Introduction

Under the background of global warming, the “low-carbon economy” based on low energy consumption and low output of greenhouse gas has become a hotspot. Converting the emitted CO2, the major greenhouse gas, into valuable chemicals is a subject that many researchers have been working on. Methanol, as an important industrial commodity [1] as well as an ideal clean synthetic fuel [2], becomes the preferred target product for CO2 conversion. CO2 hydrogenation to methanol is expected to solve both environmental and energy issues at the same time. Therefore, much attention has been attracted on this research field.

At present, Cu/ZnO/Al2O3 catalyst has been extensively studied for CO2 hydrogenation and industrialized for several decades in methanol production from syngas [3]. Although Cu/ZnO/Al2O3 shows excellent performance using syngas as feed gas, it is less active in CO2 hydrogenation to methanol [4]. In order to improve the activity, researchers have conducted in-depth researches such as adjusting the size of Cu clusters [5, 6], optimizing the preparation methods [7,8,9] and adding promoters [10, 11]. Recently, a layer of metastable ZnO was clearly observed on the top of Cu particles for some Cu/ZnO catalysts with high performance for CO2 hydrogenation [12, 13]. This ZnO layer can not only form ZnOx species acting as cocatalysts but also can protect the Cu from reshaping and sintering during the reaction [12]. This kind of catalysts with the oxide-metal structure was first proposed as “inverse catalysts” by Rodriguez et al.[14].

Due to the unique structure of the inverse catalysts, they showed better performance than conventional catalysts in some reactions, such as CO oxidation [15], partial oxidation of methanol [16] and water–gas shift reaction [14]. For CO2 hydrogenation to methanol, some inverse catalysts, e.g. ZnO/Cu(100) [17] and MnOx/Co3O4 [18], exhibited superior activity than that of Cu/ZnO (000ī) and CoOx/MnO2. In order to further investigate the influence of composition on the properties and catalytic performance of the powder ZnO/Cu catalysts, we used microemulsion method to prepare a series of ZnO/Cu catalysts and investigated their performance for CO2 hydrogenation to methanol. Our results indicated that the inverse ZnO/Cu showed higher methanol selectivity than the conventional Cu/ZnO sample.

Experimental

Catalysts preparation

The catalysts were prepared with a microemulsion method. First, two microemulsions with the same composition were prepared. Both of them contained 32% wt. of Trition-x-100 as the surfactant, 20% of n-hexanol as the cosurfactant, 38% of n-heptane as the oil phase and 10% water. In one of the microemulsion, an appropriate amount of copper nitrate was dissolved in the water phase, while an appropriate amount of precipitating agent (tetramethylammonium hydroxide, TMAH) was added to the other microemulsion. After stirring for 1 h, the TMAH-containing microemulsion was added dropwise to the microemulsion containing copper nitrate. The mixture was kept at 25 °C for 24 h to complete the precipitate. Then, the precipitate was separated with a centrifuge for 3 times and washed with methanol. After dried in an oven at 100 °C for 24 h and calcined at 350 °C in a muffle for 2 h, CuO was finally obtained.

ZnO/CuO precursors were prepared in the same procedure except that the above-obtained CuO and zinc nitrate were added to the water phase in one of the microemulsions. After precipitation, centrifugal separation, drying, calcination at 350 °C for 2 h, and finally reduced under pure hydrogen at 300 °C for 2 h, ZnO/Cu catalysts were obtained. A series of ZnO/Cu catalysts were obtained where n(Zn)/n(Cu) equals to 1:9, 2:8, 3:7, 4:6 and 5:5.

For comparison, CuO/ZnO sample, with a n(Zn)/n(Cu) of 4:6, was also prepared using the same method. After reducing under pure hydrogen at 300 °C for 2 h, Cu/ZnO (6:4) was obtained.

Catalysts characterization

N2 adsorption–desorption was measured at liquid nitrogen temperature (77 K) on a Quantachrome Autosorb-1-C-TCD-MS instrument. Prior to the tests, evacuating of the samples for 6 h at 200 °C was carried out. Multipoint Brunauer–Emmett–Teller (BET) model and Barrett-Joyner-Halenda (BJH) model were applied to calculate the specific surface areas (SBET) and the pore size distributions.

The phase analysis of the catalyst was carried out on a Brucker Advance D8 X-ray powder diffractometer (XRD). Cu-Kα with a wavelength of 0.1506 nm was used as the radiation source. The instrument was operating at 40 kV and 40 mA, with a scanning range of 20°–80° (0.0167°/step).

X-ray photoelectron spectroscopies were collected on a VG Mulitilab 2000 instrument (Thermal Electron Corporation) under 2 × 10–6 Pa. Monochromatic Al Kα X-ray operating at 15 kV and 20 mA was used as the radiation source. Samples were first reduced in pure H2 at 300 °C for 2 h. After cooled down to room temperature in pure H2, they were sealed in the glass sample bottles without air. Finally, the samples were transferred into the XPS instrument for and measurement. The C1s binding energy (284.6 eV) was used as a standard to calibrate the shift of the binding energy caused by the charge effect.

TEM images were obtained on FEI Tecnai G2 F20 electron microscope under 200 kV. Scanning electron microscope (SEM) images of the sample were acquired on a Hitachi-48005 instrument equipped with energy dispersive X-ray analysis (EDX) detector.

The reducibility of the catalysts was tested on a multifunctional characterization analyzer AMI-200 (Zeton Altamira Corporation). 25 mg of the catalyst was purged at 150 °C for 1 h under argon (25 mL/min) to remove the adsorbed water. After cooling down to 50 °C, 10% H2/90% Ar with a flow rate of 25 mL/min was used to reduce the sample. H2-TPR data was collected with the temperature rising to 400 °C at a heating ramp of 10 °C/min.

H2-TPD was performed on the Micromeritics Instruments Corporation AutoChem II 2920 instrument to determine the Cu surface area and the dispersion of Cu. 0.2 g of a sample was first reduced in a pure hydrogen (35 mL/min) atmosphere. The temperature was raised from room temperature to 300 °C at a heating rate of 10 °C/min and held at 300 °C for 2 h. After cooling down to -50 °C and holding at this temperature for 1 h in pure hydrogen, pure Ar (25 mL/min) was introduced into the system. Then, the temperature was raised from -50 °C to 150 °C in the Ar at a heating rate of 10 °C/min. Assuming that the chemisorption of H2 on Cu atoms is carried out according to Cu:H = 2:1 [19, 20], the specific copper surface area (SCu) and the degree of dispersion (D) were calculated as follows:

$${\text{S}}_{{{\text{Cu}}}} = {4} \times {\text{H}}_{{2}} {\text{uptake}} \times {\text{N}}_{{{\text{av}}}} /({1}.{47} \times {1}0^{{{19}}} )({\text{m}}^{{2}} - {\text{Cu}}/{\text{g}})$$

In this equation, Nav is the Avogadro’s constant, 1.47 × 1019 is the value of Cu atoms per square meter [19, 21].

$${\text{Percentage dispersion }}\left( {\text{D}} \right) = \frac{{{\text{amount of Cu atoms on surface }}\left( {{\text{moles}}} \right)}}{{{\text{total amount of Cu atoms in the sample}}\left( {{\text{moles}}} \right)}} \times 100{{\% }} = \frac{{{\text{H}}_{2} {\text{ uptake}} \times 4\left( {{\text{moles}}} \right)}}{{{\text{total amount of Cu atoms in the sample}}\left( {{\text{moles}}} \right)}} \times 100{{\% }}$$

Catalytic performance measurements

CO2 hydrogenation to methanol activity measurements were carried out in a stainless-steel high-pressure reaction device. 0.1 g of a sample was loaded in the fix-bed reactor with a diameter of 6 mm. Before the activity tests, the sample was first reduced in pure H2 with a GHSV of 2.4 L h−1 g−1 at 300 °C for 2 h under atmospheric pressure and then cooled to room temperature. The typical gas (n(CO2)/n(H2) = 1:3) with a GHSV of 3.6 L h−1 g−1 was injected into the device. The catalytic performance test was conducted at 2 MPa and 250 °C and lasted for 50 h for each sample.

The final gas components were analyzed online by Agilent Micro GC 3000A gas chromatograph (TCD detector). The methanol and water components in the product were collected by a 0 °C cold trap, and analyzed offline with an Agilent GC 4890D (FID detector).

Results and discussion

Textural properties

Fig. 1 displays the nitrogen adsorption–desorption isotherms and pore sizes distributions of the ZnO/Cu catalysts. All samples, showed a type IV isotherm, which was the typical feature of mesoporous materials. It could also be seen that all samples had a H3 type hysteresis loop, resulted from the capillary condensation inside slit-like pores formed by aggregation of flaky particles [22]. From Fig. 1b, one can observe that ZnO/Cu (5:5) and ZnO/Cu (4:6) samples had two distinct peaks, indicating a bimodal pore size distribution. The pore structure parameters, including the BET specific area (SBET), total pore volume (Vp) and pore size (DBJH) are listed in Table 1. ZnO/Cu (3:7) had the largest specific surface area, which was beneficial to the dispersion of active components. However, due to the small pore size, the pore volume was relatively small. ZnO/Cu(4:6) had the largest pore volume and pore size.

Fig. 1
figure 1

N2 adsorption–desorption isotherms (a) and the BJH pore size distribution (b) at 77 K for the reduced ZnO/Cu catalysts

Full size image
Table 1 Physicochemical properties and particle sizes of ZnO and Cu for ZnO/Cu catalysts
Full size table

XRD measurements

The XRD diffraction patterns for ZnO/Cu catalysts were depicted in Fig. 2. As mentioned in the preparation section, the samples were reduced at 300 °C and ZnO could only be reduced at higher than 400 °C [23]. Therefore, only ZnO and Cu species could be detected for all ZnO/Cu catalysts. With the increase of Cu content, the intensity of the Cu diffraction peaks was gradually enhanced, while the characteristic diffraction peaks of ZnO were gradually weakened. The particle sizes of ZnO and Cu were calculated using the Scherrer equation at 2θ = 36.2 °C and 43.3°. The data are shown in Table 1. Before the reduction process, the particle size of CuO was about 14 nm, obtaining from the XRD pattern (Fig. S1). After the reduction, relatively large Cu particles were formed because of the coalescence of copper component [24]. Cu particle sizes followed this trend: ZnO/Cu(4:6) < ZnO/Cu(5:5) < ZnO/Cu(3:7) < ZnO/Cu(2:8) < ZnO/Cu(1:9). ZnO/Cu(4:6) catalyst had the minimum Cu particle size in the series, probably suggesting the strongest interaction between Cu and ZnO, as observed similarly for the CuO-CeO2 systems [25]. No obvious difference for ZnO particles for all the ZnO/Cu samples was observed, as the particle sizes calculated using the Scherrer equation usually have an error of 1 nm [26]. The results indicated that microemulsion method is an effective method to tune particle size of Cu in the catalysts. Moreover, for ZnO/Cu(4:6) sample, the particle sizes of ZnO (20.7 nm) and Cu (19.8 nm) are quite similar. It has been reported that the interface of Cu–ZnO samples has some connection with the particle size of the two components. Although it was impossible to get precise quantification of the interface due to the non-ideal spherical shape, the authors considered that the more similar the particle sizes were, the larger was the interface at the perimeter [27]. Out of these this factor, ZnO/Cu (4:6) sample seemed to have the largest interface. For the other catalysts, the interface showed a decrease trend with the increase of Cu particle sizes. As reported previously, CO2 mainly adsorbs on the surface of ZnO [11], while H2 dissociatively adsorbs on the surface of Cu and transfers H atoms to ZnO via spillover. Large interface, i.e. large contact area between ZnO and Cu, can facilitate the spillover of H atoms and thus the formation of methanol.

Fig. 2
figure 2

XRD diffraction patterns of ZnO/Cu catalysts

Full size image

In order to make sure whether CuZn alloy was formed in the samples, the enlarged part of the XRD patterns from 42–45 o is shown in Fig. 2b. Clearly, compared with pure Cu pattern, a slight shift of 2θ towards a lower angle can be noticed. Cu lattice constant was calculated using \({\text{a}}_{\text{Cu}}=\frac{\sqrt{3}}{2\text{sin}}\theta\uplambda\) (cubic structure, orientation (111)). The obtained aCu is shown in Table 1. The lattice constant of pure Cu (JCPDS Cu 04–0836) was also calculated using the same method and the value is 3.616 Å. The Cu lattice parameters of all the prepared samples were larger than that of pure Cu, indicating an expansion of the Cu lattice [28]. This should be attributed to the fact that Zn2+ entered into the Cu lattice to form Cu–Zn alloy, as reported by Valant [29] and Nakamura [30], since Zn2+ (0.60 Å) is bigger than Cu2+ (0.57 Å) [29]. Generally, the more Zn2+ entered into the Cu lattice, the larger the cell parameters of the resulted catalysts would be. Therefore, in the series, ZnO/Cu(4:6) had the most amount of CuZn alloy, followed by ZnO/Cu(5:5), ZnO/Cu(3:7), ZnO/Cu(2:8) and ZnO/Cu(1:9).

SEM and TEM

SEM (A) and TEM (B) images of ZnO/Cu (4:6) sample are shown in Fig. 3 and the EDS results are displayed in Table 2. From Table 2, one can see that at point a, Zn element accounts for the majority; while at point b, Cu element accounts for the majority. Combing with the SEM image of the sample, it suggested that the small particles is probably correspond to the ZnO species. Meanwhile, the flake-like structure was probably assigned to Cu particles. Hence, this finding implied that ZnO particles were relatively well-dispersed on the surface of flake-like Cu. The TEM image of the catalyst is consistent with the SEM results. The particle size of Zn in both SEM and TEM images is about 20 nm, in accordance with the XRD results. However, the particle size of Cu seems a little difficult to distinguish due to the aggregation.

Fig. 3
figure 3

SEM (a) and TEM (b) images of ZnO/Cu (4:6) sample

Full size image
Table 2 EDS analysis for ZnO/Cu (4:6) sample
Full size table

XPS measurements

The XPS measurements were performed to determine the valence state of Zn and Cu on the surface and the results are shown in Fig. 4. The binding energy for Zn 2p3/2 was at about 1021.5 eV, which agrees well with reported Zn2+ [13]. There are two main peaks for Cu 2p XPS spectrum. The peaks at binding energy (BE) = 932–933 eV and BE = 952–953 eV should be ascribed to Cu2p3/2 and Cu2p1/2. For CuO sample, the obvious shake-up satellite peak at 937–945 eV was the characteristic of Cu2+ [31]. After the reduction process, the significant decrease or the absence of the shake-up peak strongly suggested that most cupric species had been reduced to a low oxidation or metallic state. Since the binding energy of Cu+ and Cu0 are almost the same, XPS spectra could hardly give a distinguished identification. However, our previous study showed that Cu+ only existed for a short time during the reduction process [32]. There were some other reports on similar catalysts confirmed the presence of only Cu0 on the surface using Cu LMM spectra [27, 33]. Therefore, we could conclude that Cu0 was the dominant species on the surface after the reduction. Table 3 lists the Cu/Zn molar ratios both on the surface and in the bulk of the samples given by XPS and Inductively Coupled Plasma (ICP) results. It was obvious that the Cu/Zn ratios on the surface of the catalysts were significantly lower than that in the bulk phase, suggesting that only a little amount of Cu presented on the surface in the prepared catalysts, while large amount of Cu remained in the bulk.

Fig. 4
figure 4

Zn 2p and Cu2p XPS spectra of CuO and the reduced ZnO/Cu catalysts

Full size image
Table 3 Cu/Zn molar ratios, TPR and TPD results for ZnO/Cu catalysts
Full size table

The reducibility of the catalysts

The TPR profiles of the ZnO/Cu catalysts with different Zn/Cu ratios were presented in Fig. 5. For pure CuO sample, two reduction peaks at about 221 °C and 269 °C were observed, indicating the presence of two different types of CuO. Many researchers have extensively studied the reducibility of CuO. The lower temperature reduction peak was commonly ascribed to the reduction of highly dispersed CuO, while the higher reduction peak was ascribed to the reduction of bulk CuO interacting with the support [34, 35]. Furthermore, the two reduction peaks have also been described as stepwise reduction of copper, which is Cu2+ to Cu+ and Cu+ to metallic Cu [36]. In order to further analyze the reducibility of ZnO/Cu catalysts, the profiles were deconvoluted using the Gaussian Fitting. The reduction temperatures of CuO are shown in Table 3. All ZnO/Cu samples could be divided into two peaks, also suggesting the existence of two different kinds of CuO species or the two reduction steps. After adding ZnO, the position of the CuO reduction peaks shifted to the low temperatures, suggesting that the interaction between ZnO and Cu promoted the reduction of CuO. The most notable peak shift could be seen for ZnO/Cu(4:6) sample, indicating the strongest interaction, which is in accordance with the results of XRD. Meanwhile, the position of the reduction peaks of ZnO/Cu(1:9) did not change much, probably because the content of ZnO was too low to have any effects.

Fig. 5
figure 5

H2-TPR profiles for ZnO/Cu catalysts

Full size image

H2-TPD

The copper surface area (SCu) was determined using the H2-TPD method. Although N2O measurement has been used for about 30 years to calculate the Cu surface area of the Cu/ZnO catalysts, recently, it has been confirmed that this method was not accurate because N2O measured the Cu surface area as well as the oxygen vacancies on the ZnO surface [20, 37, 38], whereas the real Cu surface area should be accurately determined by H2-TPD [20, 37, 39]. Fig. 6 gives the profiles for ZnO/Cu samples. As reported previously, pure ZnO did not absorb any hydrogen in our entire test range [20]. Usually, the desorption temperature of H2 from the ZnO surface was in the range of 300–500 °C [40, 41]. However, Cu species exhibited one desorption peak in the range of 30–60 °C, which can be assigned to the desorbing results of H atoms from metallic Cu surface [42,43,44]. For all the ZnO/Cu profiles, only one desorption peak at about 30 °C can be detected, which is quite similar to the desorption H2 from metallic Cu surface. Based on the hydrogen desorption amount, we calculated Cu dispersion (D) and SCu, which are shown in Table 3. All the ZnO/Cu samples had rather small Cu surface areas and Cu dispersion due to the unique inverse structure. It was observed that for ZnO/Cu catalysts, with the increase of ZnO content from 10 to 50%, the exposed Cu surface area increased from 1.6 m2/g to 4.0 m2/g. However, the dispersion of Cu species had a different trend. It increased from 0.39 to 1.46 when ZnO content increased from 10 to 40%. When further increased ZnO content to 50%, a decrease of the dispersion of Cu species to 0.86 can be observed.

Fig. 6
figure 6

H2-TPD profiles for ZnO/Cu samples

Full size image

Catalytic and stability performance

Fig. 7 displays the stabilities of CO2 conversion with time on stream for inverse ZnO/Cu catalysts. For all the ZnO/Cu catalysts, no obvious decrease of conversion had been noticed for 50 h, suggesting no deactivation had taken place on them. CO and CH3OH selectivity as well as CH3OH yield are shown in Table 4. When the ZnO loading increased from 10 to 40%, CO2 conversion and CH3OH yield increased significantly. However, further increasing ZnO loading to 50%, CO2 conversion and CH3OH yield decreased to 8.9% and 2.4 mmol g−1 h−1. Previously, Cu/ZnO had been studied for methanol synthesis with different Cu/Zn ratios. Valant [29] reported that Cu/ZnO, with a Zn loading of 62%, exhibited the highest activity. The methanol formation rate was about 1.8 mmol g−1 h−1 at 3.0 MPa and 250 °C. Here, our results showed the best catalyst had a 2.8 mmol g−1 h−1 methanol yield, much higher than the reported one even at a lower reaction pressure. We also tested the Cu/ZnO catalyst with a Cu loading of 60%. CH3OH yield was 1.8 mmol g−1 h−1, much lower than the inverse ZnO/Cu sample with the same composition.

Fig. 7
figure 7

Stabilities of CO2 conversion with time on stream over the catalysts. Reaction conditions: 2.0 MPa, 250 °C, VH2:VCO2 = 3:1, space velocity = 3.6 Lˑh−1ˑg−1

Full size image
Table 4 Catalytic performance over ZnO/Cu catalysts at 2.0 MPa and 250 °C, V(H2):V(CO2) = 3:1, space velocity = 3.6 L h−1 g−1
Full size table

Discussion

Although many researchers have focused on the catalytic CO2 hydrogenation using Cu–ZnO based catalysts in recent years, no agreement regarding the active phase of catalysts has been made. Some researchers believed that there was a strong interaction between the activity and the CuZn alloy [39, 45]. Some argued that CO2 conversion was related to the Cu surface area [40, 46]. Still, others held the opinion that ZnO-Cu interfacial sites facilitated the methanol synthesis reaction [13, 47].

As shown in Fig. 8, methanol yield did not follow the trend of Cu surface area. This is also supported by the observation that Cu on its own is a poor catalyst for methanol synthesis from CO2/H2 [48]. Therefore, the surface area of Cu is not the factor for the methanol synthesis activity. Fig. 8 also shows the relationship between the Cu lattice constant and the yield of methanol. Clearly, a linear relationship can be detected. According to the Vegard's law [49], a linear relation exists between the crystal lattice constant of an alloy and the concentrations of the constituent elements. Therefore, it is reasonable to deduce that there is a linear relationship between the methanol yield and the CuZn alloy content.

Fig. 8
figure 8

The relationship between methanol yield and the exposed Cu surface area (obtained from the H2-TPD results), Cu lattice constant (obtained from the XRD results) Reaction conditions: 2.0 MPa, 250 °C, VH2:VCO2 = 3:1, space velocity = 3.6 L h−1 g−1

Full size image

Furthermore, the unique inverse structure inhabited the formation of highly dispersed Cu, of which the amount was much less than that of the conventional Cu/ZnO samples [50]. According to previous reports on the mechanism of CO2 hydrogenation, H/D exchange experiments showed that the main side reaction (reverse water gas shift side reaction, CO2 + H2 → CO + H2O) proceeded on clean Cu surface sites [2, 51]. On CexCuyO [25], Cu/ZrO2 and Ga2O3/Cu/ZrO2 [52] catalysts, highly dispersed Cu sites were also considered the main active phases for reverse water gas shift reaction. Because of the lower amount of highly dispersed Cu sites on the surface, ZnO/Cu(4:6) sample had a lower CO selectivity and thus a higher methanol selectivity than Cu/ZnO (6:4).

Conclusions

Inverse ZnO/Cu catalysts were prepared with different Zn/Cu molar ratios and tested for CO2 hydrogenation to methanol. The inverse structure made the Cu/Zn molar ratios on the surface much lower than in the bulk phase. CuZn alloy was formed in the samples and there is a linear relationship between the methanol yield and the CuZn alloy content. Highly dispersed CuO accounted for only a small amount on the surface, while large amount CuO was in the bulk form. ZnO/Cu(4:6) sample had the smallest Cu particle size and lowest CuO reduction peaks, indicating the strongest interaction between ZnO and CuO. In addition, the unique inverse structure inhabited the formation of highly dispersed Cu sites, which is helpful for the production of CO. As a result, ZnO/Cu(4:6) exhibited the highest CH3OH yield (2.8 mmol g−1 h−1) at 2.0 MPa and 250 °C. From the results, we reckoned that the inverse structure and the tuning of CuZn alloy content would be an important topic to guide the preparation of ZnO/Cu catalysts.