How is CO₂ hydrogenated to ethanol on metal–organic framework HKUST-1? Microscopic insights from density-functional theory calculations

Abstract

Thermocatalytic hydrogenation of CO₂ to multi-carbon chemicals (C₂₊) has received considerable interest to reduce CO₂ footprint and mitigate global warming. Comprising Cu paddle-wheel clusters, a metal–organic framework (MOF) namely HKUST-1 has been experimentally reported as a promising catalyst for CO₂ hydrogenation to ethanol under ambient conditions with the assistance of non-thermal plasma (NTP). Yet, there lacks microscopic understanding of the active center, reaction pathway and product selectivity. In this study, we conduct density-functional theory calculations to quantitatively and explicitly elucidate the fundamental mechanism involved. NTP is revealed to be responsible for H₂ dissociation, while the defective HKUST-1 with exposed Cu atoms accounts for highly selective CO₂ hydrogenation to ethanol via facile *CHOH–CO coupling, with *CHOH adsorbed on the Cu atoms and CO from the gas phase. The strong binding between the carbonyl C atoms in C₂ intermediates and Cu atoms, and the high stability of *CH₃CHOH intermediate, contribute to the higher selectivity of ethanol over acetaldehyde and ethylene, respectively. From bottom-up, this computational study provides deep microscopic insights into the catalytic mechanism of CO₂ hydrogenation to C₂ products on HKUST-1, and it will facilitate the design of new MOFs for efficient CO₂ conversion and other important chemical transformations.

---

1. Introduction

Mitigation of CO₂ emissions has become a gigantic challenge in the 21st century due to the increasing combustion of fossil fuels driven by modern industry and transportation.^1–3 Direct CO₂ hydrogenation to high value-added products is regarded as an efficient solution, as it reduces the carbon footprint and meanwhile produces renewable fuels and chemicals.^4–7 Compared with mono-carbon (C₁) chemicals such as carbon monoxide, methane and methanol, multi-carbon (C₂₊) chemicals possess a higher economic value and energy density.^8–10 Specifically, higher alcohols (i.e., alcohols containing two or more carbon atoms) have a greater variety of applications as liquid fuels, fuel additives and reaction intermediates to synthesize commodity and specialty products in the chemical, pharmaceutical and energy sectors.^8,11–13 Among higher alcohols, ethanol, as an important solvent, an industrial building block and a promising renewable fuel, has attracted great interest.

Direct production of ethanol from CO₂ and H₂ is challenging, as CO₂ is chemically inert with C=O bonding energy of 806 kJ mol⁻¹ (ref. 14) and it is also thermodynamically very stable with formation enthalpy of −394 kJ mol⁻¹.¹⁵ In addition, CO₂ hydrogenation requires the availability of H atoms from H₂ dissociation.⁹ Another obstacle for C₂₊ production from CO₂ is that it involves uncontrollable C–C bond formation with a high energy barrier. Furthermore, many bifurcation routes are involved, which are much more complicated than C₁ production. As a consequence, CO₂ hydrogenation to ethanol is mostly conducted at a high temperature and always ends up with a low selectivity.^16–18

Experimentally, non-thermal plasma (NTP) is an effective tool in thermocatalysis, as it facilitates activating source gases and overcoming both kinetic and thermodynamic limitations of chemical reactions to produce desired products.^19,20 Especially, NTP has been utilized for CO₂ conversion due to the intrinsic obstacles mentioned above. With a mean electron energy between 1 and 10 eV,^21,22 the high-energy electrons generated by NTP can excite CO₂ and H₂ molecules to activate or even break their chemical bonds, because only 5.5 eV is required to break a C=O bond²² and the H–H bond energy is even lower than the C=O bond energy.²³ Therefore, NTP contributes to both CO₂ activation and H₂ dissociation, thus promoting thermocatalytic CO₂ hydrogenation.

Metal–organic frameworks (MOFs), assembled from metal clusters and bridging organic linkers, have served as efficient catalysts for many reactions including CO₂ hydrogenation due to their porous structures, large surface areas and tunable compositions and functionalities.^21,24–32 However, most of the current studies for CO₂ hydrogenation using MOFs have been focused on C₁ products, and successful CO₂ hydrogenation to ethanol requires complicated structural modification to construct active sites.^30–32 Recently, Zou et al. reported successful NTP-assisted ethanol production by direct CO₂ hydrogenation with high selectivity on a Cu-MOF under ambient conditions.²¹ This Cu-MOF was synthesized by partially reducing HKUST-1 (Cu₃(BTC)₂, BTC = 1,3,5-benzene tricarboxylate) with Cu^II paddle-wheel clusters to generate Cu^I atoms. However, the mechanism behind improvement of CO₂ conversion, C–C coupling and origin of high ethanol selectivity is still unclear. Since Cu paddle-wheel clusters are very common in MOFs and can be conveniently modified for property manipulation,^33–36 deep understanding of their catalytic performance is of great significance for the broader application of MOFs on CO₂ hydrogenation and other important reactions.

In this study, we conduct in-depth theoretical investigation into the fundamental mechanism of CO₂ hydrogenation to ethanol on HKUST-1 using density functional theory (DFT) calculations. Following the computational methods briefly described below, we first present microscopic insight into the improvement of CO₂ conversion by untreated HKUST-1 with CO as the main product; then different intermediates from CO along C₁ pathways are examined; next, possible C–C coupling pathways are explored on the defective HKUST-1; finally, the origin of the higher selectivity of ethanol over other C₂ products (e.g., acetaldehyde and ethylene) is discussed. The findings from this work provide quantitative and explicit elucidation on the important mechanism that was elusive in the experiment, thus being conducive to the development of new MOFs for efficient CO₂ conversion.

2. Computational methods

Experimentally, divalent Cu atoms in HKUST-1 were partially reduced to monovalent by thermal treatment at 200 °C in the presence of methanol,²¹ which is a typical defect engineering approach to remove organic linkers in MOFs.^36,37 This thermal treatment could remove one BTC linker and the exposed Cu atom might coordinate with a H atom or methoxy group due to the existence of methanol. On this basis, we built four clusters to mimic the plausible catalysts, as shown in Fig. 1, including the original paddle-wheel (o_PW), defective paddle-wheel (d_PW), hydrogen-coordinated paddle-wheel (H_PW) and methoxy-coordinated paddle-wheel (CH₃O_PW). The natural population analysis (NPA) charges of two Cu atoms are approximately identical: 0.924/0.925 on o_PW, and a decrease to 0.858/0.878 on d_PW upon removing one BDC linker. After being coordinated with one H atom, both Cu atoms possess less NPA charge (0.788/0.800). On CH₃O_PW, one Cu atom is oxidized by methoxy group and its NPA charge increases to 0.935, while the other one is almost unchanged in NPA charge (0.837). Therefore, all three modified clusters (d_PW, H_PW and CH₃O_PW) could exist after thermal treatment²¹ and are thus considered here. To saturate the cleaved bonds, CN groups were added and fixed on all the clusters during DFT calculations.

Fig. 1 Original, defective, H-coordinated and methoxy-coordinated Cu paddle-wheel clusters (o_PW, d_PW, H_PW and OH_PW), where the NPA charges of Cu atoms are indicated.

DFT calculations were performed using Gaussian 16.^38,39 All the clusters were optimized at the B3LYP hybrid functional in conjunction with Grimme's D3BJ dispersion correction.⁴⁰ The main group atoms including C, N, O and H were described by using the def2-SVP basis set⁴¹ and the Stuttgart-Dresden (SDD) pseudopotential was used for Cu. Vibrational frequencies were calculated at the same level of theory as optimizations, in order to provide thermal corrections and verify the nature of stationary points (no imaginary frequency for an optimized intermediate and only one imaginary frequency for a transition state). To verify a transition state connecting directly to reactant and product, intrinsic reaction coordinate (IRC) calculations were conducted. The electronic energies and NPA charges were estimated through natural bond orbital (NBO) calculations using the M06L hybrid functional.⁴² The Gibbs energies (G) were estimated from

G = E^ele + G^cor (1)

where E^ele and G^cor represent the electronic energy and thermal correction energy. Temperature was 298.15 K, in accordance with the ambient conditions in experiments.²¹

3. Results and discussion

3.1. Improvement of CO₂ conversion

In experiments, CO₂ hydrogenation was feasible with the assistance of NTP in the absence of any catalyst and CO was found to be the main product with selectivity of 77.6%.²¹ After the untreated original HKUST-1 (i.e., o_PW cluster) was placed in an NTP discharge zone, CO₂ conversion improved greatly from 11.9% to 32.3%, indicating a great contribution of o_PW to the improvement of CO₂ conversion. Consequently, we consider the possibility of o_PW to facilitate H₂ dissociation or CO₂ activation. As evidenced by the positive adsorption Gibbs energy ΔG_ads in Fig. S1,† CO₂ cannot adsorb stably on the Cu atom of o_PW. This agrees well with the low CO₂ adsorption capacity experimentally reported on HKUST-1.^43,44 In addition, CO₂ activation is hindered on o_PW, indicated by the trivial NPA charge of CO₂ (0.053) on o_PW. For H₂ adsorption, ΔG_ads is even more positive, and furthermore the negligible NPA charge of H₂ (0.050) indicates no activation of the H–H bond. Apart from this, due to the lack of adjacent exposed Cu atom, H₂ dissociation is not feasible. On the other hand, direct CO₂ hydrogenation by H₂ to generate formic acid (HCOOH) on o_PW would overcome a high Gibbs energy barrier (ΔG_b) of 76.80 kcal mol⁻¹ (Fig. S2†), almost the same as 78.28 kcal mol⁻¹ in the gas phase (Fig. 2a and S3†). Therefore, the experimentally observed improvement of CO₂ conversion by o_PW is not accomplished through promoting H₂ dissociation or CO₂ activation. Given that H₂ can be readily dissociated into H atoms by NTP-generated high energy electrons, the initial CO₂ hydrogenation is significantly accelerated with ΔG_b of only 23.68 kcal mol⁻¹ for CO₂ + H → HCOO without a catalyst, as shown in Fig. 2a, which is much lower than that for direct CO₂ hydrogenation by H₂ gas. This hydrogenation is further boosted on o_PW with a much lower ΔG_b of only 8.36 kcal mol⁻¹. Therefore, although o_PW cannot promote direct CO₂ hydrogenation by H₂, it significantly decreases the activation barrier of CO₂ + H → HCOO, leading to the great improvement of CO₂ conversion as observed in experiments.

Fig. 2 (a) Gibbs energy profile of initial CO₂ hydrogenation under different conditions (H₂ & gas phase: direct hydrogenation by H₂ in the gas phase; H & gas phase: hydrogenation by atomic H in the gas phase; H & o_PW: hydrogenation by atomic H on the original HKUST-1); (b) adsorption Gibbs energies and (c) structures of *H, *CO and *OC on the four paddle-wheel clusters.

3.2. C₁ pathways from CO

The presence of original HKUST-1 greatly improved CO₂ conversion; however, CO was the main product like the situation in non-catalytic reaction and the production of ethanol was negligible.²¹ To elucidate the mechanism, we examine different C₁ intermediates (i.e., C₁ pathways) from CO on the four clusters. As illustrated in Fig. 2b and c, the adsorption sites for CO and H on o_PW are the same, and H adsorption is stronger than CO adsorption with a negative ΔG_ads of −9.66 kcal mol⁻¹. In addition, *CO hydrogenation to either *CHO or *COH is not favorable because *CO strongly repels H atom from the gas phase and no rational kinetic pathway exists (Fig. S4†). Therefore, CO hydrogenation on o_PW is difficult to proceed. Moreover, we also consider CO dimerization on o_PW. Although there are two open Cu atoms on o_PW, their open directions are opposite, making C–C coupling via the Langmuir–Hinshelwood mechanism⁴⁵ impossible. Therefore, the Eley–Rideal mechanism⁴⁶ is the only possibility for C–C coupling on o_PW, where only one CO molecule is adsorbed on the Cu atom and then coupled with another CO molecule from the gas phase. However, we find that the adsorbed *CO intermediate strongly repels the gaseous CO molecule, indicating that CO dimerization cannot be realized on o_PW. Therefore, CO adsorption, hydrogenation and dimerization are all impeded on o_PW, and the untreated original HKUST-1 cannot provide a catalytic effect on CO conversion. Consequently, CO is the main product on o_PW and its further reaction is very unfavorable.

Subsequently, we turn our attention to d_PW, H_PW and CH₃O_PW clusters. First, H and CO adsorption is examined. From Fig. 2c, it is seen that H adsorption on d_PW is extremely favorable, with ΔG_ads of −30.55 kcal mol⁻¹, much lower than CO adsorption with both *CO (Cu–C interaction, −16.27 kcal mol⁻¹) and *OC (Cu–O interaction, −7.65 kcal mol⁻¹) configurations. After H adsorption on d_PW, the cluster is identical to H_PW, on which the additional H adsorption is less favorable than CO adsorption with *CO configuration (ΔG_ads = −14.54 kcal mol⁻¹), while formation of the *OC configuration is endothermic. As a result, the exposed Cu atoms can bind with CO molecules, avoiding blockage by additional H atoms. Since d_PW is extremely easy to hydrogenate to H_PW in an H-rich environment, we take H_PW as the hydrogenated d_PW and thereafter combine the study of d_PW and H_PW. On CH₃O_PW, H adsorption is also stronger than CO adsorption with *CO configuration, but the difference is relatively smaller, resulting in competition of adsorption site. Since the unoccupied Cu atom is highly exposed, the co-adsorption of H and CO is plausible. After H adsorption on CH₃O_PW, ΔG_ads for additional CO adsorption on the Cu atom is −12.78 kcal mol⁻¹. However, the Cu–C distance is 2.54 Å (Fig. S5†), much longer than that in the *CO + *H co-adsorption structure on d_PW (2.05 Å), indicating that the existence of methoxy group reduces CO attraction to the Cu atom.

Next, we investigate CO hydrogenation on d_PW, where the isomers of several intermediates are considered from both thermodynamic and kinetic perspectives. As shown in Fig. 3, starting from *H + *CO, a surmountable ΔG_b of 22.87 kcal mol⁻¹ needs to be overcome for *H–*CO coupling to *CHO_1, where the C atom is coordinated to one Cu atom and the O atom faces outward. Since the C atom is closer to *H, formation of *COH is apparently more difficult than *CHO_1 from *H + *CO and thus excluded. We further examine the possibility of *CO hydrogenation by atomic H via the Eley–Rideal mechanism. However, this is endothermic with a Gibbs reaction energy (ΔG_r) of 18.33 kcal mol⁻¹, and *COH + *H is much less stable than *CHO + *H by 36.67 kcal mol⁻¹, as shown in Fig. S6.† Therefore, only *CHO is considered in the following discussion. After the formation of *CHO_1, the unoccupied Cu atom is not well coordinated and primed to adsorb another H atom, which is spontaneous with ΔG_ads of −19.47 kcal mol⁻¹. However, the adjacent *H and *CHO_1 strongly repel each other and thus cannot couple to form *CH₂O. Instead, *CHO_1 is hydrogenated to *CHOH_1 via the Eley–Rideal mechanism with ΔG_b of 13.91 kcal mol⁻¹. On the other hand, there exists a more stable isomer of *CHO, where the H atom faces outward, denoted as *CHO_2. The transformation from *CHO_1 to *CHO_2 requires a 180° rotation with a negligible energy barrier of only 3.14 kcal mol⁻¹. The further H adsorption is also spontaneous (−17.59 kcal mol⁻¹) and the subsequent *H–*CHO_2 coupling via the Langmuir–Hinshelwood mechanism only needs to overcome a very low ΔG_b of 4.12 kcal mol⁻¹ to generate *CHOH_2, which is thermodynamically more stable than *CHOH_1. Therefore, we identify *CHO_2 and *CHOH_2 as dominant configurations for *CHO and *CHOH, respectively. Theoretically, *CHOH_2 can be hydrogenated to *CH₂OH via the Eley–Rideal mechanism, which is the key intermediate for methanol production. However, the C atom in *CHOH_2 repels the H atom and no rational kinetic pathway is found to form either *CH₂OH or *CH by further dehydration. Therefore, further hydrogenation of *CHOH_2 is suppressed on d_PW, which accounts for the experimental observation that both CH₄ and CH₃OH as C₁ products exhibited extremely low selectivity.²¹Fig. 3 summarizes the evolution of Gibbs energy and intermediate structures along C₁ pathways on d_PW, and the detailed ΔG_r and ΔG_b of the elementary steps are listed in Table 1.

Fig. 3 Evolution of Gibbs energy and intermediate structures along C₁ pathways on d_PW.

Table 1 Reaction Gibbs energies (ΔG_r) and Gibbs energy barriers (ΔG_b) of the elementary steps along C₁ pathways on d_PW

	*H + *CO → *CHO_1	*CHO_1 + H(g) → *CHOH_1	*CHO_1 → *CHO_2	*CHO_2 + H(g) → *CHOH_2
ΔGr (kcal mol^−1)	1.12	−6.11	−0.57	−11.86
ΔGb (kcal mol^−1)	22.87	13.91	3.14	4.12

---

We also investigate the C₁ pathway on CH₃O_PW by exploring *CHO formation from *H + *CO co-adsorption structure. However, *CHO is found to be unstable, as the H atom tends to be attracted by the methoxy group rather than *CO. As a result, the methoxy group couples with *H to form CH₃OH, and in the end *CHO + CH₃O_PW evolves into CH₃OH + CO + d_PW, as illustrated in Fig. S7.† This is because CO adsorption on the Cu atom is oxidation via electron transfer from CO to Cu. Compared with CO in the gas phase, the NPA charge of the C atom increases by 0.136 and 0.082 in the *H + *CO co-adsorption structure on d_PW and CH₃O_PW, respectively, which results in stronger CO adsorption on d_PW than CH₃O_PW, as discussed above. For the C atom, *CO hydrogenation to *CHO is a reduction process and its higher oxidation state leads to stronger attraction to *H. Apparently, CH₃O_PW is unstable in CO hydrogenation because the methoxy group tends to detach from the Cu atom and finally both Cu atoms are exposed, just as d_PW. Therefore, after the analysis of all four clusters, we conclude that only the defective Cu-paddle wheel (d_PW) is stable and serves as an active catalyst for C–C coupling, as further considered below.

3.3. C–C coupling pathways

In their experimental study, Zou et al. assumed that C–C coupling might occur between adsorbed CH₂O and CO,²¹ where the two intermediates were bound to Cu atoms via Cu–O interaction parallelly along with *OCH₂–*OC coupling, as illustrated in Fig. 4a; however, there was no solid proof and further validation. Therefore, we evaluate the possibility of this coupling pathway first. The geometry of CH₂O is trigonally planar, where the C atom is sp² hybridized and less favorable to be attracted to the Cu atom than the carboxyl O atom. As shown in Fig. S8a,† CH₂O binds to Cu (i.e., *OCH₂) via Cu–O interaction rather than Cu–C interaction, which would cause a problem. That is, the Cu–C bond needs to be broken to realize conversion from *CHO to *OCH₂, which undergoes a drastic structural change with a huge energy obstacle for HCHO formation. The difficult formation of *OCH₂ also accounts for why no formaldehyde (HCHO) was found experimentally.²¹ On the other hand, Cu–O interaction is much weaker than Cu–C interaction as illustrated in Fig. 2b. Therefore, CO is less likely to bind with Cu than *OC to enable *OCH₂–*OC coupling. To validate our analysis, we investigated *OCH₂ + *OC co-adsorption structure and found that the CO molecule cannot adsorb on d_PW after optimization; instead, it stays in the gas phase and is located far away from *OCH₂ with C–C distance over 5.53 Å (Fig. 4c). This result is quite different from the assumption by Zou et al.²¹ On this basis, we infer that *OCH₂–*OC coupling is not likely to occur and thus exclude it.

Fig. 4 Schematics of (a) *OCH₂–*OC and (b) *CO–*CO coupling pathways; co-adsorption structures of (c) *OCH₂ + *OC and (d) *CO + *CO on d_PW.

As one of the most common C–C coupling pathways on Cu-based catalysts, CO dimerization is then investigated. On d_PW, it is expected that two CO molecules co-adsorb on the two exposed Cu atoms separately and stand side by side with an appropriate C–C distance for CO dimerization, as shown in Fig. 4b. However, the C–C distance in the optimized co-adsorption structure is 6.13 Å (Fig. 4d), which is too long for CO dimerization, and the adsorption of the second CO molecule is relatively weak with ΔG_ads of only −2.00 kcal mol⁻¹. On the other hand, as the adsorption of H atom is inevitable with highly negative adsorption energy (Fig. 2b), we further examine CO dimerization on d_PW with the existence of *H. As shown in Fig. S8b,† after geometry optimization, two *CO stay in an ‘end to end’ position and even farther apart. Consequently, *CO–*CO coupling is not practically feasible on d_PW and also excluded.

As discussed in Section 3.2, *CHO and *CHOH intermediates can be generated easily on d_PW, therefore, *CHO–*CO and *CHOH–*CO coupling might be possible pathways. Since there are two exposed Cu atoms on d_PW, we first consider the Langmuir–Hinshelwood mechanism, where one Cu atom is bound with *CHO or *CHOH and the other Cu atom with *CO before coupling. However, the situation for *CHO + *CO or *CHOH + *CO co-adsorption is similar to that for *CO + *CO, where *CHO or *CHOH and *CO are bound to different Cu atoms and too far apart with C–C distance over 5 Å, as shown in Fig. 5a and 6a. Therefore, C–C coupling via the Langmuir–Hinshelwood mechanism is unlikely to proceed on d_PW. Next, we consider the Eley–Rideal mechanism, where *CHO or *CHOH intermediate is bound to one Cu atom while a CO molecule approaches them from the gas phase. As shown in Fig. 5e, *CHO–CO coupling is slightly endothermic under ambient conditions with ΔG_b of 18.17 kcal mol⁻¹. Fig. 5b–d show the structures of the initial state, transition state and final state along the *CHO–CO coupling pathway, respectively. It can be seen that during the coupling, the Cu atom attracts the CO molecule to form a new Cu–C bond, while the Cu–C bond between Cu and *CHO breaks, followed by *CHO moving towards *CO to form a C–C bond. However, compared with *CHO–CO coupling, the exothermic *CHO hydrogenation (*CHO + *H → *CHOH) is more favorable with a lower ΔG_b of 13.91 kcal mol⁻¹ (Table 1). Therefore, *CHO–CO coupling is suppressed by hydrogenation.

Fig. 5 Structures of (a) *CHO + *CO co-adsorption, and (b) initial state, (c) transition state and (d) final state along the *CHO–CO coupling pathway. (e) Total energy profile along the *CHO–CO coupling pathway.

Fig. 6 Structures of (a) *CHOH + *CO co-adsorption, and (b) initial state, (c) transition state and (d) final state along the *CHOH–CO coupling pathway. (e) Total energy profile along the *CHOH–CO coupling pathway.

From the discussion in Section 3.2, *CHOH hydrogenation is unfavorable due to the repelling of atomic H in the gas phase, which is conducive to *CHOH–CO coupling. Similar to *CHO–CO coupling, *CHOH–CO coupling leads to the formation of a new Cu–C bond between Cu and CO, while the Cu–C bond between Cu and *CHOH breaks (Fig. 6b–d). As shown in Fig. 6e, *CHOH–CO coupling is exothermic with ΔG_b of 11.18 kcal mol⁻¹, which is lower than that in *CHO–CO coupling. Therefore, *CHOH–CO coupling is more favorable than *CHO–CO coupling both thermodynamically and kinetically, and thus we consider *CHOH–CO coupling as the dominant C–C coupling mechanism. *CHOHCO forms as a result of *CHOH–CO coupling. Interestingly, *CHOCO can be easily hydrogenated to *CHOHCO, where *H is formed spontaneously upon H adsorption on the exposed Cu atom, with ΔG_b of only 7.19 kcal mol⁻¹. Therefore, both *CHO–CO and *CHOH–CO coupling pathways lead to the formation of *CHOHCO.

3.4. Selectivity of C₂ products

Theoretically, hydrogenation along different pathways may occur after the formation of *CHOHCO, ending up with different C₂ products including ethanol (CH₃CH₂OH). Especially, acetaldehyde (CH₃CHO) and ethylene (CH₂CH₂) are two main byproducts in ethanol production from CO/CO₂ hydrogenation.^47–50 However, only CH₃CH₂OH was detected in the experiment by Zou et al.²¹ Therefore, it is instructive to investigate why the byproducts were suppressed. First, we calculated the desorption Gibbs energies (ΔG_des) and the numbers of electrons (|N_e|) transferred between adsorbates and d_PW for CH₃CHO, CH₃CH₂OH and CH₂CH₂. As shown in Fig. 7a, all three C₂ products are weakly adsorbed on d_PW, with ΔG_des lower than 15.59 kcal mol⁻¹ and the electron transfer is negligible with |N_e| smaller than 0.13, indicating that they are easy to desorb from d_PW. Therefore, the suppression of CH₃CHO and CH₂CH₂ occurs before the formation of corresponding intermediates. There are two possible pathways for CH₃CHO formation: *CH₂CHO + H → *CH₃CHO and *CH₃CO + H → *CH₃CHO. We find *CH₃CO is much more stable than *CH₂CHO by 24.56 kcal mol⁻¹, indicating the higher possibility of the latter pathway. However, *CH₃CO hydrogenation tends to form intermediate *CH₃COH, which only needs to overcome a low ΔG_b of 11.30 kcal mol⁻¹ (see Fig. 7b), while the carbonyl C atom repels H atom (i.e., the carbonyl C atom is difficult to be hydrogenated). Such repulsion was also observed when we performed transition state calculations from *CHOHCO to *CHOHCHO, *CH₂OHCO to *CH₂OHCHO, and *CH₂CO to *CH₂CHO. Accordingly, we conclude that the carbonyl C atom is stably bound to the Cu atom and more difficult to hydrogenate than the carbonyl O atom, resulting in the suppression of CH₃CHO formation. For competition between CH₃CH₂OH and CH₂CH₂, the selectivity is determined by *CH₂CH₂OH, which is the precursor of both CH₃CH₂OH and CH₂CH₂; while its isomer *CH₃CHOH cannot convert to CH₂CH₂. Thermodynamically, *CH₂CH₂OH is less stable than *CH₃CHOH by 18.85 kcal mol⁻¹, and the conversion from *CH₃CHOH to *CH₂CH₂OH needs to overcome a high ΔG_b of 36.18 kcal mol⁻¹ (see Fig. 7c). Therefore, the stabilization of *CH₃CHOH over *CH₂CH₂OH results in a higher selectivity of CH₃CH₂OH than CH₂CH₂.

Fig. 7 (a) Desorption Gibbs energies (ΔG_des) and the numbers of electrons transferred (|N_e|) between adsorbates (CH₃CHO, CH₃CH₂OH and CH₂CH₂) and d_PW; (b) Gibbs energy profile of *CH₂CHO, and of the initial state (IS), transition state (TS) and final state (FS) in *CH₃CO + H → *CH₃COH; (c) Gibbs energy profile of the IS, TS and FS in *CH₃CHOH → *CH₂CH₂OH.

3.5. Discussion

In experiments, CH₄ and CH₃OH could be produced from CO₂ hydrogenation with the assistance of NTP.²¹ Thus, we also consider the possible coupling between CO and *CH₂/*CH₃/*CH₂OH on d_PW. The ΔG_b values in *CH₂–CO, *CH₃–CO and *CH₂OH–CO coupling are 14.77, 15.66 and 15.76 kcal mol⁻¹, respectively (Fig. S9†), comparable to that in *CHOH–CO coupling. These reactions are in the premise that CH₂, CH₃ and CH₂OH groups can be formed in the gas phase; however, their formation is suppressed on d_PW, as discussed in Section 3.2, and they need being adsorbed on d_PW in order to couple with CO. These processes involve much more uncertainty than the facile adsorption of CO/H and the formation of *CHOH on d_PW. In addition, as the exposed Cu atoms on d_PW are largely occupied by *CO and *H with highly negative ΔG_ads, the adsorption of CH₂/CH₃/CH₂OH groups is suppressed. Therefore, we infer that *CHOH–CO is the dominant C–C coupling on d_PW.

In this work, we provide a guideline and fundamental understanding in CO₂ hydrogenation to ethanol on Cu paddle-wheel clusters, with the primary focus on the C–C coupling mechanism. However, the practical experimental conditions are complicated and the structural complexity of HKUST-1 is difficult to be fully captured by small clusters. Large clusters incorporating structural complexity are highly desired to gain better understanding of reaction mechanisms. During reaction, *CHOHCO hydrogenation to ethanol, especially the removal of OH group, is possible and further investigation is needed. Nevertheless, such investigation involves over 20 different intermediates and requires considering a quite large number of elementary steps, with prohibitively high computational cost. In this context, experimental characterization is instrumental to identify key C₂ intermediates, effectively narrow down calculation tasks and greatly facilitate catalyst design.

It is interesting to note that while HKUST-1 is a commonly examined MOF, a portion of MOFs containing Cu paddle-wheel (Cu₂(COO)₄) clusters are experimentally available. These Cu-MOFs are expected to have comparable catalytic activity to HKUST-1 for CO₂ hydrogenation. In addition, there exist Cu-MOFs with different configurations surrounding Cu₂(COO)₄ clusters. For example, two pyridine groups in Cu-HMOF⁵¹ are bound to a Cu₂(COO)₄ cluster, and three neighboring Cu₂(COO)₄ clusters in Cu₂PDC₂-MOF are connected by one 3,5-pyridyl group.⁵² The chemical coordination environment of Cu₂(COO)₄ clusters is crucial to the formation, adsorption and stabilization of intermediates and transition states, thus these Cu-MOFs might exhibit unique catalytic activity towards CO₂ hydrogenation. In the literature, certain other Cu-MOFs are also catalytically active for CO₂ hydrogenation. Bimetallic Cu^I₂ centres of Zr₁₂-bpdc-Cu MOF³⁰ were found to catalyse direct C–C coupling between CHO and methanol. Furthermore, inspired by the successful CO₂ hydrogenation to ethanol in Cu^II(H_xPO₄)_y@Ru-UiO MOF,³¹ where the valence state of Cu was modulated by Ru, introducing a different metal to Cu-MOFs³⁷ may serve as a new strategy for further development of MOFs for CO₂ hydrogenation.

4. Conclusions

Through DFT calculations, we have unraveled the microscopic mechanism for ethanol production by CO₂ hydrogenation on HKUST-1. H₂ dissociation is accomplished with the assistance of NTP, which generates high-energy electrons to break H–H bond. Although untreated HKUST-1 cannot catalyze H₂ dissociation, it greatly facilitates CO₂ hydrogenation to *HCOO with the existence of NTP-induced H atoms. With exposed Cu atoms, defective HKUST-1 is catalytically active for CO hydrogenation and C–C coupling. *CHOH–CO coupling via the Eley–Rideal mechanism is identified as the dominant C–C coupling pathway, leading to ethanol production on the defective HKUST-1. The production of acetaldehyde and ethylene is suppressed due to difficult hydrogenation of carbonyl C atoms in C₂ intermediates and the higher stability of *CH₃CHOH over *CH₂CH₂OH, respectively. Through comprehensive theoretical investigation, the fundamental mechanism of CO₂ hydrogenation to ethanol on HKUST-1 is well elucidated. We hope the bottom-up insights would be useful for the design of new MOF-based catalysts for CO₂ conversion into multi-carbon chemicals.

Data availability

The data supporting this article have been included as part of the ESI.†

Conflicts of interest

The authors declare no conflict of interest.

Acknowledgements

We gratefully acknowledge the A*STAR LCER-FI project (LCERFI01-0033 U2102d2006), the Ministry of Education of Singapore and the National University of Singapore (C-261-000-207-532/C-261-000-777-532 and R-279-000-574-114) for financial support.

References

1. S. Saeidi, S. Najari, V. Hessel, K. Wilson, F. J. Keil, P. Concepción, S. L. Suib and A. E. Rodrigues, Prog. Energy Combust. Sci., 2021, 85, 100905.
2. S. Shafiee and E. Topal, Energy Policy, 2009, 37, 181–189.
3. V. G. Dovì, F. Friedler, D. Huisingh and J. J. Klemeš, J. Clean. Prod., 2009, 17, 889–895.
4. R.-P. Ye, J. Ding, W. Gong, M. D. Argyle, Q. Zhong, Y. Wang, C. K. Russell, Z. Xu, A. G. Russell, Q. Li, M. Fan and Y.-G. Yao, Nat. Commun., 2019, 10, 5698.
5. P. Gao, S. Li, X. Bu, S. Dang, Z. Liu, H. Wang, L. Zhong, M. Qiu, C. Yang, J. Cai, W. Wei and Y. Sun, Nat. Chem., 2017, 9, 1019–1024.
6. S. Saeidi, S. Najari, F. Fazlollahi, M. K. Nikoo, F. Sefidkon, J. J. Klemeš and L. L. Baxter, Renew. Sustain. Energy Rev., 2017, 80, 1292–1311.
7. X. Xiaoding and J. A. Moulijn, Energy Fuels, 1996, 10, 305–325.
8. B. B. Asare Bediako, Q. Qian and B. Han, Acc. Chem. Res., 2021, 54, 2467–2476.
9. Y. Sheng, M. V. Polynski, M. K. Eswaran, B. Zhang, A. M. H. Lim, L. Zhang, J. Jiang, W. Liu and S. M. Kozlov, Appl. Catal., B, 2024, 343, 123550.
10. B. Zhang and J. Jiang, Energy Environ. Mater., 2024, 7, e12738.
11. R. G. Herman, Catal. Today, 2000, 55, 233–245.
12. H. T. Luk, C. Mondelli, D. C. Ferré, J. A. Stewart and J. Pérez-Ramírez, Chem. Soc. Rev., 2017, 46, 1358–1426.
13. V. R. Surisetty, A. K. Dalai and J. Kozinski, Appl. Catal., A, 2011, 404, 1–11.
14. S. Zhang, Z. Xia, Y. Zou, F. Cao, Y. Liu, Y. Ma and Y. Qu, J. Am. Chem. Soc., 2019, 141, 11353–11357.
15. P. J. Linstrom and E. W. G. Mallard, NIST Chemistry WebBook, NIST Standard Reference Database, National Institute of Standards and Technology, Gaithersburg MD, 20899, 2024.
16. S. Zhang, X. Liu, Z. Shao, H. Wang and Y. Sun, J. Catal., 2020, 382, 86–96.
17. S. S. Ali, S. S. Ali and N. Tabassum, J. Environ. Chem. Eng., 2022, 10, 106962.
18. L. Wang, L. Wang, J. Zhang, X. Liu, H. Wang, W. Zhang, Q. Yang, J. Ma, X. Dong, S. J. Yoo, J.-G. Kim, X. Meng and F.-S. Xiao, Angew. Chem., Int. Ed., 2018, 57, 6104–6108.
19. E. C. Neyts, K. Ostrikov, M. K. Sunkara and A. Bogaerts, Chem. Rev., 2015, 115, 13408–13446.
20. A. Bogaerts, X. Tu, J. C. Whitehead, G. Centi, L. Lefferts, O. Guaitella, F. Azzolina-Jury, H.-H. Kim, A. B. Murphy, W. F. Schneider, T. Nozaki, J. C. Hicks, A. Rousseau, F. Thevenet, A. Khacef and M. Carreon, J. Phys. D: Appl. Phys., 2020, 53, 443001.
21. N. Zou, J. Chen, T. Qiu and Y. Zheng, J. Mater. Chem. A, 2023, 11, 10766–10775.
22. B. Ashford and X. Tu, Curr. Opin. Green Sustainable Chem., 2017, 3, 45–49.
23. D. R. Lide, CRC Handbook of Chemistry and Physics, CRC press, 2004.
24. S.-L. Hou, J. Dong, X.-Y. Zhao, X.-S. Li, F.-Y. Ren, J. Zhao and B. Zhao, Angew. Chem., Int. Ed., 2023, 62, e202305213.
25. S. Shao, C. Cui, Z. Tang and G. Li, Nano Res., 2022, 15, 10110–10133.
26. R. Krishnan, K. Yang, K. Hashem and J. Jiang, Mol. Catal., 2022, 527, 112407.
27. K. Yang and J. Jiang, J. Mater. Chem. A, 2020, 8, 22802–22815.
28. B. Rungtaweevoranit, J. Baek, J. R. Araujo, B. S. Archanjo, K. M. Choi, O. M. Yaghi and G. A. Somorjai, Nano Lett., 2016, 16, 7645–7649.
29. Z. Li, T. M. Rayder, L. Luo, J. A. Byers and C.-K. Tsung, J. Am. Chem. Soc., 2018, 140, 8082–8085.
30. B. An, Z. Li, Y. Song, J. Zhang, L. Zeng, C. Wang and W. Lin, Nat. Catal., 2019, 2, 709–717.
31. L. Zeng, Z. Wang, Y. Wang, J. Wang, Y. Guo, H. Hu, X. He, C. Wang and W. Lin, J. Am. Chem. Soc., 2020, 142, 75–79.
32. S.-C. Qi, Z.-H. Yang, R.-R. Zhu, X.-J. Lu, D.-M. Xue, X.-Q. Liu and L.-B. Sun, J. Mater. Chem. A, 2021, 9, 24510–24516.
33. S. Bureekaew, S. Amirjalayer and R. Schmid, J. Mater. Chem., 2012, 22, 10249–10254.
34. V. Paredes-García, R. C. Santana, R. Madrid, A. Vega, E. Spodine and D. Venegas-Yazigi, Inorg. Chem., 2013, 52, 8369–8377.
35. S. J. Jenniefer and P. T. Muthiah, Chem. Cent. J., 2013, 7, 35.
36. K. Hashem, R. Krishnan, K. Yang, B. A. Anjali, Y. Zhang and J. Jiang, Phys. Chem. Chem. Phys., 2024, 26, 7109–7123.
37. I. Agirrezabal-Telleria, I. Luz, M. A. Ortuño, M. Oregui-Bengoechea, I. Gandarias, N. López, M. A. Lail and M. Soukri, Nat. Commun., 2019, 10, 2076.
38. R. Dennington, T. A. Keith and J. M. Millam, GaussView 6.0, Semichem Inc., Shawnee Mission KS, 2016.
39. M. J. Frisch, G. W. Trucks, H. B. Schlegel, G. E. Scuseria, M. A. Robb, J. R. Cheeseman, G. Scalmani, V. Barone, G. A. Petersson, H. Nakatsuji, X. Li, M. Caricato, A. V. Marenich, J. Bloino, B. G. Janesko, R. Gomperts, B. Mennucci, H. P. Hratchian, J. V. Ortiz, A. F. Izmaylov, J. L. Sonnenberg, D. Williams-Young, F. Ding, F. Lipparini, F. Egidi, J. Goings, B. Peng, A. Petrone, T. Henderson, D. Ranasinghe, V. G. Zakrzewski, J. Gao, N. Rega, G. Zheng, W. Liang, M. Hada, M. Ehara, K. Toyota, R. Fukuda, J. Hasegawa, M. Ishida, T. Nakajima, Y. Honda, O. Kitao, H. Nakai, T. Vreven, K. Throssell, J. A. Montgomery Jr, J. E. Peralta, F. Ogliaro, M. J. Bearpark, J. J. Heyd, E. N. Brothers, K. N. Kudin, V. N. Staroverov, T. A. Keith, R. Kobayashi, J. Normand, K. Raghavachari, A. P. Rendell, J. C. Burant, S. S. Iyengar, J. Tomasi, M. Cossi, J. M. Millam, M. Klene, C. Adamo, R. Cammi, J. W. Ochterski, R. L. Martin, K. Morokuma, O. Farkas, J. B. Foresman and D. J. Fox, Gaussian 16 Rev. A.03. Wallingford, CT, 2016.
40. S. Grimme, S. Ehrlich and L. Goerigk, J. Comput. Chem., 2011, 32, 1456–1465.
41. F. Weigend and R. Ahlrichs, Phys. Chem. Chem. Phys., 2005, 7, 3297–3305.
42. Y. Zhao and D. G. Truhlar, J. Chem. Phys., 2006, 125, 194101.
43. M. Mohamedali, A. Henni and H. Ibrahim, Microporous Mesoporous Mater., 2019, 284, 98–110.
44. C. Zhu, Z. Zhang, B. Wang, Y. Chen, H. Wang, X. Chen, H. Zhang, N. Sun, W. Wei and Y. Sun, Microporous Mesoporous Mater., 2016, 226, 476–481.
45. L. R. L. Ting, O. Piqué, S. Y. Lim, M. Tanhaei, F. Calle-Vallejo and B. S. Yeo, ACS Catal., 2020, 10, 4059–4069.
46. Y. Wang, B. Li, B. Xue, N. Libretto, Z. Xie, H. Shen, C. Wang, D. Raciti, N. Marinkovic, H. Zong, W. Xie, Z. Li, G. Zhou, J. Vitek, J. G. Chen, J. Miller, G. Wang and C. Wang, Sci. Adv., 2023, 9, eade3557.
47. Y. Wang, H. Luo, D. Liang and X. Bao, J. Catal., 2000, 196, 46–55.
48. C. Jia, J. Gao, Y. Dai, J. Zhang and Y. Yang, J. Energy Chem., 2016, 25, 1027–1037.
49. W.-F. Xiong, D.-H. Si, H.-F. Li, X. Song, T. Wang, Y.-B. Huang, T.-F. Liu, T. Zhang and R. Cao, J. Am. Chem. Soc., 2024, 146, 289–297.
50. Y. Ouyang, L. Shi, X. Bai, C. Ling, Q. Li and J. Wang, ACS Catal., 2023, 13, 15448–15456.
51. Z.-X. Xu, X. Wu, J. Liu, Y. Kang and J. Zhang, CrystEngComm, 2015, 17, 6107–6109.
52. T. Furuta, I. Kanoya, H. Sakai and M. Hosoe, Polymers, 2011, 3, 2133–2141.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4ta08052a

---