Adsorption Dynamics of CO on Silica Supported CuOx Clusters

State University, Fargo, North Dakota 58108, United States. J. Phys. Chem. C , 2012, 116 (9), pp 5792–5801. DOI: 10.1021/jp211500s. Publication ...
0 downloads 14 Views 3MB Size
Download PDF
Article pubs.acs.org/JPCC

Adsorption Dynamics of CO on Silica Supported CuOx Clusters: Utilizing Electron Beam Lithography To Study Methanol Synthesis Model Systems M. Komarneni, J. Shan, A. Chakradhar, E. Kadossov,† S. Cabrini,‡ and U. Burghaus* Department of Chemistry and Biochemistry, North Dakota State University, Fargo, North Dakota 58108, United States S Supporting Information *

ABSTRACT: Electron beam lithography was used to nanofabricate 12 and 63 nm Cu clusters supported on silica (model nanoarray catalysts). The Cu clusters could reversibly be oxidized and reduced at ultrahigh vacuum conditions. The chemical activity of these clusters was probed by Auger and X-ray photoelectron spectroscopy, thermal desorption spectroscopy, and molecular beam scattering. CO was used as the probe molecule. Scanning electron microscopy was used to obtain cluster size distributions. CO adsorption is molecular and nonactivated. CO binding energies on the oxidic clusters are larger than for the metallic clusters. Adsorption transients, recorded as a function of surface temperature and CO impact energy, are consistent with precursor models, as expected from the so-called capture zone model (CZM). Cluster size effects are evident, as predicted by the CZM. However, unexpectedly, the CO saturation coverage does not simply scale with the cluster area but depends also on the rim length of the deposits. Metallic Cu clusters are more reactive than oxidic clusters, in part not only due to the cluster size effect but apparently also because of the electronic effect.

1. INTRODUCTION 1.1. Motivation. The use of nanoparticles for promoting catalytic processes is a major thrust in both academia and industry. How the size, rim, shape, and the chemical composition of metal clusters affect their chemical/catalytic activity are among the major questions currently explored.1 EBL (electron beam lithography) allows one to nanofabricate nearly monodisperse cluster samples, forming a regular pattern on a substrate (so-called model nanoarray catalysts).2−6 EBL predetermines not only the cluster size and shape but also the rim length and height of the structures. Therefore, a timeconsuming characterization of the sample’s morphology may not be stringent; however, the samples studied here have been inspected by SEM (scanning electron microscopy). In addition, spectroscopic tools such as Auger electron spectroscopy (AES) and X-ray photoelectron spectroscopy (XPS) were used. Although the smallest size of the clusters which can be made covering a macroscopic surface is still technically limited (here we studied 12 and 63 nm Cu structures), cluster size effects can be studied in a unique way by utilizing EBL. Silica is conveniently used to support nanostructures. In addition, SiO2 supported catalysts are of some practical interest for methanol steam reforming7 and hydrogenation reactions.8 Blind experiments on clean silica can be found e.g. in refs 9 and 10 and are additionally included here. Furthermore, Cu PVD (physical vapor deposition) clusters on silica supports were studied before (see e.g. refs 11 and 12 and references therein); these PVD results will serve as a reference and are therefore included, too. Considering copper clusters, a scientifically interesting complication arises from the fact that different oxides may form, even at UHV (ultrahigh vacuum) conditions. Unfortunately, this may lead to the formation of © 2012 American Chemical Society

mixed cluster samples consisting of Cu, CuO, and Cu2O moieties. In this study, however, different oxides have been characterized both before and after each kinetics/dynamics experiment by means of XPS and AES peak analysis. Thus, the chemical state of the Cu clusters was well-documented, which allowed to compare directly the chemical activity of metalliclike and oxide-like Cu clusters. The electronic structure of clusters may affect kinetics and dynamics properties.13,14 In addition to gaining a better understanding of mechanistic aspects of gas−surface interactions with Cu clusters, many applications for Cu as a less expensive and readily available catalyst are well-known (e.g., methanol synthesis,15,16 water splitting catalysis under visible light irradiation,17 water-gas shift reaction,18 low-temperature CO oxidation19) as well as applications in materials science (e.g., surface plasmon resonance,20 solar cells,21 high-Tc superconductors22). Therefore, in this study CO was used as the adsorbate since it is relevant for a number of catalytic processes, and it is a standard surface science probe molecule. A related project about CO2 adsorption will be presented elsewhere. The gas-to-surface energy transfer processes (adsorption dynamics) were characterized using molecular beam scattering techniques. Utilizing beam scattering techniques is particularly useful for the study of small clusters with a large dispersion (small total coverage) since the beam can directly be focused onto the surface while keeping the background pressure low. 1.2. Very Brief Literature Survey. Because of the importance of the copper system for a variety of applications, Received: November 29, 2011 Revised: February 13, 2012 Published: February 15, 2012 5792

dx.doi.org/10.1021/jp211500s | J. Phys. Chem. C 2012, 116, 5792−5801

The Journal of Physical Chemistry C

Article

direction (for beam scattering experiments) and one aligned with the beam (for time-of-flight, TOF, beam characterization). In addition, a double pass cylindrical mirror analyzer equipped with an electron gun for Auger electron spectroscopy (AES) and an X-ray source for XPS (all from Perkin-Elmer including an upgrade from RBD) is mounted on the scattering chamber. Furthermore, a sputter gun, home-built metal evaporator, an electron beam metal evaporator, and an atomic hydrogen source are mounted on the scattering chamber. The impact energy, Ei, of the CO molecules could be varied within 0.09− 0.98 eV by using a pure beam and by seeding 3% CO in He, combined with a variation of the nozzle temperature within 300−750 K (see ref 41). The impact energies have been measured by TOF and are in perfect agreement with calculated values. The beam flux was determined by measuring the equivalent beam pressure using the mass spectrometer aligned with the beam and amounts to F = (1.8 ± 0.1) × 1013 CO molecules cm−2 s−1 for the seeded CO beam. The beam flux was kept constant (within 6%). Initial adsorption probabilities, S0, and coverage-dependent adsorption probabilities, S(Θ), were collected using the King and Wells type uptake technique.42 The molecular beam was directed perpendicular to the surface. For a single experiment, the uncertainties in S0 amount to ±0.05, as estimated from the signal-to-noise ratio of the transients. Averaging dropped the uncertainties in S0 to ±0.02. The sample temperature, Ts, could be reduced down to 90 K by bubbling He gas through a liquid nitrogen containing dewar on which the sample was mounted. The reading of the thermocouple was calibrated in situ within ±5 K by recording the condensation peaks of alkanes. The heating rate did amount to 1.7 K/s. For XPS, the Mg Kα line (at 1253.6 eV) was used with a pass energy of 100 eV (survey spectrum) and 50 eV (individual peaks) of the analyzer. The XPS spectra were referenced with respect to the Si 2p line at a binding energy of 103.4 eV.43 The electron energy for AES amounted typically to 2 keV with a modulation voltage of 2 eV. 2.2. Sample Fabrication. The nanofabrication procedures of the EBL (electron beam lithography) cluster samples at Molecular Foundry (Lawrence Berkeley National Laboratory) is detailed in ref 5. Cu cluster samples with ds = 12 nm and dl = 63 nm diameter clusters were made with a lattice constant of as = 100 nm and al = 150 nm, respectively. A h = 5 ± 2 nm thick Cu layer was used in the fabrication process and a 5 by 5 mm area on a 10 by 10 mm silica support was covered with the clusters. SEM images have been collected with a Zeiss Ultra 60 (equipped with a field emission gun). The FESEM has a nominal resolution of ∼1.2 nm with 20 keV for the electron beam acceleration voltage. 2.3. Sample Cleaning. Cleaning electron beam lithography (EBL) samples without destroying the morphology and/or changing irreversibly the chemical composition is a nontrivial task, as discussed e.g. in the pioneering works of Somorjai and Kasemo et al. on EBL samples.2,4 The 12 and 63 nm Cu/SiO2 model nanoarray catalysts studied here were initially cleaned by annealing in a flux of atomic hydrogen (p(H2) = 7 × 10−8 mbar, Ts = 480 K) using a commercial thermal hydrogen doser, following earlier procedures of our group.5,44 This removes carbon containing impurities as judged by AES. In some cases, sulfur impurities were present which could be removed by annealing in 2 × 10−5 mbar of molecular oxygen at Ts = 500 K.

the literature is overwhelming; i.e., only a few prior projects directly related to this study are briefly summarized. More and more specific references are given in the Results and Discussion section. The Cu cluster literature appears to be dominated by studies on various Cu nanoparticle powders, synthesized by a variety of “wet chemistry”/electrochemistry techniques.20,23−25 However, most of these copper nanoparticles are capped with passivating coatings or ligands. Typical supported powders were considered and characterized by XRD (X-ray diffraction), XPS, IR (infrared spectroscopy), and SEM. Most investigations on powder samples focus on surface reactions; however, kinetics/dynamics adsorption studies of small molecules on these systems (at UHV) are rare. The traditional Blyholder model for CO adsorption on metal surfaces has to be modified for metal oxides.26 The differences in the electronic structure lead to larger binding energies of CO also on CuOx oxides, as determined by TDS e.g. for Cu2O(100).27 CO adsorbs/desorbs molecularly on Cu metal oxide single crystals, and no CO2 formation was seen.27 However, CO2 formation has been detected for both silica supported Cu2O nanoparticles powder23 and copper single crystals.28 1.3. Capture Zone Model. Beam scattering experiments on model catalysts are reviewed e.g. in refs 29−31, and a few key beam scattering projects are cited as refs 32−35. Most experiments compiled in our study have been conducted at an adsorption temperature where the coverages of CO on the support is negligible. However, even in that case, the lifetime of the probe molecule on the surface can be sufficiently long enough, allowing diffusion of the molecules from the support to the deposits (similarly to inverse spillover effect). So-called “capture/collection zone models” (short CZM) have been applied by Boudart et al.36 and others (see e.g. refs 29 and 37) to account for these kinetic effects. Therefore, one has to consider at least the following adsorption/desorption pathways: (1) thermal desorption from the deposits and/or the support, (2) direct adsorption on the support and/or the metal deposits, (3) trapping of the adsorbates on the support and subsequent diffusion to the deposits (CZM), and (4) trapping of the adsorbates on the deposits and subsequent diffusion to the support (similarly to spillover effects). The third adsorption pathway generates a so-called capture (diffusion) zone surrounding the metal deposits. The size of the capture zone is given by the ratio of the lifetime of the adsorbates on the support (or in the precursor state) to the diffusion (hopping) time. Depending on the binding energies, the size of the capture zone is significant even at rather large temperatures.29,36,37 The adsorption dynamics of CO and CO2 was studied on Cu/ZnO and copper single crystal systems before (see e.g. refs 37−40). Surface chemistry studies on Cu (oxide) cluster systems, in particular with molecular beam scattering techniques,37 are, however, still rather rare. In addition, the concept of utilizing electron beam lithography (EBL) has still not been applied very often.2−6

2. EXPERIMENTAL PROCEDURES 2.1. Experimental Setup. The measurements have been conducted at NDSU by a home-built, triply differentially pumped molecular beam system (see ref 41 for details). The supersonic beam is attached to a scattering chamber which houses two mass filters: one perpendicular to the beam 5793

dx.doi.org/10.1021/jp211500s | J. Phys. Chem. C 2012, 116, 5792−5801

The Journal of Physical Chemistry C

Article

2.4. Reduction of Copper Oxide. Metallic and oxidic-like Cu clusters were studied. The reduction of Cu oxides to the metallic state was performed by annealing the CuOx/SiO2 sample in molecular hydrogen at 420 K (p(H2) = 5 × 10−5 mbar). The chemical state of copper was monitored by XPS. The absence of oxide satellite features indicates elemental copper, as discussed in more detail below. The Cu cluster samples were remarkably stable for weeks at UHV conditions judged by well reproducible XPS, AES, and beam scattering data. The samples have never been annealed above 500 K avoiding sintering of the clusters. Note that the Tammann temperature of Cu amounts to 678 K.45 Sintering effects on Au EBL samples have been studied in prior work of our group.46 In this case, beam scattering data were distinctly affected by sintering effects (which reduces S0) only for annealing temperatures above 600 K. These effects have not been observed here for the Cu clusters. 2.5. Blind Experiments. For the nanofabrication, silica wafers were used as supports. The cleaning procedures of the model nanoarray catalysts may also affect the chemical composition of the support which was determined by AES and XPS before and after each beam scattering experiment. In addition, blind experiments were collected on silica supports which underwent cleaning procedures identical to those employed for the EBL samples. Furthermore, fully oxidized and partially reduced silica supports were studied. In both cases, no significant uptake of CO was seen in beam scattering experiments at the lowest surface temperature. More details are described below, and some blind data are depicted together with the data on the Cu clusters.

3. RESULTS AND DISCUSSION 3.1. Sample Morphology. The sample morphology was studied by SEM. A typical example for 12 nm Cu clusters is shown in Figure 1A. The dots are separated by as = 100 nm forming a predetermined rectangular pattern. When enlarging one Cu dot in the SEM image (see inset of Figure 1), it appears just as a rather structureless spot. This most likely does not reflect the actual shape of the clusters, but results from the resolution and contrast limits of the SEM. Figure 1B,C depicts the size distribution of the clusters obtained by analyzing a 1.4 μm by 1 μm section (using the software tool Pixcavator). Accordingly, the most likely cluster size (diameter) for the small (s) and large (l) clusters amounts to ds = 11.9 ± 0.3 nm and dl = 63 ± 2 nm, respectively, with a lattice constant of as = 100 ± 1 nm and al = 150 ± 4 nm. Using EBL, the cluster size distribution is very narrow, e.g., 0.3/11.9 ≈ 2.5% using the fwhm of the Gaussian fit shown in Figure 1B as an error estimate. The calculated Cu coverage (Cu area vs support area) amounts to 0.011 ML (= 1/4πd2/a2) and 0.139 ML (assuming circular clusters) when going from small 12 nm to large 63 nm clusters, respectively. This corresponds to a factor of 12.6 ± 1.0. (The uncertainties in the theoretical predictions are related to the cluster size variations.) Thus, when comparing CO saturation coverages, Θsat, for these two cluster sizes, Θsat should increase by that factor if the adsorption dynamics is dominated by terrace sites. Relative Θsat can be determined using e.g. beam scattering data (see section 3.4b). However, the total rim length is given by ρAπd with A as the total area covered by the clusters and ρ for the cluster density (particles per area). Therefore, the ratio (large/small clusters) of the rim length amounts to dlρl/dsρs = 2.3 ± 0.4. Thus, if rim

Figure 1. (A) SEM image of the 12 nm Cu EBL (electron beam lithography) model array catalyst. (Inset shows an enlarged section of the image. The original contrast was modified.) Particle size distributions for (B) 12 nm and (C) 63 nm clusters. (The cluster diameter is a diameter equivalent to the area of circular clusters.)

effects dominate the dynamics, Θsat should increase only by a factor of 2.3 when increasing the clusters′ size from 12 to 63 nm. (We assume here that the rim length and surface area are proportional to the number of adsorption sites and that the rim is a one-dimensional (1D) structure. The former is trivial. The latter may be a good assumption considering the small aspect ratio of the clusters. The cluster density (=1/a2) decreases from ρs = 1 × 10−4/nm2 to ρl = 4 × 10−5/nm2 when comparing the 12 and 63 nm cluster samples.) If we assume that CO adsorbs on both rim and terrace sites, then Θsat should increase by a factor of 6 ± 0.8 for the cluster 5794

dx.doi.org/10.1021/jp211500s | J. Phys. Chem. C 2012, 116, 5792−5801

The Journal of Physical Chemistry C

Article

sizes considered here. For this estimate we assume now a 2D rim with a height of h = 5 nm (see section 2.2). This probably overestimates the area of rim sites. (The total area of terrace and rim sites is given by πAρd(1/4d + h).) Similarly, the calculated ratio of the rim area to cluster size is 1.66 (= 4h/d) and 0.32 for 12 and 63 nm clusters, respectively. Therefore, if rim adsorption sites have special properties, these should be most distinct for the smaller cluster sample which has additionally a greater cluster density. EBL samples have the advantage for model studies that these simple estimates are possible without very time-consuming e.g. STM (scanning tunneling microscopy) sample characterization. 3.2. Spectroscopic Characterization of the Chemical State of the Cu Clusters. Figures 2 and 3 summarize the

Figure 3. Oxidic and metallic Cu clusters can be prepared, as evident from this XPS scan of 63 nm EBL samples. (For data on the 12 nm sample see Supporting Information.)

Table 1. 2p3/2 Cu XPS Line Positions for Cu and Cu Oxidesa this work

other studies

system

ref

Cu

932.7

Cu2O

932.8

CuO

934

932.8 933.1−932.7 932.6 932.6 932.3 933.0−932.4 932.4 932.4 933.8 933.9 935.5−933.8 933.5 933.6

single crystal Cu/Zr metallic Cu single crystal thin films Cu2O/SiO2 nanocrystals powder single crystal thin films CuO/SiO2 nanocrystals powder

48 43 47 48 73 43 74 47 48 73 43 74 47

a

The Cu2O data refers to 12 nm cluster sample, and the rest is for the 63 nm sample.

a. Brief Literature Survey. The XPS literature on copper systems is extensive. In our study, XPS has only been used to document the state of the catalyst before and after the reactivity tests with CO. Briefly, despite different positions of the LMM AES line for each of the oxides, bulk CuO (powders, single crystals) is characterized by intense Cu(2p) XPS shakeup satellites and a broad O(1s) peak.47 In contrast, Cu2O and metallic Cu show weaker/no Cu(2p) satellites and a narrower O(1s) peak. An assignment of the oxidation state depending on the LMM AES line position, which is affected by charging of the support, is experimentally very challenging. Unfortunately, whether or not the inspection of the LMM AES signature is required to distinguish Cu2O and Cu appears to be a controversial matter in the literature. According to e.g. ref 48, the Cu(2p) shakeup satellites are caused by charge transfer into 3d bands which are filled in the case of Cu0. Therefore, small Cu(2p) satellite peaks would be characteristic of Cu2O, and their complete absence would indicate Cu0 (see ref 48). The same holds true for Cu clusters, but here the XPS line positions depend also on the clusters′ morphology (primarily on coverage) and support which makes an exact assignment of the Cu oxidation state, based on XPS/AES peak positions, even more difficult.43,49 In this study, we label Cu cluster samples as

Figure 2. (A) AES survey scan of the 63 nm Cu cluster sample before and after cleaning. (B) XPS survey of the clean 63 nm Cu/SiO2 sample. Inset depicts the Si 2p region.

sample characterization using AES and XPS. Table 1 summarizes the XPS line positions. Additional spectroscopic data for the 12 nm Cu cluster sample are given in the Supporting Information. The data gathered on the smaller clusters have naturally a smaller signal-to-noise (S/N) ratio due to the smaller total metal coverage. (See also ref 5 about 12 nm Au EBL clusters.) 5795

dx.doi.org/10.1021/jp211500s | J. Phys. Chem. C 2012, 116, 5792−5801

The Journal of Physical Chemistry C

Article

nearly metallic (Cu0-like) if no XPS satellite peaks were seen and as fully oxidized (CuO-like) if very intense satellite peaks were present. A few data points (for 12 nm clusters) were collected for an intermediate case where weak XPS satellite peaks were evident (see Supporting Information). b. AES/XPS Survey Scans. AES survey scans of the as received and cleaned 63 nm Cu dot samples are given in Figure 2A; Figure 2B depicts XPS of the cleaned sample. Besides support peaks and Cu lines, no other structures were evident after cleaning the samples. These data are in agreement with earlier experiments (see e.g. refs 50−57) and library data.58,59 The inset depicts the XPS Si 2p region. For fully oxidized silica only one peak is present, whereas partially reduced silica would show two features (see e.g. ref 59 and Supporting Information). Thus, we conclude that the cleaning procedure did not reduce the support. Although blind beam scattering data have been collected for clean silica and partially reduced silica supports, for all beam scattering experiments the Si region did consist of only one peak characteristic of silica. Thus, the cleaning procedures affected the small and highly reactive Cu clusters but modified the support to a far lesser extent. This result may be important and encouraging, at least for other studies on silica supported cluster systems, since silicon (Si) is more reactive than silica, e.g., for probe molecules such as hydrogen. Silica (SiO2), however, is chemically fairly inert, which allows one to distinguish support and cluster effects more easily. The XPS data are rather noisy due to the small overall Cu coverage. This does, however, not prevent to determine the oxidation state (metallic vs oxidic). c. Oxidation State of the Cu Clusters. Figure 3 depicts typical examples of the Cu 2p XPS region of the silica supported 63 nm Cu clusters. Results for the smaller cluster sample are very similar except a worse S/N ratio (see Supporting Information). The results are in good agreement with the literature (see Table 1). Depending on the sample treatment, nearly metallic Cu clusters, partially oxidized, and fully oxidized Cu clusters could be prepared. According to XPS/AES data the oxidation/reduction of the clusters was reversible. 3.3. Kinetics. Figure 4 summarizes CO TDS data on 63 nm Cu EBL samples for metallic (panel A) and fully oxidized (panel B) Cu clusters. (CO was dosed by the molecular beam system; the intensities are given in arbitrary units.) For the metallic clusters (Figure 4A) a very broad feature is evident which shifts slightly to smaller temperatures with increasing exposure. Obviously, the TDS signal is weak due to the small overall Cu coverage which results in rather noisy curves. The peak width and shift would be consistent with kinetically distinct adsorption sites, repulsive interactions, or a combination of both. The total CO coverage even on the larger Cu clusters is small which renders large effects of lateral interactions unlikely. Small clusters have, however, a large rimto-surface area ratio (section 3.1) which supports the conclusion that mostly different adsorption sites are responsible for the peak broadening. TDS peak positions and binding energies (estimated using the Redhead equation and a prefactor of 1 × 1013/s) are given in Table 2. The TDS peak width is similar to prior data collected for PVD Cu clusters.11,37 The TDS peak position is, however, significantly lower. The only obvious explanation is morphological differences in details of the clusters′ structure and/or electronic properties. Similar deviations in kinetics data were also seen when comparing Au PVD and Au EBL cluster systems.5 As already noted in prior

Figure 4. TDS spectra of CO adsorbed on silica-supported (A) metallic (B) oxidic clusters (63 nm clusters). CO was dosed with the molecular beam.

Table 2. Binding Energies of CO Determined from TDS Peak Positions As Compared to Literature Valuesa Tpeak (K) Cu0

CuO

203 430

132−143 (α) 195 K (β) 125 K

Ed (kJ/mol) 49.4 107.5 58.6 54.3 33.6−36.5 50.2 31.7

system

ref

Cu(311)

75

Cu(110) Cu(111) 63 nm EBL clusters 63 nm EBL clusters

76 77 this work this work

a The uncertainty in temperature calibration amounts to ±5 K, which results in an uncertainty of ±1.3 kJ/mol in binding energies.

studies (e.g., ref 37), a clear assignment of TDS features to certain crystallographic phases is rarely possible. Similar TDS experiments on oxidic Cu clusters are depicted in Figure 4B. Very broad TDS peaks, which may consist of two structures (α and β peaks), are evident. We are not aware of kinetics data on oxidic Cu clusters, but on oxide single crystals such as TiO2 and ZnO (see refs 60 and 61) also two CO TDS peaks were detected. The standard model is to assign the different TDS peaks to different types of adsorption sites. For example, the higher temperature (larger binding energy) feature 5796

dx.doi.org/10.1021/jp211500s | J. Phys. Chem. C 2012, 116, 5792−5801

The Journal of Physical Chemistry C

Article

may be related to defect sites (rim sites, edges, etc.) and the low temperature peak to pristine (terrace) sites. Similarly, the binding energies of CO on oxygen and Cu sites of the oxidic clusters would be different. On iron oxide clusters (see e.g. refs 62 and 63), different CO TDS features were assigned to different Fe oxides. We cannot definitely exclude that different oxide species may be present; however, this appears rather unlikely considering the XPS results. We should mention that we did not detect any CO2 formation in TDS and beam scattering experiments with CO. 3.4. Adsorption Dynamics. Figures 5−9 summarize the molecular beam scattering results on the silica supported EBL

Figure 6. S0 of CO vs impact energy for 12 and 63 nm Cu EBL clusters as well as for selected Cu PVD depositions (from ref 11). Ts = 95 K.

Figure 7. S0 of CO vs coverage at constant impact energy (Ei = 0.39 eV) and surface temperature (Ts = 95 K). Data for metallic EBL and PVD (from ref 11) Cu clusters are shown. (The line acts as a guide for the eye. The Cu coverage is used here and not the number of adsorption sites.)

knowledge, only one beam scattering project on that system is reported in the literature.11 The Cu PVD clusters did remain metallic at UHV conditions in this prior beam scattering study;11 oxidic clusters were not considered. The adsorption dynamics of CO could be modeled with the CZM (see section 1.3). S0 decreases with impact energy, as expected for molecular and nonactivated adsorption. In refs 37 and 40 a study about CO adsorption on ZnO(0001) supported Cu PVD clusters, a crossover from direct adsorption dynamics (Langmuirian-like) to more precursor assisted adsorption dynamics (Kisliuk-like), has been observed, which clearly depends on the size of the metal deposits. Some of the observed trends are consistent with simple CZM, but interestingly not the shape of the S(Θ) curves, especially at small Cu coverage and at large impact energies. The results suggest either uncommonly large diffusion activation energies or nonthermal precursor states which would shorten the lifetime of trapped particles and hence weaken the effect of precursor states.

Figure 5. CO adsorption transients on (A) metallic and (B) oxidic clusters. Data for small (12 nm) and large (63 nm) EBL clusters are shown. In addition, blind experiments on the clean supports are included (Ei = 0.39 eV; Ts = 95 K).

cluster samples in comparison with blind experiments on the clean supports and Cu PVD cluster data.11 Data for Cu and CuO-like EBL clusters are shown. a. Brief Literature Survey. The PVD deposition of Cu on silica has been studied before with spectroscopic and kinetic techniques.15,16,64 The morphology of Cu PVD clusters was characterized using STM.50−53 The data indicate that Cu PVD clusters nucleate on the silica support for low Cu exposures, forming highly dispersed and small clusters. This nucleation phase is followed by a cluster growth regime. To the best of our 5797

dx.doi.org/10.1021/jp211500s | J. Phys. Chem. C 2012, 116, 5792−5801

The Journal of Physical Chemistry C

Article

Figure 8. S0 of CO vs impact energy for 12 and 63 nm metallic and oxidic EBL clusters (Ts = 95 K).

Clean copper surfaces were studied with molecular beam scattering techniques as well.65 We are not aware of molecular beam scattering data on Cu oxide single crystals, supported CuOx clusters, or any Cu EBL system. b. Examples of Typical CO Adsorption Transients: Rim vs Terrace Sites. The solid lines in Figure 5 show typical examples of CO adsorption transients (CO pressure vs exposure time) of the silica-supported EBL clusters, together with blind experiments on silica and partially reduced silica. The top panel displays results for metallic and the bottom one for oxidic Cu EBL clusters. The curves are normalized such that 1 − S(t) vs t is depicted, with t for the exposure time by the seeded molecular beam and S(t) for the adsorption probability. Therefore, the adsorption probabilities can directly be read from these graphs. For example, S0 can be determined at t = 0 where the CO beam started to strike the surface. Integrating these transients give one the CO coverage dependent adsorption probability, S(Θ) (see section 3.4e). Evidently, the transients approach the saturation level (where S = 0 or 1 − S = 1) much slower on the supported cluster samples than on the bare supports. Thus, the clusters indeed affect the adsorption dynamics and CO adsorbs on the clusters and/or along the clusters′ rim. The area above the transient and below the saturation level equals the amount of adsorbed probe molecules. Therefore, it is evident that the CO coverage on the bare supports is negligible (for silica and reduced silica) at the given adsorption temperature. Furthermore, with the measured CO flux (section 2.1), the total CO uptake (or saturation coverage Θsat) amounts to Θsat = ∫ FS(t) dt ≈ 1.8 × 1014 molecules/cm2 (or 1.8 × 1014/ 1.9 × 1015 = 0.09 ML) for 63 nm Cu clusters, respectively. (Using the bulk lattice constant of Cu (3.6 Å), the calculated Cu atom density of e.g. a (111) plane amounts to 1.9 × 1015 atoms/cm2.) The experimental saturation coverage (0.09 ML) appears reasonable since the calculated geometrical Cu coverage of the 63 nm clusters would be 0.139 ML (see section 3.1) and the calculated CO coverage amounts to 0.07 ML = (0.139 × 0.52). (A saturation coverage of 0.52 ML for CO adsorption on Cu(111) was determined experimentally.66) More interesting, when comparing large and small clusters, the experimental ratio of Θsat,l/Θsat,s amounts to 5.5 ± 0.9 and 4.5 ± 0.7 for metallic and oxidic clusters, respectively. (The

Figure 9. Coverage-dependent adsorption probabilities for (A) metallic and (B) oxidic 63 nm EBL clusters. The impact energy was varied, as indicated; the surface temperature was constant at 95 K.

experimental uncertainties are obtained from averaging independent experimental runs.) As described in section 3.1, ratios of 2.3 ± 0.4, 12.6 ± 1.0, and 6.0 ± 0.8 would be expected from simple geometrical considerations, when the gas-phase species adsorb only along the rim sites, terrace sites, or both (rim + terrace), respectively. This result allows for the following conclusions: Nearly the same saturation coverage ratios were obtained independent of the oxidation state of the clusters. Therefore, it appears that geometrically similar adsorption sites on both cluster types may be present. A direct comparison with the TDS data (Figure 4) is unfortunately difficult since determining absolute intensities in different TDS runs is problematic. The experimentally determined ratio matches within uncertainties the prediction for adsorption on rim sites plus terrace sites. Therefore, CO indeed also adsorbs on the clusters′ rim, as perhaps expected, which can, however, experimentally be distinguished from pure rim effects using EBL samples. This result may also imply that the rim sites on Cu clusters do not show special behavior besides simple geometrical effects. Even if the CO molecules are dosed with a molecular beam system, the saturation coverage is determined by the adsorption kinetics. Therefore, more precisely, we should note that the rim sites on the EBL clusters provide additional adsorption sites 5798

dx.doi.org/10.1021/jp211500s | J. Phys. Chem. C 2012, 116, 5792−5801

The Journal of Physical Chemistry C

Article

This trend is consistent with simple mass matching models when taking site blocking effects into account. A CO molecule can interact only with Cu adsorption sites on the metallic clusters, but with Cu and O adsorption sites on the oxidic clusters. The CO-to-Cu mass match is worse than the CO-to-O mass match. This should result in more efficient gas-to-surface energy transfer processes (and larger S0) in the latter case. Therefore, simple mass-matching models would predict larger S0 values for the oxidic clusters, contrary to the observation. However, this may suggest that CO does not adsorb on oxygen sites of the oxidic clusters. Since the adsorption probability is basically the ratio of adsorption rate and impinging rate, site blocking effects (or poisoning) result in smaller S0 values. The O sites on the oxides where CO cannot adsorb would act as poisoned adsorption sites decreasing S0, as indeed observed. (With these assumptions, the available number of adsorption sites on the oxidic clusters is simply smaller than on the metallic clusters.) Obviously this is a simple model. CO adsorption on oxidic structures may in principle also be influenced by defects such as oxygen vacancy sites or even carbonate formation. When considering cluster size effects for the oxidic clusters, it becomes evident that S0 (within experimental uncertainties) is independent of cluster size for the oxide (Figure 8) but depends on cluster size for the metal clusters (Figures 7 and 8). The database for the oxides is limited, and electronic effects should be present. However, when oxidizing Cu clusters, also the clusters′ size should increase. The crystallographic phase which dominates the clusters is unknown, but the unit cell of Cu(111) amounts to 6.5 Å2 whereas the unit cell of bulk CuO is with 16 Å2 (= a × b, see ref 72) much larger (by a factor of 2.5). Therefore, when oxidizing the clusters, according to Figure 7, the capture zones would most likely already overlap; i.e., no effect of cluster size on S0 would indeed be expected for the oxides. In addition, the CO binding energies on the oxide is larger than for the metal (Figure 4); i.e., the precursor lifetimes and therefore the capture zones will be larger and overlap early. e. Coverage-Dependent Adsorption Probabilities. Integrating the adsorption transients allows for calculating a relative coverage, and hence the coverage, Θ, dependent adsorption probabilities, S(Θ), are available. Figure 9 depicts S(Θ) as a function of Ei for CO adsorption on 63 nm Cu clusters (Figure 9A) and CuO-like clusters (Figure 9B) at constant Ts. The coverage was normalized to 1 ML, which corresponds to 1.8 × 1014 molecules/cm2 (see section 3.4b). At low Ei, S(Θ) obeys reasonably well the traditional Kisliuk shape. Here, S(Θ) remains initially large up to saturation of the catalyst where S(Θ) drops to zero. This type of curve shape indicates the effect of precursor states, as expected from the capture zone model. For, large Ei, S(Θ) decreases about linearly with Θ; i.e., Langmuirian-like adsorption dynamics is seen. This crossover from precursor-mediated Kisliuk-like dynamics to direct Langmuirian-like dynamics is commonly seen for CO also on planar catalysts and simply reflects the decrease in the trapping probability in the precursor state with increasing Ei.70

which, however, appear not to behave kinetically much different than terrace sites. For example, if the kinetics is dominated only by rim sites (as it is for Au clusters12,67), a quite different ratio of the saturation coverages would be expected. Variations in S0 are seen depending on the cluster size and oxidation state, as described next. c. Energy Dependence of the Initial Adsorption Probability: Cluster Size Effects. In Figure 6, the initial adsorption probability, S0, of CO is shown at constant adsorption temperature and as a function of impact energy, Ei. This graph summarizes results for metallic EBL and PVD clusters (from ref 11) of different cluster sizes or Cu exposure. (The uncertainties are identical for all data points; however, only a few error bars are explicitly depicted, not to clutter the figures. The lines are presented as a guide for the eye.) In all cases, S0 decreases with Ei, consistent with nonactivated and molecular adsorption. This is in agreement with the absence of carbon after adsorption/desorption cycles, as verified by XPS/AES. Overall and in particular for small Ei, the adsorption probabilities of CO are large, as commonly observed for that probe molecule on a variety of other systems (see e.g. refs 68−70). Cluster size effects are consistently evident in several data sets. The arrow in Figure 6, for example, highlights that S0 increases with increasing cluster size (while keeping all other parameters constant). Or, Figure 7 shows that S0 of the Cu clusters increases with cluster size (or coverage). Here S0 of Cu PVD data (ref 11) as a function of Cu exposure and S0 of small and large EBL Cu clusters are compared. The CO impact energy and surface temperature were kept constant. Importantly, the results obtained on Cu EBL and Cu PVD clusters match. For large exposures (PVD), S0 appears to approach the values determined in other studies65,68 for Cu single crystals or Cu films. The same trend has been seen for ZnO (ref 37) supported Cu PVD clusters. In other words, Cu clusters with a total coverage in the percent range are as reactive as Cu single crystals. This is a typical result of the capture zone effect (see e.g. refs 36 and 71) mentioned already in the Introduction. Even if the coverage of the probe molecules on the bare support is negligible, the probe molecules are trapped on the support and diffuse from their landing site to the metal clusters where they adsorb. Therefore, the adsorption probabilities for supported cluster systems are larger than it may be expected from their geometrical size. An enhancement in reactivity by cluster systems, as characterized by measuring e.g. S0 values, is therefore expected. The initial increase in S0 (Figure 7) with cluster size/ coverage is caused by the increase in available adsorption sites. However, for large cluster densities (large Cu exposure in the PVD experiments), the capture zones around the clusters will overlap. At that point, increasing the cluster size and dispersion cannot increase S0 further (Figure 7). Therefore, the initial adsorption probability in S0 vs exposure plots levels out. d. Effect of Oxidation State on Initial Adsorption Probability. Figure 8 compares S0 of metallic and oxidic EBL clusters as a function of impact energy and parametric in cluster size. Differences in S0 are evident, depending on the oxidation state of the clusters: metallic Cu clusters are more reactive than oxidic Cu clusters (with respect to S0 of CO). This trend can most clearly be seen for large Ei, where details in the adsorption dynamics are naturally more important than at low impact energies.

4. CONCLUSIONS Kinetics (TDS), dynamics (adsorption probabilities), microscopic (SEM), and spectroscopic (AES, XPS) techniques were used to characterize the reactivity of Cu and CuOx clusters toward CO adsorption. The silica supported 12 and 63 nm cluster samples were fabricated using EBL (electron beam lithography). 5799

dx.doi.org/10.1021/jp211500s | J. Phys. Chem. C 2012, 116, 5792−5801

The Journal of Physical Chemistry C

Article

(7) Kobayashi, H.; Takezawa, N.; Minochi, C. J. Catal. 1981, 69, 487−494. (8) Ossipoff, N. J.; Cant, N. W. J. Catal. 1994, 148, 125−133. (9) Funk, S.; Goering, J.; Burghaus, U. Appl. Surf. Sci. 2008, 254, 5271−5275. (10) Funk, S.; Nurkic, T.; Burghaus, U. Appl. Surf. Sci. 2007, 253, 4860−4865. (11) Komarneni, M.; Shan, J.; Burghaus, U. J. Phys. Chem. C 2011, 115, 16590−16597. (12) Shan, J.; Komarneni, M.; Burghaus, U. Chem. Phys. Lett. 2011, 517, 59−61. (13) Kadossov, E.; Burghaus, U. Surf. Interface Anal. 2008, 40, 893− 898. (14) Kadossov, E.; Goering, J.; Burghaus, U. Surf. Sci. 2008, 602, 811−818. (15) Hansen, J. B. Handbook of Heterogeneous Catalysis; VCH: Weinheim, 1997; Vol. 1. (16) Askgaard, T. S.; Norskov, J. K.; Ovesen, C. V.; Stoltze, P. J. Catal. 1995, 156, 229−335. (17) Hara, M.; Kondo, T.; Komoda, M.; Ikeda, S.; Kondo, J. N.; Domen, K.; Hara, M.; Shinohara, K.; Tanaka, A. Chem. Commun. 1998, 357−358. (18) Newsome, D. S. Catal. Rev.Sci. Eng. 1980, 21, 275−318. (19) Sadykov, V. A.; Tikhov, S. F.; Bulgakov, N. N.; Gerasev, A. P. Catal. Today 2009, 144, 324−331. (20) Chan, G. H.; Zhao, J.; Hicks, E. M.; Schatz, G. C.; Duyne, R. P. V Nano Lett. 2007, 7, 1947−1952. (21) Musa, A. O.; Akomolafe, T.; Carter, M. J. Sol. Energy Mater. Sol. Cells 1998, 51, 305−316. (22) Bednorz, J. G.; Mueller, K. A. Z. Phys. B 1986, 64, 189−200. (23) White, B.; Yin, M.; Hall, A.; Le, D.; Stolbov, S.; Rahman, T.; Turro, N.; O’Brien, S. Nano Lett. 2006, 6, 2095−2098. (24) Premkumar, T.; Geckeler, K. E. J. Phys. Chem. Solids 2006, 67, 1451−1456. (25) Oetelaar, L. C. A.; Partridge, A.; Stapel, P. J. A.; Flipse, C. F. J.; Brongersma, H. H. J. Phys. Chem. B 1998, 102, 9541−9549. (26) Cox, P. A. Transition Metal Oxides; Clarendon Press: Oxford, 1995. (27) Cox, D. F.; Schulz, K. H. Surf. Sci. 1991, 249, 138−148. (28) Sueyoshi, T.; Sasaki, T.; Iwasawa, Y. Surf. Sci. 1995, 343, 1−16. (29) Henry, C. R. Surf. Sci. Rep. 1998, 31, 235−268. (30) Libuda, J.; Freund, H. J. Surf. Sci. Rep. 2005, 57, 157−298. (31) Freund, H. J.; Roberts, M. W. Surf. Sci. Rep. 1996, 25, 225−234. (32) Kim, T. S.; Gong, J.; Ojifinni, R. A.; White, J. M.; Mullins, C. B. J. Am. Chem. Soc. 2006, 128, 6282−6300. (33) Díaz, C.; Olsen, R. A.; Auerbach, D. J.; Kroes, G. J. Phys. Chem. Chem. Phys. 2010, 12, 6499−6532. (34) Wodtke, A. M.; Matsiev, D.; Auerbach, D. J. Prog. Surf. Sci. 2008, 83, 123−182. (35) Juurlink, L. B. F.; Killelea, D. R.; Utz, A. L. Prog. Surf. Sci. 2009, 84, 69−74. (36) Rumpf, F.; Poppa, H.; Boudard, M. Langmuir 1988, 4, 722−728. (37) Wang, J.; Burghaus, U. J. Chem. Phys. 2005, 123, 184716−12. (38) Funk, S.; Hokkanen, B.; Wang, J.; Burghaus, U.; Bozzolo, G. H.; Garces, J. E. Surf. Sci. 2006, 600, 583−590. (39) Wang, J.; Burghaus, U. Chem. Phys. Lett. 2005, 403, 42−46. (40) Wang, J.; Johnson, E.; Burghaus, U. Chem. Phys. Lett. 2005, 410, 131−135. (41) Wang, J.; Burghaus, U. J. Chem. Phys. 2005, 122, 044705−11. (42) Rettner, C. T.; DeLouise, L. A.; Auerbach, D. J. J. Chem. Phys. 1986, 85, 1131−1148. (43) Espinos, J. P.; Morales, J.; Barranco, A.; Caballero, A.; Holgado, J. P.; Gonzalez-Elipe, A. R. J. Phys. Chem. B 2002, 106, 6921−6929. (44) Kadossov, E.; Justin, J.; Lu, M.; Rosenmann, D.; Ocola, L. E.; Cabrini, S.; Burghaus, U. Chem. Phys. Lett. 2009, 483, 250−253. (45) Golunski, S. E. Platinum Met. Rev. 2007, 51, 162−204. (46) Kadossov, E.; Burghaus, U. Catal. Lett. 2010, 134, 228−232. (47) Poulston, S.; Parlett, P. M.; Stone, P.; Bowker, M. Surf. Interface Anal. 1996, 24, 811−820.

Interestingly, the saturation coverage of CO, as determined by measuring adsorption transients (Figure 5) with a molecular beam system, does not simply scale with the area of the clusters, but rim effects clearly increase the saturation coverage. The initial adsorption probability of CO, S0, decreases with increasing impact energy (Figure 6), and XPS/AES do not indicate carbon residuals after CO experiments. Thus, we conclude nonactivated molecular adsorption. S0 of Cu clusters depends on cluster size (Figure 7), which can be described in the framework of the capture zone model (CZM). Interestingly, differences in the adsorption dynamics depending on the oxidation state of the clusters are evident (Figures 8 and 9). S0 of CO on the metallic clusters are larger than for the oxidic clusters. In addition, cluster size effects are only distinct for the metallic clusters. Although we cannot rule out electronic effects, simple mass matching models would predict the first observation. The second one would be consistent with a cluster size expansion (in the oxidation process) of the oxidic clusters and larger binding energies.

■

ASSOCIATED CONTENT

S Supporting Information *

AES and XPS data of the 12 nm Cu model nanoarray catalyst. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]; Fax: 701.231.8831; Web address: www.uweburghaus.us. Present Address †

Xplosafe, LLC., Stillwater, OK 74074.

Notes

The authors declare no competing financial interest. ‡ Nanofabrication Facility, Molecular Foundry, LBNL, Berkeley, CA 94720.

■

ACKNOWLEDGMENTS Financial support by an NSF-CAREER award (CHE-0743932) is acknowledged by NDSU. A supplemental equipment grant (for acquiring the second-hand XPS system) from the Chemical Sciences, Geosciences, and Biosciences Division, Office of Basic Energy Sciences, Office of Science, U.S. Department of Energy, is acknowledged (Project DE-FG0208ER15987) as well. Work at the Molecular Foundry was supported by the Office of Science, Office of Basic Energy Sciences, of the U.S. Department of Energy under contract no. DE-AC02-05CH11231.

■

REFERENCES

(1) Liu, C. J.; Burghaus, U.; Besenbacher, F.; Wang, Z. L. ACS Nano 2010, 4, 5517−5526. (2) Ribeiro, F. H.; Somorjai, G. A. Recl. Trav. Chim. Pays-Bas 1994, 113, 419−422. (3) Jacobs, P. W.; Riberio, F. H.; Somorjai, G. A.; Wind, S. J. Catal. Lett. 1996, 37, 131−136. (4) Wong, K.; Johansson, S.; Kasemo, B. Faraday Discuss. 1996, 105, 237−246. (5) Kadossov, E.; Cabrini, S.; Burghaus, U. J. Mol. Catal. A: Chem. 2010, 321, 101−109. (6) Johanek, V.; Laurin, M.; Grant, A. W.; Kasemo, B.; Henry, C. R.; Libuda, J. Science 2004, 304, 1639−1643. 5800

dx.doi.org/10.1021/jp211500s | J. Phys. Chem. C 2012, 116, 5792−5801

The Journal of Physical Chemistry C

Article

(48) Panzer, G.; Egert, B.; Schmidt, H. P. Surf. Sci. 1985, 151, 400− 408. (49) Moretti, G.; Rossi, S. D.; Ferraris, G. Appl. Surf. Sci. 1990, 45, 341−349. (50) Xu, X.; Vesecky, S. M.; Goodman, D. W. Science 1992, 258, 788−792. (51) Xu, X.; He, J. W.; Goodman, D. W. Surf. Sci. 1993, 284, 103− 108. (52) Xu, X.; Goodman, D. W. J. Phys. Chem. 1993, 97, 683−691. (53) Xu, X.; Goodman, D. W. Appl. Phys. Lett. 1992, 61, 1799−1805. (54) Piercea, D. E.; Burnsa, R. P.; Gabrielb, K. A. Thin Solid Films 1991, 206, 340−345. (55) Dumas, P.; Tobina, R. G.; Richardsa, P. L. Surf. Sci. 1986, 171, 579−583. (56) Cechal, J.; Polcak, J.; Kolibal, M.; Babor, P.; Sikola, T. Surf. Sci. 2010, 256, 3636−3641. (57) Zhou, J. B.; Gustaffsson, T.; Garfunkel, E. Surf. Sci. 1997, 372, 21−27. (58) Hedberg, C. L. Handbook of Auger Electron Spectroscopy; Physical Electronics, Inc.: Minneapolis, 1970. (59) Moulder, J. F.; Stickle, W. F.; Sobol, P. E.; Bomben, K. D. Handbook of X-ray Photoelectron Spectroscopy; Perkin-Elmer Corporation: Eden Prairie, MN, 1992. (60) Linsebigler, A.; Lu, G.; Yates, J. T. J. Chem. Phys. 1995, 103, 9438−9443. (61) Henderson, M. A. Surf. Sci. 1998, 400, 203−212. (62) Kadossov, E.; Funk, S.; Burghaus, U. Catal. Lett. 2007, 120, 179−183. (63) Lemire, C.; Meyer, R.; Henrich, V. E.; Shaikhutdinov, S.; Freund, H. J. Surf. Sci. 2004, 572, 103−114. (64) Bowker, M.; Houghton, H.; Waugh, K. C. J. Chem. Soc., Faraday Trans. 1981, 77, 3023−3026. (65) Kunat, M.; Boas, C.; Becker, T.; Burghaus, U.; Wöll, C. Surf. Sci. 2001, 474, 114−119. (66) Bartels, L.; Meyer, G.; Rieder, K. H. Surf. Sci. Lett. 1999, 432, L621−629. (67) Haruta, M. Catal. Today 1997, 36, 153−171. (68) Kneitz, S.; Gemeinhardt, J.; Koschel, H.; Held, G.; Steinrück, H. P. Surf. Sci. 1999, 433, 27−33. (69) Kunat, M.; Burghaus, U. Surf. Sci. 2003, 544, 170−178. (70) Burghaus, U.; Ding, J.; Weinberg, W. H. Surf. Sci. 1997, 384, L869−876. (71) Bowker, M. Surf. Rev. Lett. 1994, 4, 549−552. (72) Zheng, X. G.; Suzuki, M.; Xu, C. N. Mater. Res. Bull. 1998, 33, 605−608. (73) Barreca, D.; Gasparotto, A.; Tondello, E. Surf. Sci. Spectra 2007, 14, 41−51. (74) Wu, C.-K.; Yin, M.; O’Brien, S.; Koberstein, J. T. Chem. Mater. 2006, 18, 6054−6058. (75) Fu, S. S.; Somorjai, G. A. Surf. Sci. 1992, 262, 68−73. (76) Wachs, I. E.; Madix, R. J. J. Catal. 1978, 53, 208−220. (77) Kessler, J.; Thieme, F. Surf. Sci. 1977, 67, 405−415.

5801

dx.doi.org/10.1021/jp211500s | J. Phys. Chem. C 2012, 116, 5792−5801