CO2 Hydrogenation to Olefin-Rich Hydrocarbons Over Fe-Cu Bimetallic Catalysts: An Investigation of Fe-Cu Interaction and Surface Species

Previously, we reported a strong Fe-Cu synergy in CO2 hydrogenation to olefin-rich C2 + hydrocarbons over the γ-Al2O3 supported bimetallic Fe-Cu catalysts. In this work, we aimed to clarify such a synergy by investigating the catalyst structure, Fe-Cu interaction, and catalyst surface properties through a series of characterizations. H2-TPR results showed that the addition of Cu made both Fe and Cu easier to reduce via the strong interaction between Fe and Cu. It was further confirmed by X-ray absorption spectroscopy (XAS) and TEM, which showed the presence of metallic Fe and Fe-Cu alloy phases in the reduced Fe-Cu(0.17) catalyst induced by Cu addition. By correlating TPD results with the reaction performance, we found that the addition of Cu enhanced both the moderately and strongly adsorbed H2 and CO2 species, consequently enhanced CO2 conversion and C2 + selectivity. Adding K increased the adsorbed-CO2/adsorbed-H2 ratio by greatly enhancing the moderately and strongly adsorbed CO2 and slightly suppressing the moderately and strongly adsorbed H2, resulting in a significantly increased O/P ratio in the produced hydrocarbons. The product distribution analysis and in situ DRIFTS suggested that CO2 hydrogenation over the Fe-Cu catalyst involved both an indirect route with CO as the primary product and a direct route to higher hydrocarbons.


INTRODUCTION
Waste CO 2 can be utilized as a carbon source to synthesize valuable chemicals and transportation fuels. Recently, using renewable energy to conduct catalytic CO 2 conversion is attracting increased attention, as it has the potential on mitigating both CO 2 emissions and the fossil fuel dependence (Song, 2006;Wang et al., 2011;Hu et al., 2013a;Meylan et al., 2015). CO 2 is a highly thermodynamically stable molecule (ΔG° −394 kJ/mol). Thus, a high energy input is normally required for CO 2 conversion, especially when CO 2 is the solo reactant. However, such an energy demand could be much reduced and CO 2 conversion would become thermodynamically easier when one or more molecules with high Gibbs free energy such as hydrogen (ΔG° 0 kJ/mol) could co-react with CO 2 (Song, 2006). The synthesis of oxygenates (Jiang et al., 2015;Sun et al., 2015;Prašnikar et al., 2019;Prašnikar et al., 2021) [such as methanol over Cu-based catalysts (Huš et al., 2017;Prašnikar et al., 2019;Prašnikar et al., 2021)] and hydrocarbons (Willauer et al., 2013;Satthawong et al., 2015;Zhang et al., 2015) from CO 2 and H 2 , where CO 2 is captured from flue gas and H 2 is produced using renewable energy sources (e.g., such as solar or wind) (Pontzen et al., 2011), is one promising approach for sustainable production of chemicals and fuels. Among them, C 2 -C 4 light olefins are primary building blocks for various chemicals and polymers. Therefore, developing effective catalysts for selective conversion of CO 2 to light olefins is highly desired (Riedel et al., 1999;Song, 2002;Song, 2006;Centi and Perathoner, 2009;Wang et al., 2011).
Whether hydrocarbons from CO 2 hydrogenation are through a direct or indirect mechanism is under debate. Many believe in the indirect mechanism because CO 2 hydrogenation is considered as a modified version of FTS, where CO 2 is converted to hydrocarbons through a process involving two steps: CO 2 is first reduced to CO (which is the so-called RWGS reaction) and the formed CO is then converted to hydrocarbons through FTS (Lee et al., 2009;Saeidi et al., 2014). It is also suggested by some literature that CO 2 is directly hydrogenated to methane, which contributes to the majority of hydrocarbons formation (Fujita et al., 1987;Schild et al., 1991;Wei and Jinlong, 2010).
In our previous work, we discovered a strong bimetallic Fe-Cu synergy in CO 2 conversion and C 2 + hydrocarbons formation when Cu was added to Fe at a certain composition. Adding Cu to Fe enhanced the formation of CO and C 2 + hydrocarbons and suppressed CH 4 formation. The production of C 2 -C 4 olefin-rich hydrocarbons was promoted dramatically when K was incorporated into the Fe-Cu bimetallic catalysts, in which, the C 2 -C 4 olefin to paraffin ratio reached as much as 5.2 (Wang et al., 2018). The DFT simulation suggested that the addition of Cu dramatically shifted the reaction pathway. On mono-metallic Cu and Fe catalysts, the main pathway included the formation of CO and subsequent CO hydrogenation to CH* species (Riedel et al., 2001;Satthawong et al., 2015). In contrast, over the Fe-Cu bimetallic catalysts, a new and completely different reaction pathway was favored, which did not go through CO as the intermediate (Nie et al., 2016;Nie et al., 2017). As a result, more C 2 + hydrocarbons were produced, exhibiting a strong Fe-Cu synergetic effect.
To better understand such Fe-Cu synergetic effects observed in our previous work, in this study, we performed a detailed catalyst characterization and investigated the relationship between the physical structure and catalytic performance of Fe-Cu catalysts in CO 2 hydrogenation to C 2 + hydrocarbons. A series of fresh, reduced, and spent Fe-Cu catalysts with different Cu/Fe atomic ratios were investigated by various characterization techniques including H 2 temperature-programmed reduction (H 2 -TPR), transmission electron microscopy (TEM), scanning transmission electron microscopy/energy dispersive X-ray spectroscopy (STEM/ EDS), Fe K-edge X-ray absorption spectroscopy (XAS), in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS), and temperature-programmed desorption of hydrogen and carbon dioxide (H 2 -TPD and CO 2 -TPD). In combination with the catalytic reaction data, we systematically examined the mechanistic roles of Fe-Cu bimetallic composition and K promoter, and their influences on CO 2 conversion and the formation of olefin-rich C 2 + hydrocarbons.

Catalyst Preparation
A pore-filling incipient wetness impregnation (IWI) method was used to prepare Fe, Cu monometallic, and Fe-Cu bimetallic catalysts by adding an aqueous solution of Fe(NO 3 ) 3 ·9H 2 O (Aldrich, 99.99%) and Cu(NO 3 ) 2 ·2.5H 2 O (Alfa Aesar ≥98%) onto the γ-alumina (Sasol PURALOX TH 100/150) support. The total metal loading (Fe + Cu) was kept at 15 wt% (based on the support weight). To obtain the desired Cu/(Fe + Cu) atomic ratios, the Fe and Cu concentrations were adjusted. After impregnation, the catalysts were dried at 383 K for 3 h and then calcined in an electric furnace at 673 K for 2 h. The detailed procedure was reported elsewhere (Wang et al., 2018). The K-promoted catalysts were prepared by using a sequential impregnation in two steps. First, K was impregnated onto the alumina support using K 2 CO 3 aqueous solution, which was dried at 383 K overnight. Then Fe and Cu were co-impregnated on the K-loaded alumina support using the mixed Fe and Cu aqueous solution, followed by drying and calcination under the same conditions. The Cu/(Fe + Cu) atomic ratios varied in the range of 0.0-1.0. The catalysts are termed as Fe-Cu(X)/K(Y)/Al 2 O 3 , where X and Y represent the Cu/(Fe + Cu) and K/Fe atomic ratios, respectively.
0.2 g of catalyst mixed with amorphous SiO 2 (75-250 μm particle size, about 0.5 g) was loaded into a stainless-steel tube, which had an internal diameter of 6 mm (The catalyst bed volume was about 1.2 cm 3 ). The catalyst was pretreated at 673 K with H 2 (50 ml min -1 ) for 2 h. After cooling down to 573 K, the feed gas, CO 2 /H 2 /Ar (24/72/4, vol%, 99.995%) was then introduced and the system pressure was increased to 1.1 MPa (GHSV 3600 ml g -cat -1 h -1 ). The reactor system was maintained at 573 K for the catalytic reaction. The gas flow rate and reactor pressure were regulated with mass flow controllers and a backpressure regulator, respectively. An Agilent 3000 Micro-GC with TCD was used for the online analysis of gas products including Ar, CO, CH 4 , and CO 2 while the gas-phase hydrocarbon products were analyzed online with SRI 8610C GC-FID.

Catalyst Passivation
After CO 2 hydrogenation at 573 K and 1.1 MPa for 20 h, the catalyst was cooled down to 298 K under the feed gas and subsequently passivated with 1 vol% O 2 /He (purity >99.999%) at a flow rate of 30 ml min -1 . The area of O 2 peak in the effluent measured by the Agilent Micro-GC-TCD was used as an indicator for the catalyst passivation. When the O 2 peak area became constant, the catalyst passivation was complete. Then the catalyst was collected for characterization, which hereafter is denoted as spent catalyst. Freshly reduced catalysts after the typical H 2 -pretreatment procedure were also preserved through the same passivating procedure, which are denoted as reduced catalysts.

Catalyst Characterization
Hydrogen Temperature-Programmed Reduction Hydrogen temperature-programmed reduction (H 2 -TPR) was carried out at ambient pressure using a Micromeritics AutoChem 2910 instrument. About 100 mg of catalyst was loaded in a U-shaped quartz tube reactor (held by quartz wool). To remove the adsorbed species on the surface, the catalyst was pretreated at 393 K (10 K min -1 ) under Ar (purity >99.999%, 25 ml min -1 ) for 1 h prior to reduction. After being cooled down to 323 K, the carrier gas was subsequently switched to 10 vol% H 2 /Ar at 20 ml min -1 and sustained throughout the analysis. After a stable baseline was established, the temperature program was started by increasing the temperature to 1173 K at a ramp of 10 K min -1 and then maintained at 1173 K for 30 min. For the quantitative analysis of the H 2 consumption, the TCD signal was calibrated with the standard Ag 2 O powder (100%).

Fe K-Edge X-Ray Absorption Spectroscopy
Fe K-edge X-ray absorption spectroscopy (XAS) was performed at the insertion-device beamline of the Materials Research Collaborative Access Team (MRCAT) at the Advanced Photon Source (APS), Argonne National Laboratory. In brief, the X-ray passed through a cryogenically cooled double-crystal Si (111) monochromator in conjunction with an uncoated glass mirror to minimize the presence of higher harmonics, and focused onto the sample. The XAS data were collected in a transmission mode using I 0 and I ionization chambers filled with N 2 /He and 20% Ar/ N 2 , respectively.
In situ measurements were conducted using a continuous-flow XAS reactor cell (18 in. long, 0.75 in. diam.). Calcined Fe/Al 2 O 3 and Fe-Cu/Al 2 O 3 catalyst samples were pressed into a cylindrical holder with a thickness chosen to give a total absorbance (μ t ) at the Fe K-edge of about 2.0 and an edge step (Δμ t ) of ca. 0.5, and then reduced in a stream of H 2 at 673 K for 30 min followed by CO 2 /H 2 at 573 K for 30 min. After treatment, X-ray absorption spectra at Fe K-edge were obtained at ambient temperature. The spectrum of the Fe foil was acquired simultaneously with those of Fe-based catalysts for energy calibration. The obtained Fe K-edge XAS was analyzed in a conventional manner including background subtraction and normalization followed by Fourier filtering. The X-ray absorption near-edge structure (XANES) and extended X-ray absorption fine structure (EXAFS) data were recorded near the K-edge of Fe and treated using the Athena software package.

High Resolution Transmission Electron Microscopy
The reduced Fe-Cu catalysts were analyzed by the transmission electron microscopy (TEM, Talos, 80 kV, FEI Company). Reduced samples were first collected after the passivation and then ultrasonically dispersed in ethanol. A few droplets of ethanol suspension were dripped onto a carbon-coated copper grid followed by drying at ambient temperature. Typically, about 30 micrograms were taken for each sample. Alloys of Fe-Cu particle size histograms were generated by counting at least 200 particles.
The morphological properties and elemental distribution of the reduced Fe-Cu catalysts were examined by the scanning transmission electron microscope in conjunction with the energy-dispersive X-ray spectroscopy (STEM/EDS) using the FEI Titan 60-300 microscope equipped with a spherical aberration corrector on the probe-forming lens at an accelerating voltage of 300 kV with a monochromatic beam and spherical aberration correction, providing sub-angstrom imaging resolution. EDS maps were acquired using a highangle annular dark-field (HAADF) detector. A collection angle of 51-300 mrad was used for the HAADF-STEM imaging. A camera length of 115 mm, beam current of 45 pA, and beam convergence of 30 mrad were used for STEM image acquisition. The sample preparation procedure was the same as that for the TEM analysis. Noteworthy, an Au-coated molybdenum grid, with Mo washer and clip was applied instead of the traditional copper grid to avoid the interference of Cu signal from the copper grid in both mapping and quantifying.

In situ Diffuse Reflectance Infrared Fourier Transform Spectroscopy
The possible adsorbed surface species during CO 2 hydrogenation over the mono-and bi-metallic Fe-Cu catalysts were investigated by in situ DRIFTS over a Nicolet Nexus 470 spectrometer equipped with a smart collector environmental chamber. About 20 mg of sample was mounted on the chamber for each measurement. The sample was initially purged by N 2 (50 ml min -1 ) at 373 K for 1 h, followed by the reduction in a UHP H 2 flow (50 ml min -1 ) at 523 K (ramping rate, 10 K min -1 ) for 1 h. The sample was purged again by N 2 at this temperature for 30 min. An interferogram was then collected at this condition and used as the background. After that, the CO 2 /H 2 /Ar (24/72/4, vol%) mixture gas was introduced at a flow rate of 50 ml min -1 to initiate the reaction. The IR spectra were then recorded at the reaction time of 10, 20, and 30 min, respectively. The spectra were acquired by 128 scans at a resolution of 4 cm -1 . OMNIC ver. 7 software was used for data collection and processing.
It should be pointed out that our target is to differentiate the formed and adsorbed species over the catalyst surface during the reaction through this series of experiments, rather than the change of the catalyst in the conditions of H 2 /N 2 purge. Thus, the interferogram of the catalyst sample itself was used as the background instead of KBr.

Temperature-Programmed Desorption
Micromeritics AutoChem 2910 was used to perform the temperature-programmed desorption (TPD) experiments at ambient pressure. For H 2 -TPD, typically, about 150 mg of the calcined catalyst was placed in a U-shaped quartz tube reactor and reduced at 673 K (5 K min -1 ) under a 50 ml min -1 H 2 flow (purity >99.999%) for 2 h. The reactor was then cooled down to 200 K under H 2 flow using an isopropanol-liquid-N 2 mixture to prevent desorption of weakly adsorbed hydrogen. To remove H 2 gas remained within the reactor and the gas line, the system was then purged with 0.95 vol% Ar/He flow (purity >99.999%, 30 ml min -1 ) to remove residual gaseous H 2 . When the H 2 pressure (observed from an on-line mass spectrometer) became constant, the H 2 -TPD experiment was performed by raising the reactor temperature from ∼280 to 1173 K at a ramp of 10 K min -1 under 0.95 vol% Ar/He (purity >99.999%) flow (30 ml min -1 ). Here, Ar was used as an internal standard for the quantitative analysis of the effluent gas detected by an on-line mass spectroscopy (AMETEK Dycor Dymaxion Mass Spectrometer DM200M).
In CO 2 -TPD, after the catalyst was reduced at 673 K, the temperature was reduced to 573 K under H 2 flow, and then the catalyst was purged with 0.95 vol% Ar/He (purity >99.999%, 30 ml min -1 ) for 30 min. CO 2 adsorption was then performed at 573 K by introducing 12 vol% CO 2 /He mixture gas at a flow rate of 30 ml min -1 for 1 h, followed by cooling down to room temperature. The catalyst was purged with 0.95 vol% Ar/He (purity >99.999%, 30 ml min -1 ) again to remove any trace of gaseous CO 2 until CO 2 pressure became constant as indicated by the mass spectrometer. The CO 2 -TPD experiment was carried out using the same procedure as that for H 2 -TPD.

Reducibility of Supported Fe-Cu Catalysts
The H 2 -TPR profiles of the calcined Fe-Cu/Al 2 O 3 catalysts with different Cu/(Fe + Cu) atomic ratios are shown in Figure 1. A major reduction peak was observed at 604 K for the calcined Fe/ Al 2 O 3 catalyst along with two smaller peaks at 735 and 961 K. The reduction of iron oxide usually involves three sequential steps of Fe 2 O 3 to Fe 3 O 4 , Fe 3 O 4 to FeO, and FeO to metallic Fe (Lögdberg et al., 2009;Jozwiak et al., 2007), occurring at different temperatures. The reduction temperatures can also vary with the particle size, morphology, and the interaction between iron oxide phase and the alumina support. Nonetheless, the two smaller peaks at higher temperatures (735 and 961 K) and its low degree of reduction (41%, Figure 2) indicates the strong interaction between iron oxide and Al 2 O 3 support in the calcined Fe catalyst, resulting in the formation of large amount of reduction-resistant Fe-Al-O species (Lögdberg et al., 2009).
With the addition of Cu, all the reduction peaks of Fe-Cu(X)/ Al 2 O 3 catalysts shifted to lower temperature. The more Cu amount, the lower reduction temperature. It suggests that the addition of Cu made the reduction of iron oxide easier. It is worth mentioning that the reduction peaks of Fe-Cu(0.75)/Al 2 O 3 sample located at lower temperatures than those for the Cu/ Al 2 O 3 sample, indicating that the presence of Fe also promoted the reduction of Cu species. Furthermore, the first reduction peak at 453-547 K for the Fe-Cu(X)/Al 2 O 3 samples involved the reduction of both Fe and Cu oxide species, which may form Fe-Cu alloy after reduction. No definitive peak corresponding to the Fe-Cu alloy was observed because its amount was low or it was overlapped with the reduction peak of Fe and Cu.
Based on H 2 consumption, we calculated the degree of reduction (DR) for all the samples. The DR of the supported metal oxides is defined as the ratio of the actual measured H 2 consumption to the total theoretical H 2 consumption based on the composition of metal oxides. The actual measured H 2 consumption was calculated by integrating the TPR profile, while the theoretical H 2 consumption was the amount of H 2 required for the reduction of the supported Fe and Cu oxides toward their metallic states according to the stoichiometry of the equations shown below (Eqs 1, 2).
(2) Figure 2 shows the DR as a function of the Cu/(Fe + Cu) atomic ratio. The DR increased with the increase of Cu content in the catalyst (the blue line). Because the Cu/Al 2 O 3 catalyst itself had a high DR of 98.8%, the increased DR could be attributed to Cu alone and/or the synergistic effect of Fe and Cu. Assuming there was no interaction between Fe and Cu, we could calculate the DR for each Fe-Cu(X)/Al 2 O 3 sample by adding their corresponding DRs in proportion to their composition (the black line). Obviously, all the measured DRs of Fe-Cu(X)/ Al 2 O 3 samples were higher than the sum of the individual DR when using the Fe/Al 2 O 3 and Cu/Al 2 O 3 separately, confirming the Fe-Cu synergy when combining Fe and Cu in the Fe-Cu(X)/ Al 2 O 3 catalyst.
To further compare the Fe-Cu synergy among the Fe-Cu(X)/ Al 2 O 3 samples, we used the DR difference (ΔDR) as an indicator to the Fe-Cu synergy, which was calculated by subtracting the summed DR value (the black line) from the measured DR value (the blue line), which is also plotted in Figure 2 (the red line). As seen, the Fe-Cu synergy was almost the same for the Fe-Cu(0.05)/ Al 2 O 3 and Fe-Cu(0.17)/Al 2 O 3 samples and showed the highest among all samples. On further increasing the Cu content, the content of Fe decreased as the total metal loading was fixed. The DR contribution from Cu became larger, whereas the DR difference decreased. As a result, a decreasing trend in the Fe-Cu synergy was obtained. Figure 3 shows the XAS data for the reduced Fe/Al 2 O 3 and Fe-Cu/Al 2 O 3 catalysts with the Fe 2 O 3 , Fe 3 O 4 , and Fe foil as the reference compounds. The Fe K-edges of the reduced Fe/Al 2 O 3 and Fe-Cu/Al 2 O 3 catalysts were observed at 7,133.6 and 7,133.0 eV, respectively, while the K-edge value for Fe foil was 7,112.0 eV. The adsorption energies of the Fe 2 O 3 and Fe 3 O 4 were at 7,134.6 and 7,133.5 eV, respectively, which reflect electronic transitions from 1s orbitals to unoccupied p-d hybrid molecular orbitals (Li et al., 2001). The XANES spectra ( Figure 3A) of the reduced Fe/ Al 2 O 3 and Fe-Cu/Al 2 O 3 catalysts were similar to that of the Fe 3 O 4 sample, especially the pre-edge peaks of the reduced Fe/Al 2 O 3 and Fe-Cu/Al 2 O 3 catalysts as shown in Supplementary Figure S1. The results reveal that iron oxide was not reduced to Fe 0 under the conditions used for the XAS analysis.

Fe K-Edge XAS
As shown in Supplementary Figure S1, the peak in the preedge region of the reduced Fe-Cu(0.17)/Al 2 O 3 was broadened relative to that of the reduced Fe/Al 2 O 3 catalyst at 7,114.4 eV. The peak in the pre-edge region of Fe-Cu(0.17)/Al 2 O 3 shifted to lower energy, while there was also an increase in intensity at higher energy. This shift suggests that there was a small amount of metallic Fe in the reduced Fe-Cu(0.17)/Al 2 O 3 catalyst. Similar changes in the pre-edge region occurred for reduced Fe-Cu(0.5)/ Al 2 O 3 , but with less contribution from metallic Fe. It is consistent with the catalyst composition where the Fe-Cu(0.17) sample contains more Fe content than Fe-Cu(0.5) sample.
Similar to the XANES profiles, the EXAFS spectra in Figure 3B also suggests that the reduced Fe/Al 2 O 3 catalyst was more like Fe 3 O 4 . It had an Fe-O peak (phase uncorrected distance) at 1.47 Å, and an Fe-O-Fe scattering peak at 2.82 Å (phase uncorrected distance). Over the reduced Fe-Cu(0.17)/ Al 2 O 3 catalyst, besides the Fe-O and Fe-O-Fe(Cu) distances, a new distance of 2.36 Å emerged, which can be assigned to the metallic Fe-Fe(Cu) distance. It is a clear evidence showing that the addition of Cu increased the reduction degree of Fe in the Fe-Cu bimetallic catalyst. Similarly, the radial distances from the reduced Fe-Cu(0.5)/Al 2 O 3 catalyst were not much different from the reduced Fe-Cu(0.17)/Al 2 O 3 catalyst, implying that the types of the reduced elements are similar, but the amounts are different likely due to the difference in the Fe-Cu compositions. Due to the low contrast between the alumina support and Fe-Cu elements, the particles were hardly discerned directly by eyes. We had to use the elemental analysis to confirm the distribution of the particles, which were widely dispersed over the support as shown in Supplementary Figure S2. The average particle size  Frontiers in Chemical Engineering | www.frontiersin.org October 2021 | Volume 3 | Article 708014 6 mainly lied in the range of 4-7 nm ( Figure 4D). The lattice space was measured to be 2.01 Å and 2.53 Å ( Figure 4C), which corresponds to FeCu alloy (110) and Fe 3 O 4 (311) (Omar, 2016;Wei et al., 2017), respectively. The alloy lattice space matches the d-spacing of the FeCu alloy from the XRD results reported in our prior work (Wang et al., 2018). It should be noted that the d-spacing values for the Fe-Cu alloys are similar. Additionally, according to the binary phase diagrams of Fe and Cu, there is no specific ratio of Fe and Cu in alloys. Thus, it is impossible to distinguish them both in XRD and TEM.

Morphological Properties of Fe and Cu Species and Element Distribution
The presence of Fe-Cu alloy particles was also confirmed by Z-contrast in the high-angle annular dark-field image (HAADF) of the reduced Fe-Cu(0.17)/Al 2 O 3 catalyst as shown in Figure 5. The corresponding EDS mapping images for Al, O, Fe, and Cu elements are shown in Supplementary Figure S2. Both Al and O species distributed uniformly across the sample island, while Fe and Cu species existed in close proximity but at different relative concentrations. Evidently, Fe and Cu were well-dispersed, which is consistent with the observations from the TEM images. Comparing the Fe and Cu distributions, we observed that most Fe and Cu elements were located at the same places, especially at the large Fe areas, indicating the combination of Fe and Cu. It was further confirmed by measuring the d-spacing. The magnified inset in Figure 5A shows a representative particle with a relatively clear lattice fringe alignment. The measured average d-spacing was ca. 2.01 Å, which is in accordance with the results from both TEM images and XRD patterns (Wang et al., 2018). Thus, these particles are representative Fe-Cu alloy particles.
To confirm the alloy formation, these eight representative alloy particles (with the d-spacing 2.01 Å) in the HADDF and EDS images (the same area of TEM images) were scanned by the energy-dispersive X-ray spectroscopy mapping technique in STEM mode (STEM/EDS). This technique allows for the identification of the elemental compositions of individual bimetallic nanoparticles on the catalyst surface. Their compositions are listed in Table 1. The Cu/(Fe + Cu) ratios were in the range of 0.17-0.48, more centered between 0.2 and 0.3 with the mean at 0.29. It is consistent with the best Cu/(Fe + Cu) ratio for the Fe-Cu alloy composition suggested by the DFT simulation (Nie et al., 2016;Nie et al., 2017). Therefore, the STEM/EDS results demonstrated a uniform distribution of Fe and Cu in the nanosized alloy particles after reduction, which is in good agreement with the XRD results over the same sample (Wang et al., 2018).

Surface Adsorbed Species
To investigate the surface species on the bimetallic Fe-Cu catalysts during CO 2 hydrogenation, in situ diffuse reflectance FT-IR spectra were recorded after the CO 2 hydrogenation reaction was initiated at 523 K and atmospheric pressure for 30 min. Before the reaction, all samples were reduced with pure  H 2 at 523 K for 1 h. The obtained DRIFT spectra are shown in Figure 6. As seen, IR bands at 1,377, 1,634, 1,760, 1,942, and 2,120 cm -1 were observed within the range of 1,200-2,200 cm -1 over the Fe-Cu(0.17) catalyst, while the Fe-Cu(0.5) sample displayed the IR bands at 1,525, 1,646, 1,760, and 1942 cm -1 . The bands at 1,634 and 1,646 cm -1 can be attributed to the adsorbed bicarbonates with different vibrational modes (Busca and Lorenzelli, 1982;Rethwisch and Dumesic, 1986;Gopal et al., 1987), while the 1,525 cm -1 IR band is the characteristic of carbonates (Peri, 1966;Parkyns, 1971;Morterra and Orio, 1990;Turek et al., 1992;Morterra and Magnacca, 1996). These IR bands confirm the formation of adsorbed bicarbonate and carbonate species on both bimetallic Fe-Cu catalysts during CO 2 hydrogenation. The 1,942 and 2,120 cm -1 bands can be attributed to the adsorbed CO species (Bando et al., 1997;Hadjiivanov et al., 2001), which were not observed on either Cu/Al 2 O 3 or Fe/Al 2 O 3 catalysts. It indicates that the combination of Fe and Cu promoted the RWGS reaction, generating more adsorbed CO species. Although Cu/Al 2 O 3 can promote the RWGS reaction, the formed CO species are quickly released into the gas phase due to the weak interaction between Cu and CO. As a result, such two bands were not observed over the Cu/Al 2 O 3 catalyst.
Besides, the IR bands at 1,377, 2,854, and 2,932 cm -1 represent different vibrational modes of adsorbed formate species (Millar et al., 1991;Clarke and Bell, 1995;Cabilla, 2003) and the 1760 cm -1 IR band is the characteristic of formic acid (Shimanouchi, 1972). The observation of the adsorbed formate and formic acid species provides an evidence for one potential CO 2 hydrogenation pathway on the bimetallic Fe-Cu/Al 2 O 3 catalysts that the reaction could go through a direct hydrogenation via formate, then formic acid as main intermediates to C 2 + hydrocarbons, which was also proposed on the basis of DFT simulation (Nie et al., 2016;Nie et al., 2017).
The intensities of these IR bands on the Fe-Cu(0.17) catalyst were much stronger than those of the Fe-Cu(0.5) catalyst, indicating that the Fe-Cu synergy in the Fe-Cu(0.17) catalyst is stronger than that in the Fe-Cu(0.5) catalyst in terms of CO 2 hydrogenation. It is in good agreement with H 2 -TPR results shown in Figure 2. On the contrast, the Fe/Al 2 O 3 catalyst only showed a broad IR peak at 1,300-1,700 cm -1 along with a band at ∼2,932 cm -1 .
Besides the IR spectra after 30 min of CO 2 hydrogenation reaction, we also recorded the DRIFT spectra at different reaction time including 10, 20, and 30 min over the Fe-Cu(0.17)/Al 2 O 3 catalyst, which is shown in Supplementary Figure S3. No change was detected in these spectra, revealing that the adsorbed species are stable.  Figure S4. The H 2 desorption on the reduced Fe/Al 2 O 3 catalyst mainly occurred at 800 K with several minor desorption peaks at  Compared to the Fe/Al 2 O 3 catalyst, the combination of Fe and a small amount of Cu (X 0.17) enhanced the H 2 desorption at 800 K. Upon further increasing the Cu content to X 0.5, however, the H 2 desorption peak at 800 K became smaller. On the other hand, the desorption peak at 380-600 K was much promoted with the increase of Cu amount. Interestingly, the H 2 -TPD profile of the reduced Cu/Al 2 O 3 catalyst was significantly different from those of the Fe-based catalysts, but close to that of the Al 2 O 3 support, suggesting that Cu has low ability in H 2 adsorption. The H 2 desorption in the range of 600-900 K on the Cu catalyst was negligible.

H 2 Adsorption Properties
To quantitatively analyze the H 2 adsorption amount, the H 2 -TPD peaks were deconvoluted using 50% Gaussian plus 50% Lorentzian functions. As shown in Figure 8 with the reduced Fe-Cu(0.17)/Al 2 O 3 catalyst as a typical example, the final curve generated from the summation of the deconvoluted peaks fitted the experimental data well. The deconvoluted peaks were arbitrarily denoted as I-IV from low to high temperature, representing four different types of adsorbed hydrogen species on the catalyst surface. Types I and II represent the weakly adsorbed H 2 at lower temperatures (below 600 K), which can be ascribed to atomic hydrogen adsorbed on the metal surface (Bozso et al., 1977). While the H 2 desorption observed at higher temperatures (>600 K, type III as moderately adsorbed H 2 and type IV as strongly adsorbed H 2 ) are the H 2 species adsorbed on the metallic Fe 0 and/or Fe-Cu alloy surface decorated with unreduced metal oxide (Zowtiak and Bartholomew, 1983;Bartholomew, 1990).
Based on the peak deconvolution, we were able to quantitatively estimate the amount of different types of adsorbed H 2 by integrating each peak, which are listed in Supplementary Table S1 for all Fe-Cu(X)/Al 2 O 3 samples. Figure 9A shows the amounts of type III + IV adsorbed H 2 on the Fe-Cu(X)/Al 2 O 3 catalysts. Adding a small amount of Cu greatly promoted the amount of type III + IV adsorbed H 2 , especially at the Cu/(Fe + Cu) atomic ratio of 0.10 and 0.17, the amount of type III + IV adsorbed H 2 of which was 238.1 and 240.0 μmol g -cat -1 , respectively, about 50% higher than that of the Fe/Al 2 O 3 catalyst. Further increasing Cu content, however, resulted in the decrease of type III + IV adsorbed H 2 . When the Cu/(Fe + Cu) ratio was 0.75, the amount of type III + IV adsorbed H 2 became 113.5 μmol g -cat -1 , which was even lower than that of the Fe/Al 2 O 3 catalyst. Such a trend is consistent with the change in the Fe-Cu synergy characterized by the degree of reduction of the Fe-Cu catalyst with the addition of Cu shown in Figure 2.
We then correlated the space-time yield (STY) of C 2 + hydrocarbons with each type of adsorbed H 2 species. The amount of moderately plus strongly adsorbed H 2 (type III + IV) showed a good relationship with C 2 + hydrocarbons production ( Figure 9B). Such a strong correlation evidently  Frontiers in Chemical Engineering | www.frontiersin.org October 2021 | Volume 3 | Article 708014 9 displays that the moderately and strongly adsorbed H 2 species played an important role in the formation of C 2 + hydrocarbons. In other words, the Fe-Cu synergy in CO 2 hydrogenation could be attributed to the enhancement in H 2 adsorption induced by the combination of Fe and Cu.

CO 2 Adsorption Properties
The effect of Fe-Cu combination on the CO 2 adsorption properties of the reduced Fe-Cu(X)/Al 2 O 3 catalysts was examined by CO 2 -TPD, which is presented in Figure 10. The γ-Al 2 O 3 support was also evaluated as a reference. It showed a CO 2 desorption peak at 350 K, which is known as the decomposition of the surface bicarbonate species (Pan et al., 2008). Over the monometallic Fe catalyst, another strong CO 2 desorption peak was observed at around 480 K, which can be ascribed to the decomposition of the surface carbonate species (Pan et al., 2008). With the addition of Cu, another new small CO 2 desorption peak at 550-700 K emerged over the Fe-Cu bimetallic catalysts, which can be referred to the bidentate carbonate species (Pan et al., 2008). The observed bicarbonate and carbonate species are in a good agreement with the DRIFTS results. In contrast, the monometallic Cu catalyst exhibited a similar profile to the γ-Al 2 O 3 support, indicating that Cu itself does not contribute to CO 2 adsorption. Therefore, the new peak induced by the addition of Cu could be attributed to the Fe-Cu strong interaction and/or Fe-Cu alloys. Correlating to CO 2 hydrogenation performance, it also suggests that the observed Fe-Cu synergy is because the combination of Fe and Cu enhanced CO 2 adsorption.
It is projected that the addition of K could greatly enhance CO 2 adsorption due to its basic nature. Consequently, we also examined the CO 2 desorption behavior of K-promoted catalysts, i.e., the Fe-Cu(0.17)/K(Y)/Al 2 O 3 catalysts. As shown in Figure 11A, compared to the Fe-Cu(0.17) catalyst, the addition of K increased all types of the adsorbed CO 2 species with desorption peaks at around 350 K, 450 K, and 550-700 K. Additionally, a new desorption peak at higher temperatures (>700 K) emerged, which could be ascribed to the adsorbed CO 2 on potassium ferrite (K 2 Fe 2 O 4 ) (Zowtiak and Bartholomew, 1983). Similar desorption peak after K addition was also observed on the Fe-Zn-K catalysts (Zhang et al., 2015).
To find out the relationship between the adsorbed CO 2 species and CO 2 hydrogenation reaction, the desorption profiles of the Fe-Cu(0.17)/K(Y)/Al 2 O 3 catalysts were deconvoluted and quantitatively analyzed. Using the Fe-Cu(0.17)/K(0.3)/Al 2 O 3 catalyst as an example ( Figure 11B), four peaks denoted as types I-IV were identified from low to high temperatures, suggesting four different CO 2 adsorption states on the catalyst surface. The quantities of deconvoluted peaks for all CO 2 -TPD profiles are summarized in Supplementary Table S2.
As described above, the moderately adsorbed CO 2 (Type III at 550-700 K) is related to the Fe-Cu interaction. The addition of K changed the moderately adsorbed CO 2 , suggesting that K also interacted with Fe and Cu. In addition, it is responsible for the emergence of strongly adsorbed CO 2 (Type IV at 600-900 K), as no such peak was observed over the corresponding catalyst without K addition. It is mainly because K could increase the surface basicity by donating the electron to the Fe-based catalyst (i.e., acting as an electronic promoter) (Dorner et al., 2010). Thus, we correlated the STY of C 2 + hydrocarbons with each type of adsorbed CO 2 . It was found that the C 2 + hydrocarbons production rate is dependent on the amount of moderately plus strongly adsorbed CO 2 (type III + IV) (Supplementary Figure S6). It suggests that the moderately and strongly adsorbed CO 2 may be responsible for the formation of C 2 + hydrocarbons. It also means the Fe-Cu synergy comes from the Fe-Cu strong interaction, which enhances CO 2 adsorption (both amount and strength) on the catalyst surface.

Effect of H 2 and CO 2 Adsorption Properties on Olefin Formation
In CO 2 hydrogenation, we observed that the olefin selectivity increased significantly with the addition of K, resulting in an increased O/P (olefin/paraffin) ratio, which may be related to the change in the H 2 and CO 2 adsorption properties of the catalysts with K. Consequently, the H 2 -TPD experiments were also performed over the Fe-Cu(0.17)/K(Y)/Al 2 O 3 catalysts, and the results are presented in Supplementary Figure S5. The quantitative results via peak deconvolution are listed in Supplementary Table S1. Adding a small amount of K (K/Fe 0.3) enhanced H 2 adsorption on Fe-Cu(0.17) catalyst, especially the weakly adsorbed H 2 , due likely to the electronic enhancement of K to metals. However, further increasing K content (K/Fe 0.5 and 1.0) resulted in the decrease of the moderately and strongly adsorbed H 2 , due likely to that overloaded K could cover the metal surface for H 2 adsorption.
As discussed in the H 2 adsorption properties and CO 2 adsorption properties sections, both the moderately and strongly adsorbed H 2 species (Type III + IV) and the moderately and strongly adsorbed CO 2 (Type III + IV) showed good relationship to the formation of C 2 + hydrocarbons. On the other hand, in CO 2 hydrogenation, the production of paraffins requires more H 2 than olefins. Hence we used the ratio of the amount of the moderately and strongly adsorbed CO 2 over the amount of the moderately and strongly adsorbed H 2 to correlate with the olefin/paraffin ratio (O/P) obtained from CO 2 hydrogenation over the Fe-Cu(0.17)/K(Y)/ Al 2 O 3 catalysts. The result is presented in Figure 12. FIGURE 12 | Relationship between the olefin/paraffin ratio and the ratio of the amount of type III + IV adsorbed CO 2 over the amount of type III + IV adsorbed H 2 .
Frontiers in Chemical Engineering | www.frontiersin.org October 2021 | Volume 3 | Article 708014 Here we observed a clear and very strong relationship between the O/P ratio and the ratio of the amount of the moderately and strongly adsorbed CO 2 over the amount of the moderately and strongly adsorbed H 2 with the R-square value of more than 0.99. It reflects that both the moderately and strongly adsorbed CO 2 and H 2 played a critical role in determining the olefin selectivity in CO 2 hydrogenation. In other words, it is possible to obtain more olefins than paraffins by controlling the ratio of the moderately and strongly adsorbed CO 2 over the moderately and strongly adsorbed H 2 when we design a proper catalyst for CO 2 hydrogenation to higher hydrocarbons.
When we compared the Fe-Cu catalyst with relevant studies of Fe-based catalyst, we found that this CO 2 /H 2 adsorption behavior is the key to light olefin formation. For instance, Zhang et al. reported a promotion of Fe-Zn-K catalyst on CO 2 conversion, but they reported Zn addition had almost no effect on light olefin selectivity. And they also found that adding Zn decreased CO 2 uptake (Zhang et al., 2015). These results proved that Zn acted more like a structural promoter that can help Fe to be better dispersed and reduced. It is not an actual synergetic effect. We will further discuss the importance of light olefin formation and its role in the whole CO 2 hydrogenation process in next section.
The comprehensive investigation on the relationship between the adsorbed CO 2 and H 2 species and the catalyst activity and product selectivity shall thus be conducted to improve our fundamental understanding of CO 2 hydrogenation to higher hydrocarbons over the Fe-Cu bimetallic catalysts.

Pathways of CO 2 Hydrogenation
To investigate the pathways of CO 2 hydrogenation, the Fe-Cu(0.17)/K(0.3)/Al 2 O 3 catalyst was tested over a wide range of residence time (W/F) at 573 K and 1.1 MPa. The CO 2 conversion as a function of W/F is shown in Supplementary Figure S7. Clearly, the CO 2 conversion increased with increasing the residence time and gradually became constant at high W/F values. Figure 13 shows the CO, CH 4 , and C 2 + hydrocarbons selectivities as a function of CO 2 conversion. With the increase of CO 2 conversion, the CO selectivity decreased while the selectivity of C 2 + hydrocarbons increased, indicating that CO is a primary product while C 2 + hydrocarbons are produced from CO, probably via a process similar to FTS (Riedel et al., 2001). It is noticed that at the CO 2 conversion less than 10%, the CO selectivity, CH 4 selectivity, and C 2 + hydrocarbons selectivity were almost constant at 90, 4, and 6%, respectively. It hints that small parts of CH 4 and C 2 + hydrocarbons may come from direct hydrogenation of CO 2 without through CO intermediate (Riedel et al., 2001). In addition, CH 4 selectivity did not change much with the increase of CO 2 conversion whereas the CO selectivity decreased sharply, suggesting that CH 4 formation is not closely related to CO formation.
Among the C 2 + hydrocarbons, the paraffin selectivity increased drastically with the increase of CO 2 conversion, while the olefin selectivity only increased a bit, leading to a rapid decrease in the C 2 -C 4 olefin/paraffin ratio ( Figure 14). It suggests that light olefins could be the primary product from CO 2 hydrogenation, and then are further hydrogenated to paraffins. Hence, to achieve olefin-rich hydrocarbons, it would be crucial to suppress the consequent hydrogenation of olefins by tailoring the surface coverage of CO 2 and H 2 in CO 2 hydrogenation.

CONCLUSION
This work focused on catalyst characterizations to better understand the Fe-Cu synergistic effect in CO 2 hydrogenation, where the Fe-Cu(0.17) catalyst showed the best performance. Adding Cu to Fe promoted the reduction of iron oxide to Fe 0 and the formation of Fe-Cu alloy as evidenced by H 2 -TPR, high resolution TEM, and XAS. It also influenced the adsorbed surface species during CO 2 hydrogenation as observed by in situ DRIFTS. Furthermore, the addition of Cu promoted the adsorption of the moderately and strongly adsorbed H 2 species and moderately adsorbed CO 2 species, which were responsible for the increased CO 2 conversion and C 2 + selectivity. Adding K FIGURE 14 | C 2 -C 4 olefin/paraffin ratio as a function of CO 2 conversion from CO 2 hydrogenation over the Fe-Cu(0.17)/K (0.3)/Al 2 O 3 catalyst at 573 K and 1.1 MPa.
Frontiers in Chemical Engineering | www.frontiersin.org October 2021 | Volume 3 | Article 708014 further increased the moderately and strongly adsorbed CO 2 , but slightly reduced the moderately and strongly adsorbed H 2 , resulting in an increased adsorbed-CO 2 /adsorbed-H 2 ratio and consequently producing olefin-rich hydrocarbons with a significantly increased O/P ratio. Both the in situ DRIFTS and the product distribution analysis suggested that C 2 + hydrocarbons were largely produced from consequent reduction of CO via the FT-like process, along with a small portion formed via a direct CO 2 hydrogenation pathway.

DATA AVAILABILITY STATEMENT
The original contributions presented in the study are included in the article/Supplementary Material; further inquiries can be directed to the corresponding authors.