Performance of CO₂ conversion to hydrocarbons on cobalt ferrite

Initially, cobalt ferrite precursors with different Fe/Co ratios (FC1, FC2 and FC5) were synthesized by a simple hydrothermal method and then pre-treated under the in situ reaction atmosphere (H₂/CO₂=4) in the light at 300 °C for 10 h. The activity measurements of three samples were carried out to explore the performance of the CO₂ hydrogenation to high-valued hydrocarbons with or without light. For FC2, CH₄ and C₂−C₄ hydrocarbons could be produced above 240 °C under dark. Solar C₂−C₄ hydrocarbons were produced at 200 °C and significantly increased above 260 °C. An impressive C₂−C₄ selectivity of 29.8 % with a noteworthy production rate of 1.1 mmol g⁻¹ h⁻¹ and 12.9 % CO₂ conversion efficiency were achieved at 300 °C (Figure 1a,b). In comparison with the other two samples, FC1 was inferior to FC2 with respect to the production of C₂-C₄ at the same conditions, while neither CH₄ nor C₂-C₄ hydrocarbons were detected for FC5, which also had the lowest CO production rate (Figure S2). The relating works on CO₂ hydrogenation to C₂−C₄ production are shown in Table S1, the selectivity of C₂−C₄ hydrocarbons of FC2, even conducted at atmospheric pressure in a flow mode, was comparable to those observed under higher pressures. A typical product distribution measured by GC is shown in Figure S3.

FC2 was further used for conducting stability test. It displayed excellent stability during the long-term test under ambient pressure and illumination (Figure 1c,d). FC2 was also examined at a much harsher condition with pressure (3 MPa) in a strictly iso-thermal annular gap microreactor under industrially relevant conditions. As shown in Figure S4, CH₄ and C₂−C₄ hydrocarbons could be produced with higher selectivity than that under lower pressure below 300 °C, and liquid C₅₊ species were generated when the operating temperature was raised to 350 °C. The excellent CO₂ reduction ability manifested the significant potential for industrial applications.

Identification of active sites

Co and Fe were distributed uniformly in the three fresh samples (FC1, FC2 and FC5) and the atomic ratios of Fe to Co were consistent with the trend of their original feed values (Figure S5 and Table S2). The FC2 matched well with the cubic spinel structure CoFe₂O₄ (JCPDS #22-1086) in PXRD patterns,¹³ as well as FC1 (Figure S6). FC5 had both CoFe₂O₄ and Fe₂O₃, and the later was further confirmed by the Fe−O stretching mode at 485 cm⁻¹ via ATR-FTIR spectrum (Figure S7).¹⁴ In addition, all the as-synthesized catalysts intensively absorbed light across UV/Vis-NIR region, here 250–2500 nm (Figure S8). An increased Co component resulted in a promoted optical absorption. After solar CO₂ hydrogenation reactions, CoFe alloy and CoO integrated with CoFe₂O₄ nanospinel were detected in the spent cobalt-rich FC1 (Figure S9). Both CoFe alloy and CoFe₂O₄ nanospinel were measured in the spent FC2 (Figure 2a).¹⁵ And CoFe alloy could not be detected in the spent iron-rich FC5 (Figure S9). Considering the best performance of FC2, it was reasonable to infer that heterostructured CoFe alloy/CoFe₂O₄ nanocomposite undertook a critical role in the conversion of CO₂ to hydrocarbons.

To identify the structure and composition, the TEM and HRTEM images of the spent FC2 are shown in Figure S10 and Figure 2b,c. The lattice spacings of 0.286 nm and 0.242 nm, observed on a single spent FC2 particle, corresponded to the (100) of CoFe nanoalloy and the (222) of CoFe₂O₄.¹⁶ The HRTEM results indicate the CoFe closely adhered to the CoFe₂O₄, forming the composite heterostructure. The particle size of fresh and spent FC2 showed a little shrinkage due to its in situ reduction and the formation of CoFe alloy, as shown in Figure S10.

STEM-EELS analysis shows the EEL O K-edge, Fe L-edge and Co L-edge spectra in Figure 2d,e, respectively. The L₂ (L₃) features in the metal spectra arise from 2p_1/2 (2p_3/2)→3d transitions, while the O K-edge comes from the transitions from the O 1s→2p states, which are hybridized with Fe 3d (Co 3d) orbitals due to covalent Fe−O (Co−O) bonding.¹⁷ The O K pre-peak signal was not detected at some points (from Point 1 to Point 4), further confirming the partial reduction of the cobalt ferrite and the formation of a metallic alloy phase.¹⁸ Both of Fe L₃/L₂ and Co L₃/L₂ white-line intensities were recorded across the sample (Table S3), indicating the uniform distribution of metal Fe and Co in the particle. So the metallic component was CoFe alloy instead of the single phase of Co or Fe. The STEM-EELS line scan results of a CoFe nanoalloy pre-fabricated by H₂ reducing are shown in Figure S11. There was no O pre-peak observed in the spectra, which in turn confirmed the in situ reduction of CoFe₂O₄ during the reaction. The average Fe L3/L2 and Co L3/L2 white line intensity ratios of the pre-fabricated CoFe alloy and spent FC2 are summarized in Table S4. Moreover, the element mapping results showed that Fe and Co elements were distributed uniformly in the pre-fabricated CoFe alloy (Figure S12). The atomic ratio of Co : Fe in the pre-fabricated CoFe alloy was 1 : 1.72 obtained from EDX analysis, in accordance to the ICP result of the fresh FC2 (Table S2), so we can speculate that the ratio of Co and Fe in CoFe alloy is around 1 : 1.7.

To simulate the reaction condition, chemical states of the catalyst in atmospheric H₂ or CO₂+H₂ were analyzed by in situ XPS under light irradiation. Obviously, the catalyst surface within the detection limit, around several nanometers, remained a metal oxide layer according to the Fe 2p and Co 2p spectra, however, they measured in both atmospheres showed an increased proportion of bivalent component over the time on stream (Figure 2f–h, Figure S13 and Tables S5 and S6).¹⁹ Meanwhile, in the O 1s spectra, the signal diagnostic of the lattice oxygen exhibited a decreased intensity versus time due to the removal of lattice oxygen atoms in the reductive H₂+CO₂ and H₂ atmosphere. Both of above analysis indicated the gradual in situ reduction process of FC2 on stream. The illustration of cobalt ferrite structural transformation is shown in Figure 2i.

In situ XAS was conducted both in H₂ and H₂+CO₂ atmospheres to further explore the structural evolution of the cobalt ferrite. Heating samples at 300 °C was employed to match the reaction temperature, and at 350 °C to simulate photothermal heating. In Fe and Co XANES spectra (Figure 3a,b), reduction of cobalt ferrite was indicated qualitatively by a gradual intensity increase in the pre-edge feature (ca. 7113 and 7712 eV, respectively) and an intensity decrease of the white-line feature (ca. 7130 and 7725 eV, respectively). Fe showed some initial reduction (albeit slow) at 300 °C under H₂, virtually no change was observed in the Co spectra, suggesting that the Fe sites were more active than Co sites to be reduced. As the temperature was increased to 350 °C, both Co and Fe in the oxide were transformed from cationic states to metallic form (Fe⁰ and Co⁰) to produce CoFe alloy. After CO₂ was purged into the reaction, reduction of Fe or Co was to some extent inhibited, consistent with the in situ XPS result.

In Fe and Co Fourier transform EXAFS spectra (Figure 3c,d), the peak located at ca. 1.4 Å corresponded to the nearest-neighbor metal-oxygen (M−O) bonding, i.e., Fe−O or Co−O. And the peak located at ca. 2.6 Å corresponded primarily to the metal-metal scattering interactions M(O_h)−O−M(O_h), i.e., Fe(O_h)−O−M(O_h) or Co(O_h)−O−M(O_h), where M could be Fe or Co. Another peak around 3.2 Å represented the interaction of Fe(T_d)−O−M(T_d) or Co(T_d)−O−M(T_d).²⁰ By comparing the relative strength of M(O_h)−O−M(O_h) to M(T_d)−O−M(T_d), it could conclude that Fe atoms occupied both O_h and T_d positions, and most Co atoms were located at the O_h centers, indicative of an inverse spinel structure for initial FC2.²⁰ Very little change was observed in either peak of the M−O and the M(O_h)−O−M(O_h) following reduction at 300 °C. While obvious reductions of M(T_d)−O−M(T_d) were observed in both Co and Fe spectra, therefore, the cation reductions mainly occurred at the O_h centers at first. Accordingly, more Fe cations in T_d sites than Co cations and the more sensitive capacity of Fe to construct interactions with H₂, thus, possibly resulted in the CO production at low-temperature regions. On increasing the temperature to 350 °C, M−O and M−(O_h)−O−M(O_h) peaks began to decrease in intensity, while the M−M interactions in CoFe alloy, located at ca. 2.2 Å, were rapidly increased in intensity. Consistent results were obtained through the coordination numbers (CNs) for representative scattering (Figures S14–S16 and Tables S7 and S8).

Based on the structural and electronic characterizations, how CoFe₂O₄ was reconstructed after the activation procedure was figured out. Fe³⁺ and Co²⁺ in the lattice of CoFe₂O₄ were collapsed and aggregated to form the CoFe alloy phase in the reductive atmosphere, resulting in the active CoFe/CoFe₂O₄ heterostructure. Meanwhile, the surface of the catalyst intensively interacted with the CO₂ that could prevent the complete reduction and remained oxidative state.

Synergistic effect of CoFe alloy and CoFe₂O₄ in CO₂ hydrogenation

The synergistic effects of the two components in the in situ formed CoFe/CoFe₂O₄ heterostructure should play essential roles in solar C₂−C₄ hydrocarbons production.²¹ Pure CoFe nanoalloy fabricated from CoFe₂O₄, as proven by PXRD results (Figure S17), was first investigated as a comparison. Even though CoFe alloy had better light absorption property than cobalt ferrite (Figure S18), little amounts of CO (0.085 mmol g⁻¹ h⁻¹) and CH₄ (0.002 mmol g⁻¹ h⁻¹) and no C₂−C₄ hydrocarbons were produced under light condition (Figure S19), far from the performance of CoFe/CoFe₂O₄. It was reasonable to infer pure CoFe alloy might be difficult to initiate the RWGS reaction for CO formation, thus the successive hydrogenation was inevitably limited. These results were consistent with the inferior C₂−C₄ hydrocarbons production of FC1 compared to that of FC2, because of more proportion of CoFe alloy in the spent FC1.

Obvious pieces of evidence of synergistic effect between CoFe₂O₄ and CoFe alloy were obtained from the CO₂ adsorption behaviors. When exposure CO₂ on the pristine CoFe₂O₄ at room temperature (Figure 4a), infrared fingerprint modes at 1220 and 1413, and 1620 cm⁻¹ assigned to the adsorbed CO₂ species were observed in DRIFTS.²² The intermediates gradually transformed to formates when the temperature increased to 200 °C with emerging peaks at 1520 and 1320 cm⁻¹, indicating the easy activation of CO₂ over CoFe₂O₄. In sharp contrast, no signals belonging to adsorbed species were detected over pure CoFe alloy (Figure 4b). Even introducing the reactive gases (H₂/CO₂=4), the signals of adsorbed CO₂, formate intermediates or hydrocarbon products were not observed from room temperature to 300 °C (Figure 4b). The CO₂-TPD results further confirmed above conclusions (Figure 4c). Two desorption peaks were observed for FC2 and CoFe/CoFe₂O₄, while no signals appeared for CoFe alloy. CO₂ adsorption and activation were the critical steps before subsequent hydrogenation procedures.²³ Therefore, the relatively poor CO₂ adsorption capacity resulted in the inferior CO₂ hydrogenation activity over CoFe alloy. And these results hinted that CoFe₂O₄ was an essential component for the first RWGS step, and CoFe alloy should participate in the subsequent CO hydrogenation processes.²⁴ Their synergies drove the reaction to produce C₂−C₄ under the ambient condition with light irradiation.

Reaction pathway

The in situ DRIFTS were applied to provide a molecular-level understanding of CO₂ activation over CoFe/CoFe₂O₄. In the presence of both H₂ and CO₂, the peak intensity of adsorbed CO₂ decreased with the rising temperature and disappeared around 200 °C (Figure 4d). New vibrational modes at 1354 and 1595 cm⁻¹ associated with the formation of carboxylate species emerged,^{22c, 25} implying an effective conversion of adsorbed CO₂ species in the hydrogenation process. In addition, several modes at 1372, 1384, 1555 and 1580 cm⁻¹ appeared beyond 220 °C, which attributed to the newly formed formate species.²⁶ Carboxylate and formates gradually transformed to CO and continued the FTS process, leading to longer chain hydrocarbons.²⁷ Concurrently, C−H stretching vibrations of methyl and methylene groups in the region of 2500–3000 cm⁻¹ belonging to various hydrocarbon species appeared beyond 200 °C and grew with the temperature, consistent with the results of temperature-dependent C₂−C₄ production activities.²⁸

Accordingly, the collected results of the catalytic performance and DRIFTS studies allowed us to draw the possible reaction pathways over the CoFe/CoFe₂O₄ nanocomposite involved in the hydrogenation of CO₂ to CO, CH₄ and C₂−C₄, as shown in Figure 4e. Under the reductive atmosphere, partial CoFe₂O₄ was initially reduced by H₂ to form the CoFe/CoFe₂O₄ heterostructured nanocomposite. Then CO₂ adsorbed on the surface of CoFe₂O₄, and reacted with H₂ to make formate, which further hydrogenated to generate CO and H₂O (the RWGS pathway). Subsequently, the CoFe alloy in the interface of CoFe/CoFe₂O₄ heterostructure nanocomposite will contribute to C−C coupling in the chain growth of hydrocarbons.²⁹

The role of solar energy

The advantage of light-assisted CO₂ hydrogenation on FC2 has been illustrated by the above results. To investigate the role of solar energy, hydrogen programmed temperature reduction (H₂-TPR) measurements were conducted and three peaks in the 100–400 °C were recorded in both light and dark conditions (Figure 4f). Thereinto, peaks around 170 and 280 °C in the dark shifted towards lower temperatures at 160 and 230 °C respectively in the light, indicating solar light enabled the reduction of CoFe₂O₄ at a lower external temperature. This might be ascribed to the solar light-induced photothermal effect, causing the improvement of local temperature. In addition, the performance of FC2 at higher temperature (400 °C) with the 17.9 % selectivity of C₂−C₄ and 13.0 % CO₂ conversion, was slightly lower than that at 300 °C under solar light (Figure S20), further revealing the importance of the solar-induced thermal heating effect.^{8g, 30}

The temperature at the surface of the catalyst during the reaction was monitored using an infrared (IR) camera (Figure S21), which illustrated a approximate 17–19 °C increase under light illumination. To further explore the contribution of the photothermal effect to the reaction process, a 2D reactor model was built in an in-house MATLAB®-based multi-physics simulation environment. Continuum methods were applied to map light irradiation and heat transport properties, and therewith estimated the bulk temperature distribution in the catalyst bed. Sensitivity analysis showed that the maximum bulk temperature rise over the reactor body can be expected at about 20 °C depending mostly on the optical and thermal properties of the catalyst bed (Figure S22). Our simulation results represented a qualitative indication of the heating effect of the incident light itself over the catalyst bed, although the estimated extent of the bulk temperature rise cannot fully explain the observed differences between light and dark. In our case, the photo-induced local heating might fluctuate significantly and result in an inhomogeneous local temperature distribution depending on the microstructure of the catalyst bed, but our simulation indeed indicated the significance of such photothermal heating, a topic that deserves more attention in future studies. Consequently, the solar energy not only facilitated the in situ formation of CoFe/CoFe₂O₄, but also benefited the CO₂ hydrogenation reaction through the photothermal effect.