Roughening of Copper (100) at Elevated CO Pressure: Cu Adatom and Cluster Formation Enable CO Dissociation.

Carbon monoxide participates in many copper-catalyzed reactions, which makes CO-induced structural changes of Cu catalysts key for important industrial processes. We have studied the interaction of carbon monoxide with the Cu(100) single crystal termination at 120, 200, and 300 K by means of low-energy electron diffraction (LEED), temperature-programmed desorption (TPD), X-ray photoelectron spectroscopy (XPS), polarization-modulation infrared reflection absorption spectroscopy (PM-IRAS), and density functional theory (DFT) calculations. The absorption band of CO (2082-2112 cm-1) at elevated gas pressure (up to 5 mbar) and at 200/300 K was found at a higher wavenumber than the characteristic band of the c(2 × 2)CO structure and was consistent with CO adsorbed on low-coordinated Cu atoms. The combined PM-IRAS/DFT analysis revealed that exposure to CO induced surface roughening through the formation of Cu adatoms and clusters on the (100) terraces. The roughened surface seemed surprisingly active for CO dissociation, which indicates its unique catalytic properties.

Introduction

Both carbon monoxide and copper are key players in the industrial synthesis of methanol. The latter is produced from a mixture of CO, CO2, and H2, typically over a (nominal) Cu/ZnO/Al2O3 catalyst at 510–530 K and 50–100 bar, in a process developed by ICI in 1966.[1] It is commonly agreed on that CO2 is the main carbon source of methanol[2] even though methanol synthesis from CO on Cu/ZnO is also feasible.[3] Moreover, density functional theory (DFT) suggested that the reaction may proceed through a CO intermediate on Cu/CeO2 catalysts.[4] Cu is known to be mainly responsible for the catalytic activity; however, a fine-tuning by potential Zn-alloying and by the ZnO support is crucial.[5−7] The exact role of CO in methanol synthesis is still debated, and particularly, its interaction with Cu at elevated pressure deserves further attention. CO interaction with Cu is also important for the electrochemical CO2 reduction to CO and further reduction to valuable C2 products catalyzed by Cu.[8−11] A promising approach toward these and other complex industrially important processes is via well-defined model studies.[12−15]

The adsorption of CO on the (100) single crystal surface of Cu in UHV at cryogenic temperature has been well studied by various experimental techniques, such as infrared reflection absorption spectroscopy (IRAS),[16−19] photoemission spectroscopy (PES),[20−23] low-energy electron diffraction (LEED),[24−26] pump–probe,[27−29] and high-resolution electron energy loss spectroscopy (HREELS).[26] However, there are discrepancies among the few experimental studies of CO adsorption at room (and higher) temperature and higher gas pressure, which may originate from significant variations in experimental conditions. For example, Taylor and Pritchard[30] reported a single IR-absorption band at 2071 cm–1 assigned to CO adsorbed on top of Cu atoms, upon exposing Cu(100) at 300 K to 0.44 mbar CO. Quite differently, Truong et al.[31] reported one prominent band at 2086 cm–1 and a shoulder between 2095 and 2110 cm–1, when exposing Cu(100) at 265 K to 0.57 mbar CO. The first feature was assigned to CO adsorbed on top sites, whereas the second one was ascribed to CO adsorbed at stepped sites.

In these early studies, the adsorbate-induced reconstruction of the Cu substrate[32] was not considered, despite being known for other metals. For example, CO induces restructuring of stepped platinum surfaces[33] and supported metal nanoparticles, as shown by electron microscopy,[34] infrared spectroscopy,[35] or scanning tunneling microscopy (STM).[36] Very recent high-pressure (HP-) STM and near ambient pressure X-ray photoelectron spectroscopy (NAP-XPS) investigations of Cu(111),[37] Cu(110),[38] and Cu(100)[39] single crystals by Salmeron and co-workers revealed reconstruction of Cu surfaces upon exposure to mbar pressure of CO at 300 K. On all single crystal surfaces the formation of Cu clusters was observed with, e.g., 3, 5, or 19 atoms, as well as of three atom wide (elongated) clusters. For Cu(110), IRAS experiments were performed at 0.02 Torr, with adsorbed CO detected at 2099 cm–1 (CO adsorbed at the end atom (coordination number = 6) of short (few nm) linear Cu clusters), and at 2084 cm–1 (CO adsorbed in the middle (CN = 7) of linear clusters).[38]

Herein, we present a combined LEED/temperature-programmed desorption (TPD)/XPS/PM-IRAS/DFT study of CO/Cu(100), including both in situ and ex situ studies of the (ongoing) surface roughening of CO at room temperature and near-atmospheric pressure, employing a dedicated UHV–high-pressure cell combination.[40,41] Special caution was taken to guarantee a clean sample environment (gold-coated cell) and purified gases. Apart from Cu cluster formation, which may be important for methanol synthesis by maintaining a rough active surface and counteracting sintering, the modified surface also exhibited unexpected activity for CO dissociation.

Experimental and Computational Details

All experiments were performed in a custom-built UHV chamber, which has been previously described in detail.[40,41] It has been successfully used for various investigations such as CO adsorption/hydrogenation and methanol decomposition/oxidation on (sputtered) Pd(111) and supported Pd nanoparticles,[42−45] methanol steam re-forming on bimetallic PdZn model catalyst,[46−50] and CO adsorption on Ir.[51] In brief, the chamber comprises two parts, a classical UHV chamber and a UHV-compatible high-pressure cell (“Rupprechter-design”[40,41]). The upper UHV section allows sample preparation and characterization by standard surface science techniques such as ion bombardment and annealing, XPS, LEED, and TPD. Afterward, the sample can be transferred under UHV to the lower section, a gold-coated spectroscopic high-pressure cell, in which polarization-modulation IRAS (PM-IRAS) experiments can be performed in a pressure range from UHV to 1 bar. All mbar gas pressure exposures herein were performed in the well-baked PM-IRAS high-pressure cell. The gold-coated walls counteract the adsorption of potential contaminations that could be subsequently replaced from the walls. Room temperature UHV transfer of a Cu(100) sample to the high-pressure cell, its exposure to UHV for 1 h, and transfer back to the UHV section did not affect the TPD spectrum of CO, with only minute amounts of “transfer carbon” detected in the XPS spectrum of the sample (see below).

For the current study, a disk-shaped (8 mm diameter, 2 mm thickness) 5N-purity copper (100) single crystal (roughness and orientation accuracies better than 30 nm and 1°, respectively) was purchased from MaTeck. It was mounted via a tantalum wire cage on two molybdenum rods, which allowed resistive heating and cooling by liquid N2. The temperature was monitored by a chromel–alumel thermocouple spot-welded to the back of the sample. The Cu single crystal was cleaned by repeated cycles of Ar+ bombardment for 15 min (1 × 10–5 mbar, 1.5 kV, 3.6 × 10–6 A sputter current) and annealing to 800 K for 5 min. When required, the sample was annealed to 600 K in 1 × 10–7 mbar oxygen before sputtering, to remove carbon contamination. For other high-pressure cell designs, refer to refs (52−54).

Surface order and cleanliness of the sample were confirmed by LEED (SPECS ErLEED 150) and XPS, respectively. X-ray radiation from a nonmonochromated dual anode Mg/Al source (SPECS XR 50) was impinging on the sample at the angle of 55° with respect to the surface normal, and photoelectrons were detected at normal emission by a hemispherical energy analyzer (SPECS PHOIBOS 150). For the TPD experiments, the sample was heated with a linear temperature ramp of 1 K s–1 (Eurotherm 3216 PID controller). The desorption products were detected by a differentially pumped mass spectrometer (Pfeiffer PrismaPlus 220) through a nozzle closely facing the sample, in order to minimize the detection of species desorbing from the sample mount and to prevent readsorption of molecules. Research grade CO (99.997%) from an aluminum lecture bottle (Messer Austria) was admitted to the experimental chamber only after passing through a carbonyl purifying cartridge (Entegris GateKeeper 300 KF) and a cold trap filled with a liquid N2/ethanol mixture at 170 K. PM-IRAS (Bruker IFS 66v/S FTIR spectrometer, liquid N2-cooled HgCdTe detector, HINDS PEM-90 ZnSe photoelastic modulator) was used to measure the vibrational properties of adsorbed CO species. Using polarization-modulation enables differentiation between surface adsorbate and gas-phase contributions and acquisition of surface vibrational spectra from UHV to atmospheric pressure.[55−58] The PM-IRAS spectrum of the clean sample, prior to admitting CO to the spectroscopic cell, was used in all cases to normalize the spectra (not shown). PM-IRAS spectra were fitted with multiple Lorentzian lines, and XPS spectra were fitted with the CasaXPS software after linear-background subtraction. The peaks were fitted with mixed Gaussian/Lorentzian line shapes, and fwhm and peak positions were left unconstrained.

The VASP software was used to perform periodic calculations with the rPBE exchange correlation functional, which is particularly suitable for studies of adsorption on metal surfaces. The eigenstates of the valence electrons were calculated using plane-wave basis sets with a cutoff of 400 eV. First-order Methfessel–Paxton smearing of 0.1 eV was applied to the occupation numbers. The presence of core electrons was accounted for via the projector augmented wave technique. Monkhorst–Pack mesh of k-points (5 × 5 × 1) was used to sample the reciprocal space in the slab calculations. Geometry optimization was performed until the forces on all atoms were less than 0.2 eV nm–1. Vibrational frequencies were calculated in the harmonic approximation by displacing C and O atoms by ±3 pm in all three Cartesian directions, which yielded values within 3 cm–1 from those obtained using density functional perturbation theory. The surfaces were modeled using a 6 layer p(4 × 4) periodic slab hexagonal supercell of 1275 × 1275 pm2. The separation between adjacent slabs exceeded 1 nm.

The average adsorption free energies of the CO species were calculated as follows:In this equation, G[N × CO/substrate] is the energy of the substrate (i.e., surface, Cu1/Cu(100), Cu5/Cu(100), kinked surface) with N of CO molecule(s) adsorbed, G[substrate] is the energy of the substrate, and G[COgas] is the energy of the gas-phase CO. For the Cu5/Cu(100), on which 1–5 CO molecules are adsorbed, the differential adsorption energies were calculated as follows:

Finally, the formation free energies of Cu (x = 1, ..., 5) clusters were calculated as follows:Here, G[N × CO/Cu/Cu(100)] is the free energy of the substrate with N adsorbed CO molecule(s), G[Cu(100)] is the energy of the bare surface, G[Cu] is the energy of the single Cu atom in Cu bulk, and G[COgas] is the free energy of the gas-phase CO. The free energy calculations were done at T = 200 K and included zero-point energy corrections. In particular, we used the ideal gas approximation and p(CO) = 0.1 mbar to evaluate the free energy of gas-phase CO. With these definitions, negative adsorption free energies and cluster formation free energies correspond to exothermic processes.

Results and Discussion

CO Adsorption on Cu(100) under UHV Conditions

Figure displays LEED results. After cleaning the Cu single crystal prior to each experiment, a LEED pattern characteristic of the (100) surface was observed (Figure a), and no contaminants were detected by XPS (i.e., in the O 1s and C 1s binding energy (BE) regions, Figure S1). Thereafter, the low-temperature adsorption of CO on Cu(100) was addressed, and several benchmark experiments were performed by LEED, XPS, TPD, and PM-IRAS. In all these experiments, the adsorbate layer was prepared by cooling the clean copper single crystal from 300 to 120 K in 1 × 10–7 mbar CO. The cooling took approximately 9 min; therefore, the sample was exposed to ∼40 L (1 L = 1 × 10–6 Torr s) of carbon monoxide. After evacuation, a sharp c(2 × 2) LEED pattern, corresponding to 0.5 ML CO coverage, was observed (Figure b).

Figure 1

LEED patterns of (a) the clean and ordered as-prepared Cu(100) surface at room temperature, (b) the c(2 × 2)CO overlayer at 120 K after ∼40 L CO exposure, and (c) the Cu surface disordered after 40 min exposure to 1 mbar CO at room temperature. The beam energy was 157 eV.

The c(2 × 2)CO overlayer formed on the Cu surface was also examined by XPS, and the spectra for O 1s and C 1s binding energy regions are shown in Figure a. The left side of Figure displays the O 1s difference spectra (between adsorbate-covered and bare surfaces) to exclude contributions of the high-kinetic-energy tail of the Cu LMM Auger triplet. In both O 1s and (on the right side) C 1s spectra, along with the two main (“adiabatic”) peaks at 533.2 and 286.2 eV, respectively, also satellite features appeared at higher BE upon CO adsorption (Figure a). Such “giant satellites” are well-known to be characteristic of weakly chemisorbed CO.[20−23,59] Taking advantage of the known CO coverage of 0.5 ML, confirmed by LEED, we have used the XPS peak area for quantitative calibration, required to determine the coverage of oxygen and carbon species in subsequent XPS experiments.

Figure 2

XPS spectra of the O 1s and C 1s binding energy regions of (a) the c(2 × 2)CO overlayer at 120 K after ∼40 L CO exposure; (b) the clean Cu(100) surface after UHV transfer to the HP cell (keeping the sample there for 60 min), and UHV transfer back; (c) the clean Cu(100) surface after UHV transfer to the HP cell (keeping the sample there in 1 mbar Ar for 40 min), and UHV transfer back; (d) the same after exposure to 1 mbar CO at 300 K for 40 min and evacuation; and (e) after exposing the surface of (d) to 8 × 10–7 mbar CO at 510 K for 10 min. All spectra acquired in UHV: (a) at 120 K and (b–e) at 300 K.

TPD results are collected in Figure . Before examining the c(2 × 2)CO overlayer on the Cu surface by TPD, we have acquired a background TPD in order to determine the amount of potential contaminants adsorbing from the residual gas (base pressure 2 × 10–10 mbar), which appeared to be negligible ∼0.01 ML traces of water and CO (see Supporting Information Figure S2). As shown in Figure a, CO of the c(2 × 2) overlayer has a peak maximum at 170 K. Desorption below 140 K is due to CO desorbing from the Ta heating wires, as observed previously using the same experimental setup.[60] On the basis of the Redhead approximation[61] and assuming a desorption rate prefactor of 1 × 1013 s–1, the activation energy of desorption is estimated to be 44 kJ mol–1 for the main peak in Figure a, in good agreement with the value reported for CO on Cu(100).[62] We further observed that the adsorption was fully reversible, so that the TPD spectrum of the c(2 × 2)CO structure could be repeated several times (see Supporting Information Figure S2). We also performed a TPD experiment after preparing a CO adsorbate layer on the sputtered, defect-rich, Cu(100) surface. As shown in Figure b, an additional desorption feature appeared at 200 K, due to CO desorbing from defect sites created by ion bombardment. An activation energy of desorption of 53 kJ mol–1 was calculated for the defect sites. Note that also for the defect-rich surface all CO desorbed below 220 K.

Figure 3

All TPD spectra were taken after cooling the Cu(100) sample from 300 to 120 K in 1 × 10–7 mbar CO, i.e., ∼40 L CO exposure: (a) CO desorption from the c(2 × 2)CO overlayer prepared on the well-ordered Cu(100) surface and (b) on the ion-bombarded defective Cu surface. (c) TPD experiments performed after transferring the clean sample in UHV to the HP cell (keeping the sample there in UHV or 1 mbar Ar at 300 K) and UHV transfer back. Spectrum d was taken after Cu(100) exposure to 1 mbar CO at 300 K for 40 min with CO and CO2 signals enlarged 10 times. The TPD spectrum in part e was acquired after exposing the surface after treatment described in part d to 8 × 10–7 mbar CO at 510 K for 10 min.

The low-temperature CO adsorption on Cu(100) at 1 × 10–6 mbar CO pressure was also investigated by PM-IRAS (Figure ), and the band at 2082 cm–1 at 120 K was attributed to the c(2 × 2)CO overlayer. When the sample was heated to more than 140 K, the CO coverage decreased and the combined effect of reduced dipole–dipole and chemical interaction red-shifted the band, as previously discussed[18,19,63] (see Figure a). At 160 K, the band position at ∼2072 cm–1 was in line with the data by Ryberg.[19] Heating to higher temperature (180 K) led to a further red-shift to 2060 cm–1. Above 170 K (the TPD desorption maximum in Figure a), the peak became very small.

Figure 4

PM-IRAS results of (a) the as-prepared well-ordered Cu(100) surface and (b) the sputtered Cu surface, cooled from 300 to 120 K in 1 × 10–6 mbar CO.

In turn, PM-IRAS spectra of CO adsorbed at low temperature on sputtered Cu(100) showed a (weaker) band at 2082 cm–1, assigned to CO adsorbed on well-ordered terraces, and a shoulder at 2092 cm–1 (Figure b).

The latter higher-frequency PM-IRAS peak, apparently related to the sputtering, was conclusively assigned on the basis of DFT simulations (Figure and Table ). The same holds for CO adsorbed on various sites and Cu clusters. The selection of the sites was based on previous STM observations of surface roughening by the Salmeron group.[39] The absolute values of experimental and calculated CO vibrational wavenumbers, ν̃, differ due to unaccounted anharmonic contributions to vibrational frequencies in the simulations and inherent approximations of the RPBE exchange-correlation functional. In order to facilitate comparison of experimental values and computational results, we consider the difference between CO vibrational frequencies on a given site X and on ordered Cu(100), Δν̃[X] = ν̃[X] – ν̃[Cu(100)]. For example, the PM-IRAS spectrum of CO adsorbed on sputtered Cu(100) exhibits a shoulder at Δν̃ = 2092–2082 cm–1 = 10 cm–1, which compares well to simulation of CO adsorbed on step edges, Δν̃ = 2019–2011 cm–1 = 8 cm–1. The difference between calculated CO adsorption energies on step and terrace sites, 41–30 = 11 kJ mol–1, also agrees well with the respective difference obtained from TPD analysis, 53–44 = 9 kJ mol–1. The adsorption of CO on low-coordinated Cu sites is also in line with adsorption studies on polycrystalline Cu films (∼2105 cm–1 or 2090–2093 cm–1).[30,64−68]

Figure 5

Calculated structures of CO adsorbed on various Cu sites: (a) (100) terrace, (b) step, (c) kink atom, (d) Cu1/Cu(100), and (e) Cu5/Cu(100).

Table 1

Calculated Free Energies (G, kJ/mol) and Vibrational Frequencies (ν̃, cm–1) for CO Adsorbed on Various Sites of Cu(100) Surfaces

site	Gads	Gforma	ν̃	Δν̃DFTb	Δν̃expc
Cu(100)	–30		2011	0	0
step	–41		2019	8	10
kink	–44		2034	23	n.a.
Cu1/Cu(100)	–56	4	2046	35	30
Cu5/Cu(100)d	–44	–118	3 × 2030,e 2046, 1973	3 × 19,e 35, −38	19

Gform values of Cu1/Cu(100) and Cu5/Cu(100) calculated in the absence of CO are 60 and 160 kJ mol–1, respectively.

The calculated difference between CO vibrational frequency on the given site and on Cu(100).

The experimental difference between CO vibrational frequency on the given site and on Cu(100). Experimental assignments based on literature and the current DFT results.

Gads and Gform are given for Cu5 cluster with 5 adsorbed molecules. The stability of this structure is confirmed by significant differential adsorption energy of the 5th CO molecule, Gadsdiff = −36 kJ mol–1 (Table S1).

Three calculated frequencies at 2032, 2030, and 2029 cm–1 are combined in one peak (see Table S2 for further details).

Adsorption of CO on Cu(100) at mbar Pressure

After collecting UHV benchmark data at low temperature, CO adsorption at elevated temperature and at near atmospheric pressure was addressed (Figure ). Because of the well-known issue of nickel or iron contamination when working with CO in this pressure regime (originating from the gas storage/manifold hardware[56]), special cleaning precautions were made (see the Experimental and Computational Details section). Indications of a contaminated Cu surface would include unusual low-wavenumber bands for adsorbed CO in IRAS (from 2020 to 2060 cm–1),[69] CO desorption features at about 360 K in TPD, and detection of the Fe or Ni 2p doublet in postexposure XPS analysis.[2,30] None of this was observed, ruling out Fe or Ni contamination.

Figure 6

(a) PM-IRAS spectra on “Cu(100)” recorded in 0.1 mbar CO at 200, 225, and 300 K. The features at 2082 and 2112 cm–1 are assigned to CO adsorbed on terraces and CO/Cu1 complexes, respectively. (b) Time evolution of PM-IRAS spectra recorded in 0.1 mbar CO at 300 K. The features at 2093 and 2103 cm–1 are assigned to CO adsorbed at Cu clusters and adatoms, respectively.

Two bands at 2082 and 2112 cm–1 were measured by PM-IRAS in 0.1 mbar CO at 200 K (Figure a). The first is due to the c(2 × 2)CO overlayer on Cu(100) terraces. On the basis of DFT results (Figure and Table ), the second band is attributed to CO adsorbed on Cu1 adatoms on the Cu(100) surface, whose calculated Δν̃DFT = 35 cm–1 agrees well with the experimental blue-shift of Δν̃exp = 30 cm–1. The formation of such CO/Cu1 is only slightly endothermic, Gform= 4 kJ mol–1, which suggests that such complexes may form through the detachment of Cu1 atoms from kink sites upon CO adsorption, and then they may migrate over the surface (Figure ). Such detachment would be favored by higher CO pressure (which explains the absence of this band in spectra in Figure ) and higher temperature (which explains the higher band intensity at 225 and 300 K in Figure a). Previously, these bands were assigned to CO adsorption on terraces and defects, but no rationalization of the variation of their relative intensity upon increasing the CO pressure from 1.3 × 10–5 to 0.57 mbar at 265 K was provided.[31]

Figure 7

Schematic CO-induced surface roughening: (a) CO on a kink atom with coordination number CN = 6, (b) on an adatom Cu1 with CN = 4, and (c) on top of a Cu5 cluster atom with CN = 5. According to the proposed mechanism for cluster formation, a CO molecule adsorbed on a kink Cu atom lowers its binding energy (a) and forms a Cu1CO complex, which is mobile on the surface (b). The latter can then coalesce with other adatoms forming a Cu cluster (c).

At higher temperature, not only can more Cu1CO complexes detach from the steps, but their diffusion rate is also expected to increase with a possible coalescence into Cu clusters via Ostwald ripening.[70] When the temperature was increased to 300 K, the band of CO covering the Cu(100) terraces became barely noticeable and red-shifted to 2074 cm–1 (Figure a). The peak characteristic of CO/Cu1 was also red-shifted by almost the same magnitude to 2103 cm–1 at 300 K.

In addition, a new band appeared at 2093 cm–1, blue-shifted (Δν̃exp = 2093–2074) by 19 cm–1. This scenario is in reasonable agreement with CO adsorbed on Cu5 clusters formed by coalescence of Cu1 adatoms (Figure , Table , and Figure ). In particular, DFT predicted CO on Cu5 clusters giving three different peaks at 2046, 2030, and 1973 cm–1 accounting for one, three, and one CO molecule, respectively. Since the signal at 2030 cm–1 (Δν̃DFT = 19) corresponds to three vibrations (2032, 2030, 2029 cm–1), we expect this band to be the most evident in the IR spectrum, matching exactly with the experimental band (Δν̃exp = 19). The formation energy of such Cu5 clusters is calculated to change from highly endothermic, Gform = 110 kJ mol–1, at low CO pressure, to highly exothermic, Gform = −118 kJ mol–1, at high CO pressure, when they accommodate 5 adsorbed CO molecules (Table ). This assignment agrees with the HP-STM characterization of such clusters on Cu(100) (0.25 mbar CO pressure) by Salmeron and co-workers.[39] When exposing a Cu(110) surface to 0.03 mbar CO at room temperature,[38] an absorption band at 2099 cm–1 was observed, assigned to CO adsorbed on Cu “end” atoms (CN = 6) of short linear Cu clusters. Although a straightforward comparison is difficult for different single crystal surfaces, the CO peaks at 2093, 2099, and 2103 cm–1 are all in the “ballpark” of low-coordinated Cu sites.

The formation of clusters from Cu1 was followed by a time-dependent PM-IRAS experiment in 0.1 mbar CO at 300 K (Figure b). The peak at 2093 cm–1 assigned to Cu clusters increased, whereas the 2103 cm–1 peak assigned to CO/Cu1 decreased in intensity. This indicates that Cu clusters are formed by coalescence of Cu1.

The Cu(100) surface reconstruction was further confirmed by two PM-IRAS experiments at 1 and 5 mbar CO pressure, see Figure . At 1 mbar (Figure a), a main band was observed at 2095 cm–1 and was assigned to CO adsorbed on Cu clusters. The shoulder at 2102 cm–1 was again attributed to CO/Cu1 complexes. At this pressure, the Cu1 species disappeared faster and Cu clusters formed in a shorter time. A dynamical equilibrium between detachment of Cu1 adatoms from steps and Cu cluster formation explains why no strong time-dependent changes were observed. At 5 mbar (Figure b), only the band of CO adsorbed on Cu clusters was detected due to the faster coalescence of Cu1 adatoms. Finally, at this elevated pressure, an equilibrium state was reached, and the detachment of Cu1 atoms from step edges and their coalescence to Cu clusters led to the formation of a rough Cu surface.

Figure 8

Time evolution of PM-IRAS spectra of “Cu(100)” recorded in (a) 1 mbar CO and (b) 5 mbar CO at 300 K.

After exposure to 1 mbar CO at room temperature, the high-pressure spectroscopic cell was evacuated; the sample was transferred in UHV to the upper UHV chamber, and a thorough postreaction analysis was carried out by LEED, XPS, and TPD. Using LEED (Figure c), an ordered pattern could not be observed anymore, and the background was brighter than in a previous analysis (the LEED and camera settings were the same as those of Figure a,b), suggesting surface reconstruction (roughening).

Using XPS at 300 K and 3 × 10–10 mbar (Figure d), an oxygen species at 530.2 eV (coverage of 0.38 ML) and a carbon species at 284.5 eV (0.27 ML) were detected. Note that no O 1s and a much smaller C 1s signals were detected after exposing the sample at 300 K for 40 min to 1 mbar Ar instead of CO (Figure c).

On the basis of refs (37) and (71), the 530.2 eV species was assigned to adsorbed O. Adsorbed molecular CO can be excluded as origin because of the different binding energy and absence of satellite peaks (cf., Figure a; the same holds for C 1s discussed below). Note also that the XPS spectrum was acquired in UHV and that TPD had indicated that even on a roughened Cu surface CO desorbed around 200 K (Figure b). Water contamination can also be excluded since XPS was acquired in UHV at 300 K and the water O 1s peak would be around 531.5–532.5 eV,[72] very different from the observed 530.2 eV (adsorbed O) species.

The C 1s peak at 284.5 eV clearly does not originate from adsorbed CO but could, in principle, be due to adventitious carbon. However, as discussed above, the gold-coated walls and clean gases strongly limit that (cf., Figures b,c and 3b,c). Furthermore, exposing, e.g., Pd(111), which does not dissociate CO, to 1 mbar CO for longer times did not produce such a pronounced carbon peak (see, e.g., ref (73)).

The straightforward interpretation of the presence of oxygen and carbon is that CO dissociation, CO* → C* + O* (ref (74)), had occurred, which is not favorable on the pristine single crystalline surface (CO adsorption on Cu(100) in UHV is reversible). This would demonstrate the effect of high-pressure CO and surface roughening on the intrinsic catalytic properties of Cu. It is in line with studies of Cu nanoclusters on alumina[75] and of the stepped Cu(211) surface,[76] indicating that both were active toward CO dissociation.

To further examine the identity of the O 1s and C 1s species on the CO-induced roughened surface (after exposure to 1 mbar CO at 300 K for 40 min), the surface was cooled from 300 to 120 K in 1 × 10–7 mbar CO (∼40 L CO) and characterized by TPD. As shown in Figure d, hardly any CO had adsorbed and only a tiny amount of CO2 desorbed (CO at 450 K is a fragment of CO2). Although spectra were enlarged by a factor of 10, the amount of CO desorbing from the roughened Cu surface was negligible (∼5%) when compared to that desorbing from the as-prepared clean sample (Figure a). The surface remained O/C covered, as indicated by a follow-up TPD (Figure S3) showing very little CO adsorption and no CO2 desorption.

Even Cu sites free from O and C (∼0.35 ML) seemed not energetically favorable for CO adsorption because of neighboring O and C adatoms. In a following experiment, the Cu surface (still) covered by adsorbed O and adsorbed C was exposed to 8 × 10–7 mbar CO at 510 K for 10 min. The CO reacted with most of the adsorbed O, whereas the C 1s signal remained nearly the same (Figure e). After this treatment, part of the surface was free from adsorbed O and therefore able to adsorb CO again, as indicated by CO-TPD (Figure e).

Altogether, the control experiments suggested that CO dissociation had occurred. All detected species are summarized in Table . No changes in the oxidation state of (metallic) copper were observed throughout the experiments (Figure S4).

Table 2

Summary of Binding and Vibrational Energies of the Detected Speciesa

species	XPS O 1s BE/eV	XPS C 1s BE/eV	PM-IRAS/cm–1
Cu(100)-c(2 × 2)CO	538.3, 534.5, 533.2	293.2, 288.9, 286.2	2084
CO/Cu defect sites			2092–2112
carbon		284.5
oxygen	530.2–530.6

Species in italics are satellites.

Nevertheless, although we have used purified gases and carried out all high-pressure exposures in the gold-coated cell, mbar pressure experiments may still encounter problems with contaminations. One could argue that the blind test with Ar may not be conclusive, as mbars of CO may replace contaminations from the chamber walls more effectively than Ar does. However, as mentioned, exposing Pd(111), which does not dissociate CO, to 1 mbar CO for longer times did not produce such a pronounced carbon peak or any O 1s species,[73] supporting the current picture of CO dissociation on roughened Cu.

Summary and Conclusions

The room temperature adsorption of CO under mbar gas pressure caused reconstruction of the Cu(100) surface, which was characterized in situ and ex situ by LEED, XPS, TPD, PM-IRAS, and DFT.

Compared to the c(2 × 2)CO overlayer on smooth Cu(100) terraces, CO on step sites was found to adsorb more strongly by ΔE = 53–44 = 9 kJ mol–1 and was characterized by a vibrational band at a higher wavenumber Δν̃exp = 2092–2082 cm–1 = 10 cm–1 (at 1 × 10–7 mbar pressure). These findings are in line with DFT simulations, which yielded ΔE = 11 kJ mol–1 and Δν̃DFT = 8 cm–1 for these adsorption sites.

At 0.1 mbar CO pressure and 200–300 K, Cu1 adatoms on Cu(100) were detected through characteristic vibrational bands. The formation of Cu1 was calculated to be only slightly endothermic when CO is adsorbed, Gform = 4 kJ mol–1. Thus, the formation of CO/Cu1 complexes is affected by the CO pressure and temperature, as observed experimentally. The vibrational band of CO adsorbed on Cu1 adatoms was found to be Δν̃exp = 2112–2082 cm–1 = 30 cm–1 higher than on Cu(100) terraces, which is in line with the respective calculated value of Δν̃DFT = 35 cm–1.

At 300 K, Cu1 adatoms and Cu clusters coexisted in dynamic equilibrium, which was shifted toward clusters at 1 and 5 mbar CO pressure. DFT simulations of Cu5 clusters on Cu(100) showed that these clusters can accommodate 5 CO molecules and that their formation is highly exothermic, Gform = −118 kJ mol–1 under high CO pressure. CO adsorbed on Cu5 clusters was calculated to preferentially vibrate at Δν̃DFT = 19 cm–1, which favorably agrees with the experimentally observed band at 2093 cm–1 (Δν̃exp = 19 cm–1).

XPS and TPD analysis suggested the Cu(100) surface roughened by mbar CO exposure to be active for CO dissociation. After exposing Cu(100) to 1 mbar CO at room temperature, XPS revealed the presence of carbon, at BE 284.5 eV, and adsorbed oxygen, at BE 530.2 eV. The adsorbed oxygen could react with CO gas at 510 K, leaving C unaffected. Since all high-pressure exposures were performed with purified gases and in a gold-coated cell, contamination effects should be very limited.

The observed reconstruction of Cu surfaces induced by mbar CO exposure may have important implications on processes relying on copper-based catalysts and CO-containing reaction mixtures, e.g., by promoting the formation of low-coordinated active copper sites and/or increasing the surface density of the latter.