English

• CO₂ emission caused by human activities results in global warming and ecological issues [1, 2]. The conversion of CO₂ into value-added hydrocarbon is a promising way to eliminate CO₂ [3]. Hydrogenation of CO₂ to methanol (CH₃OH) is one of the most promising approaches for its transformation [1, 4]. Moreover, CH₃OH can be directly used to synthesize olefin and high value-added hydrocarbons through zeolite catalysis [5-8]. By coupling of methanol-synthesis and methanol-to-hydrocarbons reaction with a bifunctional catalyst can realize the direct conversion of CO₂ to hydrocarbon products, e.g., gasoline (C₅-C₁₁), lower olefin (C₂-C₄) and aromatic hydrocarbon [5, 9]. It contains two consecutive processes: methanol synthesis from CO₂ through metal-oxide catalysts and methanol to hydrocarbons through zeolites catalysts [10]. It is an essential prerequisite that the reaction temperature of methanol to hydrocarbons should be over 340 ℃ [9, 10], which is thermodynamically restrained for methanol synthesis from CO₂ and H₂. The CH₃OH selectivity of conventionally Cu-based catalysts is lower than 5% over 320 ℃ [5, 7]. Obviously, significant and effective catalysts are required for this strategy. As Oliver Martin reported, the methanol selectivity of In₂O₃ with cubic bixbyite-type can be tuned up to 100% from 200 ℃ to 300 ℃ and over 25% at 340 ℃, which was closed to thermodynamic limit [11, 12]. Further researches showed In₂O₃ had emerged as an outstanding catalytic system for CO₂ hydrogenation to methanol with highly selectivity under industrially relevant conditions [12].

  In₂O₃ is known as a conductive transparent layer and thin-film transistors [13, 14]. There are two representative crystal structures, namely cubic bixbyite-type (denoted as c-In₂O₃) and rhombohedral corundum-type (denoted as rh-In₂O₃) [14]. As the most studied structure, c-In₂O₃ was easily lost oxygen atoms to form defective surface with oxygen vacancy [11, 15, 16]. These surface sites assist CO₂ activation and hydrogenation by stabilizing the key reaction intermediates, such as surface bound formate (HCOO), dioxymethylene (H₂COO), and H₂CO species; the hydrogenation of the latter gives surface methoxy species, supposedly via the rate-determining step [15, 17]. H₂ can re-generate the vacancy which is replenished during methanol formation. This catalytic cycle may play an important role on the formation of methanol from CO₂ [12, 18, 19]. Due to the excellent CH₃OH selectivity of In₂O₃, CO₂ had been successfully hydrogenated to lower olefin through In₂O₃/SAPO-34 catalyst [10] and aromatic hydrocarbon through In₂O₃/HZSM-5 catalyst [14]. Except for CO₂ hydrogenation, c-In₂O₃ also can be used for methanol steam reforming, water-gas shift reaction, formaldehyde reforming and so on [20, 21].

  However, there is little knowledge about the intrinsic adsorption and catalytic property of rh-In₂O₃ except the electronic and structural characterization by theoretical calculations [13, 22]. The prevalent scientific interest is the transition between these two phases as a case of ccp (cubic close-packed structure) to hcp (hexagonal closed-packed) transition at high-pressure and high-temperature condition [14, 23, 24]. As a meta-stability structure, rh-In₂O₃ was considered to be unstable at high temperature and pressure. However, Simon Penner et al. found rh-In₂O₃ remained stable over 430 ℃ in He or CO₂ atmospheres [25]. Thus stability of rh-In₂O₃ is satisfied for most of catalysis reactions. Rh-In₂O₃ have been proved that this nano-rhombodedra are terminated by (012) facets due to the geometric arrangement of face-sharing octahedral [26]. There is little information about the properties of surface chemistry, e.g., CO₂ hydrogenation activity, adsorption-desorption properties or reaction mechanisms.

  Herein, the c-In₂O₃ and rh-In₂O₃ were synthesized for comparative study of CO₂ hydrogenation to elucidate the structure-activity relationship. The crystal phase of catalysts was characterized by X-ray diffraction (XRD) and shown in Fig. 1a. The diffraction peaks of black line were well matched with JCPDS No. 06-0416, which can be indexed to cubic bixbyite-type structure In₂O₃ (c-In₂O₃) [11]. Another diffraction pattern could be assigned to rh-In₂O₃, which was well matched with JCPDS No. 22-0336 [27]. There were no impurity peaks, indicating the single phase of the product. The average crystal size of c-In₂O₃ and rh-In₂O₃ calculated by Scherer equation in Fig. 1a were 19.8 and 24.5 nm, respectively.

  Figure 1

  

  Figure 1.  Structural characterizations of c-In₂O₃ and rh-In₂O₃, (a) XRD, (b) XRD of precursor before calcination. Diagrammatic sketch of c-In₂O₃ (c) and (d) rh-In₂O₃, (e) Raman spectra.

  DownLoad: Full-Size Img PowerPoint

  During synthetic process, the difference between the catalysts was the solvent, i.e., methanol-based route for rh-In₂O₃ and water-based route for c-In₂O₃. As shown in Fig. 1b, In(OH)₃ was formed in the water-based route and then decomposed to c-In₂O₃ after calcination. In contrast, InOOH was formed in the methanol-based route and then rh-In₂O₃ was obtained afterthermal decomposition of InOOH. During the calcination, the decomposition of In(OH)₃ was accompanied by water vapor and toward the stable phase. But for methanol-based route, InOOH with the residual methanol favored a transition coordination sphere of In³⁺ ions by avoiding fast hydrolysis and condensation, which evolved to the rhombohedral corundum-type [28, 29]. The formation of precursors was critical for the successful synthesis of c-In₂O₃ and rh-In₂O₃. The rh-In₂O₃ was terminatedby (012) plane and thepresence of ammonia/methanol had strongly influence on the surface energy of the exposed facet [26]. However, the thermodynamically stable facet of c-In₂O₃ (110) face with surface energy of 0.969 J/m² was formed in the presence of water [11].

  As Table S1 (Supporting information) and diagrammatic sketch (Figs. 1c and d) showed, c-In₂O₃ was Ia space group with lattice constant a = b = c = 10.126 Å and described as an oxygen-vacancy fluorite structure with O^2- anions missing in an ordered way [28]. There were two types of In³⁺ ions in c-In₂O₃ located at the end of a face-diagonal and a body-diagonal, which occupied octahedral and trigonal prismatic interstices of the O^2- anions [28, 30]. The lattice parameters of rh-In₂O₃ with RC space group were a = b = 5.491 Å and c = 14.526 Å. In comparison with c-In₂O₃, rh-In₂O₃ showed different atomic arrangement with a rather regular octahedron, including trigonal bi-prism coordinated In³⁺ with O^2- anions [26]. This structural difference was mainly due to the O^2- anions lattice change and In³⁺ ions shifted from octahedral and trigonal prismatic to trigonal bi-prism sites.

  Raman was also conducted to understand the structural difference and the result was shown in Fig. 1e. The representative Raman spectra of the indium oxides (black line) was corresponded to c-In₂O₃. The peaks at 303, 495 and 629cm^-1 were associated with InO₆ octahedral, 133 and 363cm^-1 were associated with the bending vibrations of In₄O₄ and stretching vibrations of In-O-In [31]. Rh-In₂O₃ also exhibited the typical Raman-active phonons. The peaks at 161, 218 and 267cm^-1 were assigned to the symmetric stretching vibration of O-In-O, 377cm^-1 was the symmetric bending vibration of O-In-O [31]. The other frequencies were caused by the symmetric stretching and symmetric bending vibration of O-In-O [32].

  Transmission electron microscopy (TEM) showed the average diameter of c-In₂O₃ and rh-In₂O₃ was around 17.9 nm and 24.9 nm, respectively. The summary of particle size distribution was showed in Fig. S1 (Supporting information). High resolution transmission electron microscopy (HRTEM) of c-In₂O₃ (Fig. S2 in Supporting information) exhibited the characteristic spacing of 0.505, 0.417 and 0.264 nm, corresponding to the (010), (211) and (104) faces of c-In₂O₃, respectively [33]. The spacing of 0.475 and 0.221 nm of rh-In₂O₃ (Fig. S2) well matched with the representative lattice distance of (211) and (222) faces of rh-In₂O₃ [32]. The selected area electron diffraction patterns were presented in Fig. S3 (Supporting information). In both catalysts, the discrete spots indicated well crystallinity and the distinct diffraction rings can be ascribed to many tiny crystallites.

  X-ray photoelectron spectroscopy (XPS) was used to understand the surface chemical state and the results were enclosed in Fig. S4 (Supporting information). The binding energy was calibrated using C 1s energy of 285 eV in Fig. S5 (Supporting information). As Fig. S4 showed, the broad and symmetric peak locked at 444.6 eV and 452.2 eV were belonged to In 3d_5/2 and 3d_3/2 of In₂O₃ [27, 34]. Negligible binding energy value difference for both catalysts suggested similar surface chemical electronic property. O 1s core level of catalysts was composed of two components and the intensity was determined by curve fitting procedure. 530.1 eV can be assigned to lattice oxygen (O lattice) and 531.7 eV can be assigned to oxygen defect (O defect) [2]. This oxygen defects may be the sub-coordinated In atom that was activated by thermal calcination. The oxygen vacancy concentration was the fraction of surface oxygen atoms adjacent to a defect calculated from the de-convoluted O 1s XPS signal. It revealed that the oxygen vacancy amount of c-In₂O₃ (24.5%) was more than the amount of rh-In₂O₃ (21.2%).

  The CO₂ hydrogenation performance was shown in Fig. 2. Both catalysts showed CO₂ catalytic activity and highly methanol selectivity at all tested temperatures. In addition to CH₃OH formation, these catalysts also generated CO as a primary byproduct (The detailed products information was enclosed in Fig. S6 in Supporting information). However, CO₂ conversion over catalysts under 300 ℃ was lower than 5% and gradually increased with reaction temperature increasing (Fig. 2a). Meanwhile, the selectivity of CH₃OH decreased and CO gradually became the dominate product, which was thermodynamically favorable with the increasing reaction temperature. In Fig. S6, the presence of CH₄ for c-In₂O₃ and absence for rh-In₂O₃ at high reaction temperature indicated the different hydrogenation ability. The hydrocarbon products of rh-In₂O₃ was merely methanol and c-In₂O₃ can further hydrogenate CO₂ to CH₄ at high temperature, suggesting higher hydrogenation ability of c-In₂O₃.

  Figure 2

  

  Figure 2.  Catalytic performance of the catalysts. (a) catalytic performance at the reaction temperatures from 260 ℃ to 360 ℃ with H₂/CO₂ = 4/1, 4 MPa and 16, 000 mL g^-1 h^-1. (b) Arrhenius plot for CO₂ conversion. (c) Evolution of the methanol STY with time on stream. (d) The effect of pressure on CO₂ hydrogenation. The reaction condition: 340 ℃, H₂/CO₂ = 4/1 and 16, 000 mL g^-1 h^-1. (e) The effect of GHSV on CO₂ hydrogenation. The reaction condition: 340 ℃, H₂/CO₂ = 4/1 and 4 MPa.

  DownLoad: Full-Size Img PowerPoint

  C-In₂O₃ showed higher CO₂ conversion than rh-In₂O₃. At 360 ℃ (Fig. 2a), CO₂ conversion of c-In₂O₃ was 17.02%, which was about three times higher than rh-In₂O₃. It was worthwhile mentioned that the specific surface area obtained from N₂-sorption (Fig. S7 in Supporting information) was 24.8 m²/g for c-In₂O₃ and 22.8 m²/g for rh-In₂O₃. The similar surface area for the two catalysts indicated that the catalytic difference was not the result of surface area. In addition, the rh-In₂O₃ showed higher methanol selectivity than c-In₂O₃. At 260 ℃, the methanol selectivity of both catalysts were all over 80% and decreased gradually with the temperature increasing. Especially at 340 ℃, the methanol selectivity of rh-In₂O₃ maintained approximate 30%. But the value of c-In₂O₃ was only 20%. As Table S2 showed, the two catalytic processes of methanol synthesis and methanol to hydrocarbon can combined into one integrated process and directly hydrocarbon molecule from CO₂ and H₂, which need >30% methanol selectivity at the optimized temperature over 340 ℃ [21]. Thus, both c-In₂O₃ and rh-In₂O₃ showed wide temperature range for high methanol selectivity and promising catalysts for hydrocarbon production coupling with zeolite catalysts. The apparent activation energy for CO₂ conversion was 55.33 and 55.39 kJ/mol over c-In₂O₃ and rh-In₂O₃, respectively in Fig. 2b. The quite close activation energy may indicate the same CO₂ hydrogenation mechanism, which was further confirmed by the following results of in-situ DRIFTS.

  The space-time yield (STY) of methanol over the catalysts was also shown in Fig. 2c. C-In₂O₃ generated about 3.0 mmol g^-1 h^-1 CH₃OH at 340 ℃, which was higher than 1.8 mmol g^-1 h^-1 of rh-In₂O₃. The stability of both catalysts showed slight deactivation after 12 h reaction on stream. The XRD patterns of used catalysts was shown in Fig. S8 (Supporting information). In compared with Fig. 1, the intensities of peaks in both catalysts were decreased. However, the diffraction peaks of both samples were still sharp and intense, indicating their highly crystalline nature. No impurity peaks were observed, confirming the high purity of the used catalysts. Moreover, the transformation between c-In₂O₃ and rh-In₂O₃ was not observed, which showed the structural stability under reaction condition. Moreover, the influence of experimental conditions on CH₃OH selectivity was also evaluated (Figs. 2d and e), which showed higher gas hourly space velocity (GHSV) and pressure were beneficial for CH₃OH selectivity. And rh-In₂O₃ had higher CH₃OH selectivity during all experimental conditions.

  Temperature-programmed reduction (TPR) in hydrogen (H₂-TPR) was carried out to understand the reducibility of catalysts and the results were shown in Fig. 3a. Prior to the bulk reduction into metallic indium over 580 ℃ (c-In₂O₃) or 626 ℃ (rh-In₂O₃), there was one primary reduction peak [11, 17]. The diverse reduction peak between c-In₂O₃ (212 ℃) and rh-In₂O₃ (307 ℃) was determined by the reducible oxygen atoms and exposing facets [35]. The reduction peak of rh-In₂O₃ mainly was associated with lattice oxygen atoms from (012) planes [27]. The reduction peak of c-In₂O₃ at 212 ℃ was associated with oxygen reduction from (110) planes [22, 23]. Oxygen vacancy in In₂O₃-based catalysts was considered to be the active sites for CO₂ hydrogenation [11]. TPR profiles suggested that the oxygen of c-In₂O₃ was more active and easy to release than rh-In₂O₃. Thus, oxygen vacancy was easily generated in c-In₂O₃, which was linked with the CO₂ catalytic activity.

  Figure 3

  

  Figure 3.  Temperature programmed reduction and desorption of c-In₂O₃ and rh-In₂O₃. (a) H₂-TPR, (b) CO₂-TPD, (c) CH₃OH-TPD and (d) CO-TPD.

  DownLoad: Full-Size Img PowerPoint

  Temperature-programmed desorption (TPD) of CO₂ (CO₂-TPD) in Fig. 3b was used to understand the interaction between CO₂ and In₂O₃. CO₂ adsorbed as bicarbonate bridging two In atoms around oxygen vacancy sites and carbonate formed over oxygen atom [11, 15, 16, 19]. Aside from the peak of physically adsorbed CO₂ appeared at approximately 100 ℃, two additional peaks were observed in both catalysts. The desorption peaks at approximate 379 and 454 ℃ in c-In₂O₃ can be assigned to bicarbonate and carbonate, respectively. In-situ DRIFTS later found the transformation from bicarbonate to carbonate during heating-up. CO₂ desorption peaks of rh-In₂O₃ were appeared at 275 and 325 ℃. Lower desorption temperature of rh-In₂O₃ suggested weaker basicity and interaction between CO₂ and rh-In₂O₃ than c-In₂O₃. Moreover, c-In₂O₃ with a large CO₂ desorption area featured higher density of adsorption sites. This strong and high density of basicity sites of c-In₂O₃ was conducive to enhance the CO₂ conversion.

  The reason influenced the products selectivity can be explored by TPD experiment by using the main byproducts (CH₃OH and CO) as adsorbates 33. Before the TPD measurements, the samples were treated at 300 ℃ and then cooling to room temperature. CH₃OH was introduced through Ar bubbling at 60 ℃ and the adsorption procedure was last 2 h. As Fig. 3c showed, CH₃OH desorbed from rh-In₂O₃ was more difficult than c-In₂O₃ (peak at 384 ℃ versus 250 ℃), suggesting stronger CH₃OH adsorption ability of rh-In₂O₃. In addition, the methanol adsorption area over rh-In₂O₃ was far larger than that of c-In₂O₃, which showed the more adsorptive sites for CH₃OH over rh-In₂O₃.

  CO-TPD result was shown in Fig. 3d. c-In₂O₃ showed weak interaction with CO due to the absence of chemical desorption peak at high temperature except the physical desorption peak at about 100 ℃. For rh-In₂O₃, there was a width desorption peak at about 432 ℃ aside from the physical adsorption peak. CO₂ hydrogenation experiments indicated higher CH₃OH selectivity of rh-In₂O₃ than c-In₂O₃, which could be partially explained by the results of by-products TPD experiment. For c-In₂O₃, CO was more easily desorbed from the surface than CH₃OH due to the weaker interaction between CO and c-In₂O₃. The rh-In₂O₃ showed CH₃OH desorbed at a lower temperature (384 ℃ and showed in Fig. 3c) than CO (432 ℃ and showed in Fig. 3d), which may be beneficial for CH₃OH selectivity improvement over rh-In₂O₃.

  To further investigate the interaction between CO₂ and catalysts, in-situ DRIFTS experiment of CO₂ adsorption were carried out. It represented the characteristic peaks of adsorbed CO₂ on the surface at 40 ℃ and reaction intermediates along with the temperature increasing. There were three strong peaks located at 1356, 1420, 1649 cm^-1 and a weak peak at 1227 cm^-1 (Fig. 4a and Table S3 in Supporting information). The intensity of these peaks gradually strengthened after 30 min exposure.In detail, the peak at 1649, 1420 and 1227 cm^-1 could be identified as asymmetric stretching vibration υ_as(O—C—O), bending vibration δ(OH) and symmetric stretching vibration υ_s(O—C—O) of bicarbonate species (HCO₃^-) (as summarized in Table S1), respectively [36]. The other one at 1356 cm^-1 can be assigned to υ_s(O—C—O) of carbonate (CO₃^2-) species and υ_as(O—C—O) of carbonate may hide or overlap in the peak shoulder of bicarbonate species (1400-1500 cm^-1). The surface of In₂O₃ contained lots of hydroxyl (-OH group) and easily reacted with CO₂ to form bicarbonate species as Scheme S1 (Supporting information) briefly illustrated [15, 37].

  Figure 4

  

  Figure 4.  In-situ DRIFTS spectra of CO₂ adsorption and conversion over the catalysts, (a) c-In₂O₃ and (b) rh-In₂O₃. HCOOH adsorption and conversion over the catalysts, (c) c-In₂O₃ and (d) rh-In₂O₃.

  DownLoad: Full-Size Img PowerPoint

  Once reached 150 ℃, it can be clearly seen that the vibration of bicarbonate weakened and the vibration of carbonate strengthened due to the υ_as(O—C—O) at 1501 cm^-1 of carbonate appeared. This transformation meant bicarbonate turned to carbonate and was significantly enhanced until bicarbonate disappeared as the temperature up to 300 ℃. Meanwhile, a new peak at 2077 cm^-1 that assigned to adsorbed CO species appeared [25]. This CO species suggested CO₂ can be directly reduced to CO by In₂O₃. This result was consistent with the density functional theory (DFT) simulation of previous reports, which indicated CO₂ reacted with -OH on the In₂O₃ surface to form a bicarbonate specie [15, 16]. CO₂ gradually turned to CO and leaved O atom quenching the oxygen vacancy, which could be regenerated by hydrogen or thermal treatment, as the temperature increased. The similar adsorption and transformation phenomena could be also found in rh-In₂O₃ (Fig. 4b and Table S3). It was worthwhile mentioned that the intensity of carbonate for rh-In₂O₃ was much lower and even disappeared at 300 ℃ compared to c-In₂O₃, which was consistent with the weak interaction between CO₂ and rh-In₂O₃ resulted from CO₂-TPD experiments.

  There was an urgent need but it was a significant challenge to directly study the reaction between H₂ or CO₂/H₂ and In₂O₃ by in-situ DRIFTS. Because the interaction of In₂O₃ with H₂ strongly decreased the IR diffuse reflection and disturbed the surface stability (Fig. S9 in Supporting information), which prevented the reliable study by in situ DRIFTS [11, 38, 39]. Thus, the indirect strategy of HCOOH reaction was usually adopted instead. The mechanism of methanol synthesis from CO₂ hydrogenation on the In₂O₃ surface was formate-path as the previous reports, which meant CO₂ firstly hydrogenation to formate (HCOO^-) and subsequently to methoxy and CH₃OH [11, 15, 16, 33]. The interaction of formate and catalysts can be used to indirectly testify CO₂ hydrogenation mechanism.

  After treatment with formic acid and Ar sweeping at 40 ℃, c-In₂O₃ showed typical stretching vibration of formate species in Fig. 4c. The characteristic absorption bands at 1619, 1563, 1384 and 1359 cm^-1 can be assigned to formate [40]. The band at 1359 cm^-1 was υ_s(O—C—O) vibration and 1384 cm^-1 was δ(CH) characteristic vibration of formate species [40]. 1619 and 1563 cm^-1 were υ_as(O—C—O) vibration which indicated different adsorption sections of formate species, i.e., 1619 cm^-1 was characteristic of the vibration of unsymmetrical formate and 1563 cm^-1 was the symmetrical bidentate formate [40], as illustrated in Scheme S2 (Supporting information).

  Further investigation explored the temperature dependence of the surface coverage with unsymmetrical and symmetrical bidentate formate species. At 100 ℃, the adsorption bands of unsymmetrical formate species at 1619 cm^-1 decreased. On the contrary, the symmetrical bidentate formate species at 1563 cm^-1 enhanced. This opposite change for these formates on the surface could be indicative of the transformation of unsymmetrical to symmetrical bidentate formate species. The unsymmetrical formate was metastable state and easily changed to symmetrical bidentate formate under thermal treatment [41].

  When temperature increased to 200 ℃, the unsymmetrical bidentate formate became obscured and symmetrical bidentate formate remained stable. The band at 1171 cm^-1 can be assigned to υ(C—O) of methoxy (CH₃O) species, 1730 cm^-1 assigned to υ(C—O) vibration of aldehyde and the other two bands between 2300–2400 assigned to CO₂ gas phase [42, 43]. These new bands became stronger and the HCOO species vanished as the temperature increased to 250 ℃. These new species suggested the formate transformation on c-In₂O₃, which contained decomposition and hydrogenation reaction. Aldehyde and carbon dioxide were decomposed from formate species. The hydrogen was generated from the reaction between In-H and proton of formic acid gas (Scheme S2) [44]. Methoxyl, which was apparently as the precursor of methanol, may come from the hydrogenation of HCOO species.

  As for rh-In₂O₃, it was easily found the bands were similar to c-In₂O₃ in spite of some peaks shift in Fig. 4d. The peak position of all bands were summarized in Table S4 (Supporting information). At 250 ℃, the formate species on rh-In₂O₃ was completely disappeared and even formed the negative spikes, suggesting that formic acid was easily decomposed over rh-In₂O₃. Overall, the oxygen vacancy on the In₂O₃ surface assists CO₂ activation and hydrogenation and also stabilized the key intermediates involved in methanol formation. In addition, methanol formation replenishes the oxygen vacancy sites whereas H₂ helps to generate the vacancies. The cycle between the perfect and defective states of the surface catalyzes the formation of methanol from CO₂ hydrogenation [15, 16].

  In₂O₃ showed significant advantage in CH₃OH synthesis at high temperature. The active sites of In₂O₃ catalyst was ascribed to oxygen vacancy sites under reaction conditions. These surface sites assisted CO₂ activation and hydrogenation by stabilizing the key reaction intermediates, such as surface bound formate (HCOO), and further hydrogenation to CH₃OH. This mechanism suggested that oxygen vacancy sites could be replenished during CO₂ hydrogenation. As a reducible oxide, In₂O₃ was different from conventional binary or ternary metal-metal oxide catalysts. CO₂ activation, H₂ activation and CH₃OH synthesis can be completed by In₂O₃, not the combination or interface of metal-metal oxide catalysts. This feature successfully restrained the CO formation and enhanced the CH₃OH selectivity. Both c-In₂O₃ and rh-In₂O₃ catalysts showed catalytic activity and thermal stability at high reaction temperature. And the activation energy and reaction mechanism of CO₂ hydrogenation was identical. The results of in-situ DRIFTS showed CO can be directly generated from CO₂ on the surface of In₂O₃. This redox reaction was not sustainable because the oxygen vacancy was gradually vanished by CO₂. So the cyclic creation and annihilation of oxygen vacancy was necessary for the continuous methanol formation from CO₂ hydrogenation. The intermediate formate was used to indirectly explore the reaction mechanism, which showed formate can decompose to CO₂ and hydrogenate to CH₃OH. Meanwhile, the absence of CO signal may be ascribed to its very low concentration, indicating the generation of CO from formate can be well restrained on In₂O₃. The oxygen vacancy on the In₂O₃ were beneficial for CO₂ activation and stabilization of the key intermediates. CH₃OH generation replenished the oxygen vacancy and H₂ regenerated the vacancy to achieve the catalytic cycle. Thus, the generation of CO and CH₃OH on the surface of In₂O₃ can separate to two pathways that redoxpath reaction from CO₂ to CO and formate-path reaction from CO₂ to CH₃OH. CO₂ and H₂ favored to form formate and then hydrogenation to CH₃OH on the surface of In₂O₃, suggesting the high CH₃OH selectivity.

  However, due to the thermodynamic restrictions, CO still was the main product at high temperature. With the higher reducibility and basicity, c-In₂O₃ showed higher CO₂ conversion and CH₃OH productivity than rh-In₂O₃. And the adsorbed CH₃OH molecules desorbed at lower temperature than that of the adsorbed CO molecules on rh-In₂O₃. In contrast, CO was more easily desorbed from the surface than CH₃OH on c-In₂O₃. So the rh-In₂O₃ showed higher CH₃OH selectivity during CO₂ hydrogenation compared with the c-In₂O₃.

  Above all, the comparative study of c-In₂O₃ and rh-In₂O₃ for CO₂ hydrogenation was carried out. C-In₂O₃ showed higher CO₂ conversion activity than rh-In₂O₃, which was ascribed to the higher reducibility and basicity within c-In₂O₃ and associated with the generation of the oxygen vacancy. rh-In₂O₃ showed higher methanol selectivity than c-In₂O₃. Although different structures showed different CO₂ hydrogenation performance, the hydrogenation mechanism was identical for the two catalysts. CO₂ can be directly reduced to CO through redox mechanism and CO₂ hydrogenation to CH₃OH was through formate-path. This work demonstrated the fundamental understanding of the structureactivity relationship for rh-In₂O₃ and c-In₂O₃ catalysts and offered some hint for superior catalytic system for CO₂ hydrogenation with high CH₃OH selectivity over 340 ℃.

  Declaration of competing interest

  The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

  Acknowledgments

  The work was financially supported by the National Natural Science Foundation of China (No. 21878116) and Natural Science Foundation of Hubei Province (No. 2019CFA070). The authors thank the Analysis and Testing Center of Huazhong University of Science and Technology for analytical support.

  Appendix A. Supplementary data

  Supplementary material related to this article can be found, inthe online version, at doi:https://doi.org/10.1016/j.cclet.2020.05.031.

  
  
• 1. [1]

     S. Li, Y. Wang, B. Yang, L. Guo, Appl. Catal. A: Gen. 571(2019) 51-60. doi: 10.1016/j.apcata.2018.12.008

  2. [2]

     B. Yang, L. Wang, Z. Hua, L. Guo, Ind. Eng. Chem. Res. 58(2019) 9838-9843. doi: 10.1021/acs.iecr.9b01316

  3. [3]

     B. Yang, W. Deng, L. Guo, T. Ishihara, Chin. J. Catal. 41(2020) 1348-1359. doi: 10.1016/S1872-2067(20)63605-1

  4. [4]

     R.W. Dorner, D.R. Hardy, F.W. Williams, H.D. Willauer, Energy Environ. Sci. 3(2010) 884. doi: 10.1039/c001514h

  5. [5]

     Z. Li, J. Wang, Y. Qu, et al., ACS Catal. 12(2017) 8544-8548.

  6. [6]

     J. Wei, Q. Ge, R. Yao, et al., Nat. Commun. 8(2017) 15174. doi: 10.1038/ncomms15174

  7. [7]

     L. Qi, Y. Wei, L. Xu, Z. Liu, ACS Catal. 5(2015) 3973-3982. doi: 10.1021/acscatal.5b00654

  8. [8]

     X. Zhu, J.P. Hofmann, B. Mezari, et al., ACS Catal. 6(2016) 2163-2177. doi: 10.1021/acscatal.5b02480

  9. [9]

     K. Cheng, B. Gu, X. Liu, et al., Angew. Chem., Int. Ed. 55(2016) 4725-4728. doi: 10.1002/anie.201601208

  10. [10]

      P. Gao, S. Dang, S. Li, et al., ACS Catal. 8(2017) 571-578.

  11. [11]

      O. Martin, A.J. Martin, C. Mondelli, et al., Angew. Chem., Int. Ed. 55(2016) 6261-6265. doi: 10.1002/anie.201600943

  12. [12]

      M.S. Frei, C. Mondelli, R. García-Muelas, et al., Nat. Commun. 10(2019) 3377. doi: 10.1038/s41467-019-11349-9

  13. [13]

      J.M. Yang, Z.P. Qi, Y.S. Kang, L. Qing, W. Sun, Chin. Chem. Lett. 27(2016) 492-496. doi: 10.1016/j.cclet.2015.12.031

  14. [14]

      P. Gao, S. Li, X. Bu, et al., Nat. Chem. 9(2017) 1019-1024. doi: 10.1038/nchem.2794

  15. [15]

      A. Tsoukalou, P. Abdala, D. Stoian, et al., J. Am. Chem. Soc. 141(2019) 13497-13505. doi: 10.1021/jacs.9b04873

  16. [16]

      M.F. Bekheet, M.R. Schwarz, S. Lauterbach, et al., Angew. Chem., Int. Ed. 52(2013) 6531-6535. doi: 10.1002/anie.201300644

  17. [17]

      A. Gurlo, Angew. Chem. Int. Ed. 49(2010) 5610-5612. doi: 10.1002/anie.201000488

  18. [18]

      T. Bielz, H. Lorenz, P. Amann, B. Klötzer, S. Penner, J. Phys. Chem. C 115(2011) 6622-6628.

  19. [19]

      J. Wang, H. Wang, P. Hu, Sci. China Chem. 61(2017) 336-343.

  20. [20]

      S. Shu, D. Yu, Y. Wang, et al., J. Cryst. Growth 312(2010) 3111-3116. doi: 10.1016/j.jcrysgro.2010.07.060

  21. [21]

      E.M. Kock, M. Kogler, C. Zhuo, et al., Phys. Chem. Chem. Phys.19(2017) 19407-19419. doi: 10.1039/C7CP03632A

  22. [22]

      J. Ye, C. Liu, D. Mei, Q. Ge, ACS Catal. 3(2013) 1296-1306. doi: 10.1021/cs400132a

  23. [23]

      J. Ye, C. Liu, Q. Ge, J. Phys. Chem. C 116(2012) 7817-7825. doi: 10.1021/jp3004773

  24. [24]

      N. Rui, Z. Wang, K. Sun, et al., Appl. Catal. B 218(2017) 488-497. doi: 10.1016/j.apcatb.2017.06.069

  25. [25]

      M. Neumann, D. Teschner, A. Knop-Gericke, W. Reschetilowski, M. Armbrüster, J. Catal. 340(2016) 49-59. doi: 10.1016/j.jcat.2016.05.006

  26. [26]

      J. Ye, Q. Ge, C. Liu, Chem. Eng. Sci. 135(2015) 193-201. doi: 10.1016/j.ces.2015.04.034

  27. [27]

      A. Gurlo, S. Lauterbach, G. Miehe, H. Kleebe, R. Riedel, J. Phys. Chem. C 112(2008) 9209-9213.

  28. [28]

      K. Reyes-Gil, E. Reyes-García, D. Raftery, J. Phys. Chem. C 111(2007) 14579-14588. doi: 10.1021/jp072831y

  29. [29]

      M. Epifani, P. Siciliano, A. Gurlo, N. Barsan, U. Weimar, J. Am. Chem. Soc. 126(2004) 4078-4079. doi: 10.1021/ja0318075

  30. [30]

      A. Gurlo, G. Miehe, R. Riedel, Chem. Commun. (Camb.) 19(2009) 2747-2749.

  31. [31]

      A. Gurlo, P. Kroll, R. Riedel, Chem. Eur. J. 14(2008) 3306-3310. doi: 10.1002/chem.200701830

  32. [32]

      J.A. Sans, R. Vilaplana, D. Errandonea, et al., Nanotechnology 28(2017)205701. doi: 10.1088/1361-6528/aa6a3f

  33. [33]

      D. Yu, S. Yu, S. Zhang, et al., Adv. Funct. Mater. 13(2003) 497-501. doi: 10.1002/adfm.200304303

  34. [34]

      M.S. Frei, M. Capdevila-Cortada, R. García-Muelas, et al., J. Catal. 361(2018) 313-321. doi: 10.1016/j.jcat.2018.03.014

  35. [35]

      W. Yin, D. Esposito, S. Yang, et al., J. Phys. Chem. C 114(2010) 13234-13240.

  36. [36]

      F. Liao, Y. Huang, J. Ge, et al., Angew. Chem., Int. Ed. 50(2011) 2162-2165. doi: 10.1002/anie.201007108

  37. [37]

      J.C. Matsubu, S. Zhang, L. DeRita, et al., Nat. Chem. 9(2017) 120-127. doi: 10.1038/nchem.2607

  38. [38]

      T. Bielz, H. Lorenz, W. Jochum, et al., J. Phys. Chem. C 114(2010) 9022-9029.

  39. [39]

      E.M. Köck, M. Kogler, M. Grünbacher, et al., J. Phys. Chem. C 120(2016) 15272-15281. doi: 10.1021/acs.jpcc.6b04982

  40. [40]

      G. Popova, T. Andrushkevich, Y. Chesalov, E. Stoyanov, React. Kinet. Catal. Lett. 41(2000) 805-811.

  41. [41]

      S. Kattel, B. Yan, Y. Yang, J.G. Chen, P. Liu, J. Am. Chem. Soc. 138(2016) 12440-12450. doi: 10.1021/jacs.6b05791

  42. [42]

      S.V. Ryazantsev, V.I. Feldman, L. Khriachtchev, J. Am. Chem. Soc. 139(2017) 9551-9557. doi: 10.1021/jacs.7b02605

  43. [43]

      K. Larmier, W.C. Liao, S. Tada, et al., Angew. Chem., Int. Ed. 56(2017) 2318-2323. doi: 10.1002/anie.201610166

  44. [44]

      S.T. Teklemichael, M.D. McCluskey, J. Phys. Chem. C 116(2012) 17248-17251. doi: 10.1021/jp303835a

• 

  Figure 1  Structural characterizations of c-In₂O₃ and rh-In₂O₃, (a) XRD, (b) XRD of precursor before calcination. Diagrammatic sketch of c-In₂O₃ (c) and (d) rh-In₂O₃, (e) Raman spectra.

  下载: 全尺寸图片 幻灯片
  

  Figure 2  Catalytic performance of the catalysts. (a) catalytic performance at the reaction temperatures from 260 ℃ to 360 ℃ with H₂/CO₂ = 4/1, 4 MPa and 16, 000 mL g^-1 h^-1. (b) Arrhenius plot for CO₂ conversion. (c) Evolution of the methanol STY with time on stream. (d) The effect of pressure on CO₂ hydrogenation. The reaction condition: 340 ℃, H₂/CO₂ = 4/1 and 16, 000 mL g^-1 h^-1. (e) The effect of GHSV on CO₂ hydrogenation. The reaction condition: 340 ℃, H₂/CO₂ = 4/1 and 4 MPa.

  下载: 全尺寸图片 幻灯片
  

  Figure 3  Temperature programmed reduction and desorption of c-In₂O₃ and rh-In₂O₃. (a) H₂-TPR, (b) CO₂-TPD, (c) CH₃OH-TPD and (d) CO-TPD.

  下载: 全尺寸图片 幻灯片
  

  Figure 4  In-situ DRIFTS spectra of CO₂ adsorption and conversion over the catalysts, (a) c-In₂O₃ and (b) rh-In₂O₃. HCOOH adsorption and conversion over the catalysts, (c) c-In₂O₃ and (d) rh-In₂O₃.

  下载: 全尺寸图片 幻灯片
•