The Study of Reverse Water Gas Shift Reaction Activity over Di ﬀ erent Interfaces: The Design of Cu-Plate ZnO Model Catalysts

: CO 2 hydrogenation to methanol is one of the main and valuable catalytic reactions applied on Cu / ZnO-based catalysts; the interface formed through Zn migration from ZnO support to the surface of Cu nanoparticle (ZnOx-Cu NP-ZnO) has been reported to account for methanol synthesis from CO 2 hydrogenation. However, the accompanied reverse water gas shift (RWGS) reaction signiﬁcantly decreases methanol selectivity and deactivates catalysts soon. Inhibition of RWGS is thus of great importance to a ﬀ ord high yield of methanol. The clear understanding of the reactivity of RWGS reaction on both the direct contact Cu-ZnO interface and ZnOx-Cu NP-ZnO interface is essential to reveal the low methanol selectivity in CO 2 hydrogenation to methanol and look for e ﬃ cient catalysts for RWGS reaction. Cu doped plate ZnO (ZnO:XCu) model catalysts were prepared through a hydrothermal method to simulate direct contact Cu-ZnO interface and plate ZnO supported 1 wt % Cu (1Cu / ZnO) catalyst was prepared by wet impregnation for comparison in RWGS reaction. Electron paramagnetic resonance (EPR), XRD, SEM, Raman, hydrogen temperature-programmed reduction (H 2 -TPR) and CO 2 temperature-programmed desorption (CO 2 -TPD) were employed to characterize these catalysts. The characterization results conﬁrmed that Cu incorporated into ZnO lattice and ﬁnally formed direct contact Cu-ZnO interface after H 2 reduction. The catalytic performance revealed that direct contact Cu-ZnO interface displays inferior RWGS reaction reactivity at reaction temperature lower than 500 ◦ C, compared with the ZnOx-Cu NP-ZnO interface; however, it is more stable at reaction temperature higher than 500 ◦ C, enables ZnO:XCu model catalysts superior catalytic activity to that of 1Cu / ZnO. This ﬁnding will facilitate the designing of robust and e ﬃ cient catalysts for both CO 2 hydrogenation to methanol and RWGS reactions.


Introduction
In the recent years, climate change and ocean acidification resulting from the emission of greenhouse gases have received widespread attention in the environmental and energy fields [1][2][3]. Consequently, the utilization and conversion of CO 2 to valuable chemicals have been proposed to alleviate CO 2 emission [4,5]. Among them, CO 2 hydrogenation to methanol is a promising way to which inspires us to construct Cu doped plate ZnO as a model catalyst to simulate the direct contact Cu-ZnO interface [28][29][30][31]. In order to clearly investigate the reactivity of RWGS reaction on the direct contact Cu-ZnO interface, a series of Cu doped plate ZnO (ZnO:X) have been prepared including the samples with exceeding doping limit. The reactivity over both plate ZnO:X and Cu/plate ZnO model catalysts were evaluated and their structures were confirmed through SEM, electron paramagnetic resonance (EPR), XRD and Raman characterizations. Finally, the RWGS reaction performance over direct contact Cu-ZnO interface and ZnOx-Cu-ZnO interface was comparatively explored and discussed. Figure S1 presented SEM images of the pristine plate ZnO and 1Cu/ZnO catalyst after reduction. The pure plate ZnO has regular plate-like morphology and relatively large particle size. For 1Cu/ZnO sample after reduction, Cu nanoparticles were highly dispersed on the plate ZnO and the morphology of the support was kept. Accordingly, Figure 1 displayed SEM images of ZnO:XCu model catalysts after reduction. All samples exhibited plate-like morphology, indicating that plate ZnO model catalyst morphology was not changed during Cu doping while the size especially the thickness decreased significantly from 1.5 µm ( Figure S1) to submicrometer. This means it is not realistic to use HRTEM related techniques to investigate their fine structures. Thus, the elemental mappings of ZnO:XCu model catalysts were also presented as shown in Figure 1. It is obvious that although no Cu particle can be clearly observed on the plate ZnO surface, Cu element distribution can be detected in EDS mapping. This suggests that Cu may be incorporated into the bulk ZnO. However, when the amount of Cu reached 6.5 wt % as shown in Figure 1e, a few Cu nanoparticles can be observed on plate ZnO surface. This reveals that the maximum doping amount of Cu should be lower than 6.5 wt %. Therefore, these data indicate that all samples should have formed the direct contact Cu-ZnO interface; the ZnO:6.5Cu sample should form additional ZnOx-Cu NP-ZnO interface caused by ZnOx migration during the reduction in H 2 .

Catalyst Characterization
Catalysts 2019, 9, x FOR PEER REVIEW 3 of 16 ZnO interface [28][29][30][31]. In order to clearly investigate the reactivity of RWGS reaction on the direct contact Cu-ZnO interface, a series of Cu doped plate ZnO (ZnO:X) have been prepared including the samples with exceeding doping limit. The reactivity over both plate ZnO:X and Cu/plate ZnO model catalysts were evaluated and their structures were confirmed through SEM, electron paramagnetic resonance (EPR), XRD and Raman characterizations. Finally, the RWGS reaction performance over direct contact Cu-ZnO interface and ZnOx-Cu-ZnO interface was comparatively explored and discussed. Figure S1 presented SEM images of the pristine plate ZnO and 1Cu/ZnO catalyst after reduction. The pure plate ZnO has regular plate-like morphology and relatively large particle size. For 1Cu/ZnO sample after reduction, Cu nanoparticles were highly dispersed on the plate ZnO and the morphology of the support was kept. Accordingly, Figure 1 displayed SEM images of ZnO:XCu model catalysts after reduction. All samples exhibited plate-like morphology, indicating that plate ZnO model catalyst morphology was not changed during Cu doping while the size especially the thickness decreased significantly from 1.5 μm ( Figure S1) to submicrometer. This means it is not realistic to use HRTEM related techniques to investigate their fine structures. Thus, the elemental mappings of ZnO:XCu model catalysts were also presented as shown in Figure 1. It is obvious that although no Cu particle can be clearly observed on the plate ZnO surface, Cu element distribution can be detected in EDS mapping. This suggests that Cu may be incorporated into the bulk ZnO. However, when the amount of Cu reached 6.5 wt % as shown in Figure 1e, a few Cu nanoparticles can be observed on plate ZnO surface. This reveals that the maximum doping amount of Cu should be lower than 6.5 wt %. Therefore, these data indicate that all samples should have formed the direct contact Cu-ZnO interface; the ZnO:6.5Cu sample should form additional ZnOx-Cu NP-ZnO interface caused by ZnOx migration during the reduction in H2.  EPR is sensitive to the paramagnetic defects and the coordination of doped Cu 2+ ions in ZnO. Figure 2 displayed the EPR profiles of the as-synthesized pristine plate ZnO and Cu doped plate ZnO. For all catalysts, the resonance at g = 1.953 is attributed to the shallow donor centers caused by the interstitial Zn (Zn i ) and surface O vacancy (V o ) [32][33][34]. Compared with the pristine plate ZnO, this peak intensity of plate ZnO:Cu samples was relatively weaker and decreased with the increasing of Cu doping amount. This phenomenon implies the incorporation of Cu into ZnO metal framework and the substitution of Zn i [35]. Meanwhile, the anisotropic hyperfine structure was observed in Cu containing plate ZnO samples. The resonances parameter g // = 2.384 and g ⊥ = 2.095 indicate that the Cu 2+ replaces the cation sites of ZnO [36][37][38][39][40]. Furthermore, g // > g ⊥ > g e = 2.002 unveils that Cu 2+ ions are subjected to tetragonally elongated distortion and in a ligand field of tetragonal symmetry, coordinated by six ligand atoms [32,41]. However, when the amount of Cu reached 3.5 wt %, the anisotropic hyperfine structure was not obvious enough or even disappeared, which is attributed to the increase of the long-range dipolar interaction between Cu 2+ ions along the increase of the Cu doping amount [37,42,43]. Thus, the EPR results confirm that Cu 2+ was successfully doped into the ZnO lattice.
EPR is sensitive to the paramagnetic defects and the coordination of doped Cu 2+ ions in ZnO. Figure 2 displayed the EPR profiles of the as-synthesized pristine plate ZnO and Cu doped plate ZnO. For all catalysts, the resonance at g = 1.953 is attributed to the shallow donor centers caused by the interstitial Zn (Zni) and surface O vacancy (Vo) [32][33][34]. Compared with the pristine plate ZnO, this peak intensity of plate ZnO:Cu samples was relatively weaker and decreased with the increasing of Cu doping amount. This phenomenon implies the incorporation of Cu into ZnO metal framework and the substitution of Zni [35]. Meanwhile, the anisotropic hyperfine structure was observed in Cu containing plate ZnO samples. The resonances parameter g// = 2.384 and g⊥ = 2.095 indicate that the Cu 2+ replaces the cation sites of ZnO [36][37][38][39][40]. Furthermore, g// > g⊥ > ge = 2.002 unveils that Cu 2+ ions are subjected to tetragonally elongated distortion and in a ligand field of tetragonal symmetry, coordinated by six ligand atoms [32,41]. However, when the amount of Cu reached 3.5 wt %, the anisotropic hyperfine structure was not obvious enough or even disappeared, which is attributed to the increase of the long-range dipolar interaction between Cu 2+ ions along the increase of the Cu doping amount [37,42,43]. Thus, the EPR results confirm that Cu 2+ was successfully doped into the ZnO lattice. respectively. The diffraction pattern of all samples showed the orientation prefers along (101) plane and the hexagonal wurtzite structure was kept during doping. However, when the amount of Cu achieved 3.5 wt %, another two diffraction peaks appeared at approximately 2θ = 42.3 and 43.3°, attributed to the diffraction of Cu2O (200) and Cu (111), respectively. The appearance of Cu2O and elemental Cu may be due to the auto-reduction of the Cu 2+ ions adsorbed on the surface of plate ZnO during the calcination in oxygen-free environment [44]. Figure 3b displayed the enlarged diffraction peak of Cu doped ZnO (002) plane, which clearly showed a left shift and an intensity decrease of this peak with a rise in Cu concentration. Such kinds of XRD diffraction peak shift and intensity decrease depend upon the preparation conditions like doping procedures and have been studied extensively in Cu doped ZnO [30,45]. However, ZnO:6.5Cu sample showed intensity increase and right shift of the (002) diffraction peak. Such a shift at higher Cu concentration is consistent with previous reported  [44]. Figure 3b displayed the enlarged diffraction peak of Cu doped ZnO (002) plane, which clearly showed a left shift and an intensity decrease of this peak with a rise in Cu concentration. Such kinds of XRD diffraction peak shift and intensity decrease depend upon the preparation conditions like doping procedures and have been studied extensively in Cu doped ZnO [30,45]. However, ZnO:6.5Cu sample showed intensity increase and right shift of the (002) diffraction peak. Such a shift at higher Cu concentration is consistent with previous reported results on Cu doped ZnO film, but the reasons are not clear [46]. The XRD patterns of the reduced ZnO:XCu Catalysts 2020, 10, 533 5 of 16 samples were shown in Figure S2. After reduction, Cu (111) phase can be observed over ZnO:3.5Cu and ZnO:6.5Cu samples except ZnO diffraction. Both EPR and XRD results demonstrate the doping of Cu into bulk plate ZnO; the direct contact Cu-ZnO interface will be produced after reduction in H 2 .
Catalysts 2019, 9, x FOR PEER REVIEW 5 of 16 results on Cu doped ZnO film, but the reasons are not clear [46]. The XRD patterns of the reduced ZnO:XCu samples were shown in Figure S2. After reduction, Cu (111) phase can be observed over ZnO:3.5Cu and ZnO:6.5Cu samples except ZnO diffraction. Both EPR and XRD results demonstrate the doping of Cu into bulk plate ZnO; the direct contact Cu-ZnO interface will be produced after reduction in H2. is assigned to the oxygen vibration. The E1(LO) mode originates from second-order Raman scattering, which is affected by impurities and/or defects (Zni, Vo) [47]. It can be clearly seen from the inset of Figure 4 that the E high 2 mode in ZnO:XCu exhibited red shift from 439 to 428 cm −1 compared with that of pristine ZnO. Furthermore, the E high 2 showed a significant intensity decrease and line shape broadness along with the increase of Cu doping. The red shift, weakening and broadening of E high 2 are the consequence of the defect formation and local lattice distortions caused by Cu doping [31,48]. Similar characteristics can be found in the previous publications [47,48]. This indicates the incorporation of Cu into ZnO, in good agreement with EPR and XRD results. The intensity ratios of E1(LO) to E high 2 were calculated as an indicator to reflect the corresponding defect concentrations of ZnO:XCu after the doping by the phonon confinement model (PCM) [47,49]. The wavenumber of E high 2 and the intensity ratio of E1(LO) to E high 2 were listed in Table 1. The value of the intensity ratio of E1(LO) to E high 2 increased in the order of ZnO < ZnO:0.5Cu < ZnO:1Cu < ZnO:1.5Cu < ZnO:3.5Cu < ZnO:6.5Cu, increasing along with Cu amount. Then, the correlation between the intensity ratio of E1(LO) to E high 2 and the Cu content were displayed in Figure S3. With the increase of Cu amount, ZnO:XCu model catalysts possessed more structure defects. The defect amounts of catalysts kept good linear with the Cu doping amount, except for the ZnO:6.5Cu sample although it has the highest value, indicating the non-linear increase of defects. Moreover, the Raman spectra and the intensity ratio of E1(LO) to E high 2 for the reduced samples were presented in Figure S4 and Table S1, respectively. More defects can be obtained due to oxygen vacancy formation during the reduction in H2. However, it should be noted that the increment of the intensity ratio of E1(LO) to E high 2 was gradually getting smaller along with Cu increase and there was hardly any increase for ZnO:6.5Cu after reduction. This indicates that this sample has quite different defect distribution compared with other samples with low Cu concentration. is assigned to the oxygen vibration. The E 1 (LO) mode originates from second-order Raman scattering, which is affected by impurities and/or defects (Zn i , V o ) [47]. It can be clearly seen from the inset of Figure 4 that the E high 2 mode in ZnO:XCu exhibited red shift from 439 to 428 cm −1 compared with that of pristine ZnO. Furthermore, the E high 2 showed a significant intensity decrease and line shape broadness along with the increase of Cu doping. The red shift, weakening and broadening of E high 2 are the consequence of the defect formation and local lattice distortions caused by Cu doping [31,48]. Similar characteristics can be found in the previous publications [47,48]. This indicates the incorporation of Cu into ZnO, in good agreement with EPR and XRD results. The intensity ratios of E 1 (LO) to E high 2 were calculated as an indicator to reflect the corresponding defect concentrations of ZnO:XCu after the doping by the phonon confinement model (PCM) [47,49]. The wavenumber of E high 2 and the intensity ratio of E 1 (LO) to E high 2 were listed in Table 1. The value of the intensity ratio of E 1 (LO) to E high 2 increased in the order of ZnO < ZnO:0.5Cu < ZnO:1Cu < ZnO:1.5Cu < ZnO:3.5Cu < ZnO:6.5Cu, increasing along with Cu amount. Then, the correlation between the intensity ratio of E 1 (LO) to E high 2 and the Cu content were displayed in Figure S3. With the increase of Cu amount, ZnO:XCu model catalysts possessed more structure defects. The defect amounts of catalysts kept good linear with the Cu doping amount, except for the ZnO:6.5Cu sample although it has the highest value, indicating the non-linear increase of defects. Moreover, the Raman spectra and the intensity ratio of E 1 (LO) to E high 2 for the reduced samples were presented in Figure S4 and Table S1, respectively. More defects can be obtained due to oxygen vacancy formation during the reduction in H 2 . However, it should be noted that the increment of the intensity ratio of E 1 (LO) to E high 2 was gradually getting smaller along with Cu increase and there was hardly any increase for ZnO:6.5Cu after reduction. This indicates that this sample has quite different defect distribution compared with other samples with low Cu concentration.   The hydrogen temperature-programmed reduction (H2-TPR) profiles of various Cu doped ZnO were presented in Figure 5 in order to determine the reduction process of Cu species. Three reduction peaks can be deconvoluted, the reduction temperature range of 130-145 °C, 150-160 °C and 170-180 °C, corresponded to α, β and γ peak, respectively. The α peak was ascribed to the reduction of Cu 2+ → Cu + and further reduction to Cu 0 [36,50]. In addition, this behavior suggests a certain interaction between the incorporated Cu and ZnO support [51]. The β peak was attributed to the reduction of Cu + to Cu 0 due to the unreacted Cu + from the two-step reduction process of dopant Cu 2+ [36]. Compared to the lower doping samples, ZnO:3.5Cu and ZnO:6.5Cu both have a unique H2 consumption peak, γ, in the temperature region of 170-180 °C, which should be assigned as the reduction of Cu2O according to XRD results. These results indicate that all Cu species can be reduced to Cu 0 to form the direct contact Cu-ZnO interface.
Interestingly, there were two negative peaks displayed in the temperature range of 300-450 °C and 550-650 °C, respectively. The higher temperature negative peak can be attributed to hydrogen release from bulk ZnO support according to the H2 temperature-programmed desorption (H2-TPD) of pristine plate ZnO and ZnO:1Cu after reduction in H2 at 300 °C without H2 adsorption treatment, as displayed in Figures S5 and S6, respectively. Compared to the pure ZnO, the additional lower temperature negative peak probably should be related to Cu doping. In addition, Chen et al. [52] reported that the active metallic atom could dissociate molecular hydrogen into atomic hydrogen species and then migrate to the defects of support. Therefore, the negative peak of ZnO:XCu catalysts at much lower temperature can be postulated to the desorption of the dissociatively adsorbed hydrogen species on the lattice defects of ZnO support due to the spill over from the reduced Cu dopant.  The hydrogen temperature-programmed reduction (H 2 -TPR) profiles of various Cu doped ZnO were presented in Figure 5 in order to determine the reduction process of Cu species. Three reduction peaks can be deconvoluted, the reduction temperature range of 130-145 • C, 150-160 • C and 170-180 • C, corresponded to α, β and γ peak, respectively. The α peak was ascribed to the reduction of Cu 2+ → Cu + and further reduction to Cu 0 [36,50]. In addition, this behavior suggests a certain interaction between the incorporated Cu and ZnO support [51]. The β peak was attributed to the reduction of Cu + to Cu 0 due to the unreacted Cu + from the two-step reduction process of dopant Cu 2+ [36]. Compared to the lower doping samples, ZnO:3.5Cu and ZnO:6.5Cu both have a unique H 2 consumption peak, γ, in the temperature region of 170-180 • C, which should be assigned as the reduction of Cu 2 O according to XRD results. These results indicate that all Cu species can be reduced to Cu 0 to form the direct contact Cu-ZnO interface.
Interestingly, there were two negative peaks displayed in the temperature range of 300-450 • C and 550-650 • C, respectively. The higher temperature negative peak can be attributed to hydrogen release from bulk ZnO support according to the H 2 temperature-programmed desorption (H 2 -TPD) of pristine plate ZnO and ZnO:1Cu after reduction in H 2 at 300 • C without H 2 adsorption treatment, as displayed in Figures S5 and S6, respectively. Compared to the pure ZnO, the additional lower temperature negative peak probably should be related to Cu doping. In addition, Chen et al. [52] reported that the active metallic atom could dissociate molecular hydrogen into atomic hydrogen species and then migrate to the defects of support. Therefore, the negative peak of ZnO:XCu catalysts at much lower temperature can be postulated to the desorption of the dissociatively adsorbed hydrogen species on the lattice defects of ZnO support due to the spill over from the reduced Cu dopant. CO2-TPD profiles of ZnO:XCu model catalysts were shown in Figure 6 to acquire the CO2 adsorption capabilities. Four desorption peaks can be observed approximately at 84, 371, 429 and 662 °C, denoted as α, β1, β2 and γ, respectively. In general, the four desorption peaks were corresponded to the adsorption on weak basic sites (α peak), moderate basic sites (β1, β2 peak) and strong basic sites (γ peak), which were associated with the hydroxy group, metal-oxygen pairs and low coordination oxygen atoms, respectively [53,54]. In comparisons, the CO2-TPD profile of the pristine plate ZnO was also displayed in Figure 6. However, α peak was absent in the patterns of pristine plate ZnO. The presence of the α peak for all Cu doped ZnO samples indicates that the addition of Cu can improve the CO2 adsorption ability on catalyst surface. The desorption peak temperature and respective peak area of the ZnO:XCu and 1Cu/ZnO ( Figure S7) model catalysts were listed in Table 2. The total peak area of β1 plus β2 of the pristine plate ZnO was very small and hardly exhibited any activity. The desorption from peak β1 and β2 over ZnO:XCu model catalysts increased along with the amount of Cu, indicating that the incorporation of Cu prompts CO2 adsorption and activation. The increasing desorption amounts of CO2 from medium basic sites may be attributed to the electronic promotion effects of Cu dopant, more defect sites on the Cu-ZnO interface, providing additional active sites to adsorb CO2. Although Raman result demonstrates that ZnO:6.5Cu had more defects compared to that of ZnO:3.5Cu, it exhibited less CO2 desorption. This is because that the ZnO:6.5Cu sample has quite different defect distribution mode compared with other ZnO:XCu model catalysts; furthermore, only oxygen vacancies are generally considered to be responsible for CO2 adsorption other than intrinsic defects caused by lattice distortion. Besides, considering about the high CO2 desorption from 1Cu/ZnO, the lower desorption from ZnO:6.5Cu sample should be also due to small amount of Cu NPs on the doped plate ZnO surface and the different electronic properties between the doped ZnO and pure ZnO surfaces, according to the previous publications [55]. Thus, although ZnO:6.5Cu sample had a large quantity of defects with a few Cu NPs on doped ZnO surface, it still showed comparable CO2 desorption to that of ZnO:1.5Cu.  Figure 6 to acquire the CO 2 adsorption capabilities. Four desorption peaks can be observed approximately at 84, 371, 429 and 662 • C, denoted as α, β 1 , β 2 and γ, respectively. In general, the four desorption peaks were corresponded to the adsorption on weak basic sites (α peak), moderate basic sites (β 1 , β 2 peak) and strong basic sites (γ peak), which were associated with the hydroxy group, metal-oxygen pairs and low coordination oxygen atoms, respectively [53,54]. In comparisons, the CO 2 -TPD profile of the pristine plate ZnO was also displayed in Figure 6. However, α peak was absent in the patterns of pristine plate ZnO. The presence of the α peak for all Cu doped ZnO samples indicates that the addition of Cu can improve the CO 2 adsorption ability on catalyst surface. The desorption peak temperature and respective peak area of the ZnO:XCu and 1Cu/ZnO ( Figure S7) model catalysts were listed in Table 2. The total peak area of β 1 plus β 2 of the pristine plate ZnO was very small and hardly exhibited any activity. The desorption from peak β 1 and β 2 over ZnO:XCu model catalysts increased along with the amount of Cu, indicating that the incorporation of Cu prompts CO 2 adsorption and activation. The increasing desorption amounts of CO 2 from medium basic sites may be attributed to the electronic promotion effects of Cu dopant, more defect sites on the Cu-ZnO interface, providing additional active sites to adsorb CO 2 . Although Raman result demonstrates that ZnO:6.5Cu had more defects compared to that of ZnO:3.5Cu, it exhibited less CO 2 desorption. This is because that the ZnO:6.5Cu sample has quite different defect distribution mode compared with other ZnO:XCu model catalysts; furthermore, only oxygen vacancies are generally considered to be responsible for CO 2 adsorption other than intrinsic defects caused by lattice distortion. Besides, considering about the high CO 2 desorption from 1Cu/ZnO, the lower desorption from ZnO:6.5Cu sample should be also due to small amount of Cu NPs on the doped plate ZnO surface and the different electronic properties between the doped ZnO and pure ZnO surfaces, according to the previous publications [55]. Thus, although ZnO:6.5Cu sample had a large quantity of defects with a few Cu NPs on doped ZnO surface, it still showed comparable CO 2 desorption to that of ZnO:1.5Cu.

Catalytic Performance
The catalytic performance of 1Cu/ZnO and ZnO:XCu samples was evaluated. For all samples, higher CO2 conversion can be obtained at higher reaction temperature as shown in Figure 7a. The doping of Cu can significantly improve the catalytic activity and the ZnO:3.5Cu sample exhibited the highest CO2 conversion for RWGS reaction. With a further increase in the amount of Cu, the activity of ZnO:6.5Cu was suppressed and even lower than that of ZnO:3.5Cu at 550 °C. This indicates that there exists an optimum copper doping amount and a further increase in the amount of Cu would decrease the catalytic activity. Furthermore, in order to further understand the reactivity of direct contact Cu-ZnO interface on RWGS reaction, the CO formation rate per catalyst weight (Rw) on the investigated catalysts as a function of Cu amount at 450 °C were illustrated in Figure 7b. As shown, when the amount of Cu less than 3.5 wt %, CO formation rate by catalyst weight increased along with the Cu doping amount and presented almost linear correlation. This implies that higher Cu doping is beneficial to RWGS reaction. Compared with all ZnO:XCu catalysts, it is clear that 1Cu/ZnO exhibits the superior catalytic reactivity at 450 °C; however, when the reaction temperature higher than 500 °C, it is significantly lower than that of ZnO:XCu model catalysts.

Catalytic Performance
The catalytic performance of 1Cu/ZnO and ZnO:XCu samples was evaluated. For all samples, higher CO 2 conversion can be obtained at higher reaction temperature as shown in Figure 7a. The doping of Cu can significantly improve the catalytic activity and the ZnO:3.5Cu sample exhibited the highest CO 2 conversion for RWGS reaction. With a further increase in the amount of Cu, the activity of ZnO:6.5Cu was suppressed and even lower than that of ZnO:3.5Cu at 550 • C. This indicates that there exists an optimum copper doping amount and a further increase in the amount of Cu would decrease the catalytic activity. Furthermore, in order to further understand the reactivity of direct contact Cu-ZnO interface on RWGS reaction, the CO formation rate per catalyst weight (R w ) on the investigated catalysts as a function of Cu amount at 450 • C were illustrated in Figure 7b. As shown, when the amount of Cu less than 3.5 wt %, CO formation rate by catalyst weight increased along with the Cu doping amount and presented almost linear correlation. This implies that higher Cu doping is beneficial to RWGS reaction. Compared with all ZnO:XCu catalysts, it is clear that 1Cu/ZnO exhibits the superior catalytic reactivity at 450 • C; however, when the reaction temperature higher than 500 • C, it is significantly lower than that of ZnO:XCu model catalysts.

The RWGS Reactivity over Different Cu-ZnO Interfaces
The surface basic sites are pivotal to the RWGS reaction. According to the previous publications, the introduction of transition metal dopants (Cu 2+ , Ni 2+ , Fe 2+ etc.) will change the basicity of the host metal oxides, then modifying the adsorption capacity and strength of acidic CO2 [56][57][58]. Here, the desorption from medium basic sites increase along with the increase of Cu doping; however, when the amount of Cu reached 6.5 wt %, it was only close to that of ZnO:1.5Cu according to CO2-TPD. Then, the relationship between RSSA and CO2 desorption density (based on the surface area) from the medium basic sites was displayed in Figure 8 and the corresponding specific surface area of all samples was listed in Table S2. Figure 8 clearly demonstrates that the higher density of the moderately adsorbed carbon species leads to higher RSSA and plays important roles in CO formation. Interestingly, ZnO:6.5Cu has higher RSSA than that of ZnO:1.5Cu but with similar amount of carbon species desorption from the medium basic sites. EPR results indicate that doped Cu 2+ mainly replace lattice Zn 2+ sites and interstitial sites (Zni), while SEM and XRD characterization demonstrate the appearance of some Cu NPs on the surface of plate ZnO over ZnO:6.5Cu sample. Besides, it is widely accepted that the ZnOx will migrate from bulk ZnO support to the surface of Cu NPs to form ZnOx-Cu-ZnO interface during reduction in H2 [12,22], which should account for the peculiar CO2 adsorption behavior of ZnO:6.5Cu sample. Therefore, the higher RWGS reactivity of ZnO:6.5Cu compared to ZnO:1.5Cu should be due to the formation of ZnOx-Cu NP-ZnO interface on ZnO:6.5Cu. This means ZnOx-Cu NP-ZnO interface also displays apparent RWGS reactivity.

The RWGS Reactivity over Different Cu-ZnO Interfaces
The surface basic sites are pivotal to the RWGS reaction. According to the previous publications, the introduction of transition metal dopants (Cu 2+ , Ni 2+ , Fe 2+ etc.) will change the basicity of the host metal oxides, then modifying the adsorption capacity and strength of acidic CO 2 [56][57][58]. Here, the desorption from medium basic sites increase along with the increase of Cu doping; however, when the amount of Cu reached 6.5 wt %, it was only close to that of ZnO:1.5Cu according to CO 2 -TPD. Then, the relationship between R SSA and CO 2 desorption density (based on the surface area) from the medium basic sites was displayed in Figure 8 and the corresponding specific surface area of all samples was listed in Table S2. Figure 8 clearly demonstrates that the higher density of the moderately adsorbed carbon species leads to higher R SSA and plays important roles in CO formation. Interestingly, ZnO:6.5Cu has higher R SSA than that of ZnO:1.5Cu but with similar amount of carbon species desorption from the medium basic sites. EPR results indicate that doped Cu 2+ mainly replace lattice Zn 2+ sites and interstitial sites (Zn i ), while SEM and XRD characterization demonstrate the appearance of some Cu NPs on the surface of plate ZnO over ZnO:6.5Cu sample. Besides, it is widely accepted that the ZnOx will migrate from bulk ZnO support to the surface of Cu NPs to form ZnOx-Cu-ZnO interface during reduction in H 2 [12,22], which should account for the peculiar CO 2 adsorption behavior of ZnO:6.5Cu sample. Therefore, the higher RWGS reactivity of ZnO:6.5Cu compared to ZnO:1.5Cu should be due to the formation of ZnOx-Cu NP-ZnO interface on ZnO:6.5Cu. This means ZnOx-Cu NP-ZnO interface also displays apparent RWGS reactivity.
Based on our previous publications [26,27], the 1Cu/ZnO model catalyst has both the direct contact Cu-ZnO interface and ZnOx-Cu NP-ZnO interface. Here, the lower doped ZnO:XCu model catalysts only have the direct contact Cu-ZnO interface. Then, in order to further disclose the RWGS reactivity over different Cu-ZnO interfaces, the catalytic performance of the 1Cu/ZnO model catalyst was also presented in Figure 8. It is striking that R SSA (based on the SSA of ZnO plate) is far higher than that of all ZnO:XCu model catalysts. Furthermore, the slope of 1Cu/ZnO model catalyst is much steeper than that of all ZnO:XCu model catalysts, which means the adsorbed carbon species on the former model catalyst are more active than those adsorbed on the latter ones or the specific surface area of ZnOx-Cu NP-ZnO interface should be also included. Thus, the specific surface area of ZnOx-Cu NP-ZnO interface was counted to calculate the R SSA , which was about 35 Cu ·s −1 , which was still higher than that of the direct contact Cu-ZnO interface simulated by ZnO:3.5Cu model catalyst. Thus, the incorporation of Cu into bulk plate ZnO to form direct contact Cu-ZnO interface was beneficial to the RWGS reaction; however, the ZnOx-Cu NP-ZnO interface was more active for the reaction at low temperature. Based on our previous publications [26,27], the 1Cu/ZnO model catalyst has both the direct contact Cu-ZnO interface and ZnOx-Cu NP-ZnO interface. Here, the lower doped ZnO:XCu model catalysts only have the direct contact Cu-ZnO interface. Then, in order to further disclose the RWGS reactivity over different Cu-ZnO interfaces, the catalytic performance of the 1Cu/ZnO model catalyst was also presented in Figure 8. It is striking that RSSA (based on the SSA of ZnO plate) is far higher than that of all ZnO:XCu model catalysts. Furthermore, the slope of 1Cu/ZnO model catalyst is much steeper than that of all ZnO:XCu model catalysts, which means the adsorbed carbon species on the former model catalyst are more active than those adsorbed on the latter ones or the specific surface area of ZnOx-Cu NP-ZnO interface should be also included. Thus, the specific surface area of ZnOx-Cu NP-ZnO interface was counted to calculate the RSSA, which was about 35.91 μmol·m −2 Cu/ZnO·s −1 according to the specific surface area of 1Cu/ZnO model catalyst (4.94 m 2 ·g −1 cat). This value is still above that of all ZnO:XCu model catalysts. Among all ZnO:XCu model catalysts, ZnO:3.5Cu model catalyst displayed the highest RSSA (34.62 μmol·m −2 ZnO:Cu·s −1 ), which reflects the maximum CO formation rate over the simulated direct contact Cu-ZnO interface. The specific surface areas of plate ZnO and 1Cu/ZnO model catalyst were 2.93 and 4.94 m 2 ·g −1 cat, respectively; then it can be roughly estimated that the specific surface area of Cu NPs was around 2.01 m 2 Cu·g −1 cat. Considering RSSA over 1Cu/ZnO model catalyst is 35.91 μmol·m −2 Cu/ZnO·s −1 , then RSSA over ZnOx-Cu NP-ZnO interface could be calculated about 37.80 μmol·m −2 Cu·s −1 , which was still higher than that of the direct contact Cu-ZnO interface simulated by ZnO:3.5Cu model catalyst. Thus, the incorporation of Cu into bulk plate ZnO to form direct contact Cu-ZnO interface was beneficial to the RWGS reaction; however, the ZnOx-Cu NP-ZnO interface was more active for the reaction at low temperature.

The RWGS Reaction Stability over Different Cu-ZnO Interfaces
Both the direct contact Cu-ZnO interface and ZnOx-Cu NP-ZnO interface were active for the RWGS reaction, the latter one was more active than the former at temperatures lower than 500 °C. This is important in understanding the low methanol selectivity in CO2 hydrogenation to methanol. However, the high temperature was essential to the RWGS reaction; thus, the catalytic performance of these samples at 600 °C was comparatively investigated as displayed in Figure 9. As shown in Figure 9a, the initial CO2 conversion of 1Cu/ZnO catalyst was lower than that of ZnO:1Cu, even lower Figure 8. The relationship between CO formation rate R SSA and CO 2 desorption density from peak (β 1 + β 2 ) at 450 • C.

The RWGS Reaction Stability over Different Cu-ZnO Interfaces
Both the direct contact Cu-ZnO interface and ZnOx-Cu NP-ZnO interface were active for the RWGS reaction, the latter one was more active than the former at temperatures lower than 500 • C. This is important in understanding the low methanol selectivity in CO 2 hydrogenation to methanol. However, the high temperature was essential to the RWGS reaction; thus, the catalytic performance of these samples at 600 • C was comparatively investigated as displayed in Figure 9. As shown in Figure 9a, the initial CO 2 conversion of 1Cu/ZnO catalyst was lower than that of ZnO:1Cu, even lower than that of 0.5 wt % Cu doped plate ZnO at 600 • C. Furthermore, the 1Cu/ZnO catalyst deactivated very quickly after 20 h measurement, and the activity of 1Cu/ZnO decreased 35.5%, as shown in Figure 9b. As known, Cu-based oxide catalysts are easily deactivated due to the sintering of surface Cu particles especially at high temperature [59,60]. This suggests that the specific surface area of the ZnOx-Cu NP-ZnO interface in the 1Cu/ZnO model catalyst will decrease, then lose activity quickly. In comparison, all ZnO:XCu samples possessed relatively stable catalytic performance especially when the doping amount was less than 3.5 wt %. These results mean that the aggregation and sintering of active Cu species in doped samples should not exist at high reaction temperature, improving the catalytic activity and stability. In order to exclude the influences of ZnOx-Cu NP-ZnO interface induced by Cu nanoparticle formation during the reaction evaluation, the SEM and XPS investigation of ZnO:1Cu model catalyst after RWGS stability evaluation were displayed in Figure S8 and Table S3, respectively. Although a slight deformation of plate morphology appeared, no Cu nanoparticle can be observed on the surface with only slight Cu/Zn ratio increase after the stability test, indicating no apparent Cu segregation during the reaction test. Accordingly, the activity loss over the ZnO:6.5Cu catalyst reached 35.14% after reaction for 20 h, similar to that of 1Cu/ZnO model catalyst. This further proved that the additional ZnOx-Cu NP-ZnO interface would impair the stability. According to the above results, the direct contact Cu-ZnO interface is active for the RWGS reaction and more stable than that of the ZnOx-Cu NP-ZnO interface at high temperature.
interface induced by Cu nanoparticle formation during the reaction evaluation, the SEM and XPS investigation of ZnO:1Cu model catalyst after RWGS stability evaluation were displayed in Figure  S8 and Table S3, respectively. Although a slight deformation of plate morphology appeared, no Cu nanoparticle can be observed on the surface with only slight Cu/Zn ratio increase after the stability test, indicating no apparent Cu segregation during the reaction test. Accordingly, the activity loss over the ZnO:6.5Cu catalyst reached 35.14% after reaction for 20 h, similar to that of 1Cu/ZnO model catalyst. This further proved that the additional ZnOx-Cu NP-ZnO interface would impair the stability. According to the above results, the direct contact Cu-ZnO interface is active for the RWGS reaction and more stable than that of the ZnOx-Cu NP-ZnO interface at high temperature.

Synthesis of Cu Doped Plate ZnO Catalysts
Cu doped plate ZnO model catalysts were synthesized by a hydrothermal method. Firstly, the required amounts of Cu(NO3)2·3H2O (AR grade, Shanghai Aladdin) and zinc acetate dehydrate (Zn(CO2CH3)2, 6.3 g, AR grade, Shanghai Aladdin) were dissolved in 60 mL deionized water, followed by addition of hexamethylenetetramine (C6H12N4, 1.6 g, AR grade, Shanghai Aladdin). After 1 h vigorous magnetic stirring, the solution was transferred into a 100 mL Teflon-lined autoclave and maintained at 97 °C for 12 h. The resulted precipitate was collected after the autoclave cooled to room temperature, and then washed with deionized water and ethanol for several times, finally dried at 80 °C for 12 h and calcined at 350 °C for 3 h in pure N2. The final model catalysts with the doping amount were 0.5, 1.0, 1.5, 3.5 and 6.5 wt % determined by ICP analysis, denoted as ZnO:0.5Cu, ZnO:1Cu, ZnO:1.5Cu, ZnO:3.5Cu and ZnO:6.5Cu, respectively.

Synthesis of Cu/Plate ZnO Catalysts
The pristine plate ZnO was prepared according to ZnO:Cu preparation procedure without the addition of Cu(NO3)2·3H2O. 1 wt % Cu/plate ZnO model catalyst was prepared by the impregnation method. The required amount of Cu(CO2CH3)2·H2O was dissolved in deionized water, then copper-

Synthesis of Cu Doped Plate ZnO Catalysts
Cu doped plate ZnO model catalysts were synthesized by a hydrothermal method. Firstly, the required amounts of Cu(NO 3 ) 2 ·3H 2 O (AR grade, Shanghai Aladdin) and zinc acetate dehydrate (Zn(CO 2 CH 3 ) 2 , 6.3 g, AR grade, Shanghai Aladdin) were dissolved in 60 mL deionized water, followed by addition of hexamethylenetetramine (C 6 H 12 N 4 , 1.6 g, AR grade, Shanghai Aladdin). After 1 h vigorous magnetic stirring, the solution was transferred into a 100 mL Teflon-lined autoclave and maintained at 97 • C for 12 h. The resulted precipitate was collected after the autoclave cooled to room temperature, and then washed with deionized water and ethanol for several times, finally dried at 80 • C for 12 h and calcined at 350 • C for 3 h in pure N 2 . The final model catalysts with the doping amount were 0.5, 1.0, 1.5, 3.5 and 6.5 wt % determined by ICP analysis, denoted as ZnO:0.5Cu, ZnO:1Cu, ZnO:1.5Cu, ZnO:3.5Cu and ZnO:6.5Cu, respectively.

Synthesis of Cu/Plate ZnO Catalysts
The pristine plate ZnO was prepared according to ZnO:Cu preparation procedure without the addition of Cu(NO 3 ) 2 ·3H 2 O. 1 wt % Cu/plate ZnO model catalyst was prepared by the impregnation method. The required amount of Cu(CO 2 CH 3 ) 2 ·H 2 O was dissolved in deionized water, then copper-containing solution was added to 2 g of plate ZnO support, and the slurry was stirred until dry. Finally, the sample was dried at 80 • C for 12 h in the oven, followed by calcination at 350 • C for 3 h in air. The obtained catalyst with Cu loading amount was 1 wt %, labeled as 1Cu/ZnO.

Characterization Technologies
Inductively coupled plasma (ICP) analysis was utilized to determine the actual copper amounts on an inductively coupled plasma-optical emission spectroscopy (ICP-AES, Thermo Scientific FLASH 2000). Scanning electron microscopy (SEM) images and energy dispersive X-ray spectroscopy (EDS) was obtained on a scanning electron microscope (Carl Zeiss Merlin, Germany).
Electron paramagnetic resonance (EPR) spectra were recorded on a Bruker A300 spectrometer operated in the X-band frequency range at 100 K. The as-synthesized samples were loaded in EPR tubes and operated with a microwave power of 2.18 mW, modulation frequency of 100 kHz, microwave frequency of 9.33 GHz and modulation amplitude of 1.00 G.
The surface areas were collected at an analyzer (Micromeritics, ASAP 2020, USA). All corresponding samples were degassed at 200 • C for 4 h before the measurement.
X-ray photoelectron spectroscopy (XPS) was characterized by a US Thermo Fisher Scientific ESCALAB Xi+ with Al Kα (1486.6 eV) as the excitation source, and the binding energies were corrected by the C1s line at 284.6 eV.
X-ray diffraction (XRD) patterns were obtained on an X-ray diffractometer (Bruker D8 Advance, Germany) with Cu Kα radiation, in the 2θ range from 20 to 80 • . Visible Raman spectroscopy was collected on a LabRAMHR evolution laser Raman spectrometer (HORIBA, France), which was equipped with a HeNe laser (532 nm) and a CCD detector. The wavenumber region was in the range of 200-800 cm −1 .
Hydrogen temperature-programmed reduction (H 2 -TPR) of the calcined sample (100 mg) was carried out on an automated chemisorption analyzer (AutoChem II2920, Micromeritics) equipped with a TCD detector. The sample was firstly pretreated in Ar at 300 • C for 1 h and then cooled down to 50 • C in Ar. After that, the baseline was recorded in 5 vol.% H 2 /Ar at a flow rate of 30 mL·min −1 until it was stable. Finally, the sample was heated from 50 to 800 • C at a ramping rate of 5 • C·min −1 .
CO 2 temperature-programmed desorption (CO 2 -TPD) was conducted on the automated chemisorption analyzer. The calcined catalysts (200 mg) were reduced at 300 • C for 3 h in 5 vol.% H 2 /Ar with a flow rate of 30 mL·min −1 and ramping rate of 5 • C·min −1 . After cooling down to 45 • C in He, the catalysts were saturated with 20 vol.% CO 2 /Ar for 1 h. The CO 2 desorption was recorded by on-line mass spectrometer from 45 to 800 • C in He with a flow rate of 30 mL·min −1 .

Catalytic Activity Test
Catalytic activity test was carried out in a quartz tube reactor (quartz U-tube, 4 mm inner diameter) under atmosphere pressure. The catalysts (200 mg, 40-60 mesh) were loaded into the reactor and pretreated in 5 vol.% H 2 /Ar at 300 • C for 3 h prior to the reaction test. Afterwards, the catalysts were flushed with feed gas CO 2 /H 2 (1:3 molar ratio, 30 mL·min −1 ). The reaction was evaluated in the temperature range of 450 to 600 • C, and the products were analyzed by an online gas chromatography. The CO 2 conversion and CO selectivity were defined as follows: represent the concentrations of CO 2 , CO and CH 4 in the outlet gas, respectively.
Additionally, the CO formation rate was defined as follow: where R W and R SSA is the CO formation rate according to catalyst weight (µmol·g −1 ·s −1 ) and specific surface area (µmol·m −2 ·s −1 ), respectively. m is the mass of the catalysts used (g), S SSA is the specific surface area (m 2 ·g −1 ) and F CO 2 is corresponded to the molar flow rate of CO 2 (mol·s −1 ).

Conclusions
A series of Cu doped ZnO model catalysts were designed and successfully prepared to simulate the direct contact Cu-ZnO interface of the supported Cu-ZnO based catalysts, while the 1Cu/ZnO model catalyst was prepared to simulate both direct contact Cu-ZnO and ZnOx-Cu NP-ZnO interfaces. The comparative investigation indicates that the RWGS reaction reactivity on the direct contact Cu-ZnO interface was inferior to that of the ZnOx-Cu NP-ZnO interface at a temperature lower than 500 • C; however, it is more stable at a reaction temperature higher than 500 • C, which enables ZnO:XCu model catalysts superior catalytic activity to that of 1Cu/ZnO. Considering the typical reaction temperature (200-300 • C) of CO 2 hydrogenation to methanol, both direct contact Cu-ZnO and ZnOx-Cu NP-ZnO interfaces will facilitate CO formation while the latter plays vital roles. Then, the reactivity of CO hydrogenation to methanol on both interfaces will be further investigated in order to elucidate the methanol formation during CO 2 hydrogenation to methanol over Cu-ZnO based catalysts. This finding will help to design more efficient and stable catalysts for both the RWGS reaction and methanol production from CO 2 hydrogenation.