Activation of Cu(111) surface by decomposition into nanoclusters driven by CO adsorption

See allHide authors and affiliations

Science  29 Jan 2016:
Vol. 351, Issue 6272, pp. 475-478
DOI: 10.1126/science.aad8868

Nanoclusters just by adding CO

The most closely packed surfaces of transition metals are usually stable under vacuum, but during catalytic reactions, energetic changes that result from adsorbing molecules could change the surface structure. Eren et al. present an extreme example for carbon monoxide (CO) adsorption on the (111) surface of copper at very low partial pressures. The surface decomposed into small nanoclusters (most containing 3 or 19 atoms). The surface was more reactive than the original and, for example, could dissociate adsorbed water at room temperature.

Science, this issue p. 475


The (111) surface of copper (Cu), its most compact and lowest energy surface, became unstable when exposed to carbon monoxide (CO) gas. Scanning tunneling microscopy revealed that at room temperature in the pressure range 0.1 to 100 Torr, the surface decomposed into clusters decorated by CO molecules attached to edge atoms. Between 0.2 and a few Torr CO, the clusters became mobile in the scale of minutes. Density functional theory showed that the energy gain from CO binding to low-coordinated Cu atoms and the weakening of binding of Cu to neighboring atoms help drive this process. Particularly for softer metals, the optimal balance of these two effects occurs near reaction conditions. Cluster formation activated the surface for water dissociation, an important step in the water-gas shift reaction.

An extensive array of surface-sensitive characterization techniques that provide structural (e.g., electron and x-ray diffraction and scanning probe microscopy) and spectroscopic (e.g., Auger electron, x-ray photoelectron, infrared, and Raman) information (1, 2) have revealed the structure of many crystal surfaces in their pristine clean state. Most of these studies are carried out in ultrahigh vacuum (UHV), which makes it possible to control sample composition and cleanliness to better than 0.1% of a monolayer (ML). Under realistic ambient conditions, however, our knowledge is far less extensive, because the most sensitive techniques using electrons cannot operate in the presence of gases at pressures above ~10−6 Torr. Of particular interest is the structure of surfaces in dynamic equilibrium with gases at near-ambient pressure and temperature (1). Under these conditions, weakly bound adsorbates can be present in considerable densities, a situation that can also be achieved under vacuum, but only at cryogenic temperatures. The surface structures obtained in such rarefied conditions often represent kinetically frozen states and may not be representative of the structure under practical operating conditions. Here, we overcome this difficulty using high-pressure scanning tunneling microscopy (HPSTM) (38) and ambient pressure x-ray photoelectron spectroscopy (APXPS) (9, 10), which make possible the study of surfaces in the presence of gases at or near atmospheric pressures at room temperature and above.

Copper-based heterogeneous catalysts are used in reactions such as water gas-shift (WGS), methanol oxidation, methanol synthesis, and others (1117). The weaker cohesive energy of Cu compared with other metals, such as Pt (3.50 versus 5.84 eV), is an important factor that influences (18, 19) its response to CO adsorption, as we will show here. Because of its high cohesive energy, Pt(111) is stable under CO at pressures of up to at least one atmosphere (20, 21), although stepped surfaces restructure under CO due to the increased binding strength of CO at step edges (22). For Cu(111), the densest and lowest energy surface, the adsorption of CO (weaker in energy than on Pt by a factor of ~1.7) causes a much larger restructuring in the form of metal clusters formed by detachment of atoms from the steps. These clusters are mobile and adopt particularly stable sizes and shapes, including 3 and 19 atoms. With the help of density functional theory (DFT) calculations, we explain the findings as resulting from the increased adsorption energy of CO at low-coordinated Cu atoms, together with the lowering of the binding energy of the metal atoms bound to CO [experimental details of sample preparation, HPSTM imaging, and methodology used in DFT calculations are explained in detail in the supplementary materials (23)]. The cluster-covered surface is stable at room temperature (in the scale of hours) after desorption of CO when the gas phase is evacuated, and this surface is extremely active in the dissociation of water, an important step in the WGS reaction.

An image of the clean Cu(111) surface under UHV (Fig. 1A) shows micrometer-scale terraces, atomic steps, and a few screw dislocations are visible. After introduction of 0.1 Torr of CO in the chamber, a new structure was observed along the step edges, while the rest of the terrace remained atomically flat (Fig. 1B). At 0.2 Torr, the terraces became covered with nanoclusters (Fig. 1C) that increased in density with CO pressure until the clusters filled the surface (Fig. 1D and fig. S3) (23). The CO coverage, evaluated from APXPS measurements in a different chamber under identical conditions, increased from 0.06 ML at 0.1 Torr to 0.09 ML at 0.2 Torr and to 0.16 ML at 0.5 Torr of CO (24).

Fig. 1 STM images of Cu(111) showing clusters filling the terraces as a function of ambient CO pressure.

(A) In UHV [Vb = –2.4 V; It = 0.1 nA]. (Bottom inset) Atomically resolved image [Vb = 0.2 V; It = 1 nA]. (B) Under 0.1 Torr of CO clusters form at step edges [Vb = –2.5 V; It = 0.2 nA]. (C) Under 0.2 Torr of CO clusters form on the terraces [Vb = 1.85 V; It = 0.5 nA]. CO coverages in (B) and (C) are shown in the insets, as determined from the APXPS peak intensities. (D) Under 10 Torr of CO [Vb = 0.15 V; It = 0.8 nA]. A high density of clusters with adsorbed CO molecules (expanded in the inset) completely covers the surface. The bright spots, due to CO on top sites, form (2×2)-3CO and c(4×2) unit cells. Scale bars are 25 nm for (A) and (C), 2 nm for (B), and 3 nm for (D).

The structure of the clusters formed at 0.2 Torr of CO is shown in Fig. 2A. A roughly bimodal size distribution is apparent, with small clusters ~0.5 nm in diameter with poorly resolved triangular shape and larger hexagonal shape clusters ~1.5 nm in diameter. We associate the first with three Cu atom clusters with an apparent height about half that of a monatomic step, similar to the case in UHV studies (25). The larger ones we assign to 19 atom clusters forming hexagonal closed-shell structures (typically with an apparent height corresponding to a monatomic step). The 19-atom closed-shell structures (“magic number” clusters) are reported to be the building blocks for the homoepitaxial Cu growth on Cu(111) (26). Sometimes the clusters contain a few more atoms (fig. S1) (23) or come in contact with each other, forming aggregates that separate again later as a result of thermal mobility, which causes the clusters to evolve spatially in the scale of minutes, as illustrated in fig. S2 (23). Time-lapse images show the clusters forming by splitting from step edges and growing by coalescence and accretion of smaller clusters. At 1 Torr, the clusters increased further in size and density (fig. S3) (23). These clusters are not aggregates of CO molecules, because their height is close to that of the steps, whereas CO produces a contrast of only a fraction of an angstrom (27).

Fig. 2 STM images, ball models, and simulated contrast images of Cu clusters at 0.2 Torr with hexagonal and trigonal symmetries.

(A) STM image [bias voltage (Vb) = –1.5 V; tunneling current (It) = 0.2 nA]. (B) and (C) are images of the two types of clusters with C6 or C3 symmetry colored in (A). The bright center of the cluster in (B) is due to a CO molecule and to a cluster of three Cu atoms decorated by CO in (C). (D) A 19-atom cluster model from DFT calculations and simulated STM image (Vb = –2.7 V) of the cluster in (B). (E) DFT-optimized model and simulated STM image (Vb = –2.7 V) of the cluster in (C). Images are simulated with a CO-terminated tip and using a higher bias (in absolute values) than the experimental value because of the +U correction, used to match the experimental CO-binding energy on the flat surface.

High-resolution images (Fig. 2, B and C) showed that the perimeters of the clusters contain bright maxima arranged in hexagonal (C6) or trigonal (C3) symmetries, which we attribute to CO molecules. The presence of CO bound to Cu edge atoms can be rationalized by the energy gain obtained by CO adsorption on low-coordinated Cu atoms. As discussed below, DFT predicts that 12 CO molecules, one for each atom in the periphery of the 19 atom cluster, are necessary for energetic stability, and only the molecules bound to corner Cu atoms appear bright. The three-atom clusters require three CO molecules for stability. From the known CO coverage (from APXPS) and from simple cluster counting in the STM images, we conclude that all the adsorbed CO molecules are bound to the cluster edges, leaving the rest of the surface with a negligible CO coverage below 0.01 ML.

At pressures between 10 and 100 Torr, the surface was completely covered with clusters that were larger and closer to each other, making estimation of their individual sizes difficult because of finite tip-size effects. Figure 1D shows an example of the topography of the surface under 10 Torr of CO, with clusters densely covering adjacent terraces separated by monatomic steps. Unlike the case for pressure below a few Torr, the clusters are now completely covered by CO molecules, imaged as bright spots separated by distances of Embedded Image and 2 times the Cu atomic periodicity and aligned in directions forming 60° and 90° between them. A similar surface was also observed at 100 Torr of CO (fig. S4). We interpret the observed STM contrast as arising from atop site CO molecules in local (2×2)-3CO and c(4×2) geometries with coverages of 0.75 and 0.5 ML, respectively. In mixed top and bridge or hollow CO sites, the STM contrast is large only for the top sites (21, 28). The (4×4) superstructure reported at cryogenic temperatures (29) was not observed here.

The CO-promoted formation of metal clusters on Cu(111) contrasts with the case of Pt(111), where no clustering is observed, and with the stepped Pt(332) and Pt(557) surfaces, where clusters form and entirely occupy the terraces (22). The CO adsorption energy on Pt is > 1 eV, but on Cu(111), this energy is only ~0.5 eV. However, Cu has a much lower cohesive energy of 3.50 eV compared with the 5.84 eV of Pt (18, 19). The low cohesive energy of Cu has many manifestations, such as the frizzled appearance of the steps of the clean surface at room temperature caused by kink atom diffusion (30), which was not observed on Pt.

On the Cu(111) surface, we calculated the formation energy of a Cu adatom by detachment from kink sites as 0.83 eV, indicating that on a clean surface the formation of clusters is energetically unfavorable. However, the adsorption of CO on a kink site reduces the detachment energy of CO+Cu molecule-adatom couples to 0.63 eV because of the difference in CO adsorption energy on a kink site and on a Cu adatom, which we calculated as –0.77 eV and –0.96 eV, respectively. The mobility of the Cu adatoms on (111) terraces can be predicted from the calculated potential energy surface that shows a diffusion barrier of 0.14 eV (fig. S5) (23). This barrier decreases to 0.10 eV for the Cu+CO couple. The lowering of the binding energy of metal atoms by adsorbed CO, leading to their detachment from the steps and their higher mobility on the terrace, is known as the “harpooning” effect (31). The higher density and diffusion rate of Cu+CO couples on the surface is the reason for their coalescence into clusters.

A detailed look at the Cu clusters, colored in Fig. 2A, shows that two types of CO decoration motifs exist, with C6 and C3 symmetry (Fig. 2, B and C). To explain the formation, stability, and structure of these clusters, we start by placing six CO molecules on the six corner Cu atoms of a 19-atom hexagon, plus one additional CO on the top site of the center atom. The DFT calculation for this cluster provides a CO adsorption energy of 0.83 eV on the low-coordinated corner sites, much greater than the 0.47-eV adsorption energy on the flat terrace. The adsorption energy of CO in the cluster’s center is 0.47 eV, similar to the flat surface. The total energy gain from the six corner CO molecules (2.16 eV) is not enough to overcome the formation energy of a 19-atom cluster, which we calculated to be 3.61 eV (table S2) (23). Adsorption of CO molecules to each Cu periphery atom (i.e., including those in the middle of each side), however, results in an energy gain averaging 0.82 eV per CO. Hence, the formation of the 19-atom cluster with 13 CO molecules is –0.59 eV (Fig. 2B), meaning that it is energetically favorable (table S2) (23). The CO molecules on this cluster have different adsorption geometries with tilt angles of 0°, 26°, and 37° with respect to the surface normal for the central, edge, and corner CO molecules, respectively.

The observation of only six bright spots at the periphery (plus the central spot) is related to the electronic structure and tunneling probability of the different CO molecules. We illustrate this by calculating the tunneling current probability using the standard Bardeen approximation [equations 2 and 3 in (23)] and the calculated partial density of states (DOS) for CO molecules on the cluster and on the tip (fig. S6) (23). The calculation reveals that the CO molecules on the corners indeed have greater tunneling contributions than the CO molecules on the edges, qualitatively explaining the experimentally observed contrast of the STM images with the six bright spots plus one in the center, as shown in Fig. 2, B and D. We could explain the threefold symmetry of some of the 19 atom clusters by adding three Cu atoms at the center of the 19 atoms. These three low-coordinated Cu atoms, producing the bright center of the cluster images, can bind three additional CO molecules and distort the tilt angles of the peripheral CO molecules, as shown in Fig. 2, C and E [(details are shown in (23)].

Finally, we investigated the effect of clustering on surface reactivity for the WGS reaction (i.e., CO+H2O↔CO2+H2), which Cu catalyzes. Water does not adsorb on the Cu(111) surface at room temperature (Fig. 4B) (32), whereas it dissociatively adsorbs on the more active Cu(110) surface (32). Once the gas phase CO at 1 Torr was pumped away, the STM images revealed that the Cu clusters were still present, although atomic resolution could not be achieved, likely because of the absence of CO molecules adsorbed on the tip in high vacuum (Fig. 3A). In the presence of 2 × 10−9 Torr of H2O, the cluster-covered surface was very active in dissociating water, as shown by the increasing oxygen peak in both the Auger electron spectra (AES) shown in Fig. 3B, and in the XPS spectra shown in Fig. 3C. The APXPS spectrum indicates that the O peak is a result of the dissociative adsorption of H2O (Fig. 4A) and that no such peak appears after experiments at 0.1 Torr of CO because clustering of the Cu did not occur at lower CO pressures (Figs. 1B and 3B). A similar effect was also observed during exposure to CO+H2O mixtures, as shown in Fig. 4A. The pristine Cu(111) surface, on the other hand, not pre-exposed to CO, is inactive (Fig. 4B).

Fig. 3 STM images and AES spectra taken after evacuation of the CO gas phase.

(A) STM images show that the Cu clusters remain on the surface in the presence of 2 × 10−9 Torr of water 0.5 hours after evacuation [Vb = 0.5 V; It = 0.5 nA]. Expanded image in inset [Vb = 1.5 V; It = 0.2 nA]. (B) AES after evacuation of the gas-phase CO from different initial pressures. The dissociative adsorption of H2O on the cluster-covered surface is evident from the increasing intensity of the O peak. (C) Similar observation using XPS in the Synchrotron chamber.

Fig. 4 APXPS experiments of H2O adsorption on Cu(111), with and without pre-adsorption of CO.

(A) H2O adsorption after CO-induced morphological changes. From bottom: under 0.5 Torr of CO; after pump down to 2 × 10−9 Torr, mostly containing water [mass spectra shown in fig. S8 (23)]; under 0.05 Torr of water; and (top) under 0.4 Torr of a 3:1 gas mixture of CO and H2O. Water dissociation is activated by the large increase of low-coordinated sites produced by the CO-induced nanocluster formation. In the presence of both water and CO (top), only OH and H2O species are observed because of the efficient reaction of CO with atomic O (24). (B) APXPS of a pristine Cu(111) (i.e., not exposed to CO) in the presence of 0.1 Torr of H2O. The weak adsorption peaks of H2O and OH [compare with top two spectra in (A)] probably arise from adsorption on the defect sites.

Our findings open the possibility that other soft materials (e.g., Ag, Au, and Zn) can similarly undergo large reconstructions at sufficiently high pressures of CO (or other molecules). We have also demonstrated that the inactive (111) face of Cu for water dissociation, a key step in the water-gas shift reaction, becomes highly activated as a result of the CO-induced clustering. The need for this type of study to extend our understanding of the working of catalysts under operating conditions is clear.

Supplementary Materials

Materials and Methods

Supplementary Text

Figs. S1 to S8

Tables S1 to S3

References (3341)

References and Notes

  1. See supplementary materials on Science Online
Acknowledgments: This work was supported by the Office of Basic Energy Sciences (BES), Division of Materials Sciences and Engineering, of the U.S. Department of Energy (DOE) under contract no. DE-AC02-05CH11231, through the Chemical and Mechanical Properties of Surfaces, Interfaces and Nanostructures program (FWP KC3101). D.Z and L.-W.W. were supported by the Organic/Inorganic Nanocomposite Materials program (FWP KC3104). It used resources of the National Energy Research Scientific Computing Center and the Advanced Light Source, which are supported by the Office of Science of the U.S. DOE. The computation used resources from the Oak Ridge Leadership Computing Facility (OLCF), with time allocated by the Innovative and Novel Computational Impact on Theory and Experiment (INCITE) project.
View Abstract

Navigate This Article