Introduction

As the culprit for greenhouse effect, CO2, especially those in high purity released from cement manufacturing, breweries, and fuel processing facilities, can be regarded as a promising candidate to synthesize chemicals which are currently produced from fossil resources. Long-chain olefins (C4+=) are versatile industrial feedstocks for a variety of value-added products such as synthetic lubricants, high-octane gasoline, biodegradable detergents, new polymers, agricultural chemicals, coatings, and corrosion inhibitors1,2. The prevalent method for the synthesis of these olefins is based on oligomerization of ethylene which is mostly produced from petroleum resources3,4. The use of C4+= as industrial feedstocks would play a pivotal role in the development of a sustainable society if C4+= could be directly obtained from CO2 hydrogenation. To ensure the whole process carbon negative, H2 must be produced from water electrolysis powered by renewable energy instead of coal gasification or reforming of natural gas5,6. Considering that electrolysis is distributed and produced in small-scale devices, it would be attractive to perform the subsequent CO2 hydrogenation in ambient-pressure units for matching the manufacturing scales and flexible on-demand production7,8.

For CO2 hydrogenation, ambient pressure is adverse to the formation of liquid long-chain olefins based on Le Chatelier’s principle. Currently, olefins produced from CO2 hydrogenation are mainly in the gaseous range of C2-4=, where the corresponding catalysts generally comprised metal oxides for methanol synthesis and zeolites for methanol-to-olefin process9,10,11. Limited catalysts targeting on long-chain olefins were operated under high pressure12,13,14. For instance, an iron aluminum oxide exhibited high selectivity (66.8%) for long-chain olefins (C4+=) under 35 bar15. However, it is not simply a case of lowering the pressure for these catalysts if one intends to achieve ambient-pressure synthesis of long-chain olefins.

Designing a tandem process including CO-intermediate and methanol-intermediate routes represents a successful strategy for CO2 hydrogenation into long-chain products16,17,18. To carter to the ambient-pressure condition, we should choose CO-intermediate route because low-pressure benefits reverse-water-gas-shift (RWGS) reaction but suppresses methanol synthesis process8. In this regard, the challenge lies in seeking active sites for Fischer-Tropsch synthesis (FTS) under ambient pressure. C–C coupling during FTS generally includes carbide mechanism and CO insertion mechanism19,20,21,22. The carbide mechanism involves the dissociation of CO into surface carbon, the hydrogenation of surface carbon into CHx* (x = 1, 2, or 3) intermediates, the surface polymerization of CHx* (chain growth), and the hydrogenation (chain termination) to form hydrocarbon products19. The low total pressure induces low surface coverage of CHx* due to the insufficient dissociation of CO, resulting in the short-chain length and a large amount of undissociated CO molecules. If these undissociated CO molecules are utilized as the monomer unit for chain growth, namely CO insertion mechanism21, it is promising to achieve the formation of long-chain products under ambient pressure.

Herein, we report a Cu–Fe catalyst which was operated under ambient pressure with comparable C4+= selectivity to that of the state-of-the-art catalysts optimized under high pressure toward CO2 hydrogenation (Supplementary Fig. 1 and Supplementary Table 1). The catalyst activated from delafossite oxides CuFeO2 contained Cu, iron oxides, and iron carbides, which was denoted as activated CuFeO2. Iron oxides enabled RWGS reaction to produce CO, while CO underwent carbide mechanism over iron carbides and experienced CO insertion over interfacial sites between copper and iron carbides. The synergy of carbide path and CO insertion path resulted in a high C4+= selectivity of 66.9% (excluding CO) and a CO2 conversion of 27.3% under 1 bar (H2:CO2 = 3:1) with a space velocity of 2400 mL h−1 gcat−1 at 320 °C. Moreover, we found that increasing the total pressure (30 bar) was able to refresh the deactivated catalyst during CO2 hydrogenation under ambient pressure.

Results and discussion

Catalytic properties

We prepared the catalysts via a hydrothermal method. The resulting sample exhibited a typical delafossite-type structure with homogeneous distribution of Cu, Fe, and O elements, which was denoted as fresh CuFeO2 (Supplementary Figs. 2 and 3). The fresh CuFeO2 was activated via H2 reduction under 4 bar at 400 °C for 2 h, followed by being exposed to 1 bar of mixed gas (H2:CO2 = 3:1) with a space velocity of 2,400 mL h−1 gcat−1 at 320 °C for 4 h since when the conversion of CO2 became stable. The obtained catalyst was designated as activated CuFeO2. The catalytic properties of activated CuFeO2 were evaluated in a fixed-bed reactor under 1 bar (H2:CO2 = 3:1) with a space velocity of 2,400 mL h−1 gcat−1 at 320 °C for 4 h on stream. The ambient pressure inevitably resulted in high CO selectivity of 43.7%. The selectivity for C4+ hydrocarbons was 74.0% (excluding CO) with an extremely high ratio (9.4) of olefin to paraffin (o/p ratio) at a CO2 conversion of 27.3% (Fig. 1a and Supplementary Table 1). Though activated CuFeO2 was operated under ambient pressure, the C4+= selectivity was as high as 66.9%, which was comparable to that over the state-of-the-art catalysts optimized for high-pressure conversion (Supplementary Table 1). The CH4 selectivity reached 5.4%, while the selectivity for C2-3 hydrocarbons was 20.6%. Figure 1b shows the distribution of hydrocarbon products, well matching the Anderson–Schulz–Flory (ASF) distribution. The probability of chain growth (α) was calculated as 0.72, while the coefficient of determination (R2) was 0.98. As indicated by the high linearity and high CO selectivity, the reaction tandemly proceeds along a RWGS reaction and a typical FTS process obeying the ASF model. Once following ASF distribution, the carbon chain grew via one-carbon-atom process while the α value was independent on carbon numbers23. Notably, the hydrocarbon distribution of activated CuFeO2 behaved differently from that of a similar reported catalyst even under the same condition24 (Supplementary Table 2 and Supplementary Fig. 4). In addition, the pre-reduction in 1 bar of H2 did no obviously influence the catalytic activity and selectivity compared with the catalyst pre-reduced in 4 bar of H2 (Supplementary Table 3).

Fig. 1: Catalytic properties toward CO2 hydrogenation.
figure 1

a The hydrocarbon product distribution, CO2 conversion, and CO selectivity of activated CuFeO2 and activated Fe2O3. b The detailed hydrocarbon product distribution, the ASF plot, and the corresponding α value of activated CuFeO2. α is the probability of chain growth. R2 is the coefficient of determination, describing the goodness of linear fitting. Wn is the weight fraction of a product with n carbon atoms. CO2 conversion and product selectivity at c different space velocities and d different H2:CO2 ratios over activated CuFeO2. e Stability test of activated CuFeO2. For panels a and b, the reaction was conducted under 1 bar (H2:CO2 = 3:1) with a space velocity of 2400 mL h−1 gcat−1 at 320 °C after 4 h on stream. The reaction conditions for panels c–e are similar to that for panel a except for the space velocity, H2:CO2 ratio, and time on stream, respectively. The selectivity for hydrocarbon product excludes CO.

Full size image

By comparison, Fe2O3 after the same activation procedure (denoted as activated Fe2O3) was tested under 1 bar (H2:CO2 = 3:1) with a space velocity of 2,400 mL h−1 gcat−1 at 320 °C for 4 h on stream. The C4+ selectivity reached 46.0% (excluding CO) with an o/p ratio of 7.2 at a CO2 conversion of 26.6%, whereas the selectivities for CH4 and C2-3 hydrocarbons were 17.9% and 36.1%, respectively (Fig. 1a and Supplementary Table 1). The C4+= selectivity was 40.4%, lower than that (66.9%) over activated CuFeO2. As shown in Supplementary Figure 5, the distribution of hydrocarbon products for activated Fe2O3 also followed the ASF distribution with an α value of 0.59 and a high R2 value of 0.99. When the total pressure increased to 30 bar, the C4+= selectivity increased to 50.2% (Supplementary Table 1). In addition, we prepared ZnO-ZrO2 solid solution supported on a Zn-modified SAPO-34 zeolite (ZnZrO/SAPO) as a reference catalyst. This catalyst was reported as a highly selective catalyst toward light olefins under high pressure25. Considering that ZnZrO/SAPO was active at high temperature (360–400 °C), we tested its catalytic performance under 1 bar (H2:CO2 = 3:1) with a space velocity of 2,400 mL h−1 gcat−1 at 380 °C. Under this condition, the major product was light olefins (C2–C4=) with the selectivity of 61.4% instead of long-chain olefins (C4+=) with the selectivity of 3.5% (Supplementary Table 1). Moreover, the selectivities for CO and methane (excluding CO) boosted to 85.1% and 30.3%, respectively (Supplementary Table 1), because lowering the pressure suppresses methanol synthesis due to Le Chatelier’s principle.

To investigate the robustness of activated CuFeO2, we varied the space velocity and the ratio of H2 to CO2 over activated CuFeO2. When the space velocity increased from 2400 to 9600 mL h−1 gcat−1, the C4+= selectivity decreased slightly from 66.9% to 61.6% (Fig. 1c). Moreover, the selectivity for CH4, C2-3, and C4+ along with o/p ratios also exhibited slight variations, while the conversion of CO2 decreased from 27.3% to 19.8% (Fig. 1c). When the ratio of H2 to CO2 increased from 1 to 4 with the space velocity of 2,400 mL h−1 gcat−1, the C4+= selectivity dropped from 70.2% to 62.0%, while the conversion of CO2 rose from 15.6% to 29.5% (Fig. 1d). Therefore, the selectivity for long-chain olefins is insensitive to both the space velocity and the ratio of H2 to CO2, indicating high robustness of activated CuFeO2 which applied to a wide range of reaction conditions.

We further investigated the stability of activated CuFeO2 under 1 bar (H2:CO2 = 3:1) with a space velocity of 2,400 mL h−1 gcat−1 at 320 °C. After 50 h on stream, the conversion of CO2 kept stable, whereas the selectivity for long-chain olefins (C4+=), unfortunately, dropped to 57.6% (Fig. 1e). Meanwhile, the selectivity for methane increased to 14.6%. Thus, activated CuFeO2 was not stable enough during ambient-pressure hydrogenation of CO2.

Regeneration of activated CuFeO2

We notice that an elevated total pressure (30 bar) with high H2:CO2 ratio (3:1) enables the coverage of surface hydrogen on catalysts to become high enough for minimizing carbon formation26,27. Inspired by these points, we propose that increasing the total pressure serves as a promising approach to the regeneration of the catalysts. When the reaction proceeded under 1 bar at 320 °C for 50 h on stream, the C4+= selectivity declined to 55.2% (Fig. 2a). Afterward, the total pressure was increased to 30 bar. Under this condition, the C4+= selectivity reached 66.3% under 30 bar, almost equal to that (66.9%) under ambient pressure (Supplementary Table 3). Compared with the results obtained under 1 bar, the major difference lies in the significant decrease of CO selectivity (12.3%) and the appearance of oxygenated products with the selectivity of 7.5% (Supplementary Table 3). Based on Le Chatelier’s principle, the elevated pressure promotes the conversion of CO to hydrocarbons and oxygenated products, while RWGS is insensitive to pressure, accounting for the decreased selectivity for CO. The appearance of oxygenated products indicates the existence of CO insertion mechanism in the catalytic system21. After operating the catalyst under 30 bar for 4 h, we lowered the total pressure back to 1 bar. It is worth noting that the C4+= selectivity recovered to 65.2% (Fig. 2a). When we repeated the regeneration procedure, activated CuFeO2 resumed its high selectivity for C4+= every time. Therefore, elevating the total pressure represents a convenient regeneration method which does not need to switch gases or unload the catalysts from the reactor.

Fig. 2: Regeneration of activated CuFeO2.
figure 2

a Long-term stability test with regeneration over activated CuFeO2. The reaction was conducted under 1 bar (H2:CO2 = 3:1) with a space velocity of 2400 mL h−1 gcat−1 at 320 °C. The purple dash vertical lines refer to the regeneration treatment that was conducted under 30 bar (H2:CO2 = 3:1) with a space velocity of 2400 mL h−1 gcat−1 at 320 °C for 4 h on stream. TGA profiles of activated CuFeO2 b before regeneration treatment and c after regeneration treatment. Raman spectra of activated CuFeO2 d before regeneration treatment and e after regeneration treatment.

Full size image

The effect of the regeneration was investigated by means of thermogravimetric analysis (TGA) in N2 atmosphere. For activated CuFeO2 that had been operated under 1 bar for 50 h (denoted as the catalyst before regeneration), the weight was lost by 6.62 wt% (Fig. 2b). The catalyst after working under 1 bar for 50 h and subsequently under 32 bar for 4 h was denoted as the catalyst after regeneration. During TGA tests, activated CuFeO2 after regeneration lost 2.02 wt% weight, which was lower than that (6.62 wt%) before regeneration (Fig. 2c). As such, high-pressure treatment cleaned long-chain hydrocarbons to refresh the catalyst. We further conducted Raman measurements. The peaks at 1352 and 1364 cm−1 were assigned to disordered carbon (D band), while those at 1601 and 1607 cm−1 corresponded to graphite (G band) (Fig. 2d, e)28. The intensities of these two peaks for activated CuFeO2 before regeneration were higher than those for the catalyst after regeneration (Fig. 2d, e). In this case, high-pressure treatment is also able to remove disordered carbon and graphite.

In addition, we conducted Brunauer–Emmett–Teller (BET) measurements of activated CuFeO2 after different treatments. For clarity, we denote samples I, II, III, and IV as the activated CuFeO2 after reaction for 0, 10, 20 h, and regeneration, respectively. The BET surface areas of samples I, II, III, and IV were measured as 2.88, 2.62, 2.35, and 2.96 m2 g−1, respectively (Supplementary Fig. 6). In this case, the BET surface area decreased with the prolonged reaction time and recovered after regeneration. However, the variation in surface area of the samples after different treatments was not prominent, presumably consistent with the stable conversion of CO2. There must be other underlying mechanisms for catalyst deactivation and regeneration in addition to carbon deposition. To this end, we conducted energy dispersive X-ray (EDX) elemental mapping characterizations of samples I, III, and IV. We found that the spatial overlap between Cu and Fe elements of sample III is obviously smaller than that of sample I and that of sample IV (Supplementary Figs. 7–9). As such, we speculate that the deactivation and regeneration are presumably associated with the segregation and re-dispersion of Cu and Fe elements.

Structural and electronic characterizations of activated CuFeO2

To explore the nature of active sites, we resorted to multiple structural characterizations. Comparing scanning electron microscopy (SEM) images of fresh and activated CuFeO2, we found that the activation procedure etched the initial polyhedral particles into a porous structure (Supplementary Figs. 2a and 10). Figure 3a shows a high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) image of activated CuFeO2. The lattice parameters of 0.21, 0.25, and 0.22 nm were ascribed to Cu(111), Fe3O4(311), and χ-Fe5C2(11-2) facets, respectively. The interfaces between these facets marked by red lines were clearly identified. Supplementary Figure 7a shows EDX elemental mapping images, suggesting that activated CuFeO2 comprised Cu, Fe, O, and C elements. In comparison with the elemental mapping images of fresh CuFeO2, the compositional line profile of activated CuFeO2 indicated that the homogeneously distributed Cu and Fe elements underwent obvious segregation after the activation (Supplementary Fig. 7b). Moreover, the activation treatment led to intimate contact between Cu and Fe species as indicated by elemental mapping images and compositional line profile (Supplementary Fig. 7). The phase segregation was further verified by the X-ray diffraction (XRD) profile of activated CuFeO2. As shown in Fig. 3b, the characteristic XRD peaks were assigned to the phases of Hägg carbides (χ-Fe5C2) and pure Cu. The Mössbauer spectra of activated CuFeO2 showed that iron phases include χ-Fe5C2, Fe3C, and Fe3O4. Specially, χ-Fe5C2, Fe3C, and Fe3O4 occupied 73.6%, 9.8%, and 16.6% of the total iron phases (Fig. 3c and Supplementary Table 4). The iron phases of activated CuFeO2 were different from those of a similar reported catalyst which contained metallic Fe and χ-Fe5C2 without iron oxides24. In comparison, the structures of activated Fe2O3 were also investigated by means of multiple characterization techniques including SEM, XRD, and Mössbauer spectroscopy. We found that activated Fe2O3 comprised Fe3O4 and χ-Fe5C2 (Supplementary Fig. 11 and Supplementary Table 5).

Fig. 3: Structural characterizations of activated CuFeO2.
figure 3

a HAADF image of activated CuFeO2. b XRD profiles of activated CuFeO2. c Mössbauer spectra of activated CuFeO2. d Cu LMM Auger, e Fe 2p XPS, and f O 1s XPS spectra of fresh and activated CuFeO2.

Full size image

The electronic structures of activated CuFeO2 were investigated via X-ray photoelectron spectroscopy (XPS) measurements. Supplementary Figure 12 shows Cu 2p XPS spectra of fresh and activated CuFeO2. No evident satellite peaks were presented, indicating the absence of Cu2+. However, the assignment of Cu0 and Cu+ cannot be concluded only from XPS results since Cu and Cu+ have close binding energies29,30. To discriminate between Cu0 and Cu+, we turned to Cu LMM Auger measurements. The Auger spectrum of fresh CuFeO2 showed a peak at 917.0 eV which arose from Cu+ (Fig. 3d)29,30. As for activated CuFeO2, two peaks at 916.6 and 918.7 eV were observed, corresponding to Cu+ and Cu0, respectively (Fig. 3d)29,30. Based on Fe 2p XPS spectra, Fe species on fresh CuFeO2 were at the oxidation state of +3, whereas activated CuFeO2 contained Fe3+, Fe2+, and Fe0 species (Fig. 3e)31,32. As shown in Fig. 3f, O 1s spectra were deconvoluted into three peaks. Specially, the main peaks at around 529.4 and 530.0 eV were ascribed to the lattice O atoms (Olattice)9,33. The peaks at 531.0 and 531.4 eV were assigned to O atoms proximal to a defect (Odefect), while those at 532.8 eV corresponded surface hydroxyl groups (OH*)9,33.

Based on structural and electronic characterizations, we conclude how CuFeO2 reconstructed after the activation procedure. After the activation, Cu+ in the lattice of CuFeO2 collapsed and aggregated to form pure Cu phase with partially oxidized surface. Meanwhile, Fe3+ in the lattice of CuFeO2 underwent partial reduction and carbonization, resulting in the formation of Fe3O4, χ-Fe5C2, and Fe3C. In this case, a mixed states of Fe3+, Fe2+, and Fe0 existed. Therefore, activated CuFeO2 was composed of copper, iron oxides, and iron carbides. Moreover, these species were in intimate contact with each other, resulting in the formation of multiple interfacial sites such as the interface between Cu and χ-Fe5C2.

Mechanistic studies

In order to explore the origin of high selectivity for C4+= olefins under ambient pressure, we conducted temperature-programmed desorption (TPD) measurements to explore the adsorption of CO. Specially, the samples were exposed to the mixed gas (CO:He = 1:9) under 1 bar at 50 °C for 30 min, followed by being purged in He for 30 min. The TPD curves were recorded from 50 °C to 800 °C at a heating rate of 5 °C min−1. The CO-TPD profile of activated Fe2O3 shows a peak at 508 °C for dissociative adsorption of CO (Fig. 4a)34. Considering that activated Fe2O3 comprised Fe3O4 and χ-Fe5C2, we proposed the reaction scheme of activated Fe2O3 as that for Fe-based catalysts previously reported16,35,36. Specially, CO2 was hydrogenated over Fe3O4 into CO via RWGS, while the produced CO was dissociated on χ-Fe5C2 and subsequently underwent FTS process to yield product whose distribution followed ASF distribution. As for the CO-TPD profile of activated CuFeO2, two peaks at 290 and 501 °C appeared, corresponding to non-dissociative and dissociative adsorption of CO, respectively (Fig. 4a)34,37. The non-dissociative adsorption of CO* was further supported by diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) measurements using CO as a probe molecule. Compared with the DRIFTS spectrum of activated Fe2O3, the DRIFTS spectrum of activated CuFeO2 showed an additional peak at 2061 cm−1 (Supplementary Fig. 13). This peak was assigned to non-dissociative adsorption of CO*. We further conducted in-situ DRIFTS measurements of activated CuFeO2 after exposure to the mixed gas (H2:CO2 = 3:1, 1 bar) at 300 °C. As shown in Supplementary Fig. 14, the peaks for CHx, gaseous CO2, gaseous CO, and chemically adsorbed CO* were observed. These spectropic evidences prove non-dissociative adsorption of CO* which serves as a prerequisite for CO insertion.

Fig. 4: Mechanistic studies.
figure 4

a CO-TPD profiles of activated CuFeO2 and activated Fe2O3. b Absolute photoionization cross sections for acetaldehyde over activated CuFeO2. c Comparison in energy barriers of CH2 + CO over Cu-χ-Fe5C2 and χ-Fe5C2(11-2) facets. d Comparison in energy barriers of CH2 + CH2 over Cu-χ-Fe5C2 and χ-Fe5C2(11-2) facets. Transient response curves obtained during propene pulses into H2 flow over e activated CuFeO2 and f activated Fe2O3. Ratio refers to the ratio of C3H6/C3H8 peak area detected by mass spectroscopy.

Full size image

As activated CuFeO2 was composed of Cu, Fe3O4, χ-Fe5C2, and Fe3C, Fe3O4 enabled RWGS reaction to produce CO, while CO was non-dissociatively adsorbed on Cu and dissociatively adsorbed on iron carbides, subsequently experiencing FTS process. It is worth noting that the chain propagation process during FTS generally involves CO insertion and carbide mechanism, though all the possible mechanisms can result in the selectivity as predicted by ASF model38. Dissociative adsorption of CO favors carbide mechanism which is generally obeyed by C-C coupling over iron carbides38. Non-dissociative adsorption of CO allows for CO insertion mechanism, which is implied by the appearance of detectable oxygenated products under high pressure (Supplementary Table 3)38.

To identify the existence of CO insertion mechanism, we resorted to probing oxygenate intermediates which are regarded as characteristics of CO insertion, because carbide mechanism cannot account for the formation of oxygenate products via surface polymerization of CHx39. To detect the intermediate, we conducted synchrotron-based vacuum ultraviolet photoionization mass spectrometry (SVUV-PIMS). This technique is highly sensitive to low-concentration intermediates and able to identify isomers because of tunable photon energy and soft ionization40,41. The reaction proceeded under 1 bar (H2:CO2 = 3:1) with a high space velocity of 36,000 mL h−1 gcat−1 at 320 °C to shorten the contact time in case that active intermediates would be hydrogenated. The photon energy was selected at 10.40 eV to avoid the ionization of CO2 whose ionization energy is 13.77 eV42. For activated CuFeO2, a signal of mass/charge ratio (m/z) = 44 appeared in addition to stable hydrocarbon products (Supplementary Fig. 15). When the photon energy was varied from 9.10 to 10.40 eV, the point of inflection appeared at 10.25 eV which was assigned to acetaldehyde (CH3CHO) after excluding the possibilities of other species with m/z = 44 (Fig. 4b and Supplementary Table 6)43. To investigate whether C-C coupling occurs via acetaldehyde oligomerization, we conducted temperature-programmed surface reaction (TPSR) measurements of activated CuFeO2. Specially, the acetaldehyde vapor was introduced via bubbling with the mixed gas (H2:Ar = 1:9, 1 bar) with a gas-flow rate of 100 mL min−1. Then the temperature was raised to 300 °C with a rate of 10 °C min−1 and kept at 300 °C. During the whole period, no detectable products were observed as shown in Supplementary Fig. 16, thereby excluding the possibility of the acetaldehyde oligomerization. Therefore, we speculate that CO insertion occurred during the chain growth over activated CuFeO2.

Generally, interfacial sites between copper and iron carbides are regarded as the active center for CO insertion. To gain theoretical insights into the mechanisms, we conducted density functional theory (DFT) calculations. To simulate the interface between iron carbides and copper, we established χ-Fe5C2 clusters on the surface of Cu(111) (Supplementary Fig. 17). This simplified model captures the main features of the interface, though this model cannot completely reproduce the real catalyst. Cu(111) and χ-Fe5C2(11-2) were constructed for comparison (Supplementary Fig. 17). We calculated the d-band centers of these facets to qualitatively evaluate the catalytic performance. The d-band center of Cu-χ-Fe5C2 interfacial sites is −0.32 eV, which is closer to the Fermi level than that (−0.69 eV) of χ-Fe5C2(11-2) and that (−2.02 eV) of Cu(111) facets (Supplementary Fig. 18). The upward shift of d-band center empties more antibonding states which allows for accepting more electrons from CO and alkyl species, thereby facilitating CO insertion into alkyl species44,45.

To further investigate CO insertion over different surfaces, we calculated the energy barriers of C–C coupling. FTS is a complex process involving various possible steps. For simplicity, we use CH2 + CO to represent the C-C coupling via CO insertion, while CH2 + CH2 represents the carbide mechanism. CO insertion over χ-Fe5C2(11-2) is an endothermic process (ΔH = + 0.48 eV), requiring an energy barrier as high as 1.61 eV (Fig. 4c and Supplementary Fig. 19). As for Cu-χ-Fe5C2 interfacial sites, CO insertion becomes exothermic (ΔH = −0.52 eV), while the energy barrier decreases to 0.72 eV (Fig. 4c and Supplementary Fig. 20). As such, Cu-χ-Fe5C2 interface significantly promotes CO insertion relative to iron alone. In contrast, the carbide mechanism is less sensitive to interface than CO insertion, as the introduction of interface only lowers the energy barrier by 0.25 eV (versus 0.89 eV in CO insertion) (Fig. 4c, d, Supplementary Figs. 19–22).

To support the important role played by interface, we qualitatively tuned the interfacial sites by varying the ratios of Cu to Fe from 1:9 to 9:1 in Cu-Fe binary oxides. Though we cannot quantify the interfacial sites, the amount of interfacial sites should exhibit a volcano-type trend maximized at Cu:Fe = 1:1 based on rough estimation. The reaction was conducted under 1 bar (H2:CO2 = 3:1) with a space velocity of 2,400 mL h−1 gcat−1 at 320 °C after 4 h on stream. As shown in Supplementary Figure 23, the selectivity for C4+= exhibited a volcano-type trend against the Cu:Fe ratios with the maximum at Cu:Fe = 1:1 (activated CuFeO2). We can draw a qualitative conclusion that the C4+= selectivity increased with Cu-Fe interfacial sites. This result indirectly supported our claim that interfacial sites promoted C-C coupling. Therefore, the synergy of carbide path over iron carbides and CO insertion path over copper/iron carbides interfacial sites led to efficient C-C coupling into long-chain products under ambient pressure (Supplementary Fig. 24).

Given that Cu–Fe interfacial sites contribute to C–C coupling, we explain the deactivation and regeneration mechanisms. The overlap between Cu and Fe elements can roughly estimate the amount of interfacial sites. The results show that the 20-h reaction under ambient pressure led to a significant decrease in the interfacial sites between copper and iron. Moreover, the regeneration treatment redispersed Cu and Fe elements, leading to the recovery of interfacial sites. As such, the decreased C4+ selectivity with the reaction time presumably derived from the decreased interfacial sites. The regeneration not only cleaned the carbon deposits but also recovered the interfacial sites, inducing the recovery of the C4+ selectivity.

To explain the high o/p ratio of activated CuFeO2, we turned to pulse experiments for exploring whether alkenes prefer desorption or hydrogenation on the catalysts. The catalysts were operated under 1 bar (H2:CO2 = 3:1) at 320 °C for 1 h, before switching to a H2 flow. Afterward, propene which was chosen as the representative of alkenes was pulsed into the reactor. As shown in Fig. 4e, f, the ratio of C3H6/C3H8 peak area for activated CuFeO2 was 30.7, much higher than that (6.5) for activated Fe2O3. Such a high ratio for activated CuFeO2 indicates that the formation of propane was almost totally inhibited.

In conclusion, we achieved ambient-pressure hydrogenation of CO2 into long-chain olefins over activated CuFeO2. The C4+= selectivity under 1 bar at 320 °C reached as high as 66.9% which was comparable to that of the state-of-the-art catalysts optimized under high pressure (35 bar). The high C4+= selectivity under ambient pressure derives from the synergy of carbide path and CO insertion path. Our findings represent a promising approach to CO2 conversion connected with a decentralized use of renewable power-generated hydrogen. Moreover, we offer a viable method for the regeneration of deactivated catalysts under ambient-pressure hydrogenation of CO2.

Methods

Chemicals and materials

Fe(NO3)3·9H2O, Cu(NO3)2·3H2O, NaOH, Na2CO3, (NH4)2CO3, Zn(NO3)2·6H2O, Zr(NO3)2·5H2O and FeCl3 were analytical grade and purchased from Sinopharm Chemical Reagent Co., Ltd. FeCl2·4H2O and acetaldehyde aqueous solution (35%) were purchased from Sigma-Aldrich. SAPO-34 zeolites was purchased from Nankai University Catalyst Co., Ltd.

Synthesis of CuFeO2

CuFeO2 was synthesized via a hydrothermal method. Typically, 2.42 g of Cu(NO3)2·3H2O and 4.04 g of Fe(NO3)3·9H2O were added to 40 mL of deionized (DI) water with stirring to form a clear solution. In the above solution, 40 mL of 5 M NaOH aqueous solution was then added dropwise under stirring at room temperature. After stirring for 30 min, 1 mL of propionaldehyde was added as the reducing agent. The mixture was transferred to 100-mL Teflon-lined stainless steel autoclave and kept at 180 °C for 24 h. The product was separated by centrifugation, washed twice with DI water, and dried overnight at 60 °C.

Synthesis of activated CuFeO2

Five hundred milligrams of CuFeO2 was reduced in 4 bar of a pure H2 flow with a flow rate of 100 mL min−1 at 400 °C for 2 h, followed by being exposed to 1 bar of mixed gas (H2:CO2 = 3:1) with a space velocity of 2,400 mL h−1 gcat−1 at 320 °C for 4 h.

Synthesis of Fe2O3

Fe2O3 was synthesized via co-precipitation method. Typically, 8.08 g of Fe(NO3)3·9H2O was added to 40 mL of DI water under stirring until the formation of a clear solution. In the above solution, 20 mL of 2 M Na2CO3 aqueous solution was then added dropwise under stirring at room temperature. After being aged for 1 h, the turbid liquid was filtrated, washed twice with DI water, and dried overnight at 60 °C. The resulting powders were calcinated in muffle furnace at 350 °C for 4 h.

Synthesis of activated Fe2O3

Five hundred milligrams of Fe2O3 was reduced in 4 bar of a pure H2 flow with a flow rate of 100 mL min−1 at 400 °C for 2 h, followed by being exposed to 1 bar of mixed gas (H2:CO2 = 3:1) with a space velocity of 2,400 mL h−1 gcat−1 at 320 °C for 4 h.

Synthesis of ZnZrO/SAPO

ZnZrO was synthesized via co-precipitation method, Typically, 1.325 g of Zn(NO3)2·6H2O and 5.66 g of Zr(NO3)4·5H2O were dissolved in 70 mL of DI water at 70 °C. 50 mL of 0.625 M (NH4)2CO3 aqueous solution was added dropwise under vigorous stirring at 70 °C. The pH value of the solution was kept at about 7.0. After being aged for 2 h at 70 °C, the product was separated by centrifugation, washed twice with DI water, and dried overnight at 80 °C. The resulting powders were calcinated in muffle furnace at 500 °C for 5 h. The ZnZrO/SAPO was prepared through physical mixing the ZnZrO solid solution and SAPO-34 zeolite, the mass ratio of these two components was 1:1.

Synthesis of Cu-Fe binary oxide with different ratios

copper iron binary oxide was synthesized via co-precipitation method, 2.42 g Cu(NO3)2·3H2O and Fe(NO3)3·9H2O with a given molar ratio of Cu2+:Fe3+ (9:1, 3:1, 1:3, 1:9, respectively) were dissolved in 50 mL of DI water with stirring to form a clear solution. In the above solution, 2 M Na2CO3 aqueous solution was then added dropwise under stirring at room temperature, and the pH value of final suspension was maintained at 9. After being aged for 1 h, the product was separated by centrifugation, washed twice with DI water, and dried overnight at 60 °C. The resulting powders were calcinated in muffle furnace at 350 °C for 4 h.

Catalytic tests

CO2 hydrogenation reactions were carried out in a fixed-bed reactor under 1 bar of mixed gas at 320 °C. The mixed gas contained 96 vol% H2/CO2 as reactants and 4 vol% Ar as an internal standard. Generally, the catalyst (500 mg, 20–40 meshes) diluted with powdered quartz (500 mg, 20–40 meshes) was loaded into a fixed-bed reactor with an inner diameter of 9 mm. The catalysts refer to activated CuFeO2 and activated Fe2O3.

For the tests over ZnZrO/SAPO, 200 mg of ZnZrO/SAPO was pretreated in an Ar flow with a flow rate of 30 mL min−1 at 380 °C for 1 h. Afterward, the catalyst was exposed to 1 bar of mixed gas (H2:CO2 = 3:1) with a space velocity of 2,400 mL h−1 gcat−1 at 380 °C. The catalytic data were obtained after when the reaction reached a steady state.

For the tests over Cu-Fe binary oxides with different ratios, 500 mg of the catalyst (20–40 meshes) was reduced in 4 bar of a pure H2 flow with a flow rate of 100 mL min−1 at 400 °C for 2 h. followed by being exposed to 1 bar of mixed gas (H2:CO2 = 3:1) with a space velocity of 2,400 mL h−1 gcat−1 at 320 °C for 4 h.

All of the products from the reactor were introduced in a gaseous state and analyzed with two online gas chromatographs (Shimadzu GC-2014). H2, CO, CO2, CH4, and Ar were analyzed by using a carbon molecular sieves column (TDX-1) with a thermal conductivity detector (TCD). Hydrocarbons were analyzed using a PONA capillary column with a flame ionization detector (FID). CH4 was taken as a reference bridge between TCD and FID. CO2 conversion was calculated according to an internal standard method, assuming that the amount of Ar remained constant after the reaction.

CO2 conversion was calculated on a carbon-atom basis, as follows:

$${{{{{{\rm{CO}}}}}}}_{2}\,{{{{{\rm{conversion}}}}}}=\frac{{{{{{\rm{CO}}}}}}_{{{{{\rm{2inlet}}}}}}-{{{{{\rm{CO}}}}}}_{{{{{\rm{2outlet}}}}}}}{{{{{{\rm{CO}}}}}}_{{{{{\rm{2inlet}}}}}}}\times 100 \%$$
(1)

where CO2 inlet and CO2 outlet are moles of CO2 at the inlet and outlet, respectively.

CO selectivity was calculated according to:

$${{{{{\rm{CO}}}}}}\,{{{{{\rm{selectivity}}}}}}=\frac{{{{{{\rm{CO}}}}}}_{{{{{\rm{outlet}}}}}}}{{{{{{\rm{CO}}}}}}_{{{{{\rm{2inlet}}}}}}-{{{{{\rm{CO}}}}}}_{{{{{\rm{2outlet}}}}}}}\times 100 \%$$
(2)

where COoutlet refers to moles of CO at the outlet.

The selectivity for hydrocarbon CnHm was obtained according to:

$${{{{{{\rm{C}}}}}}}_{{{{{{\rm{n}}}}}}}{{{{{{\rm{H}}}}}}}_{{{{{{\rm{m}}}}}}}\,{{{{{\rm{selectivity}}}}}}=\frac{{n}{{{{{{{\rm{C}}}}}}}_{{{{{{\rm{n}}}}}}}{{{{{{\rm{H}}}}}}}_{{{{{{\rm{m}}}}}}\,{{{{{\rm{outlet}}}}}}}}}{\mathop{\sum}\limits_{i}i{{{{{{\rm{C}}}}}}}_{{{{{{\rm{i}}}}}}}{{{{{{\rm{H}}}}}}}_{{{{{{\rm{m}}}}}}\,{{{{{\rm{outlet}}}}}}}}\times 100 \%$$
(3)

where CnHm outlet represents moles of individual hydrocarbon product at the outlet. The selectivity for oxygenates was below 1.0% and therefore was not reported in the product selectivity. The carbon balance was over 95.0%.

Mössbauer measurements

57Fe Mössbauer spectra were carried out on a Topologic 500 A spectrometer driving with a proportional counter at room temperature. The radioactive source was 57Co (Rh) moving in a constant acceleration mode. Data analyses were performed assuming a Lorentzian lineshape for computer folding and fitting.

TGA

TGA was conducted on Pyris Diamond TG-DTG in a N2 flow with the rate of 100 mL min−1 at the heating rate of 5 °C min−1 from 50 °C to 800 °C.

CO-TPD measurements

CO-TPD measurements were conducted by using a TPD instrument (AutoChem II 2920). Prior to CO-TPD, the samples were cleaned in He with a flow rate of 50 mL min−1 at 200 °C for 2 h. Then, the gas was switched to the mixed gas (CO:He = 1:9, 1 bar)with a flow rate of 20 mL min−1 at 50 °C. After CO adsorption for 30 min, the samples were purged by He with a flow rate of 50 mL min−1 at 50 °C for 30 min. The CO-TPD curves were recorded from 50 °C to 800 °C at a heating rate of 5 °C min−1.

DRIFTS spectra using CO as a probe molecule

In-situ DRIFTS experiments were conducted in an elevated-pressure cell (DiffusIR Accessory PN 041-10XX) with a Fourier transform infrared spectrometer (TENSOR II Sample Compartment) and a liquid-nitrogen-cooled MCT detector. Spectra were measured by accumulating 32 scans at a resolution of 8 cm−1. Prior to the test, the sample was flushed with He with a gas-flow rate of 30 mL min−1 at 200 °C for 30 min, followed by cooling to 25 °C. The background spectra of the sample were acquired under He flow at 25 °C. Then, 1 bar (CO:He = 1:9) with a gas-flow rate of 10 mL min−1 was allowed to flow into the cell at 25 °C for 30 min. The spectra were recorded under the mixed gas (CO:He = 1:9).

In-situ DRIFTS spectra of activated CuFeO2

In-situ DRIFTS experiments were conducted in an elevated-pressure cell (DiffusIR Accessory PN 041-10XX) with a Fourier transform infrared spectrometer (TENSOR II Sample Compartment) and a liquid-nitrogen-cooled MCT detector. Spectra were measured by accumulating 32 scans at a resolution of 8 cm−1. Prior to the test, the sample was reduced in 1 bar of H2 with a gas-flow rate of 50 mL min−1 at 300 °C for 30 min. Afterward, the sample was flushed with He with a gas-flow rate of 30 mL min−1 at 300 °C for 30 min. The background spectra of the sample were acquired under He flow at 300 °C. Then, 1 bar (H2:CO2 = 3:1) with a gas-flow rate of 20 mL min−1 was allowed to flow into the cell at 300 °C for 30 min, followed by purged with 1 bar of He with a gas-flow rate of 30 mL min−1 at 300 °C and the spectra were obtained to detect the adsorbed species on the sample.

SVUV-PIMS

SVUV-PIMS study was carried out at the combustion beamline of the National Synchrotron Radiation Laboratory at Hefei, China. A quartz reactor with a nozzle size of ~0.1 mm was designed, which was connected to the online SVUV-PIMS spectrometer. The detection limit of SVUV-PIMS is 0.1 ppm for CO2. The reaction proceeded under 1 bar (H2:CO2 = 3:1) with a space velocity of 36,000 mL h−1 gcat−1 at 320 °C.

TPSR measurement of activated CuFeO2

TPSR measurement was carried out in a AutoChem II 2920 apparatus with a mass spectrometer (Hiden HPR20). The activated CuFeO2 (500 mg, 20–40 meshes) was loaded in a quartz U-tube. After pretreatment with He at 200 °C for 30 min, the acetaldehyde vapor was introduced by bubbling with 1 bar (H2:Ar = 1:9) with a gas-flow rate of 100 mL min−1. Then the temperature was raised to 300 °C with a rate of 10 °C min−1 and kept at 300 °C.

DFT methods

DFT calculations were performed using the Vienna ab initio Simulation Package code46,47,48. The Perdew-Burke-Ernzerhof functional with generalized gradient approximation49 was used for the geometry optimizations and electronic structure calculations. The projector-augmented wave method50 was used to describe the electron-ion interactions. The atomic structures were fully relaxed by using a conjugate gradient scheme without symmetry restrictions until the maximum force on each atom was less than 0.02 eV Å−1. A vacuum space of ~20 Å along the z direction was used to separate the interactions between the neighboring slabs with an energy cutoff of 500 eV. The Cu(111) surface containing 96 Cu atoms is simulated by using a (4 × 6) supercell (10.224 Å × 13.281 Å) with four Cu layers, where the lower two layers are fixed to the bulk structure, while the upper two layers are fully relaxed. The DFT calculated lattice constants of bulk χ-Fe5C2 are (a = 11.354 Å, b = 4.413 Å, c = 4.914 Å; α = γ = 90o, β = 97.61o). The χ-Fe5C2(11-2) surface is simulated using a (1 × 1) supercell which contains 80 Fe and 32 C atoms. The Cu-χ-Fe5C2 interface is constructed from a (4 × 6) Cu(111) surface and a cluster of Fe10C4. The (3 × 3 × 1) Monkhorst-Pack mesh is used to sample the Brillouin-zone for all the electronic structure calculations. The (1 × 1 × 1) k-mesh is used for the geometry optimizations.

TPD pulse experiments

TPD pulse experiments were carried out in a AutoChem II 2920 apparatus with a mass spectrometer (Hiden HPR20). 200 mg of activated catalysts were cleaned in He with a gas-flow rate of 50 mL min−1 at 200 °C for 2 h, followed by treated in 1 bar of mixed gas (H2:CO2 = 3:1) with a flow rate of 20 mL min−1 at 320 °C. After 1-h reaction, the gas was switched to the mixed gas (H2:He = 1:9) with a flow rate of 50 mL min−1. 525 μL of diluted (C3H6:He = 1:9) was pulsed into the system every 5 min.

Characterizations

XRD patterns were recorded by using a Philips X’Pert Pro Super diffractometer with Cu-Kα radiation (λ = 1.54178 Å). XPS measurements were conducted on an ESCALAB 250 (Thermo-VG Scientific, USA) with an Al Kα X-ray source (1486.6 eV protons) in Constant Analyser Energy (CAE) mode with pass energy of 30 eV for all spectra. The values of binding energies were calibrated with the C1s peak of contaminant carbon at 284.60 eV. Raman spectra were detected by a Renishaw. RM3000 Micro-Raman system with a 514.5 nm Ar laser.