Hydroxyl-oxygen vacancy synergy over In₂O₃–ZrO₂ catalysts: mechanistic insights into CO₂ hydrogenation to methanol

Abstract

The synergistic interplay between oxygen vacancies (O_V) and hydroxyl species in In₂O₃–ZrO₂ catalysts plays a crucial role in steering CO₂ hydrogenation pathways, however, the atomic-scale interactions between these features have remained elusive. In this study, we engineered In₂O₃–ZrO₂ solid solutions via ZrO₂ aerogel phase modulation and thoroughly elucidated the surface chemistry using advanced experimental techniques, including solid-state NMR, in situ DRIFTS, and adsorption studies. The results demonstrate that three distinct hydroxyl site types on the catalyst's surface (terminal hydroxyls (μ₁-OH), bridged hydroxyls (μ₂-OH), and triply bridging hydroxyls (μ₃-OH)) are in close spatial proximity. Besides, μ₂-OH and μ₃-OH are particularly susceptible to dihydroxylation, a process that facilitates the generation of O_V that serve as anchoring sites for CO₂. These hydroxyl-vacancy ensembles effectively promote CO₂ activation to carbonate/bicarbonate species, which then undergo selective hydrogenation to methanol via a formate-mediated pathway, thus establishing a self-sustaining catalytic cycle. This work clarifies the cooperative role of vacancy coordination and hydroxyl chemistry in CO₂ activation and provides a mechanistic guide for the rational design of bimetallic oxide catalysts for CO₂ hydrogenation to methanol.

---

Guangchao Li

Dr Guangchao Li obtained his PhD in physical chemistry in 2021 at the University of Chinese Academy of Sciences (UCAS). He was a visiting PhD student at the University of Oxford (2019–2021). Later, he worked as a postdoctoral researcher at the Hong Kong Polytechnic University (PolyU). Since 2023, he has worked as a research assistant professor in the Department of Applied Biology and Chemical Technology at PolyU. His current research interests focus on the characterization and application of porous materials in energy conversion.

1 Introduction

Since the inception of the “methanol economy” and “liquid sunshine”, CO₂ hydrogenation to methanol has emerged as a pivotal technology for fossil fuel substitution strategies.^1–4 Bimetallic oxide catalysts, particularly those combining redox-active and structural-stabilizing components, have shown significant advantages in optimizing CO₂ activation kinetics and methanol selectivity. These benefits arise from their synergistic interfacial interactions and tailored electronic configurations.^5,6 However, the atomistic mechanisms driving these processes, particularly defect-mediated catalysis, remain poorly understood due to the inherent complexity of these systems. This complexity necessitates the use of model catalysts to understand underlying structure–activity relationships. This knowledge deficit motivates focused investigation of the In₂O₃–ZrO₂ system, where interfacial oxygen vacancies emerge as catalytic linchpins linking macroscopic performance to atomic-scale phenomena.

Early studies on pure In₂O₃ revealed oxygen vacancies as critical for CO₂ activation, with the coordination environments governing product selectivity.⁷ Triple-coordinated vacancies on specific crystal planes (e.g., c-In₂O₃(111)) promote methanol formation via HCOO* stabilization, whereas double-coordinated vacancies favour CO production through C–O bond cleavage.⁸ This reactivity highlights that vacancy geometry, not merely concentration, is the primary structural factor determining the selectivity of CO₂ hydrogenation pathways.⁹ Since doping ZrO₂ brings additional complexity, there is an ongoing debate about the origin of the reactive oxygen vacancies.^10,11 Blum et al. attributed these vacancies to a partial reduction of In₂O₃ at the In₂O₃/m-ZrO₂ interface.¹² While Müller et al. demonstrated that the In-O_V-Zr structure, formed by the solvation of In³⁺ into the ZrO₂ lattice, serves as the true reactive centre.¹³ Furthermore, Gong et al. reported that electron transfer modulates the electronic properties of the active sites, enhancing the selectivity toward methanol in ways that transcend the influence of vacancy concentration alone.¹⁴ These findings challenge the simple vacancy-centred model, underscoring the multifaceted nature of catalytic selectivity in these systems.

Further investigations reveal that the surface chemical complexity of In₂O₃–ZrO₂ catalysts arises not only from the dynamic evolution of oxygen vacancies but also from the diversity of surface hydroxyl species.¹⁵ Metal oxides inherently host a variety of hydroxyl groups, including terminal hydroxyls (μ₁-OH), doubly bridged hydroxyls (μ₂-OH), and triply bridged hydroxyls (μ₃-OH), whose stability and reactivity are governed by their local coordination environments.¹⁶ Under thermal pretreatment and hydrogenation conditions, weakly bound hydroxyl groups are selectively destabilized via dehydration or heterolytic H₂ dissociation, thereby revealing distinct O_V configurations.¹⁷ Moreover, the catalytic behaviour varies significantly among these hydroxyl groups. Previous studies have shown that the In₂O_3−x(OH)_y surface is mainly composed of terminally bonded In–OH⋯In and bridged In–OH→In species, with only the former exhibiting the active frustrated Lewis pair (FLP) feature. This FLP efficiently dissociates H₂ and directs the selective conversion of CO₂ into CO.^18–20 This finding suggests that different hydroxyl-vacancy combinations exhibit conformational specificity in regulating the reaction pathways. Nevertheless, there is still a gap in the understanding of the mechanistic details within the more complex In₂O₃–ZrO₂ solid solution system.

In this work, we synthesized a series of composite supports containing monoclinic zirconia and tetragonal zirconia phases by heat treating ZrO₂ aerogel precursor. Subsequently, In₂O₃ was doped onto these supports to obtain In₂O₃–ZrO₂ catalysts featuring diverse hydroxyl sites. By employing magic-angle spinning solid-state nuclear magnetic resonance (MAS SSNMR), in situ diffuse reflectance infrared Fourier transform spectroscopy (DRIFTS), and chemical adsorption analysis techniques, we thoroughly investigated the surface chemical environment of the In₂O₃–ZrO₂ catalysts and elucidated the synergistic interaction between hydroxyl groups and oxygen vacancies in the activation of CO₂ molecules. Our findings revealed that double and triple-coordinated O_V, especially those in the In–O_V–Zr configuration, serve as critical sites for CO₂ adsorption and activation. Moreover, hydroxyl groups adjacent to these asymmetric oxygen vacancies further stabilized CO₂ and facilitated subsequent hydrogenation processes. These insights provide an understanding of the reaction mechanism underlying CO₂ hydrogenation over bimetallic oxide catalysts.

2 Results and discussion

2.1 Catalytic performance for CO₂ hydrogenation

Mixed-crystal supports with varying proportions of m-ZrO₂ and t-ZrO₂ were prepared by calcining amorphous zirconia at 300, 400, 600, and 800 °C. Afterwards, four catalysts, designed as 10In400Zr300, 10In300Zr400, 10In300Zr600, and 10In300Zr800, were synthesized by wet impregnation with indium contents confirmed by ICP-OES (see full description in the Experimental section and Table S1, ESI†). It is noteworthy that the relative content of m-ZrO₂ in the catalyst increased with increasing calcination temperature of the supports (Table S1†). The CO₂ hydrogenation performance for these catalysts was evaluated over a temperature range of 250–310 °C at 5 MPa, enabling direct comparison with previous reports on In₂O₃–ZrO₂ catalysts. Prior to testing, all the catalysts were reduced at 350 °C in 10% H₂/Ar for 1 h unless otherwise noted. No methanol is formed from pure ZrO₂, and the space-time yield (STY_Methanol) of bare In₂O₃ is only 104.3 mg g_cat⁻¹ h⁻¹, which is significantly lower than that of 10In300Zr400 (302 mg g_cat⁻¹ h⁻¹) at 300 °C, 5 MPa, 12 000 mL g_cat⁻¹ h⁻¹ (Fig. S1 and Table S2†). Additionally, the In₂O₃–ZrO₂ catalyst synthesized via the one-step hydrothermal method performs worse than 10In300Zr400. This may be mostly In doped in ZrO₂, which is detrimental to methanol synthesis (Fig. S2 and S5†). Due to the competitive relationship between methanol synthesis and the reverse water-gas shift reaction, CO₂ conversion and methanol selectivity show a seesaw effect with increasing temperature (Fig. 1a).²¹ Whereas, the process of increasing temperature promotes the activation of reactants and intermediates and exhibits higher STY_Methanol (Fig. 1d). Moreover, the methanol yield is almost unchanged for up to 106 h, which is indicative of exceptional catalyst stability under prolonged reaction conditions (Fig. 1e).

Fig. 1 Catalytic performance of 10In400Zr300, 10In300Zr400, 10In300Zr600, and 10In300Zr800 in the CO₂ hydrogenation reaction. (a) The correlation between methanol selectivity and CO₂ conversion at different reaction temperatures (for the same catalyst, each data point represents a temperature, 250, 270, 290 and 310 °C in that order), (b) CO₂ conversion and product distribution, (c) Arrhenius plots for the methanol synthesis reaction kinetics and the corresponding activation energy (E_a) calculations by using Arrhenius equation. (d) Relationship between the space-time yield (STY) of methanol and the reaction temperature. (e) Long-term TOS testing of 10In300Zr400. Reaction conditions: 0.1 g catalyst, GHSV = 12 000 mL g_cat⁻¹ h⁻¹, 5 MPa, 310 °C, H₂/CO₂ = 3.

Fig. 1b illustrates the variability in catalytic activity among the four catalysts examined at 290 °C. The results indicate that the catalyst labelled 10In300Zr800 exhibits almost no activity. In contrast, 10In300Zr400 shows superior catalytic performance at this temperature, achieving 83% methanol selectivity and 8% CO₂ conversion. Interestingly, as the m-ZrO₂ content in the support increases (Table S1†), the selectivity of the In₂O₃–ZrO₂ catalyst correspondingly improves. This enhancement in selectivity may be attributed to increased electron transfer between In₂O₃ and ZrO₂. To further investigate the relationship between methanol selectivity and m-ZrO₂ content, an identical mass fraction of indium was doped onto both m-ZrO₂ and t-ZrO₂ supports. Subsequently, the heterogeneous support prepared via the one-pot method was simulated by physical mixing to facilitate a comparative analysis (Fig. S2 and S3†). Testing under identical conditions confirmed that methanol selectivity positively correlates with the m-ZrO₂ content. Furthermore, the apparent activation energy (E_a) of the methanol synthesis reaction was calculated using the Arrhenius equation (Fig. 1c). The result indicates that the apparent activation energy of the 10In300Zr400 (38.6 kJ mol⁻¹) is noticeably lower than that of 10In400Zr300 (58.9 kJ mol⁻¹), 10In300Zr600 (46.4 kJ mol⁻¹), and 10In300Zr800 (70.2 kJ mol⁻¹), also showing a similar trend as CO₂ conversion. Thus, the above results indicate that 10In300Zr400 is most effective at lowering the energy barrier of the methanol synthesis and driving CO₂ into methanol.

2.2 Catalyst structure characterization

It is well known that the crystalline phase of ZrO₂ can undergo transformation in response to changes in the external environment, particularly temperature fluctuations, as evidenced by TG-DSC, XRD and Raman profiles (Fig. S4–S8†).^14,22 Previous studies have demonstrated that surface hydroxyl condensation drives the phase evolution of ZrO₂. The strong peak at 355 °C observed in the DSC curves is directly associated with the condensation of surface hydroxyls, which facilitates the transformation of a-ZrO₂ to t-ZrO₂.^23,24 In contrast, the weak peak at 481 °C corresponds to the phase change from t-ZrO₂ to m-ZrO₂.²⁵ Raman spectroscopy with 532 nm excitation was performed on four In₂O₃–ZrO₂ samples. As shown in Fig. S8,† the appearance of Raman peaks at 146 and 269 cm⁻¹ correspond to the bulk of t-ZrO₂, whereas the m-ZrO₂ exhibits distinct Raman peaks at 178, 192, 333, and 378 cm⁻¹.²² The 10In400Zr300 sample displays a dominant tetragonal phase. Notably, the intensity of the monoclinic phase (178 cm⁻¹) increases with higher calcination temperatures, reflecting a gradual phase transition from t-ZrO₂ to m-ZrO₂. These spectral trends align with the XRD result, corroborating the structural evolution of the ZrO₂ support. Moreover, upon doping with 10 wt% In, the diffraction peaks corresponding to the (101) crystal plane (at ∼30.2°, PDF#88-1007) of both fresh and spent 10In300Zr400 shift towards higher angles relative to those of Zr400. This shift indicates the incorporation of In into the ZrO₂ lattice, resulting in the formation of InZrO_x solid solution (Fig. 2a).^13,26 Based on the Scherrer equation, the calculated particle sizes of 10In300Zr400 and 10In400Zr300 are in the range of 9–10 nm (Table S2†). The observed difference in catalytic activity between the two catalysts indicates that the crystal size has a relatively minor effect on the overall performance.

Fig. 2 Structural characterization and electronic effects. (a) XRD patterns of Zr400, fresh-10In300Zr400 and spent-10In300Zr400. (b) EDX elemental mapping and (c) HAADF-STEM images of fresh-10In300Zr400. (d) H₂-TPR profiles of the In₂O₃–ZrO₂ catalysts. XPS profiles of (e) In3d and (f) Zr3d for reduced 10In300Zr400.

Although In₂O₃ diffraction signals are obscured due to peak overlap with t-ZrO₂, aberration-corrected HAADF-STEM images enable clear visualization of its distribution and structure. On 10In300Zr400 catalyst, In₂O₃ is well dispersed, likely facilitated by the abundant surface hydroxyl groups (Fig. 2b, c and S9†).^15,27 Images acquired at higher magnification of 10In300Zr400 reveal that the In atom is positioned at the top of the Zr atomic column in the form of brighter spots (Figure S10†). This observation aligns with the XRD results, confirming that some of the indium is embedded in zirconium oxide in the form of single atoms.^28,29 Moreover, the existence of indium single atoms is still observed on the spent catalysts, suggesting that this structure was extremely stable under high reaction temperatures.

Additionally, H₂-TPR measurements were performed to evaluate the H₂ dissociation behaviour, which further elucidated the In₂O₃ morphology across the four In₂O₃–ZrO₂ catalysts (Fig. 2d and 5d). Compared with pure In₂O₃, the reduction peaks of the In₂O₃–ZrO₂ catalysts shift to higher temperatures, indicating a strong metal–support interaction (SMSI) between ZrO₂ and In₂O₃. This interaction promotes the partial reduction of In₂O_3, resulting in a more stable In–O–Zr structure and smaller In₂O₃ particles.³⁰ Also, the higher the proportion of m-ZrO₂ in the carrier, the lower the surface reduction temperature of In₂O₃. XPS analysis of the In 3d and Zr 3d spectra further supports these findings. Previous studies have shown that t-ZrO₂ and m-ZrO₂ have different electron transfer ability to In₂O₃, where the interfacial electron transfer from m-ZrO₂ to In–O–In is stronger.^14,31 In this study, we also observed that the binding energy of In exhibits a shift toward lower values, whereas the binding energy of Zr demonstrates a shift toward to higher values with increasing m-ZrO₂ content (Fig. 2e and f). This change suggests that the high electron density of In₂O₃ may contribute to the methanol selectivity. Moreover, although the H₂ dissociation ability of 10In300Zr400 is not the highest among the samples, its H₂ consumption is considerably greater than that of other systems, which promotes the formation of oxygen vacancies (Table S3†). These oxygen vacancies are essential for the effective activation and dissociation of H₂ and CO₂, as will be discussed in detail below.

2.3 Identification of surface hydroxyls and oxygen vacancies

The spatial configuration of the surface hydroxyls, which act as a key bridge between the catalyst crystal and the oxygen vacancies, directly affects the exposure of the active sites. ¹H MAS SSNMR experiments were conducted to bring out the detailed structural information on the surface hydroxyls of In₂O₃, different crystalline ZrO₂, and the corresponding In₂O₃–ZrO₂ catalysts. As illustrated in Fig. 3a and S11,† In₂O₃, ZrO_2, and In₂O₃–ZrO₂ catalysts possess three types of surface hydroxyls. Within the range of 0–10 ppm, characteristic peaks of the surface hydroxyls emerge, with chemical shifts at 0–1.5 ppm, 1.6–3.8 ppm, and 3.9–5.0 ppm corresponding to the terminal hydroxyls (μ₁-OH), doubly bridged hydroxyls (μ₂-OH), and triply bridged hydroxyls (μ₃-OH), respectively (Fig. 3b).^16,32 The signals in the high-frequency regions (5.1–8.5 ppm) are attributed to the overlapping resonance of μ₂-OH and μ₃-OH, which contribute to the formation of extensive hydrogen-bonding networks. This spatial proximity is corroborated by the 2D ¹H–¹H homonuclear correlation SSNMR spectra (Fig. 3c, S13–S15†), where strong diagonal peaks reveal the mutual interactions among μ₁-OH, μ₂-OH, and μ₃-OH. Additionally, both In₂O₃ and ZrO₂ exhibit a μ₂-OH characteristic peak at 3.2 ppm, however, in the In₂O₃–ZrO₂ system, a new peak appears at 2.0 ppm. This observation suggests a significant alteration in the hydroxyl coordination environment, attributable to the formation of an In–O(H)–Zr heterointerface. Combining XRD and aberration-corrected HAADF-STEM results, it is inferred that the dissolution of In species onto the ZrO₂ surface leads to a partial reorganization of μ₃-OH groups. This reorganization results in the formation of an In–O(H)–Zr structure characterized by bimetallic coordination, directly contributing to a reduction in the μ₃-OH content in the composite catalyst and an increase in the ratio of μ₂-OH groups. These findings confirm the critical role of surface hydroxyls in stabilizing the active indium species and in determining their spatial distribution.

Fig. 3 Characterization of surface hydroxyl configurations and their structure–activity correlations. (a) ¹H MAS SSNMR spectra of In₂O₃–ZrO₂ catalysts and ZrO₂ supports, (b) schematic representation of the three types of surface hydroxyls, (c) 2D ¹H–¹H homonuclear correlation SSNMR spectra of 10In300Zr400, (d) NH₃-TPD profiles of In₂O₃–ZrO₂ catalysts, (e) the proportion of μ₂-OH and μ₃-OH total acidity and methanol yield for CO₂ hydrogenation over In₂O₃–ZrO₂ catalysts, (f) ¹H MAS SSNMR spectra of H₂-D₂ exchange experiment for 10In300Zr400. Reaction condition: 0.1 g catalyst, GHSV = 12 000 mL g_cat⁻¹ h⁻¹, 5 MPa, 310 °C, H₂/CO₂ = 3.

By analysing the ratio of various hydroxyl groups in ¹H MAS SSNMR and the total amount of acid sites obtained from NH₃-TPD measurements, we accurately calculated the specific concentrations of different hydroxyl species in the catalysts. Upon analysing the NH₃-TPD profile through peak fitting (as detailed in Fig. 3d and Table S4†), a volcanic-like relationship emerges between the total acidity of the In₂O₃–ZrO₂ catalysts and their monoclinic phase content, with the catalyst labelled 10In300Zr400 possessing the highest concentration of acid sites (182 μmol g⁻¹). When correlating the total acidity of the catalysts with the methanol yield, a positive correlation trend akin to that between oxygen vacancy concentration and methanol yield is observed, which points to a link between oxygen vacancies and surface hydroxyls (Fig. S16†). Intriguingly, the total μ₂-OH and μ₃-OH concentrations of four In₂O₃–ZrO₂ catalysts show regular variations: 10In300Zr400 > 10In300Zr600 > 10In400Zr300 > 10In300Zr800, and this order coincides with a significant linear correlation with their methanol yield (Fig. 3e). Taking this into account, it is reasonable to speculate that μ₂-OH and μ₃-OH may be the active site for methanol synthesis. To confirm this point, we adsorbed H₂ on the catalyst surface and then switched D₂ to exchange the surface hydrogen species to obtain the ¹H SSNMR spectra (Fig. 3f). Our results show that a significant attenuation of both μ₂-OH and μ₃-OH signals in the 2.0–8.5 ppm region after D₂ treatment suggests that the two types of hydroxyl groups are highly reactive. In corroboration with this, O 1s XPS analysis shows a decrease in the hydroxyl group fraction after the reduction treatment and an increase in the oxygen vacancy fraction. Therefore, we propose that μ₂-OH and μ₃-OH are structural precursors of the active site, which may be generated by dehydroxylation to oxygen vacancies during the reaction process.

More information on the density of oxygen vacancies is unravelled in the O 1s spectra by XPS. As shown in Fig. 4a, the broad peak of the O 1s spectra can be deconvoluted into three different components. The peak at the lower binding energy (529.8 eV) is associated with the lattice oxygen of ZrO₂ (O_L). The two peaks centered at ca. 531.2 eV and 532.4 eV originate from O atoms in the proximity of an oxygen vacancy (O_V) and the chemisorbed oxygen species related to the surface hydroxyls (*OH). The radio of O_V/(O_L + O_V + OH) and OH/(O_L + O_V + OH) reflects the density of surface oxygen vacancies and the surface hydroxyls, respectively.^30,33 It is worth noting that doping In on Zr400 increases the concentration of O_V without changing *OH concentration, regarded as interfacial oxygen vacancies formed via the penetration of In into ZrO₂. Also, after reducing 10In300Zr400 by H₂, the oxygen vacancy concentration is increased from 21.4% to 26.9%, while the hydroxyl group concentration is decreased from 11.3% to 9.26% (Fig. S17†). Thus, some of the oxygen vacancies on the catalyst surface are derived from the oxidation of surface hydroxyls. Surface oxygen vacancies, as electron-rich centres, can effectively activate CO₂ molecules, and are usually regarded as important active sites for methanol synthesis by CO₂ hydrogenation. Systematic comparison of the composition–performance relationships among different catalyst systems revealed that O_V and OH concentrations in reduced catalysts follow the order 10In300Zr400 > 10In300Zr600 > 10In400Zr300 > 10In300Zr800, both exhibiting linear correlations with methanol synthesis activity (Fig. 4b and S18†). This result aligns with Liu et al.'s study,¹⁵ where air plasma treatment enhanced the hydrophilicity of ZrO₂. In the highly hydrophilic 1% Ru/DBD-ZrO₂ catalyst, CO₂ is captured at oxygen vacancies, and the H atoms from surface OH* combine with CO₂ to form HCO₃* intermediates, which further convert to HCOO*. Although their work focused on CO₂ hydrogenation to CO, we propose that oxygen vacancies and hydroxyl groups serve as critical active sites and descriptors for catalyst structure–activity relationships, given the established role of formate as a key intermediate in methanol synthesis.

Fig. 4 Characterization for revealing the relationship between oxygen vacancies and CO₂ activation. (a) Deconvoluted O 1s XPS spectra In₂O₃–ZrO₂ catalysts, (b) the proportion of Ov concentration and methanol yield for CO₂ hydrogenation over In₂O₃–ZrO₂ catalysts, (c) ³¹P MAS SSNMR spectra of TMP adsorbed on In₂O₃–ZrO₂ catalysts, (d) CO₂-TPD profiles of In₂O₃–ZrO₂ catalysts, (e) 1D ¹³C MAS NMR spectra of CO₂ adsorbed on 10In300Zr400 and 10In300Zr600. (f) 2D ¹³C–¹H heteronuclear correlation MAS SSNMR spectra of 10In300Zr400.

SSNMR is a powerful technique for the characterization of active sites on the solid materials.^34–36 As the TMP molecule is highly alkaline, TMP-assisted ³¹P MAS SSNMR analysis was adopted to distinguish the acid sites over the catalysts.³⁷ Previous studies have shown that TMP is more sensitive to Lewis acidity than Brønsted acidity, and the signal ranging from −30 to −58 ppm is attributed to TMP binding onto the LAS.³⁸ Fig. 4c shows that for all catalysts except 10In300Zr800, two peaks are observed at −50 to −60 ppm and 40 to 60 ppm, respectively. The signals at −60, −57, and −55 ppm are assigned to Lewis acid sites in 10In400Zr300, 10In300Zr600, and 10In300Zr400, respectively. Apparently, 10In300Zr400 exhibits stronger Lewis acidity than both 10In300Zr600 and 10In400Zr300, which aligns with the increased oxygen vacancy concentration observed in the O 1s spectra. The signal in the positive chemical shift range corresponds to TMPO, formed by the reaction between TMP and lattice oxygen, indicating that the In₂O₃–ZrO₂ catalyst readily forms more oxygen vacancies.³⁹

To investigate the effect of Lewis acidity/basicity of In₂O₃–ZrO₂ catalysts on the CO₂ adsorption and activation, the CO₂-TPD measurement was performed, as CO₂ molecules could bind strongly to FLP. The signals of CO₂ were collected by the mass spectrometer. As shown in Fig. 4d, CO₂ desorption peaks can be divided into three regions, 0–200 °C, 200–400 °C, and 400–900 °C, which are attributed to weak, medium-strong, and strong basic sites, respectively. In the temperature range below 400 °C, the two signals correspond to the physically adsorbed CO₂ and CO₂ adsorbed on the weakly basic site, like surface hydroxyls.^40,41 In the high temperature range of 400–900 °C, three consecutive desorption peaks at around 560, 740, and 900 °C are observed for all catalysts. These peaks are assigned to CO₂ adsorbed as bidentate carbonates (b-CO₃²⁻) and monodentate carbonates (m-CO₃²⁻) on the strongly basic sites (O_V).⁴² Compared with the CO₂-TPD profiles of pure ZrO₂, we noticed that a larger peak area in the high temperature region for 10In300Zr400, which indicates that the oxygen vacancy concentration is elevated after doping with In, and this result is also evidenced by H₂-TPR and O 1s XPS (Fig. S19†). Analogously, the existence of more surface oxygen vacancies on 10In300Zr400 is further verified. To explore the effect of H₂ pretreatment on the surface properties of the catalysts, the In₂O₃–ZrO₂ catalysts were exposed to H₂ and Ar atmospheres at 350 °C for reduction and heat treatment, respectively (Fig. S20†). Obviously, after treatment with H₂, the signal intensities in the low-temperature region (<200 °C) decrease while the peak in the high-temperature region increases, suggesting that H₂ reduction creates more oxygen vacancies by removing the surface hydroxyls.

Meanwhile, high-resolution 1D ¹³C MAS SSNMR and 2D ¹³C–¹H Heteronuclear Correlation (HETCOR) MAS SSNMR spectra have been utilized to precisely elucidate the adsorption behaviour and structure of CO₂ over the catalyst surface. In general, the electron-rich oxygen atoms in the CO₂ molecule are anchored by Lewis acids (O_V) and are immobilized on the catalyst surface in the form of bidentate carbonates (b-CO₃²⁻) and monodentate carbonates (m-CO₃²⁻). As depicted in Fig. 4e, three ¹³C signals are observed in the 158–175 ppm region after introducing ¹³CO₂ to In₂O₃–ZrO₂. The peaks from 158 to 162 ppm, 163 to 167 ppm, and 167 to 171 ppm correspond to bicarbonate, bidentate carbonate, and monodentate carbonate, respectively.⁴³ Furthermore, the 2D ¹³C–¹H HETCOR MAS SSNMR experiments reveal the spatial proximity of the adsorbed ¹³CO₂ and the three surface hydroxyls (Fig. 4f and S21†). Apparently, the b-CO₃²⁻ species in the 163–167 ppm regime exhibits strong interactions with structural hydroxyl groups of μ₂-OH and μ₃-OH, as well as their associated hydrogen-bonding network. No correlation peak corresponding to an interaction with μ₁-OH is observed. Therefore, we conclude that μ₂-OH and μ₃-OH adjacent to O_V act synergistically as proton donors to promote CO₂ protonation to HCO₃⁻ on the one hand, and by stabilizing b-CO₃²⁻ through the O–H⋯O bond on the other, thereby providing an optimized microenvironment for CO₂ hydrogenation.

2.4 Investigation of the reaction mechanism

In situ DRIFTS measurements were conducted to elucidate the synergistic effect of surface hydroxyls and oxygen vacancies in methanol synthesis. Fig. 5a presents the real-time monitoring of the catalytic process over the 10In300Zr400 catalysts under programmed temperature conditions. When the test temperature is controlled at 100 °C, a broad absorption band in the range of 1660 and 1200 cm⁻¹ is observed, which can be attributed to the formation of carbonate or bicarbonate species. Specifically, the peaks at 1276, 1300, 1662 cm⁻¹ and 1508, 1335 cm⁻¹ are assigned to bidentate carbonate species (b-CO₃²⁻) and monodentate carbonate (m-CO₃²⁻), respectively. In addition, peaks at 1446 and 1624 cm⁻¹ correspond to bidentate bicarbonate species (b-HCO₃⁻).⁴⁴ These findings suggest that CO₂ is predominantly adsorbed on the reduced catalyst as carbonates and bicarbonates, consistent with CO₂-TPD and ¹³C MAS SSNMR results. Moreover, linearly adsorbed CO (2077 cm⁻¹) is also observed at 100 °C, which may originate from the dissociation of carbonates induced by surface oxygen vacancies, as supported by the literature.⁴² As the temperature increases, the band of carbonate species and bicarbonate species gradually diminish, while new spectral features emerge. These new features correspond to the formation of bidentate formates (b-HCOO*, 1392, 1376, 1592 cm⁻¹) and methoxy (CH₃O*, 1046 and 1140 cm⁻¹).^45,46 Notably, upon reaching 250 °C, the HCOO* absorption peak intensity markedly increases, while the HCOO*/CH₃O* ratio initially rise but subsequently decline, suggesting that HCOO* acts as a key intermediate in the CO₂ hydrogenation pathway, with its conversion to CH₃O* requiring a high activation energy barrier (Fig. S22†). Furthermore, time-resolved DRIFTS analysis at 300 °C captures the dynamic behaviour of reaction intermediates (Fig. 5b). Initially observed carbonate and bicarbonate species, analogous to those at 100 °C, show rapid intensity decay within 10 minutes, while HCOO* and CH₃O* species progressively increase, indicating the conversion of carbonate/bicarbonate to formate on the In₂O₃–ZrO₂ catalysts.

Fig. 5 In situ DRIFTS spectra of CO₂ + H₂ reaction in the range of 1000–2200 cm⁻¹ over 10In300Zr400 catalyst. (a) DRIFTS spectra at different temperatures (100–300 °C) after reacting for 1 h, (b) time-resolved DRIFTS spectra at a reaction temperature of 300 °C; (c) DRIFTS spectra of H₂ reduction at 350 °C; (d) H₂-TPR profiles of ZrO₂, In₂O₃, 10In400Zr300.

To further elucidate the hydrogen dissociation mechanism on the In₂O₃–ZrO₂ catalyst, we also conducted in situ DRIFTS experiments under hydrogen-rich conditions at 350 °C and 1 MPa. As illustrated in Fig. 5c, upon exposure to a hydrogen atmosphere, the characteristic peak corresponding to surface hydroxyls (OH*, 3684 cm⁻¹) on the 10In300Zr400 catalyst demonstrates a progressive intensity reduction. Meanwhile, distinct shifts at 1707, 1584, 1533, and 1390 cm⁻¹ appear, which can be assigned to the intermediate In₂H₄, Zr–H, InInH_2, and InH species.⁴⁷ Previous studies have demonstrated that molecular hydrogen preferentially adsorbs on the In₂O₃ surface, where it readily undergoes heterolytic dissociation to form indium hydride species.¹⁹ This preferential adsorption behaviour is supported by both theoretical calculations and surface characterization techniques.^7,48 In contrast, H₂-TPR has revealed that ZrO₂ exhibits a strong tendency to interact with carbon-based species rather than hydrogen molecules, indicating its limited capability for hydrogen dissociation under similar conditions (Fig. 5d).^49,50 These collectively suggest that while hydrogen molecules remove surface hydroxyls, they simultaneously generate electron-rich InH species on the indium oxide surface. Accordingly, Zr–H species are likely formed through a hydrogen spillover mechanism from the adjacent In₂O₃ surface, rather than through direct dissociation on ZrO₂.¹¹ This hydrogen migration phenomenon is consistent with the well-documented surface mobility of hydrogen atoms in mixed oxide systems.

2.5 Discussion on hydroxyl-oxygen vacancy synergy

Based on the experimental results discussed above, Fig. 6 illustrates the proposed surface chemical environment of the In₂O₃–ZrO₂ catalyst and delineates the synergistic mechanism between hydroxyl groups and oxygen vacancies in the activation of CO₂ molecules. The crystalline phase of ZrO₂ is usually closely related to the distribution of surface hydroxyls. We found that 400 °C approaches the critical phase transition temperature of ZrO₂, and calcination at temperatures above this threshold enables the formation of mixed-crystalline supports. The 10In300Zr400 catalyst constructed on mixed-crystalline ZrO₂ by low-temperature calcination is enriched with a higher O_V concentration at the interface while simultaneously maintaining a greater density of surface hydroxyls. According to XPS and XRD results, the more severe lattice mismatch between the mixed-crystalline ZrO₂ and In₂O₃, which creates a greater number of interfacial oxygen vacancies than in pure t-ZrO₂.²⁸ Compared to the physical mixed system with the same mixed-crystal composition, the catalysts synthesised by the one-pot method exhibit higher CO₂ conversion than, confirming that in situ construction of interfacial oxygen vacancies is the key to the formation of active sites (Fig. S2†). The atomic-level dispersion of In₂O₃ is promoted through surface hydroxyls and strong metal–support interactions, thereby increasing the number of active sites. However, further elevation of calcination temperatures leads to particle coarsening and reduced hydroxyl group concentration, resulting in a reduction of hydroxyl-oxygen vacancy active sites and a decline in catalytic performance. During CO₂ hydrogenation, H₂ preferentially dissociates at In-containing structural sites, such as In nanoclusters and the In–O(H)–Zr structure. This process exposes oxygen vacancies that exist in double- and triple-bridged coordination environments. These oxygen vacancies function as strong Lewis acid sites, anchoring the oxygen atoms of CO₂, while the neighbouring μ₂-OH/μ₃-OH network forms FLP sites through hydrogen bonding. This synergistic interaction promotes the formation of carbonate and bicarbonate in a locally polar environment. The CO₂ species stabilized by the hydroxyl network are activated in the presence of the In–H active site, leading to directed hydrogenation to methanol via formate intermediate. Throughout the reaction, the depleted hydroxyl groups and oxygen vacancies are dynamically regenerated by surface remodelling under reaction conditions, thereby creating a hydroxyl-oxygen vacancy synergistic catalytic cycle.

Fig. 6 Schematic representation of CO₂ hydrogenation to methanol over In₂O₃–ZrO₂ catalysts.

3 Conclusions

In summary, a straightforward In₂O₃–ZrO₂ model catalyst was utilized to investigate the distribution of surface hydroxyls and to elucidate the synergistic interactions and transformations between oxygen vacancies and surface hydroxyls in CO₂ hydrogenation to methanol. Among the catalysts tested- 10In400Zr300, 10In300Zr400, 10In300Zr600, and 10In300Zr800 -the 10In300Zr400 catalyst exhibits a higher concentration of surface hydroxyls and oxygen vacancies. Under the operating conditions of 300 °C, 5 MPa, and 12 000 mL g_cat⁻¹ h⁻¹, it achieved superior methanol synthesis activity (302 mg g_cat⁻¹ h⁻¹). Structural characterization and mechanistic studies revealed that the FLP sites formed by surface hydroxyls and oxygen vacancies, act as centres for stabilizing CO₂, providing an abundant carbon source for the formation of methanol synthesis intermediates, HCOO* and CH₃O*. Overall, this work reinterprets the physicochemical behaviour of small molecule reactants on the catalyst surface from an acid–base perspective, offering novel insights for designing catalysts with high-density active interfacial sites for enhanced performance.

Data availability

The data supporting this article have been included as part of the ESI.†

Author contributions

Xia Li: data curation, software, formal analysis, writing–original draft, writing–review & editing. Zixia Feng: investigation, data curation, formal analysis, resources, methodology. Hanjun Lua: investigation, formal analysis. Xinlin Hong: writing– review & editing, guiding the project. Guangchao Li: writing– review & editing, guiding the project. Shik Chi Edman: discussion.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

The support from the Hong Kong Polytechnic University (PolyU P0049034, P0055259) and the Department of Science and Technology of Guangdong Province (GDSTC 2025A1515011688) is gratefully acknowledged. G. Li gratefully acknowledges support from the University Research Facility in Chemical and Environmental Analysis (UCEA) at PolyU. We thank Dr Xu Zhang from City University of Hong Kong for providing the images of the samples.

Notes and references

1. G. A. Olah, Angew. Chem., Int. Ed., 2005, 44, 2636–2639.
2. C. F. Shih, T. Zhang, J. Li and C. Bai, Joule, 2018, 2, 1925–1949.
3. P. Gao, L. Zhang, S. Li, Z. Zhou and Y. Sun, ACS Cent. Sci., 2020, 6, 1657–1670.
4. H. Wang, X. Liu, P. Niu, S. Wang, J. Shi and L. Li, Matter, 2020, 2, 1377–1413.
5. F. Tao, Chem. Soc. Rev., 2012, 41, 7977–7979.
6. X. Jiang, X. Nie, X. Guo, C. Song and J. G. Chen, Chem. Rev., 2020, 120, 7984–8034.
7. J. Ye, C. Liu, D. Mei and Q. Ge, ACS Catal., 2013, 3, 1296–1306.
8. S. Dang, B. Qin, Y. Yang, H. Wang, J. Cai, Y. Han, S. Li, P. Gao and Y. Sun, Sci. Adv., 2020, 6, eaaz2060.
9. W. Wang, K. Huo, Y. Wang, J. Xie, X. Sun, Y. He, M. Li, J. Liang, X. Gao, G. Yang, S. Lin, F. Cao, H. Jiang, M. Wu and N. Tsubaki, ACS Catal., 2024, 14, 9887–9900.
10. N. H. M. D. Dostagir, C. R. Tomuschat, K. Oshiro, M. Gao, J.-y. Hasegawa, A. Fukuoka and A. Shrotri, JACS Au, 2024, 4, 1048–1058.
11. F. Gao, Y. Wang, Y. Zhao, K. Wang, W. Guo, Z. Sun, Y. Zhu, H. He, Y. Liu and Y. Cao, ACS Catal., 2025, 15, 2785–2795.
12. X. Zhang, A. V. Kirilin, S. Rozeveld, J. H. Kang, G. Pollefeyt, D. F. Yancey, A. Chojecki, B. Vanchura and M. Blum, ACS Catal., 2022, 12, 3868–3880.
13. A. Tsoukalou, P. M. Abdala, A. Armutlulu, E. Willinger, A. Fedorov and C. R. Müller, ACS Catal., 2020, 10, 10060–10067.
14. C. Yang, C. Pei, R. Luo, S. Liu, Y. Wang, Z. Wang, Z.-J. Zhao and J. Gong, J. Am. Chem. Soc., 2020, 142, 19523–19531.
15. M. Liu, R. Zou and C.-J. Liu, Appl. Catal., B, 2025, 360, 124549.
16. Q. Han, P. Gao, L. Liang, K. Chen, A. Dong, Z. Liu, X. Han, Q. Fu and G. Hou, Anal. Chem., 2021, 93, 16769–16778.
17. L. Fu, Z. Qu, L. Zhou and Y. Ding, ACS Sustain. Chem. Eng., 2024, 12, 18412–18421.
18. G. C. Welch, R. R. S. Juan, J. D. Masuda and D. W. Stephan, Science, 2006, 314, 1124–1126.
19. K. K. Ghuman, L. B. Hoch, T. E. Wood, C. Mims, C. V. Singh and G. A. Ozin, ACS Catal., 2016, 6, 5764–5770.
20. R. Jing, X. Lu, J. Wang, J. Xiong, Y. Qiao, R. Zhang and Z. Yu, Small, 2024, 20, 2310926.
21. S. Tada, S. Kayamori, T. Honma, H. Kamei, A. Nariyuki, K. Kon, T. Toyao, K.-i. Shimizu and S. Satokawa, ACS Catal., 2018, 8, 7809–7819.
22. M. Li, Z. Feng, P. Ying, Q. Xin and C. Li, Phys. Chem. Chem. Phys., 2003, 5, 5326–5332.
23. X. Guo, D. Mao, G. Lu, S. Wang and G. Wu, Catal. Commun., 2011, 12, 1095–1098.
24. Y. Wu, J. Chen, W. Hu, K. Zhao, P. Qu, P. Shen, M. Zhao, L. Zhong and Y. Chen, J. Catal., 2019, 377, 565–576.
25. Z. Yanwei, G. Fagherazzi and S. Polizzi, J. Mater. Sci., 1995, 30, 2153–2158.
26. T. Yan, N. Li, L. Wang, W. Ran, P. N. Duchesne, L. Wan, N. T. Nguyen, L. Wang, M. Xia and G. A. Ozin, Nat. Commun., 2020, 11, 6095.
27. H. Li, W. Wang, S. Xue, J. He, C. Liu, G. Gao, S. Di, S. Wang, J. Wang, Z. Yu and L. Li, J. Am. Chem. Soc., 2024, 146, 9124–9133.
28. M. S. Frei, C. Mondelli, A. Cesarini, F. Krumeich, R. Hauert, J. A. Stewart, D. Curulla Ferré and J. Pérez-Ramírez, ACS Catal., 2019, 10, 1133–1145.
29. H. Li, S. Di, P. Niu, S. Wang, J. Wang and L. Li, Energy Environ. Sci., 2022, 15, 1601–1610.
30. Y. Zhou, L. Liu, G. Li and C. Hu, ACS Catal., 2021, 11, 7099–7113.
31. C. Shen, K. Sun, R. Zou, Q. Wu, D. Mei and C. J. Liu, Appl. Catal., B, 2025, 361, 124683.
32. J. Qin, Y. Long, F. Sun, P. P. Zhou, W. D. Wang, N. Luo and J. Ma, Angew. Chem., Int. Ed., 2021, 61, e202112907.
33. J. An, Y. Wang, J. Lu, J. Zhang, Z. Zhang, S. Xu, X. Liu, T. Zhang, M. Gocyla, M. Heggen, R. E. Dunin-Borkowski, P. Fornasiero and F. Wang, J. Am. Chem. Soc., 2018, 140, 4172–4181.
34. G. Li, C. Foo, X. Yi, W. Chen, P. Zhao, P. Gao, T. Yoskamtorn, Y. Xiao, S. Day, C. C. Tang, G. Hou, A. Zheng and S. C. E. Tsang, J. Am. Chem. Soc., 2021, 143, 8761–8771.
35. G. Li, T. Yoskamtorn, W. Chen, C. Foo, J. Zheng, C. Tang, S. Day, A. Zheng, M. M. J. Li and S. C. E. Tsang, Angew. Chem., Int. Ed., 2022, 61, e202204500.
36. G. Li, C. Foo, R. Fan, M. Zheng, Q. Wang, Y. Chu, J. Li, S. Day, P. Steadman, C. Tang, T. W. B. Lo, F. Deng and S. C. E. Tsang, Science, 2025, 387, 388–393.
37. M. Wang, L. Zheng, G. Wang, J. Cui, G. L. Guan, Y.-T. Miao, J.-F. Wu, P. Gao, F. Yang, Y. Ling, X. Luo, Q. Zhang, G. Fu, K. Cheng and Y. Wang, J. Am. Chem. Soc., 2024, 146, 14528–14538.
38. D. Rivera-Barrera and J. C. Poveda-Jaramillo, J. Solid State Chem., 2021, 294, 121862.
39. Z. Zhang, J. Wu, Z. Feng, X. Xia, J. Wen, F. Wang, X. Ning, Y. Wen, X. Ke, H. Guan and L. Peng, J. Phys. Chem. C, 2024, 129, 508–517.
40. S. Lu, H. Yang, Z. Zhou, L. Zhong, S. Li, P. Gao and Y. Sun, Chin. J. Catal., 2021, 42, 2038–2048.
41. R. Zou, C. Shen, K. Sun, X. Ma, Z. Li, M. Li and C.-J. Liu, J. Energy Chem., 2024, 93, 135–145.
42. C. Zhu, X. Wei, W. Li, Y. Pu, J. Sun, K. Tang, H. Wan, C. Ge, W. Zou and L. Dong, ACS Sustain. Chem. Eng., 2020, 8, 14397–14406.
43. Y. Fu, L. Zhang, B. Yue, X. Chen and H. He, J. Phys. Chem. C, 2018, 122, 24094–24102.
44. C. Y. Regalado Vera, N. Manavi, Z. Zhou, L.-C. Wang, W. Diao, S. Karakalos, B. Liu, K. J. Stowers, M. Zhou, H. Luo and D. Ding, Chem. Eng. J., 2021, 426, 131767.
45. L. Liu, F. Cannizzaro, A. Kaychouhi, N. Kosinov and E. J. M. Hensen, Chem. Eng. J., 2024, 494, 153204.
46. A. R. Richard and M. Fan, ACS Catal., 2017, 7, 5679–5692.
47. L. Andrews and X. Wang, Angew. Chem., 2004, 116, 1738–1741.
48. Y. Dong, K. K. Ghuman, R. Popescu, P. N. Duchesne, W. Zhou, J. Y. Y. Loh, F. M. Ali, J. Jia, D. Wang, X. Mu, C. Kübel, L. Wang, L. He, M. Ghoussoub, Q. Wang, T. E. Wood, L. M. Reyes, P. Zhang, N. P. Kherani, C. V. Singh and G. A. Ozin, Advanced Science, 2018, 5, 1700732.
49. E. L. Fornero, A. L. Bonivardi and M. A. Baltanás, J. Catal., 2015, 330, 302–310.
50. K. Li and J. G. Chen, ACS Catal., 2019, 9, 7840–7861.

---

Footnotes

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d5ta03275j

‡ These authors contributed to this paper equally.

---