Introduction

The replacement of fossil carbon resources by renewable ones is a major challenge to develop sustainable chemistry. CO₂ is a non-toxic, cheap, abundant and sustainable C₁ resource, and its transformation into valuable chemicals and fuels has attracted wide attention.¹ Formamides are important intermediates and solvents in industry and their synthesis has emerged as an intensely studied topic in chemistry.² To replace highly toxic CO used for industrial production of formamides,³ various safer carbonyl source reagents such as formic acid or formate,⁴ formaldehyde,⁵ methanol,⁶ ethers,⁷ N,N-dimethylformamide⁸ and so on have been developed.^{2a, 9} N-formylation of amines with CO₂/H₂, using preferentially homogeneous noble metal catalysts is one of the main methods for synthesis of formamides.^{2c, 10} However, from the view of industrial production, homogeneous processes are less desired,¹⁰ due to inherent difficulty in product separation and catalyst recycling. Although some heterogeneous catalysts have been reported, they suffer from the use of noble metals such as Pd,¹¹ Ru¹² and Au,¹³ or alkaline additives,^{11b, 11d, 11e, 11h} high temperature (200 °C),^13b and/or high pressure (12 MPa).¹⁴ Moreover they are not active for converting aliphatic primary amine substrates.^{14, 15} To the best of our knowledge, active heterogeneous catalysts based on non-noble metals have not yet been reported for N-formylation of aliphatic primary amines with CO₂/H₂.

In the N-formylation of amines with CO₂/H₂, HCOOH has been reported to be the key intermediate in both HCOOH and HCOOMe pathways.^11f The further hydrogenation of formamides to the corresponding methylamines is the main side reaction.^{16, 17} Therefore, to obtain the desired formamide, it is vital to control the selective hydrogenation of CO₂ to HCOOH and prevent the over-reduction of formamides to methylamines. It has been reported that HCOO* is a key intermediate to synthesize HCOOH.¹⁸ Thus, it should be feasible to construct a bifunctional interface, which is both active for the selective hydrogenation of CO₂ to HCOOH and favorable to the desorption of the formamide, for the challenging transformation of aliphatic primary amines into the corresponding formamides with CO₂/H₂. Recently, it has been reported that heterostructure CuO_x/Cu exhibited superior performance in CO₂ hydrogenation to methanol compared to single CuO_x and metallic Cu sites.¹⁹

Against this background, here we for the first time achieved the challenging N-formylation of aliphatic primary amines with CO₂/H₂ by constructing an active Cu₂O/Cu phase that not only efficiently accelerates the formation of HCOOH intermediate, but also prevent the further reduction of the formamide products. When dispersing on an Al₂O₃ matrix, more active Cu₂O/Cu phase were obtained and the agglomeration of the active sites was efficiently prevented, enabling the recycling of the active catalyst several times without obvious activity loss. We found that the O_inter at the Cu₂O/Cu interface played the crucial factor via the creation of O_inter-H moieties during CO₂ hydrogenation, favoring the selective formation of HCOO* species.²⁰ Moreover, the electron-rich Cu⁰ center is favorable to the desorption of the N-formylated molecules, thereby inhibiting the over-reduction to N-methylated amines and promoting the selectivity.

Results and Discussion

The catalytic performance for N-formylation of octylamine on catalysts pre-reduced at different temperatures and times is shown in Figure 1. Only less than 10 % of formamide was obtained over fresh unsupported CuO_x without any reductive pretreatment (CuO_x-NR). This suggests that Cu^II, being the main component in this catalyst, is less active for this reaction. This is also true for sample CuO_x-150-1h, indicating that CuO_x can hardly be reduced at 150 °C during 1 h. Upon raising the reduction temperature from 210 to 450 °C, the formaldehyde yields show a volcano-type behavior with a maximum at 250 °C. This suggests, in agreement with characterization data presented below, that deep reduction of Cu^II to Cu⁰ occurring at higher temperature lowered the catalytic activity towards formamides. The same effect was caused by prolonging the reduction time, as evident from a comparison of samples CuO_x-250-1h and CuO_x-250-3h (Figure 1a). As shown by pseudo in situ X-ray photoelectron spectroscopy (XPS) data below, the latter sample contained mainly metallic Cu in addition to a small amount of Cu^I.

Furthermore, the recyclability of sample CuO_x-250-1h was investigated. While a formamide yield of 25 % was quickly reached within 3 h over the fresh CuO_x-250-1h catalyst, but only 53 % yield was obtained after 24 h. This deactivation may be due to agglomeration of the active Cu sites to larger metallic Cu particles during the reaction (Figure S1), probably promoted by reduction of CuO_x under reaction conditions.

To suppress agglomeration of the active species during reaction, we deposited the CuO_x phase on an alumina support, which is well known for stabilizing Cu sites²¹ and has been widely used in reductive amination of CO₂.^{11j, 15a, 22} The CuAlO_x catalyst with a nominal Cu content of 7.5 wt % was prepared by the same method as the CuO_x catalyst and reduced at different temperature between 250 to 650 °C for 3 h. Due to the high catalytic performance of CuAlO_x catalysts, their catalytic performance was tested at a lower reaction temperature than that of CuO_x catalysts (Figure 1b). No formamide was formed on the non-reduced catalyst (CuAlO_x-NR) in which Cu^II is dominating. Only a trace amount of formamide was observed on sample CuAlO_x-250-3h. This can be attributed to the strong stabilization effect of Al₂O₃ on copper oxide.²¹ Raising the reduction temperature from 300 to 650 °C revealed a tendency in product yields similar to that of the unsupported CuO_x catalyst and the maximum yield was obtained at 350 °C. A shorter reduction time led to a lower product yield, as seen from a comparison of CuAlO_x samples reduced for 3 h and 2 h at 350 °C (Figure 1b). A CO₂/H₂ pressure ratio of 1/3 turned out to be optimal at a total pressure of 4 MPa (Table S1, entries 1–3). Prolonging the reaction time did not significantly improve the product yield (Table S1, entries 1 and 4). An excellent yield of 99 % was obtained when increasing the reaction temperature to 150 °C (Figure 1b). Reducing the Cu loading from 7.5 wt % to 1.9 wt % led to a continuous decrease in product yields (Figure 1b and Table S1, entries 6 and 7). No formamide products were observed after removing CO₂ from the reaction system, meaning that methanol can't react with amines directly to form formamide product (Table S1, entry 7). In addition, the desired product was obtained using toluene and 1-octane as solvents, indicating CO₂ was the real carbonyl source (Table S1, entries 8 and 9).

With the most active catalyst CuAlO_x-350-3h, the substrate scope was further investigated (Table 1). Linear- and branched-chain aliphatic primary amines with C₄–C₁₆ number of carbon atoms could be converted with good to excellent formamide yields (81–96 %). 96 % and 82 % product yields were respectively obtained with cyclic aliphatic primary amines, cyclopentamine, and phenyl-substituted aliphatic primary amine, 3-phenylpropan-1-amine. Benzyl amine and its derivatives with a series of different groups such as -Me, -CH(Me)₂, -OMe, and -F can also provide the corresponding formamide products in 77–96 % yields. Moreover, the number and position of substituted groups have little effect on the activity and no obvious electronic and spatial resistance effect were observed. This system is also active for cyclohexylmethanamine and cyclopropylmethanamine and 75 % and 98 % product yields were obtained, respectively. Besides, secondary amines with different structures could be converted effectively. 84–99 % product yields have been obtained from cyclic secondary amines such as piperidine, 4-methylpiperidine, 3,5-dimethylpiperidine, morpholine and pyrrolidine as substrates. In comparison to those, dimethylamine and N-benzylethanamine showed slightly lower reactivity, providing moderate to good product yields.

Table 1. Reductive amination of CO₂ with different amines.^[a]

• [a] Yields were determined by GC-FID with biphenyl as the external standard. [b] Yields were determined by ¹H NMR with triphenylmethane as the internal standard.

Inspired by the good substrate scope of this catalytic system in the saturated amines, the selective N-formylation of amines containing unsaturated groups with CO₂ and H₂, which is challenging because the efficient catalysts for the N-formylation are usually very active for hydrogenation of the unsaturated groups, is explored. Gladly, the substrate piperidine-4-carboxamide was converted to the desired formamides in 92 % yields, while the unsaturated groups (C≡N bond) remained (Scheme 1). Moreover, this system is also chemoselective for the molecules containing both the hydroxyl and amino groups, and only N-formylated product was formed under the optimized reaction conditions. For example, 5-aminopentan-1-ol delivered the corresponding formamide product in 96 % yields (Scheme 1).

It was notable that this methodology can be also extended to the chiral amines, which are widely applied as the synthetic intermediates for the synthesis of the drug and bioactive molecules.²³ 74–97 % yields of chiral formamides were gained when chiral amines such as (R)-N,N-dimethylpyrrolidin-3-amine, (R)-2-(methoxymethyl)pyrrolidine, (2R,6S)-2,6-dimethylmorpholine, ethyl (S)-piperidine-3-carboxylate and (R)-hexan-2-amine were used as substrates (Scheme 1). To demonstrate the utility of this methodology, the active catalyst CuAlO_x-350-3h was used for the gram-scale synthesis of formamide. The N-formylation of morpholine was taken as an example and the corresponding morpholine-4-carbaldehyde (about 1.7 g) was obtained in 95 % yield (Scheme 1).

Finally, the recyclability of the active catalyst CuAlO_x-350-3h was investigated. Compared to the most active unsupported CuO_x catalyst, the introduction of aluminum oxide strongly raised the stability and this catalyst could be consecutively used for 6 times without obvious activity loss (Figure S2). Inductively coupled plasma atomic emission spectroscopy (ICP-AES) analysis of the CuAlO_x-350-3h catalyst in before and after use revealed that the Cu content slightly decreased from 6.70 wt % in the fresh catalyst to 6.56 wt % after the 6^th run, suggesting that the catalyst is stable. To demonstrate the heterogeneous nature of the active catalyst and exclude the contribution of leached Cu, the catalyst was removed by hot filtration after 6 h when 35 % product yield was measured. During further reaction of the clear solution for another 18 h after catalyst removal, no further product was formed, which clearly indicates that the solid Cu supported on alumina is the active species.

To reasonably assess our catalytic system, the catalytic performance of the most active CuAlO_x-350-3h catalyst in our system for N-formylation of primary and secondary amines was respectively compared with that of the reported heterogeneous catalysts in the literatures. Considering the wide use of C₆–C₈ aliphatic primary amines and morpholine as typical starting materials, their N-formylation were used as the model reaction to evaluate the catalytic performance of our catalyst and the reported catalysts (Table S3). Clearly, the catalytic performance of our catalyst is comparable to the reported noble metal catalysts and even much better than most of them, especially in the case of aliphatic primary amines.

To unravel a correlation between structure and activity, different ex and in situ characterization techniques have been applied such as in situ X-ray diffraction (XRD), XPS, high-angle annular dark field scanning transmission electron microscopy (HAADF-STEM) and electron energy loss spectroscopy (EELS). N₂-adsorption and desorption analysis showed that the calcined unsupported CuO_x catalyst had a very small specific surface area (14.6 m² g⁻¹) and pore volume (0.10 cm³ g⁻¹), which even decreased after reductive pretreatment (Table S2). In contrast, these values were much higher for calcined CuAlO_x and were retained widely after pre-reduction for 3 h at 350 °C (142.1 m² g⁻¹ and 0.4878 cm³ g⁻¹, Table S2).

In situ XRD powder patters of CuO_x catalysts reduced at different temperature (Figure 2a and b) showed only CuO diffraction peaks (JCPDS, Card No. 80-1916)²⁴ at 25 and 190 °C. New peaks from Cu₂O (111 and 200) (JCPDS, Card No. 65-3288)²⁵ appeared at 210 °C (Figure 2b). At higher reduction temperature (220–350 °C), the peak intensity of metallic Cu (JCPDS, Card No. 04-0836)²⁶ and Cu₂O gradually increased while that of CuO decreased. CuO was totally reduced to Cu and Cu₂O at 230 °C. Despite the fact that the Cu₂O peaks disappeared at 250 °C, this phase may still be present in amorphous or nanocrystalline form, as suggested by pseudo in situ XPS presented below. Thus, the co-existence of both Cu and Cu₂O in the CuO_x-250-1h catalyst is possible. The XRD powder patterns of the most active supported catalyst CuAlO_x-350-3h (Figure 2c) show only weak peaks of metallic Cu(111), Cu(200), Cu(220) and Cu₂O(111),²⁵ indicating the co-existence of both Cu and Cu₂O which did not change after four consecutive runs in the catalytic reaction for 24 h at 150 °C.

Considering the sensitivity of Cu^0/I species to air,²⁷ pseudo in situ XPS measurement were performed to reveal the real active surface of the CuO_x-250-1h catalyst (Figure 2d–f). The non-reduced CuO_x-NR catalyst shows characteristic CuO peaks with a broad satellite peak centered at around 944 eV and a typical Cu 2p_3/2 binding energy of 934.5 eV from Cu^II species (Figure 2d).²⁸ This was also in agreement with the corresponding Cu LMM Auger peak at 568.6 eV which showed the typical shape observed for CuO (Figure 2e).²⁹ After treatment with H₂ at 250 °C for 1 h, the satellite peak of Cu^II disappeared while the Cu LMM peak shifted to 568.0 eV and changed its shape becoming similar to that of Cu⁰,²⁹ indicating that Cu^II species in the CuO_x-NR catalyst were reduced, most probably mainly to Cu⁰. This was also in agreement with EELS results discussed below (Figure 3). However, since still an O 1s peak was observed after reductive treatment (Figure 2f), the presence of some residual oxidic copper cannot be excluded in the CuO_x-250-1h catalyst. Increasing the reduction time at 250 °C to 2 h enhanced the changes observed after 1 h even more.

Cu2p XPS spectra of CuAlO_x reduced at different temperatures were shown in Figure S3. The strong satellite peak of Cu^II between 940–947 eV was observed in the catalyst CuAlO_x-250-3h, suggesting the presence of Cu^II species. Fitting of the Cu 2p_3/2 peak revealed another contribution around 932.4 eV that could be assigned to Cu^I/0.^29a With increasing the reduction temperature, the typical satellite peak of Cu^II gradually weakened and disappeared after treatment at 650 °C for 3 h, suggesting that all Cu^II species were reduced to Cu^I/0. Furthermore, Cu LMM Auger spectra showed two peaks around 915 and 916.8 eV in the catalyst CuAlO_x-250-3h, which can be assigned to Cu^II[30] and Cu^I,³¹ respectively, i.e., there was no Cu^I/Cu⁰ interface sites, probably reasoning for its poor activity. A new peak appeared at around 918.0 eV in the case of CuAlO_x-300-3h, CuAlO_x-350-3h, CuAlO_x-450-3h and CuAlO_x-650-3h, which can be attributed to the formation of Cu⁰.³² By fitting the Auger spectra (Figure S3), it can be calculated that the Cu^I/Cu⁰ ratios were 3.24, 1.60, 0.88 and 0.62 for the catalysts CuAlO_x-300–3h, CuAlO_x-350-3h, CuAlO_x-450-3h and CuAlO_x-650-3h, respectively. Plotting the product yields with the Cu^I/Cu⁰ ratio (Figure S3), the optimum Cu^I/Cu⁰ ratio is about 1.60 for these samples, and both increasing and decreasing this ratio lowered the product yield.

HAADF-STEM images of CuO_x-250-1h showed rather large Cu⁰ particles (Figure 3a). EELS near edge structure confirmed the co-existence of Cu⁰, Cu^I and Cu^II (Figure 3b).³³ STEM-EELS maps (Figure 3c and d) showed that the Cu⁰ particles were surrounded by an oxide layer where both Cu^I and Cu^II present different types of morphology (Figure 3d). Cu⁰ was more compact with a rather smooth surface, while Cu^I and Cu^II seem more akin a cloudy type with a rough surface in close contact with the Cu⁰ particle. Considering that pseudo in situ XPS results after reduction showed no evidence for Cu^II, it was probable that Cu^II found in the HAADF-STEM pictures of the pre-reduced catalyst stemmed from oxidation of Cu^0/I by ambient air.

HAADF-STEM image of catalyst CuAlO_x-350-3h showed a few Cu crystallites with three different lattice parameters, all corresponding roughly to metallic Cu within the margin of error (Figure S4a), although other particles seemed to be of oxide character. STEM-EDS data showed that the Cu containing particles were highly dispersed on the surface of Al₂O₃ (Figure S4b and c). Furthermore, HR-TEM image revealed an existence of Cu₂O(111) with a crystal facet distance of 0.245 nm, which was adjacent to some Cu(111) crystal facets with a distance of 0.210 nm (Figure S4d).³⁴ In connection with XRD and XPS results (Figure 2c and S3), it can be concluded that the CuAlO_x-350-3h catalyst contained Cu₂O/Cu interface sites, just like CuO_x-250-1h.

Combination of recycling experimental results and the characterizations of ICP-AES, XRD and XPS of the used CuAlO_x-350-3h catalyst revealed that the active Cu₂O/Cu sites is stable during the catalytic reaction.

To reveal the effect of Al on the surface basicity of catalysts, the CO₂-TPD patterns of CuO_x and CuAlO_x were recorded in the range of 30–600 °C (Figure S5). For CuO_x, two broad peaks of CO₂ desorption were observed in the range of 50–250 °C and 350–550 °C, suggesting that CO₂ can be adsorbed on the surface of CuO_x catalyst in two different ways. Compared to CuO_x, CuAlO_x showed similar but much stronger CO₂ desorption peaks, indicating that the addition of Al greatly increased the adsorption of CO₂. That may be one of the reasons for the better catalytic activity of CuAlO_x than CuO_x.

To understand the catalytic activity of Cu₂O/Cu sites on the N-formylation of aliphatic amines with CO₂ and the effects of oxygen at the Cu₂O/Cu interface, DFT calculations were carried out. Based on the characterization results vide supra and the reported references,³⁵ DFT calculations were performed comparatively on a pure Cu(111) surface and a Cu(111) surface modified with a Cu₂O cluster (Cu₂O/Cu(111), Figure S6). Considering the CO₂ hydrogenation to formic acid acts as the rate-determining step, two pathways were calculated: 1) CO₂+2 H→HCOO+H→HCOOH and 2) CO₂+2 H→COOH+H→HCOOH. The optimal potential energy surfaces (PESs) are shown in Figure 4. On the pure Cu(111) surface (Figure 4a), starting from co-adsorbed CO₂ and hydrogen (Figure S7), either a HCOO or a COOH intermediate can be formed via hydrogenation at the carbon or oxygen atom. Since formation of COOH had a high barrier and strong endothermic (E_a=1.78 eV, E_r=0.55 eV³⁶) this pathway had been disregarded and formation of a HCOO intermediate was favored with a barrier and reaction energy of 1.03 and −0.17 eV, respectively (Reaction 1, Figure S7). Subsequently, HCOO can be further hydrogenated to HCOOH or H₂COO, whereby formation of HCOOH was more preferred thermodynamically (0.47 eV, Figure S8) and the barrier was 1.04 eV (Reaction 2, Figure S7). According to Figure 4a, the apparent barrier and reaction energy for transforming CO₂ to HCOOH was 1.03 and 0.30 eV on Cu(111).

Although CO₂ hydrogenation on a Cu₂O(111) surface was more difficult than on Cu(111)^35b and CO₂ was weakly adsorbed on the catalyst surface (Figure S9), the CO₂ hydrogenation was successfully proceeded due to the oxygen at Cu₂O/Cu(111) surface and preferably initiated by the adsorption of the CO₂+H₂ at Cu(111) (Figure 4b). After the H₂ activation at Cu(111) with the formation of two Cu−H species, one H specie can diffuse from the threefold coordinated site on the Cu(111) surface (0.00 eV) to the O_inter at the Cu₂O/Cu interface, forming an O_inter-H moiety (−0.40 eV, Figure S10 top left, Figure 4b). Similar with that on the pure Cu(111) surface (Figure S7 and S8), the HCOO intermediate was more favorably generated by reacting Cu−H with C site of CO₂ with the barrier and reaction energy of 1.06 eV and −0.18 eV (Reaction 3, Figure S10). Afterwards, the activated H on O_inter-H moiety was transferred to HCOO, producing the desired HCOOH at Cu₂O/Cu(111) (Figure S11–S12) with barrier and reaction energy of 0.52 and 0.83 eV (Reaction 4, Figure S11). Since the reaction energy was higher than the barrier, formic acid formation was regarded as a barrierless reaction at Cu₂O/Cu(111). Based on the DFT calculations vide supra, it was clearly revealed that the apparent barrier and reaction energy for CO₂ hydrogenation to HCOOH was much lower on the Cu₂O/Cu(111) interface (0.66 and 0.25 eV) than that on the pure Cu(111) surface (1.03 and 0.30 eV). Therefore, the presence of the O_inter not only reduced the apparent barrier of HCOOH formation but also changed the reaction of HCOO to HCOOH into a barrierless process. All these results strongly supported the idea of designing a heterostructured Cu₂O/Cu catalyst.

Furthermore, the origin of the high selectivity of N-formylation of the amine on Cu₂O/Cu interface was also investigated. The charge density difference (CDD) analysis was conducted to estimate the charge alteration of the Cu site during the formation of the O_inter-H species, and the results was displayed in Figure 5a–d and Supporting Information Figure S13, where the yellow and blue areas represented the higher and lower electron densities, respectively. Compared with the bare Cu₂O/Cu model, the electron was transferred to Cu during the protonation of Cu₂O/Cu with the increase of CDD. Therefore, the desorption properties of DMF were conducted on Cu₂O/Cu and protonated Cu₂O/Cu. Clearly, the distance between the oxygen of DMF and Cu site was elongated from 2.366 Å to 2.406 Å, indicating the easier desorption of DMF on protonated Cu₂O/Cu interface (Figure 5c and d). Moreover, the desorption energy (E_des) of DMF on the Cu₂O/Cu interfaces was 0.33 eV. However, the E_des of DMF over Cu₂O(H)/Cu interface was decreased to 0.27 eV, suggesting that Cu₂O/Cu interface with one OH served as more energetic sites for DMF adsorption (Figure 5e). This desorption effect modified by the protonated Cu₂O/Cu interface makes the adsorption of N-formylated product harder and eventually contributed to their accelerated surface desorption and enhanced selectivity.

Bader charge analysis also approved that the Cu₂O/Cu interfaces with OH site displayed the highest desorption energy. As shown in Figure 5f, the Bader charge analysis of bare Cu₂O/Cu and Cu₂OH/Cu with OH were 0.73 and 0.68, signifying the electron transfer from the O to Cu after the protonation of Cu₂O/Cu during the reaction. After the adsorption of DMF, the net charge on Cu₂O/Cu and Cu₂OH/Cu was 0.63 and 0.60 respectively, indicating less charge transfer to the DMF molecules occurred at Cu₂OH/Cu site. These results demonstrated that Cu site on Cu₂OH/Cu was more electron-sufficient compared with that on Cu₂O/Cu, favoring the desorption of this DMF molecules.³⁷

To investigate the reaction intermediate, the mixtures of the model reaction after reaction for 12 h were monitored by ¹H NMR and typical signals of HCOOMe were observed at 8.09 and 3.66 ppm (Figure S14). As the HCOOH had been confirmed to be the intermediate during the CO₂ hydrogenation, both HCOOH and HCOOMe pathways were possible for the production of the N-formylated amines.^11f Moreover, the reaction of HCOOH and CH₃OH can quickly finished within 0.5 h (Scheme 2), indicating the formed HCOOH could be fast consumed and converted into HCOOMe quantitatively. In addition, the yield of N-formylated amine was quantitively obtained in the presence of HCOOMe while only 67 % yield was achieved using HCOOH as the reagent. Combined experimental results disclosed that the HCOOMe pathway displayed the major possibility for the formylation of amine using CO₂ and H₂.

Based on the DFT calculations and experimental results, we believed that there are possibly three steps occurred during the N-formylation of amines with H₂ and CO₂ at Cu₂O/Cu(111) interface (Scheme 3). Initially, facilitating by the O_inter at Cu₂O/Cu(111), H₂ was adsorbed and activated with the formation of Cu−H and Cu₂OH which reacted with CO₂ to produce HCOOH. Subsequently, the HCOOH was quickly converted to HCOOMe in methanol solution. Finally, the generated HCOOMe directly reacting with amine under heating conditions to produce the desired N-formylated amine.

Conclusion

In summary, the challenging N-formylation of aliphatic primary amines with CO₂/H₂ was achieved by catalysts containing Cu₂O/Cu interface sites. The introduction of an aluminum oxide support strongly stabilizes the active mixed valence Cu₂O/Cu structure, and the catalyst can be consecutively used for 6 times without obvious activity loss. A series of formamides with diverse structures was synthesized in high yields. In situ XRD, pseudo in situ XPS, HAADF-STEM and ELNES characterization results confirmed the co-existence of both Cu₂O and Cu sites in the active catalysts, and a study of the reaction mechanism showed that methyl formate is the possible reaction intermediate. DFT calculation revealed that, compared with a pure Cu(111) surface, CO₂ conversion on the Cu₂O/Cu(111) interface was energetically more favorable, since oxygen from the Cu₂O/Cu interface, by forming O_inter-H, does not only efficiently reduce the apparent barrier of HCOOH formation but also facilitates desorption of the N-formylated amine from the Cu₂O/Cu interface, affording high activity and selectivity.

Conflict of interest

The authors declare no conflict of interest.