CO₂ emissions caused by human activities have become a worldwide issue because of their role in climate change and the global ecological effect. The chemical transformation of CO₂ can not only mitigate the corresponding emissions but also produce value-added chemicals and fuels [1]. Hydrogenation of CO₂ to CH₃OH is one of the most promising approaches for its transformation [2-4]. The most attractive and well-investigated catalysts for this process are copper-based systems, e.g., Cu/ZnO/Al₂O₃ [5], Cu/ZrO₂ [6], and Cu/CeO₂ [2]. As an industrial catalyst, Cu/ZnO/Al₂O₃ has been extensively studied. Although many models have been designed to elucidate the CO₂ hydrogenation mechanism, the active site and valence of copper during methanol formation are still controversial [5, 7]. Nakamura et al. [8] proposed that the Cu-Zn alloy interface was the active site, while Behrens et al. [9] considered Zn decorated steps at the Cu surface as the active sites. Kasatkin et al. [10] inferred that lattice strain and defects in copper played an important role in its activity for the synthesis of methanol. Lunkenbein et al. [7] reported that ZnO layers over Cu/ZnO-based catalysts enhanced the synergetic catalytic effects between Cu and ZnO. These catalysts altered the surface chemical state through charge transfer from ZnO to Cu or by ZnO segregation around Cu⁺. The improved activity of these binary catalysts indicated that an effective interfacial interaction between ZnO and Cu is beneficial for methanol synthesis [1].

Recently, Rodriguez et al. [2] reported a copper-ceria catalyst for CO₂ hydrogenation. In addition to the higher activity and lower apparent activation energy than those of Cu/ZnO, the copper-ceria catalyst showed a different reaction pathway for the hydrogenation of CO₂ to methanol, which is very important and interesting for fundamental studies. Although the copper/ceria catalyst presented outstanding catalytic features [11], fundamental questions regarding the nature of the active species or sites for methanol formation remained unanswered. Recently, Shen et al. [12] reported that the copper-ceria structure consisted of Cu⁺-O_v-Ce³⁺ groups on the bottom layer and Cu⁰ atoms on the top layer. The low-temperature water-gas shift reaction occurred at the copper-ceria interfacial perimeter and the Cu⁺ site adsorbed CO, whereas the neighboring O_v-Ce³⁺ site dissociatively activated H₂O. Rodriguez et al. [2] suggested that the copper-ceria interface played an important role in CO₂ hydrogenation. Senanayake et al. [4] further showed that the Ce³⁺-catalyst part was closely related to the catalytic activity for CO₂ hydrogenation. In fact, the presence of redox couples (such as Cu²⁺/Cu⁺, Cu²⁺/Cu⁰, and Ce⁴⁺/Ce³⁺) and oxygen vacancies resulted in complex interfacial interactions. Moreover, copper was easily incorporated into the ceria lattice and formed a solid solution, which hindered the identification of its chemical state and valence. Using CO temperature-programmed reduction experiments, Luo et al. [13] identified three CuO species in a copper-ceria catalyst: finely dispersed CuO, bulk CuO, and Cu in the CeO₂ lattice. Moreover, the finely dispersed CuO species showed the highest activity, according to the results of CO oxidation experiments. However, Gamarra et al. [14] found that the activity for preferential oxidation of CO in H₂-rich streams was related to Cu⁺-ceria active sites.

Herein, copper-ceria solid solution (CuCeO_x) and ceria-supported copper oxide (CuO/CeO₂) samples were synthesized to investigate the effect of the copper species and valence states on CO₂ hydrogenation. The active sites of the as-prepared catalysts for CO₂ hydrogenation were systematically examined. The results of the experiments demonstrated that the Cu-O_v-Ce_x species within the as-prepared catalysts had the highest CO₂ hydrogenation activity, and revealed that higher Cu-O_v-Ce_x fractions in the catalyst corresponded to a higher turnover frequency (TOF) of CH₃OH production.

CuCeO_x-Y (Y = 1, 2, 3, and 4) samples were synthesized by the hydrothermal method, and the corresponding catalysts with different CuO loadings were denoted as CuCeO_x-1, CuCeO_x-2, CuCeO_x-3, and CuCeO_x-4, respectively. Typically, specific amounts of Cu(NO₃)₂·3H₂O, Ce(NO₃)₃·6H₂O, cetyltrimethylammonium bromide (CTAB), and urea were dissolved in H₂O. After stirring, the solution was hydrothermally treated at 353 K for 24 h and then at 393 K for 12 h. Then, the products were washed, dried at 353 K, and calcined at 823 K for 5 h. In addition, CuO/CeO₂-Y (Y = 1 and 2) samples were synthesized by incipient impregnation. Then, the products were dried at 353 K for 24 h and calcined at 823 K for 5 h. The obtained catalysts with different CuO loadings were labeled CuO/CeO₂-1 and CuO/CeO₂-2. The CeO₂ support was prepared as the preparation method of CuCeO_x-Y, without the addition of Cu(NO₃)₂·3H₂O [15]. Detailed information on the samples is provided in Table 1.

N₂ adsorption isotherms were measured on a Tristar II 3020 autosorption analyzer (Micromeritics Instrument Corporation, USA). Before the measurements, the samples were outgassed in flowing N₂ at 453 K for 6 h. Surface areas were obtained by the Brunauer-Emmett-Teller (BET) method. X-ray fluorescence (XRF) measurements were conducted on an AxiosmAX (PANalytical, Holland) spectrometer. X-ray diffraction (XRD) patterns were obtained on an XRD-7000 (Shimadzu, Japan) diffractometer with Cu K_α radiation (λ = 1.5408 Å) at 40 kV and 30 mA. For the ex situ XRD measurements, the samples were pretreated in reduction or reaction conditions at the corresponding temperatures, and then immediately transferred to test under Ar protection. The microstrain values of the samples were calculated from Williamson-Hall (W-H) plots, and the Cu crystal size was calculated by the Scherrer equation. High-resolution transmission electron microscopy (HRTEM) images were collected on a JEM-2010F (JEOL, Japan) instrument with an accelerating voltage of 200 kV. X-ray photoelectron spectroscopy (XPS) measurements were performed on an AXIS-ULTRA DLD-600W (Shimadzu, Japan) spectrometer using Al K_α radiation as the excitation source (hν = 1486.6 eV). All binding energies were calibrated by the C 1s peak at 285 eV. Before the XPS measurements, the catalysts were reduced by hydrogen at 633 K and protected by Ar atmosphere. Raman spectra were measured using a LabRAM HR800 (Horiba JobinYvon, France) spectrometer. Sample excitation was carried out using a 532 nm frequency-doubled Nd:YAG laser. Ten Raman spectra of each sample were recorded with an integration time of 30 s for each spectrum, at a laser power of 20 mW. Hydrogen temperature-programmed reduction (H₂-TPR) experiments were carried out using a VDSorb-91i (VODO, China) chemisorption analyzer equipped with a thermal conductivity detector. During the H₂-TPR measurement, 20 mg of dried sample was treated by 5 vol% H₂/Ar at a flow rate of 30 mL/min. Before the measurement, the dried sample was pretreated in Ar at 573 K for 1 h. The heating temperature was increased from 323 to 973 K at 10 K/min. H₂ consumption was monitored using a thermal conductivity detector and normalized by a copper oxide standard. Deconvolution of the recorded TPR profiles was performed with the Origin 9.0 software using Gaussian and Lorentzian functions; the areas of α, β, and γ peaks were also calculated using Origin 9.0. The overall TOF values of CH₃OH were calculated assuming that all active sites were based on copper-ceria interfaces and that the concentration was equal to the overall H₂ consumption. In situ diffuse reflection infrared Fourier transform spectroscopy (DRIFTS) spectra were collected by a Tensor II (Bruker, Germany) equipped with a mercury-cadmium-telluride (MCT) detector. Before the measurement, each catalyst was reduced in a H₂/Ar mixture at 633 K for 30 min and then purged with 20 mL/min Ar at 573 K for 20 min. Subsequently, the catalyst was cooled down to the selected temperature and the background spectrum (256 scans with a resolution of 4 cm^‒1) was obtained at the corresponding temperature in Ar flow. During the reaction, the in situ DRIFTS spectra were recorded by collecting 64 scans at the resolution of 4 cm^‒1. Then, 1 mL/min CO + 99 mL/min He, 10 mL/min CO₂ + 90 mL/min Ar, or 10 mL/min CO₂ + 30 mL/min H₂ gas mixtures were introduced, and the volume flow rate was controlled by a mass flow controller. The vibrational signatures of adsorbed CO were used to identify the various copper structures. After the pretreatment, a CO + He mixture was added and thoroughly adsorbed until the CO bands no longer changed, indicating CO-saturated adsorption on the samples.

CO₂ hydrogenation experiments were performed in a continuous-flow fixed-bed quartz tubular reactor. Prior to the test, the samples were reduced in 20 vol% H₂/Ar for 2 h at 633 K. Online analysis of the products was carried out by a gas chromatograph. The CO₂ conversion [X(CO₂)], methanol selectivity [S(products, CH₃OH)], and space-time yield (STY) of the products were determined by an internal normalization method. The reaction conditions for CO hydrogenation were similar to those of the CO₂ hydrogenation experiments, except that the reaction gas was changed from CO₂ to CO. All data were collected after continuous reaction for 2 h (unless otherwise specified). Further details on the evaluation of the catalytic activities are provided in the Supplementary Data.

A series of CuCeO_x-Y solid solution catalysts with different CuO contents were prepared by the hydrothermal method. The CuO loading was measured by XRF and the corresponding data are summarized in Table 1. CuO/CeO₂-Y catalysts were prepared as a reference by conventional impregnation, and the Cu supported amount was loaded based on the maximum and minimum CuO contents of the CuCeO_x-Y samples. The specific surface area was calculated based on the N₂ adsorption isotherms (Fig. S1), and the corresponding values are summarized in Table 1. The specific surface area of the as-prepared catalysts was lower than that of the pristine CeO₂ sample and monotonously decreased with increasing CuO content; the same effect was also observed in previous reports and may be due to microstructural changes taking place after Cu addition into CeO₂ [16].

HRTEM was used to characterize the microstructure and crystallinity of the as-prepared samples. As shown in Fig. 1(a), the lattice fringes of CuCeO_x-1 particles displayed an interplanar spacing of 0.31 nm, corresponding to the (111) planes of CeO₂ [17]. The absence of CuO phases indicated that they were highly dispersed or incorporated into the ceria lattice. To further examine the CuO species, scanning transmission electron microscopy-energy dispersive X-ray (STEM-EDX) elemental mapping of CuCeO_x-1 was carried out, and the results are shown in Fig. 1(c). The STEM-EDX elemental mapping shows the uniform distribution of the three elements (Cu, Ce, and O) in CuCeO_x-1, denoting highly dispersed copper on ceria or a copper-ceria solid solution structure. Compared with CuCeO_x-1, the HRTEM image of CuO/CeO₂-1 (Fig. 1(b)) showed not only the (111) planes of CeO₂, but also those of CuO. Further HRTEM results, elemental maps, and line-scan profiles for the other samples are shown in Figs. S2, S3, and S4, respectively. CuO nanoparticles can be clearly observed in CuO/CeO₂-2 (Fig. S2(d)). As the CuO content in CuCeO_x-Y increased, CuO nanoparticles were also observed for CuCeO_x-4 (Fig. S2(c)).

The XRD patterns of the as-prepared samples are shown in Fig. 2(a). The typical diffraction peaks of the fluorite phase of CeO₂ were observed for all samples. No obvious characteristic peaks of CuO were observed for the CuCeO_x-1 and CuCeO_x-2 samples. However, increasing the CuO content from CuCeO_x-3 to CuCeO_x-4 resulted in the gradual appearance of the characteristic CuO peaks, with increasing intensity. In addition, no segregated Cu₂O or Cu phases were identified in the XRD patterns. Turning to the CuO/CeO₂-Y samples, the distinctive diffraction peaks of CuO could be clearly observed even in CuO/CeO₂-1, due to the formation of segregated or bulk CuO on these samples, consistent with the HRTEM results (Figs. 1 and S2). The diffraction peaks of CuO were absent in CuCeO_x-1, which contained a similar CuO amount to CuO/CeO₂-1, suggesting the formation of a CuCeO_x solid solution. Moreover, the intensity of the CuO diffraction peaks of CuCeO_x-4 was much lower than that of CuO/CeO₂-2, despite the similar CuO content of the two samples, indicating a higher bulk CuO content in CuO/CeO₂-2 than CuCeO_x-4. Compared with CuO/CeO₂-Y, the absence or lower intensity of the CuO diffraction peaks in the CuCeO_x-Y samples suggested copper dispersion on ceria or copper incorporation into the ceria fluorite lattice as a copper-ceria solid solution; this was also shown by the HRTEM and elemental mapping results in Figs. 1 and S2–S4.

As shown in Table 1, the lattice parameters of all samples, calculated from the Rietveld refinement analysis of the XRD peaks, were close to those of pristine CeO₂. The copper-ceria solid solution could still form in the CuCeO_x-Y samples [16, 18]. According to the values calculated from the W-H plots, the microstrain of the CuO/CeO₂-Y samples was slightly higher than that of pristine CeO₂. However, the value calculated for the CuCeO_x-Y samples was much higher than those of CuO/CeO₂-Y or CeO₂. The different microstrain of the CuCeO_x-Y sample cannot be ascribed to the size effects of ceria particles, due to the lack of obvious particle size changes (Table 1) [16, 19], which indicates that in the CuCeO_x-Y samples copper may be partially incorporated into the ceria lattice. The incorporation of copper into the ceria lattice is expected to be closely related to the synthesis method. Copper-ceria species easily co-precipitated to form single-phase Cu-Ce by rearrangement and polycondensation of CTAB with precipitation of urea. Based on theoretical simulations, the solubility limit of CuO within ceria was previously estimated to be 15 wt% [16]. Then, the excess copper would precipitate and aggregate into bulk copper species, as observed for the CuCeO_x-3 and CuCeO_x-4 samples based on the HRTEM and XRD results (Figs. 2 and S2). In addition, copper incorporation into the ceria lattice could force out oxygen ions to form vacancies for charge balance, as will be discussed later.

Figure 2(b) shows the Raman spectra of the as-prepared samples. All catalysts showed a main peak around 460 cm^‒1, corresponding to the triply degenerate F_2g mode of fluorite CeO₂ [15]. In particular, the Raman spectra of the CuCeO_x-Y samples showed a blue shift along with full width at half maximum (FWHM) broadening, suggesting the incorporation of Cu atoms into the CeO₂ lattice and the formation of a copper-ceria solid solution. At the same time, the main peak of the samples prepared by impregnation showed negligible changes. In addition, the CuCeO_x-Y samples showed a broad band between 500 and 650 cm^‒1 (inset of Fig. 2(b)), which was related to oxygen vacancies and resulted from the substitutional incorporation of Cu ions into the ceria lattice [16]. However, CuO exhibited a monoclinic structure (C2/c space group) in which the Cu cations had four close oxygen neighbors, and CeO₂ had a cubic structure with eight oxygen neighbors for each Cu cation. Due to the size mismatch with the Ce cations, the incorporated Cu cations could not fit well in the standard positions [16, 19, 20]. The mismatch resulted in substantial structural perturbations, such as oxygen vacancies. Thus, Cu atoms directly bonded to the oxygen vacancies of ceria, forming Cu-O_v-Ce_x species. Furthermore, the blue shift of the F_2g mode of the CuCeO_x-Y samples reflected the amount of Cu incorporated into ceria and did not monotonously increase with increasing CuO content. In addition, the lowest Raman shift value of CuCeO_x-2 (442 cm^‒1) denoted maximum Cu incorporation into the ceria lattice. The Raman spectra of the samples indicated that Cu cations were easily incorporated into the ceria lattice when the CuO content was low, whereas bulk CuO tended to form at high CuO contents. The CuO/CeO₂-Y samples mainly contained bulk CuO species, which was also observed in the CuCeO_x-3 and CuCeO_x-4 samples containing high amounts of Cu.

Herein, the copper species of CuCeO_x-Y were primarily classified as Cu incorporated into ceria (Cu-O_v-Ce_x), dispersed CuO (D-CuO-CeO₂), and bulk CuO (B-CuO-CeO₂) over the CeO₂ surface, according to the HRTEM, XRD, and Raman results, as well as to a previous report [12].

CeO₂ had a wide range of applications, due to its Ce⁴⁺/Ce³⁺ redox cycle along with excellent oxygen storing and release properties under oxidizing and reducing conditions [21]. These redox properties can be further optimized by combination with precious and transition metals. All samples should be activated by H₂ reduction before the activity tests. Then, Cu²⁺ and Ce⁴⁺ could be further reduced to Cu⁺ (or Cu⁰) and Ce³⁺, respectively, in a reducing atmosphere. In order to examine the elemental distribution on the surface after reduction, the Cu 2p, Ce 3d, and O 1s states were analyzed by XPS. Fig. 3(c) shows the deconvolution of the Ce 3d core-level XPS spectra of the samples after reduction at 633 K, revealing a mixture of Ce⁴⁺ and Ce³⁺ states. Six photoelectron peaks (v, v'', v''', u, u'', u'''), corresponding to three pairs of spin-orbit split doublets, could be attributed to Ce⁴⁺. The other peaks (v₀, v', u₀, u') denoted the presence of Ce³⁺ [22]. The concentration of Ce³⁺ shown in Table 2 was calculated from the intensities of the u₀ (v₀) and u' (v') peaks, relative to the total intensity of the Ce 3d region. As the CuO content increased, the Ce³⁺ proportion in the CuO/CeO₂-Y and CuCeO_x-Y series increased from 23.05% to 33.55% and 21.69% to 36.41%, respectively. This suggested substantial ceria reduction, accompanied by the reduction of Cu²⁺ to Cu⁰ or Cu⁺ and the generation of oxygen vacancies. The reduction of ceria was also promoted by hydrogen spillover caused by copper.

The XPS analysis further confirmed the chemical states of copper, as shown in Fig. 3(a). The spectra displayed two groups of peaks at 928–939 and 950–958 eV, which can be assigned to Cu 2p_3/2 and Cu 2p_1/2, respectively. Both peaks could be deconvoluted into Cu²⁺ (934.5 and 954.2 eV for Cu 2p_3/2 and 2p_1/2, respectively) and reduced copper (932.5 and 952.5 eV for Cu 2p_3/2 and 2p_1/2, respectively) components [24]. The weak peak between 940 and 947 eV corresponded to the relevant satellites. The reduced Cu states resulted from H₂ reduction; the presence of a small quantity of Cu²⁺ could be due to air exposure and re-oxidization during the XPS measurement. To further distinguish the reduced Cu states, we examined the Cu L₃VV Auger lines. As shown in Fig. 3(b), the peaks at kinetic energies of 918.7 and 916.1 eV represented Cu⁰ and Cu⁺ species, respectively, suggesting the co-existence of Cu⁰ and Cu⁺ in all reduced catalysts. The Cu²⁺ signal at 917.3 eV was also detected [12]. It should be noted that the CuO contents in CuO/CeO₂-1 and CuCeO_x-1 were too low to obtain visible XPS signals (Fig. S5). The Cu⁺/Cu ratios were calculated from a semiquantitative analysis of the Auger L₃VV lines, and are listed in Table 2. In CuO/CeO₂-2, copper was mainly present as Cu⁰, and the proportion of Cu⁺ was only 14.79%. In contrast, the Cu⁺/Cu ratio was as high as 62.99% in CuCeO_x-2, while the corresponding ratios in CuCeO_x-3 and CuCeO_x-4 were 45.51% and 35.69%, respectively. This apparently indicated the presence of different Cu chemical states in CuO/CeO₂-Y and CuCeO_x-Y after H₂ reduction. A previous study reported that further reduction of Cu⁺ was relatively difficult after its incorporation into the ceria fluorite lattice. Moreover, based on Wagner plots, previous reports demonstrated the presence of Cu⁺ in copper-ceria catalysts [16, 19, 24]. In situ X-ray absorption near-edge spectroscopy (XANES) results also showed that the Cu²⁺ species in copper-ceria solid solutions did not reach the metallic Cu state, but their oxidation state during the H₂ reduction process varied between 0 and +1 [24].

The O 1s XPS spectra of the as-prepared samples are shown in Fig. 3(d). The deconvolution revealed three kinds of surface oxygen species, corresponding to lattice oxygen (O_M-O), adsorbed oxygen (O_ad), and oxygen vacancies (O_vac) on the ceria substrate [25]. Based on the analysis summarized in Table 2, the proportion of oxygen vacancies in CuO/CeO₂-1 and CuO/CeO₂-2 was 23.69% and 25.82%, respectively. For the CuCeO_x-Y series, the maximum O_vac proportion was 29.66% in CuCeO_x-2, and a slightly lower value was observed for the other samples. The average proportion of oxygen vacancies in the CuCeO_x-Y series was slightly higher than that of the CuO/CeO₂-Y samples, which might indicate Cu incorporation into ceria, which then improved the mobility of oxygen.

In summary, the relevant copper-ceria species could be classified as Cu-O_v-Ce_x, D-CuO-CeO₂, and B-CuO-CeO₂ on CeO₂. After reduction, the electronic structures of the copper clusters consisted of Cu⁺-O_v-Ce_x and Cu⁰ species adsorbed on the ceria surface. More importantly, different synthetic methods resulted in different site fractions and chemical states at the copper-ceria interface.

To study the reducibility and copper-ceria site fractions of the samples, we conducted H₂-TPR measurements; the results are shown in Fig. 4. Compared with pure CuO or CeO₂, which were normally reduced above 633 K, the copper-ceria interaction improved the reducibility, owing to the effect of metal-support interactions [18]. The CuCeO_x-Y reduction started at a temperature below 450 K; the lower onset reduction temperature indicated higher reducibility of CuCeO_x-Y than CuO/CeO₂-Y samples. Upon increasing the CuO content, the reducibility of the CuCeO_x-Y samples increased due to the shift of the onset reduction temperature from about 430 K for CuCeO_x-1 to 410 K for CuCeO_x-4. The overall H₂ consumptions were then calculated and are summarized in Table 3. It should be noted that the H₂ consumption originated from the reduction not only of CuO but also of CeO₂ [20], which was the reason why the H₂ consumption exceeded the theoretical amount required for the complete Cu²⁺ reduction to Cu⁰. Thus, ceria reduction should be taken into account when considering the possible redox equilibrium between copper and cerium cations in the interfacial region.

As indicated in Fig. 4, the H₂-TPR profiles of the CuCeO_x-Y samples displayed three distinct reduction peaks (α, β, and γ) corresponding to the different copper-ceria states. The γ peak, whose intensity gradually increased with increasing CuO content in CuCeO_x-3 and CuCeO_x-4, could be assigned to the B-CuO-CeO₂ species, according to the XRD results. As reported in previous studies, the reducibility of finely dispersed CuO was enhanced by ceria, which led to a lower reduction temperature that further promoted the subsequent reduction of Cu-O_v-Ce_x via hydrogen spillover [15, 23]. Thus, the α and β peaks were assigned to the reduction of the D-CuO-CeO₂ and Cu-O_v-Ce_x species, respectively. The CuO/CeO₂-Y samples showed only one broad γ peak around 520 K, whose intensity increased with increasing CuO content. The fractions of each type of site are summarized in Table 3 on the basis of the H₂-TPR intensities by peak resolution and fitting. At the lowest CuO loading amount (CuCeO_x-1), a relatively low fraction of B-CuO-CeO₂ (~2%) was present; this fraction significantly increased to ca. 52% for CuCeO_x-4, as the CuO loading increased to 17 wt%. The fraction of D-CuO-CeO₂ sites exhibited the opposite trend, with an obvious drop from 44% for CuCeO_x-1 to 14% for CuCeO_x-4. On the other hand, the Cu-O_v-Ce_x fraction first increased and then decreased, reaching its maximum at a CuO loading of 4.52 wt% (CuCeO_x-2), which was consistent with the Raman and XPS results. However, B-CuO-CeO₂ was the main phase in the CuO/CeO₂-Y samples as indicated by Fig. 4.

Despite the similar CuO content of samples prepared by different methods, e.g., CuO/CeO₂-1 and CuCeO_x-1 or CuO/CeO₂-2 and CuCeO_x-4, the H₂ consumptions of the CuO/CeO₂-Y samples were still much higher than those of the CuCeO_x-Y series, which cannot be explained solely by integral deviation or co-reduction of ceria. The oxidation state of reduced Cu varied from 0 to +1, rather than reaching the pure metallic state, and the analysis of the XPS results showed that the Cu species of the copper-ceria solid solution were almost fully reduced to a Cu₂O-like phase after the reaction [16]. However, copper was mainly present as Cu⁰ in CuO/CeO₂-Y. Herein, we propose that the Cu²⁺ species in D-CuO-CeO₂ and B-CuO-CeO₂ could be reduced to Cu⁰, while Cu-O_v-Ce_x was only reduced to Cu⁺, which was consistent with the results of XPS analysis.

The presence of surface Cu⁺ states in the reduced copper-ceria samples was also demonstrated by in situ DRIFTS with CO adsorption. As shown in Fig. 6(a), all reduced copper-ceria samples exhibited a CO absorption peak at 2112 cm^‒1, corresponding to Cu⁺-carbonyl vibrations [12, 14], which demonstrated the presence of the Cu⁺ chemical state in all samples. However, the intensity of the Cu⁺-carbonyl peak obviously depended on the synthetic method and the CuO content amount. The peak intensity was much higher for the CuCeO_x-Y than CuO/CeO₂-Y samples, indicating a higher amount of Cu⁺ in the CuCeO_x-Y catalysts. Importantly, it was found that the saturated CO adsorption matched well the trend of the Cu-O_v-Ce_x fraction in the CuCeO_x-Y samples. Previous investigations already reported that Cu⁺ is relatively difficult to reduce once incorporated into the ceria fluorite lattice [11, 16]. In other words, Cu-O_v-Ce_x can only be reduced to Cu⁺-O_v-Ce_x, whereas D-CuO-CeO₂ and B-CuO-CeO₂ can be reduced to D-Cu⁰-CeO_x and B-Cu⁰-CeO_x, respectively. Furthermore, a certain amount of Cu⁺-O_v-Ce_x was present within the CuO/CeO₂-Y samples subjected to H₂ reduction, which indicated that some copper migrated into the ceria lattice during the reduction treatment and formed Cu⁺-O_v-Ce_x.

To further test and confirm the site switching of copper under reducing and reaction conditions, we carried out ex situ XRD measurements. Representative CuO/CeO₂-1 and CuCeO_x-2 samples were selected for these tests. As shown in Fig. 5(a), the CuO/CeO₂-1 sample initially contained bulk CuO species, which were gradually reduced to metallic Cu through Cu₂O as the reduction temperature increased. This metallic species remained stable under CO₂/H₂ reaction conditions from 573 to 673 K (Fig. 3(c)), indicating that metallic copper could not be oxidized during the reaction [22]. The lattice constant of ceria increased from 5.4107 to 5.4206 Å under reducing conditions, due to partial incorporation of Cu into the ceria lattice [13, 22]. On the other hand, the CuCeO_x-2 sample showed no evidence of Cu, Cu₂O, or CuO species under all reducing or reaction conditions, because of the relative stability or high dispersion of copper incorporated into the ceria fluorite lattice (Figs. 5(b) and 4(d)). The slight increase of the lattice constant from 5.4136 to 5.4156 Å under reduction conditions was ascribed to the formation of Ce³⁺ or to oxygen generation.

The interactions of CO₂ with CuCeO_x-Y and CuO/CeO₂-Y were investigated by in situ DRIFTS at 303 K and 1 bar. As shown in Fig. 6(b), the adsorption of pure CO₂ on the polycrystalline CeO₂ surface generated mono- and bidentate carbonate (CO₃^2‒) species, based on the peaks at 1272, 1375, 1586, and 1658 cm^‒1 [27]. On the other hand, the two peaks at 1272 and 1611 cm^‒1 were ascribed to carboxylate (CO₂^δ‒) species on the CuCeO_x-Y and CuO/CeO₂-Y surface [2, 4], which indicated that the copper-ceria interface activated CO₂ as carboxylate, in addition to carbonate. The formation of these two adsorbates was mainly related to the formation of carbonate-like species on substantially different adsorption sites, such as metal and oxygen ions [27]. The carbonate species involved charge transfer from a surface oxygen to the approaching CO₂, and a molecule bend to form [O-CO₂]^2‒ complexes. The carboxylate species was similar to the [M-CO₂^δ‒] complex, whose formation was associated with the metal ions exposed on the polar surface, in addition to oxygen ions [27]. All samples, including CeO₂, contained abundant Ce³⁺ sites after H₂ reduction, which may not be the reason for the occurrence of different adsorbed species. The copper-ceria interface with abundant Ce³⁺ sites was essential to stabilize CO₂ in the formation of CO₂^δ‒, whereas Ce³⁺ ions without copper decoration only resulted in the formation of CO₃^2‒.

Figure 6(c) illustrates the results of the interaction of CO₂ and H₂ with the samples, based on in situ DRIFTS experiments at 573 K, 1 bar, and CO₂/H₂ = 1:3. The formate intermediate was observed in the CO₂ hydrogenation [2, 4]. In particular, the peaks at 1468 and 1084 cm^‒1 were assigned to non-coordinated CO₃^2‒. The other peaks were assigned to HCOO^- species, and those at 1371 and 1583 cm^‒1 corresponded to symmetric and asymmetric OCO stretching modes. The peaks at 2939 and 2848 cm^‒1 were assigned to the CH stretch modes [6, 28]. Despite the presence of the formate intermediate in the CO₂ hydrogenation, this species was likely to be just a spectator in the reaction, due to its excessive stability. A previous theoretical study had shown that the mechanism mainly involved an initial reverse water-gas shift (RWGS) reaction, followed by sequential hydrogenation of CO to CH₃OH [4]. Thus, the addition of Cu to CeO₂ created a special Cu-ceria interface that activated CO₂ to CO₂^δ‒ and enabled further hydrogenation to methanol.

CO₂ hydrogenation was performed at 4 MPa, from 523 to 643 K. Under typical reaction conditions, all samples showed catalytic activity for CO₂ hydrogenation. In addition to methanol production, these catalysts also promoted the generation of CO through the RWGS. The performance of CuCeO_x-Y and CuO/CeO₂-Y catalysts gradually improved with increasing CuO content (Fig. 7(a)). Moreover, increasing the temperature from 523 to 643 K resulted in increased CO₂ conversion and decreased CH₃OH selectivity (Fig. S7).

The CO₂ conversion activity showed a significant variation with the synthetic method and was higher for the CuCeO_x-Y samples. The maximum CO₂ conversion reached about 10.8% for CuCeO_x-4 at 643 K, which was about twice that of CuO/CeO₂-2, considering the similar CuO loading. This different activity was also observed for CuCeO_x-1 and CuO/CeO₂-1. As shown in Table S1, the methanol STY of CuCeO_x-2 was 1.13 mmol/(g·h), which was higher than that of traditional copper-ceria catalysts, although still lower than that of traditional Cu/ZnO/Al₂O₃ catalysts. The average apparent activation energy of CuCeO_x-Y was 29.307 kJ/mol, much lower than the 45.727 kJ/mol value of CuO/CeO₂-Y, as shown in Fig. 7(b). The lower apparent activation energy suggested that the CuCeO_x-Y samples had higher catalytic activity.

In order to further confirm the role of CO as intermediate during the hydrogenation of CO₂ to methanol, CO hydrogenation was also carried out over CuO/CeO₂-2, CuCeO_x-2, CuCeO_x-3, and CuCeO_x-4, and the results are shown in Fig. S8. The CO conversion increased with increasing temperature, and the performance of the CuCeO_x-Y catalysts gradually improved with increasing CuO content. Taking into account the similar CuO content of samples prepared by different methods, CuCeO_x-4 showed a higher CO conversion than CuO/CeO₂-2, which was similar to the trend observed for CO₂ hydrogenation. Moreover, methanol was the main product of CO hydrogenation in Fig. S8, confirming that CO may be an intermediate species during CO₂ hydrogenation over copper-ceria catalysts.

The overall TOF values of CH₃OH were calculated by assuming that all active sites were based on the copper-ceria interface and that the concentration was equal to the overall H₂ consumption [2, 18]. Despite the presence of three different copper-ceria interfaces whose contribution may not be same, the overall TOF of methanol production per active site per second was minimum, based on this assumption. As shown in Fig. 7(c), under typical catalytic experimental conditions the TOF of the CuO/CeO₂-Y samples increased with increasing CuO content. In comparison, the CuO content dependence of the activity of the CuCeO_x-Y catalysts exhibited a volcano-like trend, and the optimum catalysts was CuCeO_x-2, which generated 9.12 × 10^‒24 μmol CH₃OH per active site per second. In addition, the rate of methanol production on the CuCeO_x-Y catalysts was higher than that of the CuO/CeO₂-Y series, especially for CuCeO_x-1 (whose rate was ca. eight times higher than that of CuO/CeO₂-1) and CuCeO_x-4 (whose rate was ca. four times higher than that of CuO/CeO₂-2). Moreover, the active site fractions are presented as a linear plot in Fig. 7(c), which shows an obvious agreement between the trends of the TOF of CH₃OH production and the fraction of Cu-O_v-Ce_x sites plotted as a function of the Cu weight loading for the CuCeO_x-Y series. To ensure a meaningful comparison, both concentrations and rates were normalized to the overall number of active sites. According to the above analysis (such as Fig. 6(a)), Cu-O_v-Ce_x sites may form in CuO/CeO₂-1 and CuO/CeO₂-2 by site switching under reducing conditions; the corresponding fractions were estimated to be ca. 2% and 10%, respectively, by theoretical simulations based on activity data. The correlation between Cu-O_v-Ce active sites and activity in the CH₃OH synthesis indicated that these sites played a major role in CH₃OH generation. Fig. 7(d) highlights a linear correlation between the intensity of Cu⁺-carbonyl in DRIFTS and the TOF of CH₃OH production. The presence of Cu⁺ was due to the stabilization by the ceria lattice, as suggested by the above discussion, which clearly demonstrated that the active sites for the conversion of CO₂ to CH₃OH were closely related to partially reduced Cu⁺ ions in the Cu-O_v-Ce_x species.

The intimate contact between copper and ceria is the key factor influencing the performance of copper-ceria catalysts. Nevertheless, the identification of active sites is still a challenging task, due to the complexity of copper-ceria interactions, such as Cu²⁺/Cu⁺, Cu⁺/Cu⁰, and Ce⁴⁺/Ce³⁺ redox couples, oxygen vacancies, solid solution formation, and electronic interactions between ceria and copper. To better understand the activity of copper-ceria catalysts in CO₂ hydrogenation, catalysts with different copper states were prepared based on site-specific synthesis. The interactions between copper and ceria were classified as Cu²⁺-O_v-Ce_x, D-CuO-CeO₂, and B-CuO-CeO₂, and the fraction of each site type was quantified by H₂-TPR.

The reducibility and valence states of the copper-ceria interface were further clarified as discussed below. First, the copper-ceria interaction improved the reducibility, due to the effect of metal-support interactions, which may result in large electronic perturbations on the Cu metal particles supported on the CeO₂ surface and significantly facilitate CuO reduction. Moreover, ceria enhanced the reducibility of CuO and led to a lower reduction temperature, which further promoted subsequent reduction of Cu-O_v-Ce_x to Cu⁺-O_v-Ce_x via hydrogen spillover. Second, copper clusters on the ceria surface exhibited Cu⁺-O_v-Ce_x and Cu⁰ electronic structures after reduction. Third, the copper species adsorbed over ceria may migrate into the ceria lattice during reduction and form Cu⁺-O_v-Ce_x structures.

During the CO₂ hydrogenation, Cu⁺-O_v-Ce_x played an important role in methanol synthesis. CO₂ was not adsorbed on bare Cu, and strongly bound CO₃^2‒ species were identified on CeO₂. The Cu⁺-O_v-Ce_x interfacial species were beneficial for CO₂ adsorption as CO₂^δ‒, which facilitated the methanol synthesis. In addition, the results of CO hydrogenation experiments also suggested that CO may be an important intermediate during CO₂ hydrogenation over copper-ceria catalysts. Moreover, Cu⁺ species were essential for CO adsorption and could facilitate the subsequent hydrogenation to methanol. The results of in situ DRIFTS measurements confirmed that the stability of Cu⁺-carbonyl species was higher than that of the Cu⁰ and Cu²⁺-carbonyl ones. Moreover, a linear relationship was observed between the intensity of Cu⁺-carbonyl peaks and the TOF of CH₃OH formation, which further implied that the Cu⁺-O_v-Ce_x species played an important role in the methanol synthesis.

Finally, the above correlations between the structure and activity of CuO/CeO₂-Y and CuCeO_x-Y catalysts can be established based on three basic factors. The first is related to the different structure of the catalysts. The copper-ceria solid solution samples (CuCeO_x-Y) could facilitate the reduction of surface copper oxide species and achieve higher hydrogenation activity by lowering the energy for H₂ activation and generating more oxygen vacancies, due to their higher Cu-O_v-Ce_x fraction than the CuO/CeO₂-Y samples [14]. The oxygen vacancies further enhance the H₂ reactivity and the CH₃OH formation [29, 30]. The second factor is related to CO₂ activation. CO₂^δ- species stabilized by the abundant Ce³⁺ sites of Cu-O_v-Ce_x groups were essential reaction intermediates. Then, the Cu-O_v-Ce_x species could not only enhance the H₂ activity and generation of oxygen vacancies, but also activate CO₂ and promote its hydrogenation to methanol by hydrogen spillover. Ex situ XRD measurements showed that Cu⁰ cannot be oxidized by CO₂, which means that Cu⁰ cannot take part in the redox or catalytic circle [22, 31, 32]. The third factor is related to the CO intermediate, which was considered as the primary reaction mechanism of the copper-ceria interface. At this point, CO₂ was adsorbed on the abundant Ce³⁺ centers, activated to CO₂^δ-, and further hydrogenated to yield CO species through the RWGS reaction. CO was weakly adsorbed on Cu⁰ or Cu²⁺ and unable to complete the following reaction [2]. Alternatively, Cu⁺ could easily react with CO to form Cu⁺-carbonyl species and subsequently achieve the hydrogenation of CO to methanol. Accordingly, Cu⁺ ions incorporated in Cu-O_v-Ce_x species at the copper-ceria interface were proposed as the active sites for the hydrogenation of CO₂ to CH₃OH.

In summary, copper-ceria catalysts were prepared based on a site-specific synthesis, which demonstrated that copper-ceria solid solutions exhibited higher CO₂ hydrogenation activity and faster CH₃OH production than ceria-supported copper. The copper-ceria interface species were classified as Cu incorporated into ceria (Cu²⁺-O_v-Ce_x), dispersed CuO (D-CuO-CeO₂), and bulk CuO (B-CuO-CeO₂) over the CeO₂ surface. The valence state of copper was also analyzed and discussed. Once annealed in reduced gases, copper and ceria were easily reduced due to the metal-support interaction. The Cu²⁺-O_v-Ce_x species was reduced to Cu⁺-O_v-Ce_x, due to the stabilization by the ceria lattice, while B-Cu²⁺-CeO_x and D-Cu²⁺-CeO_x were reduced to B-Cu⁰-CeO_x and D-Cu⁰-CeO_x, respectively. The quantitative relationships between the concentration of copper incorporated into the ceria lattice and the CH₃OH production were also established, and showed that a higher Cu-O_v-Ce_x fraction within the as-prepared catalysts resulted in higher activity for methanol formation. The present findings provide new insights into the CO₂ hydrogenation mechanism, which can support effective strategies for catalyst design.

Supplementary data associated with this article can been found in the online version. Additional details on catalytic test, Figures of catalytic performance details, N₂ adsorption-desorption isotherms, TEM, CO adsorption, products selectivity, CO hydrogenation and stability test.

The authors thank the Analysis and Testing Center of HUST for analytical support.