ZnAl₂O₄ with three distinct morphologies (granular, plate-like and rod-like) was synthesized using different methods. (Zhu et al., 2013; Davis and Occelli, 2016; Zhang et al., 2021). The details are shown in the Supplementary Material.

The CO₂ hydrogenation to methanol was carried out in a fixed-bed reactor. The products were analyzed using an online gas chromatograph (GC) (Agilent 7890B). The details are shown in the Supplementary Material.

The catalysts were characterized by powder X-ray diffraction (XRD), scanning electron microscopy (SEM), transmission electron microscopy (TEM), H₂-temperature program reduction (H₂-TPR), CO₂-temperature program desorption (CO₂-TPR) and diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) experiments. The operation details are shown in the Supplementary Material.

Three ZnAl₂O₄ samples with distinct morphologies were evaluated for CO₂ hydrogenation to methanol and their catalytic performances are shown in Figure 2. Significant differences in the catalytic performance were observed among the three morphologies. The granular ZnAl₂O₄ (k-ZA) exhibited a reaction induction period of approximately 10 h. During this period, the product distribution changed markedly: methanol selectivity increased from 0% to 45%, accompanied by the formation of 7% dimethyl ether (CH₃OCH₃). The dimethyl ether likely originates from the dehydration of methanol on either the ZnAl₂O₄ surface or a reconstructed Al₂O₃ surface. (Renfeng et al., 2011). As the reaction proceeded, CO selectivity decreased from 100% to 47%, while CO₂ conversion gradually increased to 6%. In contrast, neither the rod-like (r-ZA) nor the plate-like (p-ZA) ZnAl₂O₄ exhibited an induction period. Their CO₂ conversions remained low at around 1.7%, with CO selectivity being nearly 100%. A further comparison of methanol formation rates revealed that the space-time yield (STY) of methanol over granular ZnAl₂O₄ increased continuously during the reaction, rising from the initial value to 0.42 mmol g_cat ^-1 h^-1 after 10 h. In comparison, the STY of methanol for the rod-like and plate-like samples remained nearly zero. It is inferred that the structure of granular ZnAl₂O₄ underwent changes during the reaction, whereas the rod-like and plate-like catalysts retained their initial structures after the reaction.

Figures 3a,b present the X-ray diffraction (XRD) patterns of the three ZnAl₂O₄ morphologies. The results indicate that all three samples exist as pure phases without detectable ZnO diffraction peaks. The characteristic diffraction peaks observed at 31.2°, 36.8°, 44.7°, 55.6°, 59.3° and 65.2° can be indexed to the (220), (311), (400), (422), (511) and (440) crystal planes, respectively. (Zawadzki, 2006). The compositional differences among the three morphologies of ZnAl₂O₄ were compared using XRF characterization. The results show that the Al/Zn atomic molar ratios for all three are approximately 2. This indicates that all three are pure-phase ZnAl₂O₄ spinel structures. Compared with the fresh catalysts, the spent catalysts of all three morphologies show no significant changes in their XRD patterns. This suggests that the reconstruction of granular ZnAl₂O₄ is likely confined to the surface and does not alter the bulk structure. Combined with the catalytic performance of the rod-like and plate-like samples, it is inferred that both their bulk and surface structures remain unchanged. Figure 3c shows that the N₂ adsorption-desorption isotherms of all catalysts can be classified as Type IV, indicating the presence of mesopores formed by the stacking of particles.

In situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS) was employed to compare the surface structures of the three ZnAl₂O₄ morphologies before and after the reaction. A gas mixture of CO₂ and N₂ (CO₂:N₂ = 1:19 mL/min) was introduced at 300 °C to investigate the changes in the surface structure. Characteristic absorption peaks near 1,339 cm^-1, corresponding to carbonate species adsorbed on ZnO, were monitored to detect the possible formation of surface ZnO. (Rethwisch and Dumesic, 1986; Zhang et al., 2021). As shown in Figure 4a, none of the three fresh catalysts exhibited the absorption peak near 1,339 cm^-1, indicating the absence of surface ZnO. In contrast, Figure 4b reveals that the spent granular ZnAl₂O₄ sample showed a characteristic absorption band at approximately 1,330 cm^-1, which is attributed to carbonates adsorbed on ZnO. This confirms the formation of ZnO on its surface. No such absorption peak was observed for the spent rod-like or plate-like ZnAl₂O₄, indicating that their spinel surface structures remained intact.

In situ DRIFTS was used to monitor the evolution of surface species on granular ZnAl₂O₄ under reaction conditions, as shown in Figure 5. After introducing the reaction gases, the fresh catalyst exhibited the gradual emergence of absorption peaks at 1,370 cm^-1, 2,878 cm^-1, 2,900 cm^-1, 2,960 cm^-1 and 2,993 cm^-1, which are attributed to surface formate species. (Tada et al., 2022; Feng et al., 2023). Peaks associated with adsorbed CO were also observed at 2,110 cm^-1 and 2,170 cm^-1. (Yang et al., 2023). However, no absorption peaks corresponding to methoxy species were detected. In contrast, the spent granular ZnAl₂O₄ catalyst displayed characteristic absorption peaks at 1,093 cm^-1 and 2,943 cm^-1, which are assigned to surface methoxy species. (Wang et al., 2017; Sun et al., 2024). According to literature results, CO₂ hydrogenation to methanol generally follows the formate pathway: CO₂ is first hydrogenated to formate (HCOO*), then successively to dioxymethylene (H₂COO*), formaldehyde (H₂CO*) and methoxy (H₃CO*), with methoxy finally hydrogenated to produce methanol. (Bowker et al., 1988; Hu et al., 2024). Among these steps, the hydrogenation of the HCOO* and H₃CO* intermediates is relatively slow and is generally considered the rate-determining step(s) for methanol synthesis, leading to their accumulation on the catalyst surface. (Meunier et al., 2023; Lin and Bhan, 2024). Based on these experimental results and literature reports, it can be concluded that the fresh granular ZnAl₂O₄ catalyst possesses a weak ability to activate H₂, preventing the further conversion of surface formate species and resulting in no methanol formation. After reaction, the formation of ZnO on the surface of granular ZnAl₂O₄ enhances its H₂ activation capability, enabling the conversion of CO₂ to methanol. These DRIFTS findings are consistent with the observed reaction performance, where the methanol formation rate was initially zero but increased steadily over time. Identical in situ DRIFTS experiments were conducted on the plate-like and rod-like ZnAl₂O₄ samples. The results in Figure 6 show that no methoxy absorption peaks were detected on these catalysts after reaction, indicating that no surface reconstruction to form ZnO occurred.

Zhou et al. investigated the effect of ZnO morphology on its performance in CO₂ hydrogenation to methanol. (Zhou et al., 2023). They found that plate-like ZnO exhibited CO₂ conversion and CH₃OH formation rates 3.8 and 2.3 times higher than those of rod-like ZnO, respectively. The plate-like ZnO demonstrated superior H₂ activation capability and CO₂ adsorption capacity compared to the rod-like morphology. The predominantly exposed (002) facet of the plate-like ZnO was identified as more favorable for CH₃OH formation. A comparison was made between the reaction performance of granular ZnAl₂O₄ and plate-like ZnO. As shown in Figures 7a,b,d, under identical reaction temperatures, the granular ZnAl₂O₄ achieved both higher CO₂ conversion and higher space-time yield of oxygenates (methanol + dimethyl ether, Oxy) than the plate-like ZnO. Specifically, the CO₂ conversion over granular ZnAl₂O₄ was approximately 1.7 times that over plate-like ZnO, while the oxygenate yield was 3 times higher. Furthermore, at similar CO₂ conversion levels, the selectivity toward oxygenates was significantly greater on granular ZnAl₂O₄ than on plate-like ZnO. Figure 7c shows that the apparent activation energy for total oxygenate formation on granular ZnAl₂O₄ was 38 kJ/mol, which is considerably lower than that for rod-like ZnO (54 kJ/mol) and close to the activation energy for methanol formation on plate-like ZnO (32 kJ/mol). Based on these results, it is concluded that the reaction induces the formation of highly active ZnO on the surface of granular ZnAl₂O₄. This generated ZnO likely exposes a high proportion of the (002) facet, thereby promoting the CO₂ hydrogenation to methanol.

The influence of the reaction temperature on the surface reconstruction process was investigated. As shown in Figures 8a–c, increasing the reaction temperature led to higher CO₂ conversion, lower oxygenate selectivity and higher CO selectivity. Furthermore, the time required for the CO and oxygenate formation to reach a steady state decreased with increasing temperature (Figure 8d).

Figure 9a shows that the duration of the reaction induction period shortened from approximately 10 h at 280 °C to about 2 h at 310 °C. These observations collectively indicate that higher temperatures promote the surface reconstruction process. The slope of the tangent line at the initial point of the oxygenate formation rate curve was determined. As shown in Figure 9b, the slope of this tangent increased continuously with temperature, further confirming that the reconstruction rate accelerates at higher temperatures. The apparent activation energy for the surface reconstruction of granular ZnAl₂O₄ was determined from an Arrhenius plot of the logarithmic tangent slopes versus the inverse temperature. The result, presented in Figure 9c, yields an activation energy of 171 kJ/mol. This high value explains why the reconstruction is difficult to initiate and support the conclusion that the process is confined to the surface, while the bulk retains the spinel structure. Thermodynamic analysis of the phase separation of ZnAl₂O₄ into ZnO and Al₂O₃ over a temperature range of 100 °C–1,000 °C is shown in Figure 9d. The calculated standard Gibbs free energy change (ΔG) is approximately 40 kJ/mol, indicating that bulk phase separation is thermodynamically unfavorable.

The reaction temperature also affected the steady-state catalytic performance of ZnAl₂O₄ in CO₂ hydrogenation. As shown in Figure 10, both the CO₂ conversion and the yield of oxygenates increased with rising temperatures. However, the product distribution showed different trends: CO selectivity increased continuously, while oxygenate selectivity decreased. This is because the reverse water-gas shift (RWGS) reaction is endothermic and thus favored at higher temperatures. The increase in overall CO₂ conversion and oxygenate yield at elevated temperatures is attributed to the general promotion of the reaction kinetics.

The influence of gas hourly space velocity (GHSV) on the surface reconstruction of ZnAl₂O₄ was subsequently investigated. As shown in Figure 11, when the GHSV was increased from 1.8 to 18 L g_cat ^-1 h^-1, the CO₂ conversion, product selectivity and the time required for the oxygenate formation rate to reach steady state were significantly shortened—from an initial 10 h–2 h. This indicates that a higher GHSV promotes the surface reconstruction process. The initial slope of the tangent to the oxygenate formation rate curve increased with rising GHSV, demonstrating that the surface reconstruction rate is accelerated at higher space velocities. The nature of the reconstruction is a chemical reaction between ZnAl₂O₄, H₂ and CO₂. Changes in reactant concentration inevitably affect the reconstruction rate. The increase in the GHSV raises the local concentration of CO₂ and H₂ per unit surface area, thereby enhancing the reaction rate and consequently accelerating the surface reconstruction.

Variations in GHSV also influence the steady-state performance of ZnAl₂O₄ in CO₂ hydrogenation. As shown in Figure 12, an increase in GHSV led to decreased CO₂ conversion and CO selectivity, while both the selectivity and yield of oxygenates increased. This is attributed to the reduced residence time of the reactant gases on the catalyst surface at higher space velocities, which lowers the proportion of CO₂ molecules undergoing reaction and thus decreases CO₂ conversion. Simultaneously, a higher GHSV can suppress methanol decomposition, resulting in lower CO selectivity.

The influence of the H₂/CO₂ feed ratio on the surface reconstruction was investigated by maintaining a fixed CO₂ flow rate while varying the H₂/CO₂ ratio. As shown in Figure 13, when the H₂/CO₂ volumetric flow ratio was increased from 1:1 to 9:1, the time required for CO₂ conversion, oxygenate selectivity and oxygenate formation rate to reach steady state gradually decreased from 10 h to approximately 5 h. Concurrently, the initial slope of the tangent to the oxygenate formation rate curve increased with a higher H₂/CO₂ ratio, indicating that the surface reconstruction rate accelerates as the ratio rises, indicating that the surface reconstruction is induced by H₂, and increasing the H₂ concentration in the reaction gas promotes the reconstruction process.

Variations in the H₂/CO₂ feed ratio also affected the steady-state reaction performance. As shown in Figure 14, as the H₂/CO₂ volumetric flow ratio increased from 1:1 to 9:1, the CO₂ conversion, oxygenate selectivity and oxygenate yield all exhibited an increasing trend, while CO selectivity gradually decreased. The increased H₂ feed shifts the equilibria of both CO₂ hydrogenation to methanol and the reverse water-gas shift reaction forward, thereby promoting CO₂ conversion. Since methanol formation requires more active hydrogen species, the increased H₂ availability favors methanol production, leading to reduced CO selectivity.

The surface structure of granular ZnAl₂O₄ underwent changes during the reaction, while the rod-like and plate-like ZnAl₂O₄ maintained their original surface structures. To investigate the cause of this differential reconstruction behavior, H₂-TPR experiments were subsequently conducted. As shown in Figure 15, the granular ZnAl₂O₄ exhibited two reduction peaks at 454 °C and 560 °C, whereas no reduction peaks were observed for the rod-like or plate-like ZnAl₂O₄, confirming the high stability of their crystal plane structures. Zhang et al. proposed a mechanism for the surface reconstruction: H₂ undergoes heterolytic cleavage on Zn-O-Al structural units, leading to the breaking of Zn-O bonds and the formation of Al-OH species during the reaction.² Adjacent Al-OH groups then dehydrate to form more stable Al₂O₃, accompanied by the formation of ZnO. (Zhang et al., 2021). It can be concluded that the surface of granular ZnAl₂O₄ is more readily reduced by H₂, allowing for reaction-induced formation of surface ZnO. In contrast, the highly stable crystal plane structures of the rod-like and plate-like ZnAl₂O₄ prevent surface structural changes under the reaction atmosphere.

The crystal plane structures of ZnAl₂O₄ are shown in Figure 1. In the (311) plane, the Zn atoms are arranged in a triangular pattern; in the (440) plane, they form a rectangular arrangement; and in the (222) plane, the Zn atoms adopt a spatial hexagonal arrangement. By comparing the different crystal planes of ZnAl₂O₄, it is inferred that planes with a hexagonal arrangement of adjacent Zn atoms are more prone to surface reconstruction. Combined with the H₂-TPR results, it is hypothesized that Zn atoms in a spatial hexagonal arrangement possess a stronger ability to activate H₂. The heterolytic cleavage of H₂ between Zn-O bonds, coupled with the breaking of Zn-O bonds in the Zn-O-Al units, weakens the stability of Zn atoms within the lattice, facilitating their migration to the surface to form ZnO.

It is crucial to analyze the structural evolution of ZnAl₂O₄ from both thermodynamic and kinetic perspectives. As shown in Figure 9d, theoretical calculations indicate that in the temperature range of 100 °C–600 °C, the ΔG for the complete bulk decomposition of ZnAl₂O₄ into ZnO and Al₂O₃ is approximately +40 kJ/mol. This significantly positive value clearly confirms that bulk decomposition is thermodynamically unfeasible. Combined with the in situ DRIFTS results, we infer that the observed structural changes are limited to surface reconstruction, specifically the formation of ZnO on the surface of ZnAl₂O₄. From a kinetic standpoint, increasing the temperature accelerates the reconstruction process by enhancing molecular collision frequency and energy. An increase in gas hourly space velocity may promote this kinetic process by improving mass transfer process. More importantly, raising the H/C feed ratio and the specific crystal planes exposed by the catalyst itself jointly regulate the kinetics of hydrogen activation. Specifically, increasing the H/C ratio enhances the coverage of surface hydrogen species, while the predominantly exposed (222) crystal plane of granular ZnAl₂O₄, which exhibits optimal heterolytic H₂ activation capability, further promotes the cleavage of Zn–O bonds and the formation of ZnO. Based on this, it is easy to understand that Pd loaded on ZnAl₂O₄ promotes the kinetic process of ZnAl₂O₄ surface reconstruction by enhancing hydrogen activation and the spillover process. (Liu et al., 2022). Therefore, these factors primarily influence the rate of ZnAl₂O₄ structural evolution within the kinetic domain, without overcoming the thermodynamic limitation of bulk decomposition. Additionally, changes in these reaction conditions simultaneously exert complex effects on the intrinsic reaction kinetics of CO₂ hydrogenation to methanol, collectively determining the final catalytic performance.