Report

Highly active copper-ceria and copper-ceria-titania catalysts for methanol synthesis from CO2

See allHide authors and affiliations

Science  01 Aug 2014:
Vol. 345, Issue 6196, pp. 546-550
DOI: 10.1126/science.1253057

Converting CO2 into methanol by catalysis

By hydrogenating CO2, scientists can transform a greenhouse gas into methanol, a desirable fuel. Graciani et al. cast copper in the role of the highly active catalyst for this reaction by putting copper particles on cerium oxide. The interface between the cerium oxide and the copper enables the reverse water-gas shift reaction that converts CO2 into CO, which reacts more readily with hydrogen to make methanol. This result takes a step forward in innovating catalysts for this environmentally friendly process.

Science, this issue p. 546

Abstract

The transformation of CO2 into alcohols or other hydrocarbon compounds is challenging because of the difficulties associated with the chemical activation of CO2 by heterogeneous catalysts. Pure metals and bimetallic systems used for this task usually have low catalytic activity. Here we present experimental and theoretical evidence for a completely different type of site for CO2 activation: a copper-ceria interface that is highly efficient for the synthesis of methanol. The combination of metal and oxide sites in the copper-ceria interface affords complementary chemical properties that lead to special reaction pathways for the CO2→CH3OH conversion.

Methanol is a key commodity used to produce acetic acid, formaldehyde, and a number of key chemical intermediates (1). It is synthesized industrially from mixtures of H2, CO2, and CO at elevated pressures (50 to 100 atm) and temperatures (450 to 600 K) with a Cu/ZnO/Al2O3 catalyst (24). Of particular interest is the synthesis of methanol from CO2 (2, 59), not only as a way to mitigate this greenhouse gas but also because of the potential use of CO2 as an alternative and economical feedstock (2, 5, 1012). The activation of CO2 and its hydrogenation to alcohols or other hydrocarbon compounds is an important approach to recycle the released CO2 (2, 5, 11, 12). This is a challenging task because of the difficulties associated with the chemical inertness of CO2 (2, 59, 13). A recent study has identified the active site for the activation of CO2 and the synthesis of methanol on Cu/ZnO/Al2O3 industrial catalysts (2). The active site consists of Cu steps decorated with Zn atoms. Cu alone interacts very poorly with CO2 (2, 69), and alloying with Zn is necessary in order to bind the reactant better and accelerate its transformation into methanol (2). Here we present experimental and theoretical evidence for a completely different type of site for CO2 activation: A Cu-ceria interface, which is highly active for the synthesis of methanol. The combination of metal and oxide centers in the Cu-ceria interface provides favorable reaction pathways for the CO2→CH3OH conversion not seen over a Cu-Zn alloy (2).

The activity for methanol synthesis of a series of catalysts that contain Cu is compared in Fig. 1A (pressure of CO2 = 0.5 atm; pressure of H2 = 4.5 atm; temperature = 500 to 600 K). In addition to methanol synthesis, these surfaces also catalyze the generation of CO through the reverse water-gas shift reaction (RWGS, fig. S1). The data for Cu(111) and Cu/ZnO(000ī) were taken from (6). Extended surfaces of pure Cu exhibit a very low activity for the hydrogenation of CO2 to methanol (68). Clean Cu(111) does not bind CO2, but under reaction conditions, the metal binds it via interactions with H adatoms to form a surface formate species (HCOO) (2, 6). The deposition of Cu nanoparticles on ZnO(000ī) produces a system that is more catalytically active than Cu(111). Cu nanoparticles still do not bind CO2 well, but catalyze subsequent steps in the CO2→CH3OH conversion better than Cu(111) (6). At a temperature of 575 K, the turnover frequency (TOF) for methanol synthesis on Cu(111) is 6.3 × 10−3 molecules per Cu site per second. If one assumes that all of the Cu atoms in Cu/ZnO(000ī) participate in the reaction, the corresponding TOF for methanol synthesis is 9.3 × 10−2 molecules per second (6), which implies an increase of ~15 times in the rate of methanol production with respect to Cu(111).

Fig. 1 Kinetics and STM studies.

(A) Arrhenius plot for methanol synthesis on Cu(111), a 0.2 ML of Cu on ZnO(000ī), a Cu(111) surface covered 20% by ceria, and a 0.1 ML of Cu on a TiO2(110) surface pre-covered 15% with ceria. In a batch reactor, the catalysts were exposed to 0.5 atm of CO2 and 4.5 atm of H2. The reported values are steady-state rates measured at 600, 575, 550, 525, and 500 K. (B) STM image of a CeOx/Cu(111) surface as prepared. (C) In situ STM image taken during exposure to 1.5 torr of H2 at 300 K after 26 hours of reaction. Scanning parameters: 0.3 nA, 1.0 V.

Previous studies have enhanced the reactivity of Cu toward CO2 through alloying Cu with Zn (2) or late transition metals (8, 9). We decided to follow a different approach, coupling Cu to a reducible oxide. The CeOx/Cu(111) system has been studied before as a catalyst for the WGS and CO oxidation reactions (13, 14). The cations of the ceria nanoparticles in contact with Cu(111) can easily alternate between 3+ and 4+ oxidation states (fig. S2). Thus, in CeOx/Cu(111), an active metal-oxide interface can have oxide centers with dynamic chemical properties. The CeO2/Cu2O/Cu(111) system was prepared by vapor-depositing Ce onto Cu(111) in an atmosphere of O2 (14, 15) to form ceria islands on the step edges of a Cu2O surface (13, 15). An in situ scanning tunneling microscope (STM) (16) image of one of these ceria islands is shown in Fig. 1B. The ceria islands in the as-prepared system are predominantly one layer thick, exhibiting rough surfaces that occasionally have a CeO2(111) termination. After exposing CeO2/Cu2O/Cu(111) to 1.5 torr of H2, the complete reduction of Cu2O to Cu and the formation of clusters of O vacancies in the ceria is achieved (Fig. 1C). Reduction in H2 increased the surface roughness of the ceria particles, creating an expanded ceria-Cu interface.

We investigated the hydrogenation of CO2 over a CeOx/Cu(111) surface in which ~20% of the Cu substrate was covered by ceria. Post-reaction surface analysis with x-ray photoelectron spectroscopy (XPS) indicated that the active phase of the catalyst was Ce2O3/Cu(111), as shown in fig. S2. The data for CeOx/Cu(111) in Fig. 1A indicate that this system is a much better catalyst for methanol synthesis than either Cu(111) or Cu/ZnO(000ī). The STM image in Fig. 1C illustrates the difficulties associated with the precise quantification of the active sites at the metal-oxide interface on a CeOx/Cu(111) surface. As a first approximation, one can assume that for the CeOx/Cu(111) system in Fig. 1A, the concentration of active sites in the metal-oxide interface will be at the most equal to the area initially covered by ceria: ~20% of the concentration existing on a clean Cu(111) surface. Given this assumption, we estimate a minimum TOF value of 1.3 molecules per active site per second for methanol synthesis on CeOx/Cu(111) at 575 K. Thus, the rate of methanol production on CeOx/Cu(111) is ~200 times faster than on Cu(111) and ~14 times faster than on Cu/ZnO(000ī). Furthermore, the apparent activation energy for methanol synthesis decreases from 25 kcal/mol on Cu(111) to 16 kcal/mol on Cu/ZnO(000ī) and 12 kcal/mol on CeOx/Cu(111).

Using ambient-pressure (AP) XPS and infrared reflection absorption spectroscopy (IRRAS), we investigated the interaction of CO2 and CO2/H2 mixtures with Cu(111), CeO2(111), and CeOx/Cu(111) surfaces at temperatures between 300 and 500 K. Pure CO2 did not adsorb on Cu(111) at these temperatures. On the other hand, the adsorption of CO2 on a CeO2(111) surface produced strongly bound carbonate (CO32–) species (17). Figure 2 shows IRRAS spectra obtained at room temperature after the exposure of CeOx/Cu(111) to CO2 at two different pressures. At 6.5 × 10−3 torr of CO2 (top spectrum), two peaks were observed at 1288 and 1610 cm−1 that indicate the presence of a carboxylate (CO2δ–) species on the surface (14, 18), and we conclude that a ceria-Cu interface activates CO2. An increase in the CO2 pressure up to 1.0 torr induced the appearance of new peaks at 1210, 1402, and 1440 cm−1 that correspond to carbonate (CO32–) species (17). At this pressure, both carbonate and carboxylate species were stable on CeOx/Cu(111) at 300 K. After adding 9.0 torr of H2 at 300 K, no changes in the spectrum were observed.

Fig. 2 IRRAS spectra at ambient pressures.

The spectra were obtained after the exposure of CeOx/Cu(111) to CO2 and H2 at the indicated pressures and temperatures. All the spectra except the one at the bottom were collected in the presence of CO2 or a CO2 + H2 mixture at the indicated pressures. ΔR/R, normalized reflectivity change.

To investigate temperature effects, the CeOx/Cu(111) system was annealed in CO2 + H2 to 500 K, which led to the appearance of infrared (IR) peaks at 1295, 1330, 1370, 1598, and 2858 cm−1. The peak at 1295 cm−1 we assigned to CO2δ–, whereas the other features we assigned to a formate (HCOO) species (19). The peaks at 1330 and 1370 cm−1 we assigned to symmetric OCO stretches, νs(OCO), whereas the peak at 1598 cm−1 is for the asymmetric OCO stretches, νas(OCO), and the peak at 2858 cm−1 is for CH stretch modes, ν(CH). After evacuating the gases from the AP cell while keeping the sample at 500 K, all of the peaks disappeared, indicating that the adsorbed species formed only under reaction conditions (fig. S3). However, the surface formate species were stable after the sample was cooled down to 300 K under CO2 + H2 and the subsequent evacuation of gases from the AP cell (fig. S3). Similar results were observed with the AP XPS experiments.

Figure 3 shows a C 1s AP-XPS spectrum obtained for a CeOx/Cu(111) surface under a CO2/H2 mixture at 473 K. The weak features at ~284 eV denote the deposition of a very small amount of C on the surface of the catalyst as a consequence of the complete decomposition of CO2. The main feature can be fitted with two peaks at 289.2 and 288.4 eV that we attributed to formates and carboxylates, respectively (14, 20), in agreement with the IRRAS results in Fig. 2. The formates were stable under ultrahigh vacuum (UHV) at temperatures between 300 and 400 K. The carboxylate species were detected in appreciable amounts only when ceria was dispersed on the Cu substrate and were not stable in UHV. The lower stability of the CO2δ– species makes for a better intermediate for methanol synthesis than the formate species, which have high stability (2, 6) and may not be efficient as transient species in the CO2→CH3OH conversion (21). Thus, the addition of CeOx nanoparticles to Cu(111) creates a metal-oxide interface that allows the adsorption and activation of CO2, opening a new reaction pathway for the synthesis of methanol.

Fig. 3 XPS data.

The C 1s AP-XPS data obtained for CeOx/Cu(111) after exposure to CO2 (30 mtorr) and H2 (270 mtorr) at 473 K.

To provide a molecular description of the methanol synthesis mechanism on our system, we performed state-of-the-art density functional theory (DFT) calculations, looking for possible intermediates and screening them to build a plausible thermodynamic reaction pathway (Fig. 4). Following previous studies for CO2 hydrogenation on clean Cu(111) (2, 6, 22), we considered reaction pathways that initially involved the formation of formate (HCOO) but abandoned them because of the high stability of this species. The experimental measurements point to the simultaneous production of methanol and CO during CO2 hydrogenation (fig. S1). The reaction path in Fig. 4 shows the best calculated route for the RWGS and methanol synthesis. The overall reaction takes place under a H-rich atmosphere. Exothermic pre-hydrogenation of the ceria creates Ce3+ centers, which facilitates CO2 adsorption (adsorption energy = –12.2 kcal/mol, “a”). On these centers, CO2 is activated in a bent conformation (“b”) and hydrogenated to yield a carboxyl OCOH species (“c”). The calculated energy barriers for these processes are in the range of 10 to 11 kcal/mol. An even lower energy barrier of 4.6 kcal/mol is associated with the OCOH → CO + OH (“d’) transformation. At this point, water is formed from adsorbed H and OH species. The exothermic adsorption-dissociation of a second H2 molecule (–9 kcal/mol) facilitates the water desorption, a process that at 0 K is still endothermic by 12.2 kcal/mol. However, taking into account the favorable entropy contribution for desorption and the reaction temperature (575 K), the calculated energy for water desorption is as low as 3 kcal/mol. Through these steps, the RWGS reaction takes place. Subsequent steps achieve the hydrogenation of CO to methanol through expected intermediates: HCO (formyl, “f”), H2CO (formaldehyde, “g”), and H3CO (methoxy, “h”). Our experimental data indicate that the RWGS reaction and methanol synthesis have similar apparent activation energies (table S1). This phenomenon has been seen before for CO2 hydrogenation on Cu(111) (7) or Cu/ZnO(000ī) (6) and suggests that both reactions go through a common intermediate (21). The DFT results summarized in Fig. 4 point to OCOH as the common intermediate. The calculated reaction barriers for the transformation of CO2 into OCOH and the first hydrogenation of CO are similar (11 to 14 kcal/mol) and close to the apparent activation energies observed in the experiments for CeOx/Cu(111) (11 to 12 kcal/mol). The calculated energy barriers for the HCO→H2CO, H2CO→H3CO, and H3CO→H3COH hydrogenations are quite small, 5.3, 3.5, and 5 kcal/mol, respectively. The final desorption of the methanol to complete the whole reaction (CO2 + 3 H2 → CH3OH + H2O) is highly endothermic (19.1 kcal/mol) but, again, taking into account the favorable entropy and temperature contributions, the energy for the desorption process is only 5.3 kcal/mol. The theoretical results in Fig. 4 indicate that the thermochemistry of the reaction steps associated with the formation of methanol on a ceria-Cu interface is predominantly downhill, with an overall exothermic process, which is not the case for the thermochemistry of the reaction on a CuZn alloy (2), the active site for Cu/ZnO/Al2O3.

Fig. 4 DFT studies.

Reaction path for methanol synthesis (MS) by CO2 hydrogenation on the CeOx/Cu(111) system. Step colors are as follows: local minima states (blue), transition states (TS) (red), and energies including the entropy contribution (gray). The calculated energy for the whole methanol synthesis reaction is indicated on the right side of the graph. The main transition states are pointed out by vertical red arrows, and the respective energy values are also shown under the arrows. The other energy barriers are indicated above the respective transition-state step. The whole reaction has been divided into two parts: The first one constitutes the RWGS (indicated by a green arrow on the top of the graph), and the second one consists of the steps leading to the methanol synthesis (MS) (red arrow on the top of the graph). The optimized structures for the main intermediates are included in the bottom part of the figure. Carboxylate species observed in the IRRAS experiments correspond to the structure shown in b. Colors: O (red), C (gray), H (white), cerium (light-beige), Cu (dark pink).

The absence of CO, HCO, H2CO, and H3CO in the in situ IR and XPS spectra is probably a consequence of the low residence time of these species on the surface under reaction conditions. CO adsorbs weakly on Cu (14, 15). After the addition of the first H to CO, the barriers for the subsequent hydrogenation steps are relatively small (<6 kcal/mol) and can be easily overcome by the high temperature (>450 K) necessary for the catalytic synthesis of methanol. Under reaction conditions, the only species observed in the IR and XPS spectra are HCOO (Figs. 2 and 3), a spectator, and the active species CO2 and OH (Figs. 2 and 3 and fig. S4). The ceria component enhances the stability of CO2 and OH on the surface of the catalyst, opening efficient routes for methanol synthesis not seen on pure Cu or CuZn alloys (2, 6).

A comparison of the results in Fig. 1A for Cu(111) and Cu/ZnO(000ī) points to an enhancement of the catalytic activity of Cu when this element is present in the form of nanoparticles. We investigated the hydrogenation of CO2 on catalysts generated by codepositing nanoparticles of Cu and ceria on TiO2(110). Previous studies have examined the properties of the CeOx/TiO2(110) mixed-metal oxide in detail (2325). Ceria nanoparticles in contact with titania form small wirelike structures and have Ce3+ and Ce4+ states of almost equal stability (23, 24). STM images show a close contact between Cu and ceria on a TiO2(110) substrate (24). Thus, the type of chemistry seen in Figs. 2 to 4 for the Cu-ceria interface of CeOx/Cu(111) also can occur in Cu/CeOx/TiO2(110). Kinetics data for methanol synthesis on a catalyst generated by depositing a ~0.1 monolayer (ML) of Cu on a TiO2(110) surface pre-covered 15% by ceria nanoparticles (Fig. 1A) show that this surface is clearly the best catalyst. At 575 K, we estimate a TOF of 8.1 molecules per active site per second. The rate of methanol production on Cu/CeOx/TiO2(110) is ~1280 times faster than on Cu(111) and ~87 times faster than on Cu/ZnO(000ī). In control experiments, we examined the synthesis of methanol on 0.1 ml of Cu supported on TiO2(110) or CeO2(111) surfaces and measured catalytic activities that were somewhat greater than that of Cu/ZnO(000ī) but much less than that of Cu/CeOx/TiO2(110) or even CeOx/Cu(111). Plain CeOx/TiO2(110) had no activity for methanol synthesis. Thus, the extremely high activity of Cu/CeOx/TiO2(110) is probably a consequence of generating a metal-oxide interface involving Cu and ceria nanoparticles.

This study illustrates the substantial benefits that can be obtained by properly tuning the properties of a metal-oxide interface in catalysts for methanol synthesis. In a metal-oxide interface, one can have adsorption/reaction sites with complementary chemical properties (14, 26), truly bifunctional sites that would be very difficult to generate on the surface of a pure metal or alloy system (2, 9, 27).

Supplementary Materials

www.sciencemag.org/content/345/6196/546/suppl/DC1

Materials and Methods

Figs. S1 to S4

References (28–37)

References and Notes

  1. Acknowledgments: The research carried out at Brookhaven National Laboratory was supported by the U.S. Department of Energy, Chemical Sciences Division (DE-AC02-98CH10886). J.E. is grateful to the Instituto de Tecnologia Venezolana para el Petroleo for support of the work carried out at the Universidad Central de Venezuela. The work performed at the University of Seville was funded by the Ministerio de Economía y Competitividad (Spain, grants MAT2012-31526 and CSD2008-0023) and European Regional Development Fund. Computational resources were provided by the Barcelona Supercomputing Center/Centro Nacional de Supercomputación (Spain). The AP-XPS spectra were acquired at the Advanced Light Source (beamline 9.3.2), which is supported by the U.S. Department of Energy under contract no. DE-AC02-05CH11231.
View Abstract

Stay Connected to Science

Navigate This Article