Report

Quantifying the promotion of Cu catalysts by ZnO for methanol synthesis

See allHide authors and affiliations

Science  20 May 2016:
Vol. 352, Issue 6288, pp. 969-974
DOI: 10.1126/science.aaf0718

How zinc helps copper make methanol

Copper nanoparticles can catalyze the formation of methanol from a mixture of CO2, CO, and H2, but adding zinc oxide nanoparticles, themselves inactive in this reaction, greatly boosts the rates. Kuld et al. measured how methanol synthesis activity varies with the coverage of zinc atoms on the copper nanoparticles, as determined experimentally and with density functional theory calculations. The ZnO nanoparticle size determined how much zinc covers the copper surface and in turn controlled the catalyst activity.

Science, this issue p. 969

Abstract

Promoter elements enhance the activity and selectivity of heterogeneous catalysts. Here, we show how methanol synthesis from synthesis gas over copper (Cu) nanoparticles is boosted by zinc oxide (ZnO) nanoparticles. By combining surface area titration, electron microscopy, activity measurement, density functional theory calculations, and modeling, we show that the promotion is related to Zn atoms migrating in the Cu surface. The Zn coverage is quantitatively described as a function of the methanol synthesis conditions and of the size-dependent thermodynamic activities of the Cu and ZnO nanoparticles. Moreover, experimental data reveal a strong interdependency of the methanol synthesis activity and the Zn coverage. These results demonstrate the size-dependent activities of nanoparticles as a general means to design synergetic functionality in binary nanoparticle systems.

Nanoparticles (NPs) used to catalyze chemical reactions generally expose a variety of surface sites (e.g., facets, steps, and corners), each with their distinct reactivity (1), and in some cases the function of catalysts based on metal NPs can be predicted with first-principles methods (2). The addition of minor amounts of suitably chosen promoter elements can further enhance or suppress surface reactivity and selectivity by orders of magnitude. In general, promotion is thought to originate from either an electronic modification of the catalytic active site or a morphological change of the NPs that enhances the abundance of the active sites, but a detailed understanding is often difficult to establish because of the lower abundance of the promoter atoms. Adding to the difficulty, numerous experimental observations have shown that the catalytic active state dynamically adjusts to the working conditions (1). That must be included to understand catalyst promotion.

Here, we focus on the catalyzed transformation of synthesis gas (mixture of H2, CO, and CO2) into methanol. Methanol is consumed at the scale of 65 million tons/year (2013) and considered as a future energy carrier (3). The typical catalyst consists of Cu NPs mixed with NPs of ZnO and Al2O3. Although Cu can function alone as a methanol synthesis catalyst, its activity is substantially boosted by the interaction with ZnO, which on its own has only negligible catalytic activity (4). The Cu-ZnO system has emerged as a prototype for studying complex promotional interactions in catalysis (1, 522), and several possible explanations have been suggested, including (i) gas-dependent morphological changes of Cu on ZnO (6, 7, 9); (ii) support-induced strain in Cu (22); and (iii) ZnOx species/layers covering part of the Cu NPs (5, 6, 9, 10, 14, 1921), often denoted as a strong metal-support interaction (SMSI). More recently, the active site was associated with step sites in the Cu surface and the Cu-ZnO synergy proposed to reflect the incorporation of Zn atoms into the Cu steps (5, 18). This picture is attractive because of its consistency with observations of fully or partly reduced Zn in the Cu surface after methanol synthesis over Zn-decorated Cu foils and single crystals and after reduction of industrial-type Cu/ZnO/Al2O3 catalysts (11, 13, 15, 19). As this dynamic phenomenon may be typical for metal-promoter interactions in heterogeneous catalysis, it is important to reinforce such a picture with a quantitative extension that enables a full description of the Cu-ZnO interaction over the examined reactions’ conditions.

We initially studied the Cu/ZnO/Al2O3 catalyst in the reduced state with high-resolution transmission electron microscopy (TEM). The reduced catalyst consists of an agglomeration of Cu, ZnO, and Al2O3 NPs in intimate contact (Fig. 1A). Moreover, Fig. 1B shows that the Cu lattice fringes extend to the projected surfaces of the NPs without changes in their spacing and structure, indicating that the Cu surfaces are in direct contact with the gas environment. The Cu surfaces can thus only be covered with a submonolayer of Zn, consistent with x-ray photoelectron spectroscopy (XPS) measurements (11). For that reason, the present work focused on describing the coverage of Zn(Ox) in the Cu surface during methanol synthesis.

Fig. 1 Electron micrographs of an industrial-type Cu/ZnO/Al2O3 methanol catalyst acquired in situ during exposure to 1 mbar H2 at 300°C.

(A) The image shows the arrangement of nanoparticles in the catalyst. (B) The image reveals a Cu NP with surfaces having direct gas accessibility.

First, the Zn coverage, θZn, at steady state was determined in the Cu surface of a Cu/ZnO/Al2O3 catalyst under the exposure to synthesis gas. The concentration in bulk Cu of Zn atoms from ZnO was obtained from the thermodynamics of the ZnO reduction reaction, and the energetic driving force for Zn to migrate to different Cu facets was determined from density functional theory (DFT) calculations, which were used instead of a simple broken-bond model (23) to improve the precision. By these calculations, θZn is estimated at different gas pressures and temperatures, representative for the methanol synthesis, for a series of Cu surfaces, including steps, edges, and corners. Different Cu surface sites are considered because the methanol synthesis reaction is structure sensitive (5, 18, 24). Formation of Zn-Zn bonds is included in the calculations, as these are important at high θZn. Moreover, the calculations are also done for Cu and ZnO NPs because their reduced size affects the thermodynamic calculations and the abundance of different Cu surface sites.

The formation of a Cu-Zn bulk alloy by incorporation of Zn atoms into Cu NPs can be described by reaction 1 or 2.ZnO(s) + CO = Zn + CO2 (1)ZnO(s) + H2 = Zn + H2O (2)Under conditions where the water-gas shift (WGS) is not in equilibrium, the kinetics of the reduction/oxidation reactions determine whether reaction 1 or 2 defines the reduction potential of the gas. Here, the CO/CO2 ratio was chosen because high CO and CO2 concentrations were generally applied in the experimental measurements to increase the kinetic importance of this reaction pair. The equilibrium constant K1 for reaction 1 is given byK1 = aZnaCO2/(aZnOaCO) = exp(–dGred/RTK) (3)where dGred is the free energy, a is the activity of the species, R is the gas constant, and TK is the temperature in Kelvin. The free energy change of reaction 1 for Zn = Zn(s) is 60.95 kJ/mol at 220°C (25). Thus, pure Zn(s) cannot form in synthesis gas at 200° to 300°C, but Zn can thermodynamically form a disordered bulk α-alloy phase with Cu (23, 26, 27). The thermodynamic activity of Zn in Cu is given by aZn = γZnXZn, where XZn is the fraction of Zn in the alloy, and the activity coefficient of Zn in Cu, γZn, is calculated from (26) as ln(γZn) = –0.736 – 2977/TK. The fraction of Zn in the alloy with Cu is thus given byLn(XZn) = –dGred/RTK – ln(γZn) – ln(aCO2/aCO) (4)Equation 4 is fulfilled when steady state is reached. The diffusion coefficient of Zn in Cu indicates that establishing equilibrium concentrations of Zn in Cu NPs takes years at 200°C, days at 250°C, and hours at 300°C (27). In Contrast, high θZn is observed within a few hours at 220°C (11).

To describe the situation where θZn reaches steady state at the Cu surface, it is convenient to use Eq. 4, although bulk equilibrium of Zn is not established in the experiment. The bulk contribution in Eq. 4, however, has to be corrected for the effect of the lower atom coordination in NPs. This change in chemical potential of ZnO, ΔμZnO, depends on the ZnO crystallite diameter, dZnO, and is given by (28)ΔμZnO = 4γZnOMZnO/(dZnOρZnO) (5)where the surface energy, γZnO, is 0.74 J/m2 (29), and MZnO and ρZnO are the molar weight and density of ZnO, respectively. The additional chemical potential of Cu NPs relative to bulk Cu is described by a similar term, except that the increased chemical potential gives rise to a lower thermodynamic driving force for Cu-Zn alloy formation. Thus, combining these considerations yieldsLn(XZn) = −dGred/RTK − ln(γZn) − ln(aCO2/aCO) + 4γZnOMZnO/(dZnOρZnORTK) − 4γCuMCu/(dCuρCuRTK) (6)where 1.95 J/m2 (30) may be used for γCu.

Finally, to calculate θZn in Cu surfaces, the equilibrium constant for Zn segregation in bulk Cu, Kseg, is calculated (see fig. S1 and tables S1 to S5). Zn will tend to segregate to the surface of Cu because atoms in surfaces have fewer nearest neighbors than in the bulk, and the bond energy is lower for Cu-Zn than for Cu-Cu bonds. By means of DFT, the segregation energies of Zn from bulk Cu to Cu(111), Cu(100), Cu(211)(edge), and Cu(532)(kink) surfaces were calculated to 23.7, 26.1, 27.0, and 29.8 kJ/mol at 0 K, respectively, and the entropy contributions to the segregation were estimated for Cu(111) and Cu(211) to be 5.8 to 7.1 J/K/mol. Because Eq. 6 provides the calculation of the bulk concentration of Zn in Cu, and the segregation energies afford the relative abundances of Zn in the bulk and at the surface of Cu, θZn can be calculated for the most abundant Cu facets, steps, and corners from Embedded Image.

Figure 2A shows the calculated θZn at Cu(111), Cu(100), Cu(211)(edge), Cu(532)(corner), and XZn at 220°C as a function of the CO/CO2 ratio. A random distribution of Zn in the Cu surface is assumed as observed experimentally for a Cu(111) surface by Sano et al. (31) up to θZn = 0.18. At higher θZn, Zn-Zn bonds are established in the Cu surface and reduce the segregation energies (fig. S2). This Zn-Zn interaction energy at Cu(111), Cu(100), and Cu(211) is estimated by DFT, and the calculations of θZn include this correction. Moreover, θZn is also estimated for a catalyst with both Cu and ZnO NPs (Fig. 2A). The calculations use NPs of ZnO, with dZnO = 87 Å determined by x-ray diffraction (XRD) and a regular cubo-octahedral shape for Cu NPs, having a size of dCu = 88 Å (which is closest to the crystal size of dCu = 85 Å, obtained by XRD). The fractions of the different types of surface sites were then obtained according to the model of Van Hardeveld and Hartog (32) (table S6). At the terraces, edges, and corners of the cubo-octahedron, θZn was estimated as the coverage at the Cu(111), Cu(100), Cu(211), and Cu(532) surface sites, respectively. Taking these coverages together results in a θZn for the Cu NP that closely follows that of the more abundant (111) facet site as a function of the CO/CO2 ratio (Fig. 2A). That is, θZn increases steeply for lower CO/CO2 ratios and more slowly at higher CO/CO2 ratios, because of the decrease in the Zn segregation energies as a result of Zn-Zn interactions. Creation of additional undercoordinated sites because of bulk defects (5) is not considered. Nevertheless, the difference between the coverages of Zn at steps and terraces is not large, as seen from Fig. 2A, and thus, the number of added Zn atoms caused by a small number of additional step sites is not expected to result in substantial changes of the total θZn.

Fig. 2 Modeling and measurements of the Zn coverages at Cu nanoparticles in a Cu/ZnO/Al2O3 methanol catalyst.

θZn at specific sites of a Cu surface and on the surface of a Cu cubo-octahedron with 88 Å in diameter and dZnO = 87 Å (closest to a catalyst with dCu = 85 Å and dZnO = 87 Å, as determined by XRD) are calculated using a model based on thermodynamics and DFT. (A) Zn solubility (XZn) in Cu and Zn coverages, including Zn-Zn interactions [apart from one of the curves for edge sites], plotted as a function of the CO/CO2 ratio at 220°C. (B) Zn coverages determined by H2-TPD compared with those predicted theoretically at 220°C (dashed line) and 280°C (solid line).

Several studies have proposed the presence of ZnOx layers or Zn attached to oxygenated species partly covering the Cu particle surface in Cu/ZnO/Al2O3 catalysts (1, 4, 5, 8, 12, 13, 15, 16, 18, 21, 22, 24, 29, 3338). Previously (11), we provided strong evidence that the identity of these species is a Cu-Zn surface alloy using XPS, N2O-RFC (reactive frontal chromatography), H2-TPD (temperature programmed desorption), and H2-TA (transient adsorption). Furthermore, Nakamura et al. (13) found a good correlation between the number of Zn atoms in the Cu surface attached to formate at 90°C and the methanol synthesis activity, strongly indicating that single Zn atoms, possibly with adsorbed species like formate, and not ZnOx overlayers are the active phase in methanol synthesis, in agreement with our present TEM images.

Then, we note that Zn in the Cu surface must be partly oxidized during the methanol synthesis reaction due to the attachment of formate or other oxygenated reaction intermediates to Zn in the Cu surface (5, 18). Formate is reported theoretically and experimentally to be the most abundant adsorbate in synthesis gas (5, 13, 18, 35, 37). Our experimental work shows that the overall adsorbate coverage at the Cu surface is so low that adsorbates do not affect the Zn stability, and hence coverage, substantially at 220°C. This finding is supported by the theoretical work (5, 18, 35) (figs. S3 and S4 and tables S7 and S8) suggesting only high coverage of formate at the few step sites and low coverage at the abundant terrace sites of Cu NPs under methanol synthesis conditions. Thus, our experimental work does not rule out that Zn is partly oxidized during the reaction, but the key point for our model is that the experimental results demonstrate that any such effect is sufficiently weak that it does not affect the Zn coverage. Therefore, we find it most probable that the main part of the Zn in Cu is in the reduced state at 220°C and above, in agreement with the theoretical studies.

However, we find high coverages of adsorbates, most probably formate (13), at the Cu surface after methanol synthesis at 1 bar and 90°C. Hence, our Zn model does not cover these conditions, but the temperatures in industrial methanol synthesis are generally higher—220° to 280°C—resulting in lower adsorbate coverages. DFT calculations also show that CO will not compete effectively with formate at Cu(211) steps with and without Zn in CO2-containing synthesis gases (18).

In our recent work (11), a method for measuring θZn in industrial-type methanol catalysts was established by means of N2O-RFC, H2-TPD, H2-TA, and XPS. The results showed that values of θZn determined by H2-TPD are proportional to those determined by the complementary methods, as seen from fig. S5 (11). In the present work, the gas-induced Zn decoration of the Cu NPs was only addressed using H2-TPD. This method was applied because the probe interacts only mildly with the surface (8) and provides reliable and reproducible values of θZn with high precision (11). A series of catalysts were treated in different CO/CO2 gas mixtures with small amounts of H2 added and then analyzed by the H2-TPD method. H2 was added to mimic a real synthesis gas, but in low concentrations (H2/COx = 0.02) to suppress possible surface-formate and methanol formation, to ensure that reduction of ZnO proceeds via CO rather than via H2 and to diminish the reverse WGS. With 2% H2 added, the change in the CO/CO2 ratio at WGS equilibrium is <0.004 at 220°C and <0.006 at 280°C. In three experiments, H2/COx ratios up to 3.17 were used. For high H2/COx ratios and non-WGS equilibrium, H2 will at some point dominate the reduction of ZnO(s), and a more complicated analysis than that applied here will be needed. In all experiments, the CO/CO2 ratios were determined by direct measurements as the average between the inlet and the outlet values.

Experimental series were initiated and terminated by a (standard) H2-TPD measurement consisting of a reduction in 1% H2 at 220°C for 5 hours, flushing in He for 1 hour, exposure of the sample to 16 bar of H2 at –33°C, cooling to –196°C, and heating in He to 210°C. In between these two standard measurements, 5 to 20 hours of pretreatments in different gases followed by H2-TPD measurements were performed. Hereby, θZn was determined using Eq. 7 (fig. S6).Embedded Image (7)Embedded Image is the hydrogen desorption capacity after a pretreatment, and Embedded Image is the weighted average of the hydrogen desorption capacity obtained from the initial and the final standard H2-TPD measurements, assuming a similar change of surface area in each treatment cycle. θZn is corrected for the value of θZn = 0.04 that was found after a standard reduction in 1% H2 at 220°C by XPS (11).

Most of the Zn measurements were performed after treatments in CO/CO2/H2(2%) gas mixtures. The applicability of the CO/CO2 ratio as descriptor and the establishment of the proposed steady state of the catalyst were addressed by several additional experiments. First, θZn was measured after pretreatments in pure CO/CO2 and H2O/H2 gases, in CO/CO2/H2(400 ppm) (parts per million) gas, and in a synthesis gas (CO/CO2/H2 = 18/6/76) at 1, 10, and 40 bar. The reduction potential of the H2O/H2 gas in terms of a CO/CO2 ratio was calculated by converting the H2O/H2 ratio to a CO/CO2 ratio using the equilibrium constant for the WGS reaction: Embedded Image. Any formate present at the Cu/Zn surface would decompose in the He purge flow at 220°C (1 hour) before the H2-TPD and ZnOx or Zn-OH species in Cu would be measured as a Zn atom (11). The data from these experiments match those obtained with CO/CO2/H2(2%) mixtures (Fig. 2B), which shows the robustness of the model and the method for determination of θZn. Adsorbates at the surfaces of Cu and the oxide support during methanol synthesis could potentially modify the stability of Zn in the surface of Cu, and hence θZn. However, the fact that treatments in synthesis gas and in CO/CO2, H2O/H2, CO/CO2/H2(400 ppm) and CO/CO2/H2(2%) gas mixtures resulted in similar deviations of θZn from the model verifies that our model is also valid at synthesis conditions, even though the model is derived for reduced Zn.Second, variation of space velocity (3 to 30 Nl g−1 h−1) did not change θZn (<5.4%). Third, experiments showed that the standard 5-hour pretreatment time provides values of θZn that are negligibly (<2%) lower than those obtained using pretreatment times of 10 or 20 hours.

The good correspondence between measured and modeled data (SD = 0.037) in Fig. 2B, obtained without any fitting or optimization, and the shapes of the θZn versus CO/CO2 model curves and the experimental data being similar, indicate that the physical principles in the model are sufficient to describe the dynamic dependency of θZn on the reduction potential of the methanol synthesis gas for an industrial-type methanol catalyst. The model describes quantitatively the interaction between Cu and ZnO over a wide range of relevant methanol synthesis gas compositions and temperatures and is a function of the nanostructural architecture of the Cu/ZnO/Al2O3 catalyst as well.

Next, to determine the promoting effect of Zn in an industrial-type methanol synthesis catalyst, the methanol activity was related experimentally to θZn. Specifically, pretreatments in H2 at different pressures were used as a convenient method to induce different θZn, as described in detail previously (11). The resulting θZn after these treatments was determined by kinetics rather than thermodynamics. The methanol activities were then measured upon heating from room temperature to 140°C (in steps of 10°C in the range 90° to 140° and back to 90°C) at ambient pressure in a CO/CO2/H2 = 18/18/64 gas mixture and at a space velocity (SV) of 30 to 33 Nl/g/h. Figure 3A shows the measured methanol exit concentrations as a function of temperature. Upon heating from 90°C, activity was absent, and, after passing 110°C, activity increased substantially with time at the different temperature plateaus up to 140°C.. A similar induction time was observed by Yang et al. (39) at 140°C, and they ascribed this phenomenon to a lack of water vapor. The most likely explanation seems to be some kind of delay in the accumulation af active species at the surface of the catalyst, which possibly could involve formate (37). Nevertheless, the activity stabilized upon the following stepwise cooling of the catalyst. After the activity measurements, θZn was measured as described in more detail in the supplementary material (fig. S7). To rank the performance of the catalyst at different θZn, the constant activities at 130°C during the cooling sequence, relative to that obtained after reduction in 1% H2, are plotted as a function of θZn in Fig. 3B.

Fig. 3 Methanol activities as a function of the Zn coverages at the Cu nanoparticles of a Cu/ZnO/Al2O3 methanol catalyst.

(A) Temperature ramp and exit methanol concentrations after pretreatment as a function of time and temperature in a 18/18/64 = CO/CO2/H2 gas mixture at ambient pressure and SV = 30 to 33 Nl/g/h. (B) Relative measured methanol exit concentrations at 130°C (temperature ramp-down) as a function of postreaction values of θZn. The varying values of θZn are obtained by pretreatments in H2 at different pressures and temperatures prior to activity tests. The dashed line is a second-order polynomial fit to the data.

Figure 3B shows that the methanol synthesis activity is strongly dependent on θZn. The activity increases by a factor of 3 for θZn in the range 0.04 to 0.47 and is fitted well by a second-order polynomial equation. The data indicate that the methanol activity may reach a maximum near θZn ≈ 0.47. Previously, Nakamura and co-workers related the methanol synthesis to θZn physically deposited onto polycrystalline Cu particles and single crystals (13, 15), and they also reported a maximum in the methanol turnover frequency at θZn ≈ 0.20. The reason for this difference is not fully understood, but as suggested by Nakamura et al. (13), the loss of activity at higher values of θZn resulted from large coverage of ZnO, in agreement with our model suggesting that ZnO is the stable phase of Zn under their conditions. In Cu/ZnO/Al2O3 catalysts, part of the Cu area is already covered by ZnO, as seen from Fig. 1A. Thus, θZn is a key parameter for understanding the activity of the industrial-type Cu/ZnO/Al2O3 methanol catalyst, and Zn-covered Cu sites at Cu facets or steps or both constitute the active site for methanol synthesis. The model enables modification of the Cu-ZnO interaction, hence optimizing the activity through control of the reaction conditions as well as tuning the sizes of Cu and ZnO. The Cu and ZnO size dependencies of θZn and the relative turnover frequency of methanol, TOFMeOH,rel, were determined from the model and using the second-order polynomial fit to the data in Fig. 3B. The results are depicted in Fig. 4A. Interestingly, the present model predicts that catalysts prepared with small ZnO particles will be very active because of an enhanced spillover of Zn to the Cu surfaces (Fig. 4A); hence, it is the thermodynamic activity of the promoter phase, ZnO, that determines promotion of the methanol synthesis activity. These effects readily explain the findings by Kurtz et al. (16) and Liao et al. (17), suggesting that higher methanol synthesis activities are obtained for Cu/ZnO catalysts with nanosized ZnO or high-energy ZnO facets relative to those of catalysts containing larger ZnO particles or low-energy ZnO facets. Similarly, Fichtl et al. (40) report that deactivation of Cu/ZnO/Al2O3 methanol synthesis catalysts cannot be related solely to the loss of Cu surface area, and they observe growth in ZnO crystallite sizes during sintering. Another, and perhaps even more surprising, implication of the model is that θZn varies with Cu particle sizes, dCu. Normally, a catalyst’s activity increases as the constituent particle size decreases because of the enlargement of the gas-accessible surface area. In contrast, the model predicts a decrease in θZn with decreasing dCu and hence lower TOF for smaller Cu NPs (Fig. 4B). This effect will tend to reduce the benefits of decreasing the Cu particle sizes in methanol catalysts. To our knowledge, this effect was not observed experimentally so far, presumably because the changes in the methanol activity with Cu particle size for Cu/ZnO catalysts were only studied for dCu > ~100 Å (38, 41).

Fig. 4 Modeling of Zn coverages at Cu nanoparticles and relative methanol turnover numbers of a Cu/ZnO/Al2O3 catalyst as a function of ZnO and Cu particle sizes.

(A) θZn (dashed line) and TOFMeOH,rel (solid line) relative to that at θZn = 0, plotted as a function of the ZnO particle diameter. (B) θZn (dashed line) and TOFMeOH,rel relative to that at dCu = 0 (θZn = 0) as a function of Cu particle diameter (solid line). The parameters in the model are dZnO = 87 Å, dCu = 88 Å, T = 220°C, and CO/CO2 = 0.5. A fit to the activity data in Fig. 3B was used to estimate TOFMeOH,rel for a given value of θZn. The uncertainties in θZn and TOFMeOH,rel were estimated to 0.037 and 0.06‐ 0.36 for θZn = 0‐ 0.3, based on the deviation of the data and the model in Fig. 2B and on the fit to the data in Fig. 3B.

An alternative approach to optimizing the performance of Cu-ZnO–based methanol catalysts is to enhance θZn in the Cu surface by increasing the reduction potential of the synthesis gas. Interestingly, a peak in the activity is observed for CO/CO2 ≈ 10, when the reduction potential of a synthesis gas is varied by substituting CO with CO2 (4, 33). This maximum in activity for high CO/CO2 ratios is surprising, because isotope labeling shows that carbon atoms in methanol formed over Cu/ZnO catalysts mainly originate from CO2 (18). The present work demonstrates that high CO/CO2 ratios results in high θZn and, thus, high activities. We suggest that a CO/CO2 ratio of ~10 combines a high θZn (~0.25 at 220°C) (see Fig. 2B) and a CO2 partial pressure of 0.4 to 0.5 bar that ensures an adequate coverage of formate (5, 18, 35) or another key precursor (37) at Zn-decorated Cu sites.

Moreover, the present model also consistently explains other dynamical effects reported on methanol catalysts. For example, shape changes observed for Cu NPs on a ZnO support were observed by in situ extended x-ray absorption fine structure and in situ TEM (6, 7, 9). In the present model, the Zn atoms incorporated into the Cu surface under reducing conditions result in a decrease in the surface energy of Cu, and this will, according to (7), contribute to the increased wetting of Cu NPs on ZnO and provides an explanation of transient activity profiles.

Supplementary Materials

www.sciencemag.org/content/352/6288/969/suppl/DC1

Materials and Methods

Figs. S1 to S7

Tables S1 to S8

References (4249)

References and Notes

  1. Acknowledgments: We thank P. G. Moses for fruitful discussions and commenting on the work at an early stage. The Center for Individual Nanoparticle Funtionality is sponsored by the Danish National Research Foundation (DNRF54).
View Abstract

Navigate This Article