these supported Fe₂ dimers have competitive activity but distinct selectivity with regard to the conventional Cu-based catalysts. It is known that Cu crystals mainly produce CO under low electrode potentials, whereas CH₄ and C₂H₄ are the main products at sufficiently high electrode potentials (about −1.0 V versus RHE in experiment) (Dai et al., 2017, Mistry et al., 2016). Previous calculations revealed that Cu(211) surface encounters ΔG_RDS = 0.74 eV for CH₄ and C₂H₄, whereas formation of CO is much more favorable with ΔG_RDS = 0.41 eV due to the relatively weak adsorption of CO on the Cu surface (adsorption energy ΔE = −1.01 eV) (Peterson et al., 2010). Differently, release of CO is prohibited on the present Fe₂ dimers that have strong adsorption energy of ΔE = −2.94 to −4.04 eV with CO molecule.

Figure 6.

Catalytic Performance for CO₂ Reduction

(A) Gibbs free energy of formation for the rate determining step (ΔG_RDS) for various C₁ and C₂ products from CO₂ reduction, and (B) competition between adsorption of a CO₂ molecule and an H^∗ species on the Fe₂ dimer anchored on various nitrogenated holey carbon monolayers.

As the electroreduction of CO₂ usually take place in the neutral aqueous condition, we further explored the solvent effect on the catalytic behavior of supported Fe₂ dimer. As a representative, we considered an explicit solvent model of Fe₂@C₂N. Our calculations show that hydrogen bonds are formed between water molecules and some adsorbed reaction intermediates (such as CO₂.CO₂^∗ and CO−CH^∗), which slightly stabilize those species on the catalyst in aqueous environment, consistent with the previous theoretical report (Zhao and Liu, 2020). The variations of CO₂ adsorption energy and Gibbs free energy of formation for elementary steps are below 0.29 eV, and the kinetic barriers of rate determining steps for various products increase by less than 0.35 eV, with regard to the model in vacuum (see Table 3 and Figure S11 for details). The predicted selectivity is consistent between the model in vacuum and in water. Therefore, free-energy calculations on electrocatalysis of CO₂ reduction using a model of catalyst in vacuum can generally predict reliable results on the trend of activity and product selectivity (De Luna et al., 2018, Zhuang et al., 2018, Li et al., 2018). Besides solvent effect, the surface charge on catalysts during the electrochemical reaction may modify the electronic states and impact the catalytic properties according to a previous theoretical report (Kim et al., 2018). Future studies with sophisticated model theory are necessary to comprehensively evaluate the catalytic performance of the proposed transition metal dimers on the nitrogenated carbon substrates.

Table 3.

Key Reaction Steps in Vacuum and Aqueous Condition

	Step	CO2→ CO2∗	CO2∗ + CO2→ CO2.CO2∗	CO.CH∗ → CO−CH∗	CO.HCOH∗ + H+ + e−→
					CO.H2COH∗	CO−CH∗ + H2O
Vacuum	ΔH Ea	−0.64 0	0.68 0.70	0.11 0.23	0.60 0.77	−0.19 0.42
Aqueous	ΔH Ea	−0.65 0	0.51 0.74	−0.03 0.58	0.31 0.93	−0.18 0.66

The reaction energy (ΔH) and kinetic barriers (E_a) for activating first and second CO₂ molecule, CO−CH^∗ coupling, and protonation of CO.HCOH^∗ to CO.H₂COH^∗ species or CO−CH^∗ with H₂O molecule in vacuum and in the aqueous condition, respectively, given in the unit of eV.

The unique geometry and favorable adsorption properties of the Fe₂ dimers immobilized on carbon substrates bring about inimitable advantages for their catalytic behavior. First, CO, as an inevitable and even dominant product of CO₂ reduction on many metal catalysts, severely limits the formation of higher-energy-density products (Zhu et al., 2014, Sarfraz et al., 2016, Peng et al., 2018), but it would be largely suppressed on the anchored Fe₂ dimers. Second, the adjacent dual metal centers and their strong binding with CO pave an efficient pathway for C−C coupling reaction; in contrast, C−C coupling only occurs on metal surfaces with homogenously distributed reaction sites when the coverage of CO is sufficiently high (Morales-Guio et al., 2018, Huang et al., 2017). Third, the difficult desorption of C₂H₄ from the Fe₂ dimers may result in superior selectivity for C₂H₅OH, which is a clean liquid fuel with high heating value. For most of the Cu based catalysts, however, the yield of C₂H₅OH is quite low compared with C₂H₄ (Liang et al., 2018).

At last, we assess the activity of these supported Fe₂ dimers for HER, which is a competing reaction against CO₂ reduction and highly affects the efficiency of CO₂ conversion (Zhu et al., 2016, Cui et al., 2017). Figure 6B plots the competition between adsorption of H^∗ species and CO₂ molecule on the Fe₂ dimers. The H^∗ adsorption energy ranges from −1.52 eV to −0.28 eV. For Fe₂@5N-V₃, Fe₂@6N-V₆, Fe₂@C₂N, and Fe₂@C₃N₄, the adsorption strength of H^∗ species is notably weaker than that of CO₂ molecule by 0.09–0.67 eV, implying that CO₂ reduction would prevail over HER on these catalysts with either high activity or superior selectivity. For Fe₂@4N-V₂, Fe₂@6N-V₄(a), and Fe₂@6N-V₄(b), the H^∗ adsorption strength is stronger than that of CO₂, which may suppress the CO₂ reduction. Combining the information in Figures 6A and 6B, we conclude that four of our considered systems are eligible for catalysis of CO₂ reduction with high activity for C−C coupling toward C₂ products. Among them, Fe₂@C₂N has remarkable selectivity for ethanol; Fe₂@5N-V₃ favors the formation of both ethanol and methane; Fe₂ on g-C₃N₄ and 6N-V₆ have lower selectivity and may generate both C₁ and C₂ products. Therefore, dimeric transition metal clusters immobilized on the nitrogenated holey carbon substrates form a category of efficient electrocatalysts for reduction of CO₂ to high-value hydrocarbons and alcohols, with desired selectivity achievable by choosing proper substrate.

Conclusion

In summary, we exploited dispersed 3d transition metal dimers for CO₂ reduction to selectively produce liquid fuels. Comprehensive first-principles calculations show that nitrogenated holey carbon materials not only serve as templates to stabilize small metal clusters but also dictate their electronic structures. Specifically, controlling the metal-substrate coupling strength allows effective modulation of both activity and product selectivity. As a consequence, the spatially confined dual reaction centers within the carbon matrix exhibit the following advantageous catalytic behavior: (1) simultaneous fixation of two CO₂ molecules, (2) prohibition of CO desorption, and (3) exclusive pathway for C−C coupling with high activity. The selectivity is tunable by choosing proper substrate materials. In particular, a Fe₂ dimer embedded in the C₂N monolayer exhibits remarkable selectivity for C₂H₅OH against the other C₁ and C₂ products as well as HER. These theoretical findings provide vital knowledge of the design rules of subnano metal clusters for converting greenhouse gas to high-energy fuels and high-value chemicals and meanwhile call for more experimental and theoretical efforts to advance the technologies for precise synthesis of atomically dispersed catalysts with well-controlled composition and structures.

Limitations of the Study

This study systematically exploited 3d transition metal dimers anchored on nitrogenated holey carbon monolayers for selective reduction of CO₂ to liquid fuels and screened suitable metal elements and carbon templates with high selectivity for ethanol. However, experimental realization of such superior subnano catalysts relies on the preparation of metal clusters with specific size supported on some given substrates, which may be challenging and requires the development of advanced synthesis methods.

Methods

All methods can be found in the accompanying Transparent Methods supplemental file.

Published: May 22, 2020