Appl. Phys. A 80, 977–986 (2005)
Applied Physics A
DOI: 10.1007/s00339-004-3125-9
Materials Science & Processing
Surface chemistry studied by in situ X-ray photoelectron spectroscopy
r. denecke
Lehrstuhl für Physikalische Chemie II, Universität Erlangen-Nürnberg, Egerlandstr. 3, 91058 Erlangen, Germany
Received: 6 June 2004/Accepted: 22 October 2004 Published online: 23 February 2005 • © Springer-Verlag 2005 ABSTRACT Time-dependent
X-ray photoelectron spectroscopy is used to study the kinetics and dynamics of simple surface reactions. Combining high-resolution core level spectroscopy with a supersonic molecular beam in one experimental setup, processes such as the dissociative adsorption of methane on both Pt(111) and Ni(111), the coadsorption of water and CO on Pt(111), and the oxidation of CO on Pt(111) have been studied. In the case of methane, the observed vibrational fine structure in C 1s spectra is used to identify the adsorbed species (CH3 ) and further thermal dehydrogenation steps. While simple dehydrogenation via CH is observed on Pt(111), a C–C coupling reaction to acetylene is found on Ni(111). In the coadsorbate phase, CO is found to be able to replace predosed water from the bilayer into multilayers. Water, in turn, leads to a site change of the CO molecules, which are preferably adsorbed at bridge sites in the presence of water, as opposed to on-top adsorption on clean Pt(111). For the truly bimolecular surface reaction, the CO oxidation on Pt(111), the ability of the molecular beam to create a relatively high CO pressure was found essential to study the kinetics of the basic step (CO + O → CO2 ) without influence of adsorption or diffusion rate. An activation energy of 0.53 eV and a preexponential factor of ∼ 5 × 106 s−1 are found. PACS 68.43.Mn;
1
79.60.Dp; 82.20.Pm
Introduction
Important aspects in surface science are the kinetics of adsorption and reaction processes. However, most of the experimental techniques used in the past required long data acquisition times and an ultrahigh vacuum environment. This was especially true for X-ray photoelectron spectroscopy (XPS), a quantitative and chemically sensitive method. The advent of high-intensity synchrotron radiation from third generation facilities and its use as excitation source has changed the situation significantly. Nowadays, high-resolution XPS has developed into a time-dependent method, which can be used to study adsorption and reaction processes in situ [1]. Another prerequisite for studying the kinetics of surface processes is the ability to control all parameters involved in a most complete manner. This is in particular true for the properties of the adsorbing gas molecules. One way of doing so u Fax: +49-9131-8528867, E-mail:
[email protected]
is to use a supersonic molecular beam, which provides a localized high (at least as compared to usual surface science standards) pressure of molecules with a well-defined kinetic energy. This beam can be switched on and off in a controlled way, thus providing a well-defined time structure to study adsorption or reaction processes. By combining high-resolution XPS at a high-flux synchrotron radiation source with a molecular beam, timedependent surface processes can be followed on a time scale of seconds [2–4]. Spectroscopic information about the species involved can be deduced from the individual binding energy of adsorbate and substrate core levels. In the case of hydrocarbons, the vibrational fine structure observed in the XP spectra can additionally be used to identify surface species [5]. From quantitative information obtained by intensity analysis, kinetic parameters can be derived [1]. In this contribution, an overview will be given of recent results obtained by using this particular experimental combination of high-resolution XPS and molecular beam. After introducing the experimental details, three examples of surface reactions are presented. First, the influence of water on CO adsorption on Pt(111) is discussed [6], a rather important coadsorption system reaching out to electrochemistry. In this part for comparison, also the adsorption kinetics of CO on pure Pt(111) will briefly be reviewed [7]. Second, the CO oxidation on Pt(111) is investigated, resulting in kinetic parameters [8]. Finally, using the tunability of the kinetic energy of the impinging molecules by the supersonic molecular beam, adsorption and subsequent thermal dehydrogenation of methane on both Pt(111) [9] and Ni(111) is studied. 2
Experimental
All experiments were performed at the BESSY II synchrotron radiation facility in Berlin, Germany, in a transportable ultrahigh vacuum (UHV) apparatus, which has already been described elsewhere [4]. Briefly, it comprises three separate chambers: (1) An analysis chamber, equipped with a hemispherical electron energy analyser (Omicron EA 125 U7 HR); synchrotron radiation from beamlines U49/1-SGM or U49/2-PGM1 enters at an angle of 50◦ with respect to the analyser lens system which lies in the plane of linear light polarization. (2) In the molecular beam chamber a collimated supersonic beam is produced by expanding the probe gas through a Mo nozzle with a diameter of 100 µm into the first of three successive differential pumping stages. The molecu-
978
Applied Physics A – Materials Science & Processing
lar beam enters the analysis chamber at an angle of 45◦ with respect to the analyser lens system; it runs through the focal point of the analyser in order to allow for in situ measurements of adsorption processes. A flag, mounted on a pneumatically operated linear feedthrough in the second stage, allows for fast (< 0.1 s) switching of the beam. CO (purity 99.997%) and methane (purity 99.9995%) used in the molecular beam were further purified by passing the gas line through a cooling trap at T ∼ 220 K in order to remove Ni carbonyls, which can be produced in the stainless steel supply tube in the case of CO, and to remove residual water in the case of CH4 . He (purity 99.9999%) as seeding gas was purified by passing through a LN2 cooling trap. (3) Finally, the preparation chamber contains a low energy electron diffraction (LEED) optics, an ion gun and a quadrupole mass spectrometer. The sample holder is mounted on a xyz manipulator, which also allows for variation of the polar and azimuthal angle orientation of the sample. Alternatively to the supersonic molecular beam, gas can also be dosed into the analysis and preparation chambers by a conventional dosing system, which in this work was used for oxygen dosing (purity 99.998%). Both the Pt(111) and Ni(111) samples (diameter 10 mm) are spot-welded to Ta wires; one sample is mounted at a time. They can be heated resistively up to 1500 K (for sample cleaning) and cooled by a LN2 cryostat to 95 K. During XPS, the sample is heated radiactively by a filament positioned at the back in order to avoid disturbing magnetic fields caused by resistive heating. The temperature is measured by a K-type thermocouple spot-welded to the edge of the sample. A linear temperature ramp can be applied by a programmable temperature controller (Eurotherm). The Pt sample was cleaned by repeated cycles of Ar+ ion bombardment (1 kV), heating in oxygen (1.3 × 10−7 mbar, 300– 800 K, approx. 2 K/s ramp) and annealing to 1300 K. After the sample was bulk-cleaned, carbon contaminants on the surface could be removed efficiently by a single cycle of heating in oxygen and annealing. Cleaning of the Ni(111) crystal was facilitated by Ar+ ion bombardment (1 kV) again, with subsequent annealing to 1250 K. The carbon deposited during the hydrocarbon experiments diffuses completely into the bulk by heating to temperatures above 1250 K, as determined by the absence of a C 1s signal in the XP spectra. XP spectra were collected at photon energies of 380 eV for C 1s and 650 eV for the O 1s region in normal emission geometry for CO oxidation and for CO and water coadsorption; for the hydrocarbon adsorption experiments the sample was oriented normal to the molecular beam, yielding an electron emission angle of 45◦ ; this is necessary to maximize the normal component of the kinetic energy of the molecules, which is the important parameter to overcome the activation barrier [10]. For time-resolved measurements, spectra in the O 1s region could be collected within a time as short as 5 s per spectrum ( E B = 527 – 536 eV) at an overall resolution of ∼ 250 meV. The C 1s spectra of methane were obtained with a nominal resolution of 140 meV, requiring 8 s per 6 eV wide spectrum. For the CO related projects, C 1s measurement times down to 1.5 s for a 6 eV wide binding energy window could be reached with a slightly lower combined resolution of 180 meV. The reproducibility of binding energy values within this study is ±30 meV; the calibration of the absolute binding
energy scale compared to other studies has an uncertainty of typically ±150 meV. For the CO used in both the coadsorption experiment with water and the CO oxidation reaction, a molecular beam without seeding and with the nozzle at room temperature was used, resulting in a kinetic energy of 0.09 eV [6, 7]. For the methane adsorption studies, a supersonic molecular beam was generated by expanding mixtures of methane and helium through the molybdenum nozzle. By varying the nozzle temperature between 300 and 1073 K, and using seeding ratios between 1.25 and 10% of CH4 in He, kinetic energies between 0.09 and 0.83 eV were achieved. These energies were measured using time-of-flight methods in a similar molecular beam setup [11]. 3 3.1
Results Coadsorption of CO and D2 O on Pt(111)
To start out, we will briefly review CO adsorption on the clean surface. On Pt(111), CO adsorbs in linear (ontop) and two-fold coordinated (bridge) sites [12]. For low coverages, agreement is found in the experimental literature that the on-top species is energetically favored as compared to the bridge site (see [7] for references); both sites are separated by an energy barrier. For a total CO coverage of 0.5 ML (1 ML is defined as one adsorbate atom or molecule per substrate surface atom) a sharp c(4x2) LEED pattern is observed [12, 13], which contains equal amounts of CO occupying on-top and bridge sites. Using this structure as a reference, which is reached by CO adsorption at a background pressure of 1.7 × 10−9 mbar and a sample temperature of T = 200 K, all CO XP spectral intensities, which are obtained in the following analysis from fitting model peak profiles to the spectra (for details see [6–8]), are rescaled and converted to coverage values given in ML units. Figure 1a shows C 1s spectra recorded during CO uptake on a clean Pt(111) surface at 125 K and a pressure of 5.3 × 10−8 mbar. One can clearly observe the subsequent evolution of two well-separated, asymmetric peaks located at 286.8 and 286.1 eV, which are attributed to CO bound in on-top and bridge sites, respectively. For low exposures, site occupation starts at the on-top sites, which are energetically favored. Measurements in the CO pressure range from 1.7 × 10−9 to 2.4 × 10−7 mbar (by employing the molecular beam) yield no significant change in the occupation ratio between on-top and bridge sites, for total CO coverages below 0.35 ML and for temperatures as low as 110 K [7]. This indicates that thermal equilibrium between adsorbates in the two sites is reached under these conditions. A temperature-dependent study of the occupation ratio thus provides information about the binding energy difference for CO in these sites. Using a lattice-gas model treating adsorbate-adsorbate interactions explicitly as parameters, this difference was determined as 95 meV [14]. This value can be directly compared to theoretical values and should help to solve the long-standing “CO/Pt(111) puzzle”: theory does not describe the bond strength on different sites correctly, yielding higher values for the bridge site, in contrast to experiments [15]. For higher total coverages above 0.35 ML, changes in the site occupation with changing pressure are observed [16].
DENECKE
Surface chemistry studied by in situ X-ray photoelectron spectroscopy
979
bridge CO is used. The results for CO adsorption are shown in Fig. 2a, left side. Adsorption starts on bridge sites, while the on-top sites are only occupied for exposures above 4 L (1 L = 1 × 10−6 Torr s, uncalibrated). The new state is populated from the beginning and saturates very quickly. The proof that both effects (site change and new state) depend on the presence of water is obtained, when the temperature is raised and D2 O desorbs. This is evident on the right side of Fig. 2a. As will discussed below, 153 K mark the desorption of water in the coadsorbate system, and at this temperature the site occupation changes back to the values expected for clean Pt(111). At the same time the new state disappears; it could be assigned to CO in hollow sites [6]. Interestingly, 153 K determined in our coadsorption experiments is not the desorption temperature of the water bilayer (which is 164 K), but rather of the multilayer species (at 152 K) [6]. This leads to the conclusion that by the presence of CO water is moved from the bilayer into multilayers. That this is indeed the case, can be seen in Fig. 2b, where results from O 1s spectra are presented after the contribution of CO has been subtracted using the coverages determined from C 1s data (Fig. 2a) (for details see [6]). Upon CO adsorption, the bilayer
C 1s spectra obtained (a) during CO uptake on clean Pt(111) at 125 K, p = 5.3 × 10−8 mbar, (b) during CO uptake under the same conditions on 1 BL D2 O on Pt(111). Panel (c) shows an example spectrum from part (b) with the model functions resulting from the fitting procedure FIGURE 1
After this short review of the pure CO adsorption, the influence of water in a coadsorbate situation will be discussed in the following. Water alone adsorbs on Pt(111) in a chemisorbed bilayer, consisting of 0.66 ML water molecules in a hexagonal structure [17]. At temperatures below 152 K, additional physisorption takes place in multilayers. In O 1s spectra, both adsorption states can be clearly resolved, with binding energies of 532.2 eV for the bilayer and 532.9 eV and higher for water in multilayers [6]. In the study presented here, the adsorption of CO on the well-defined water bilayer prepared at 152 K was investigated. As can be seen from Fig. 1b, the presence of water on the surface changes the situation remarkably. Now, the adsorption state with a C 1s binding energy of 286.1 eV is occupied first, and it is accompanied by a new state evolving at a binding energy of 285.8 eV. In order to extract quantitative information again, the intensities for the individual adsorption states are determined in a fitting procedure. Figure 1c shows an example of the resulting good agreement between model functions and raw data. The apparent asymmetry of the lines is due to an unresolved vibrational fine structure [7], similar to the discussion in Sect. 3.3. For conversion into ML units, the scaling factors derived for CO on Pt(111) have been used (see discussion above and [7]); for the new peak, the scaling factor of
FIGURE 2 (a) Quantitative results obtained from the data in Fig. 1b for CO adsorption (left side) and for thermal annealing (right side). (b) Corresponding results for the D2 O species obtained from O 1s data taken under the same conditions as in panel (a)
980
Applied Physics A – Materials Science & Processing
signal decreases and simultaneously the multilayer signal increases. Only a small amount of D2 O (∼ 0.1 BL) stays in the bilayer configuration. Upon heating, the physisorbed species desorb completely with a rate maximum at 153 K. Changing the order of adsorption, two different scenarios are observed, depending on the CO precoverage. If a complete c(4 × 2) layer of CO with a coverage of 0.5 ML is adsorbed first, only physisorption of water is found; therefore, dosing of water at 152 K does not result in any adsorption. The physisorbed water, on the other hand, does not induce any change in site occupation of the CO in the c(4 × 2) layer. If a smaller CO coverage is preadsorbed on the Pt(111) surface, the site change can be observed again [6]. This is particularly obvious, if only 0.14 ML CO is adsorbed; for this coverage only on-top sites are occupied on clean Pt(111) [7]. Figure 3 shows for this case, that D2 O can completely remove CO from on-top sites and transfers it to either bridge sites or the new adsorption state. The decrease of total CO intensity above 0.5 L is due to physisorption of water on top of the CO molecules at the temperature of 148 K used in this experiment. The additional scatterers reduce the photoelectron signal due to damping effects. If the water pressure is removed from the vacuum system, this physisorbed water desorbs again and the total CO intensity recovers (larger data points on the right side). Interestingly, also the occupation ratio between on-top and bridge/new states goes back to the value at around 0.5 L, i.e., before multilayer adsorption started. This indicates that in this case water in the layer above the CO has some influence on the site occupation, in contrast to the case on the complete c(4 × 2) layer. Concluding this section, it was shown that the presence of water modifies the adsorption behavior of CO on Pt(111) quite strongly. The observed site change should have some influence on other processes, such as the CO oxidation reaction. Indeed, it has been proposed that the low temperature reaction channel observed experimentally on Pt(111) [8, 18, 19] could be caused by coadsorbed water [20, 21]. Spectroscopic investigations in this direction are still lacking. XPS, however,
Quantitative results of the CO site occupancy for D2 O exposure ( p = 2 × 10−9 mbar) on a Pt(111) surface at 148 K precovered with 0.14 ML of CO. The largr data points on the right are taken ∼ 300 s after the end of water adsorption FIGURE 3
would not be a suitable method, due to overlapping O 1s binding energies of the species involved. 3.2
CO oxidation on Pt(111)
In surface science, the oxidation of carbon monoxide to CO2 on Pt(111) is probably the most extensively studied example of an activated Langmuir–Hinshelwood reaction [22, 23]. The common interest in this important heterogeneously catalyzed reaction is partly due to its technological relevance in automotive catalytic converters, and partly due to its relative simplicity making it an ideal model system for the understanding of surface reactions. The reaction mechanism is known to consist of four substeps [23]: (1) Adsorption of O2 and CO; (2) dissociation of O2 (ad) to 2O (ad); (3) reaction CO + O → CO2 and finally (4) desorption of CO2 . In order to investigate the kinetics of the fundamental reaction step (3) separately, it is necessary to exclude the influence of the other steps. The oxygen adsorption and dissociation process can be separated by starting out with a Pt surface covered with atomic oxygen, bound in threefold hollow sites, on which CO is dosed subsequently [24–27]. The reaction product CO2 desorbs instantaneously at the reaction temperatures of above 270 K. Therefore, the adsorption of CO remains the only limiting factor. As a common method to study reaction kinetics, mass spectrometry has been used to detect desorbing CO2 [24, 25, 28]. More recently, real time scanning tunneling microscopy (STM) experiments of Wintterlin and coworkers were reported [26, 27], resulting in an activation energy of 0.49 eV, which is found constant with respect to the oxygen partial coverage. These authors could confirm that the reaction takes place on the edges of oxygen islands. However, these experimental approaches do either not provide information about different adsorption sites or are limited to rather low temperatures (up to 274 K). Spectroscopic information on the O/CO coadsorbate phase, mainly from high-resolution electron energy loss spectroscopy (HREELS) [29, 30] and infrared absorption spectroscopy (IRAS) [31], is available from static experiments, i.e., for low temperatures, where no reaction takes place. Agreement is found, that CO adsorbs on a saturated atomic oxygen layer only on linearly coordinated (on-top) sites, whereas on the clean Pt(111) surface also twofold coordinated (bridge) sites are populated. Despite the thorough research performed so far, discrepancies in recent literature (for a detailed discussion see [8]) indicate that the CO oxidation on Pt(111) is still not completely understood and it should therefore be examined by new surface science methods. Indeed, using in situ high-resolution X-ray photoelectron spectroscopy (HR-XPS), CO oxidation has already been revisited on other surfaces, e.g. Rh(110) [32] or Pd(110) [33]. In order to study the time-dependent reaction on Pt(111), O 1s and C 1s spectra have been taken while CO was dosed by the molecular beam onto a p(2 × 2) layer of atomic oxygen, prepared by dosing molecular oxygen at 100 K, 1 × 10−7 mbar for 3 min and subsequently annealing to 300 K. To rule out effects imposed by the CO adsorption rate or by diffusion processes, a CO pressure has to be used, where the
DENECKE
Surface chemistry studied by in situ X-ray photoelectron spectroscopy
impingement rate of CO is much higher than the reaction rate. As a consequence, one would expect that bridge sites and additional on-top sites, which become available for CO adsorption after the removal of oxygen, are immediately filled and the total surface population should be saturated throughout the experiment. Thus the measured reaction rate, which is a function of the CO coverage, should reach a saturation value for high CO pressures. In the following, both the saturation of the surface population as well as the reaction rate are checked for a sample temperature of 295 K and different CO pressures. Figure 4 shows O 1s spectra taken during CO adsorption on a p(2 × 2) oxygen layer at a CO pressure of 1.3 × 10−6 mbar. Results of the quantitative analysis of this data together with the C 1s information are shown in Fig. 5. While the atomic oxygen signal decreases with reaction time, the CO bridge signal increases in an almost mirror-like fashion. Due to the relatively high pressure the CO on-top population reaches high values very quickly (∼ 5 s). The good agreement between CO coverages determined from O 1s and C 1s spectra gives confidence in the data analysis. The dashed lines to the right mark the CO site population for a system without O precoverage [16]. As a simple measure of the reaction rate, the slope of a line fitted to the oxygen signal for coverages between 0.05 and 0.15 ML is taken [8]. These values are displayed in Fig. 6 as a function of the CO gas phase pressure (filled circles, left axis). Clearly, a saturation effect is observed for pressures above ∼ 9 × 10−7 mbar, indicating that the pressure accessible by the molecular beam is indeed high enough that the reaction rate is not limited by the adsorption of CO. Note, that even for the lowest pressure studied the impingement rate of CO would nominally exceed the reaction rate. Furthermore, in the limit of zero pressure the reaction rate must go to zero. To check the saturation of the CO population with increasing pressure directly, we investigate the behaviour of the sum of CO bridge and atomic O coverage, θO , as well as the sum of the total CO and the O coverage. In Fig. 6, the averaged total
O 1s spectra taken during CO oxidation on Pt(111) at 295 K. The surface was precovered with atomic oxygen in a p(2 × 2) structure (spectrum at the bottom). CO pressure was 1.3 × 10−6 mbar
FIGURE 4
981
FIGURE 5 Quantitative results obtained from the data in Fig. 4 and corresponding C 1s data by a fitting routine. The dashed lines mark the coverages obtained for CO on clean Pt(111), adsorbed under similar conditions
CO + O population for 0.05 ML < θO < 0.15 ML is included as open squares (right axis). The data clearly show the saturation effect starting at a pressure of 2.5 × 10−7 mbar, which is significantly lower than for the reaction speed (see above). Within the margin of error an identical behaviour is encountered, if only the sum of bridge-CO and oxygen coverages is analysed [8]. The reason for the different saturation pressures in Fig. 6 could be a repulsive interaction between oxygen and CO, as proposed by Völkening et al. [27]: Since sites in the vicinity of oxygen atoms are the reactive ones, but simultaneously are energetically unfavorable for CO adsorption, the surface can be almost saturated with CO and still a considerable part of these reactive sites would be vacant. This is especially true, if the number of reactive sites is small compared to the number of unreactive adsorption sites for CO; this would be the case, if the reaction would occur only at the edges of large oxygen islands. As small differences (≤ 0.01 ML) in total occupation are not detectable in XPS, it might appear as if the surface
FIGURE 6 Pressure dependence of both reaction rate (determined from the slope in atomic oxygen O 1s data, left axis) and total CO + O coverage (right axis), for a reaction temperature of 295 K
982
Applied Physics A – Materials Science & Processing
is already saturated at a certain pressure, while the reaction speed still increases with pressure. In order to determine kinetic parameters of the CO oxidation, isothermal reaction experiments have been performed for different temperatures between 275 and 305 K, using a CO pressure of 1.3 × 10−6 mbar. The experimental values of the oxygen coverage are displayed in Fig. 7 as a function of time. In the graph, the time scale for each experiment is shifted by an offset value, such that at t = 0 all curves coincide in the point θO = 0.16 ML, which was done for the following reasons: despite equal preparation procedures of the p(2 × 2)atomic oxygen layer, the observed coverages at the start of each run exhibit some variation. In addition, an initial drop observed in O coverage due to a fast reaction channel involving disordered oxygen [8] should not influence the data analysis. For quantitative analysis of the data in Fig. 7 an appropriate rate equation has to be used. Starting with the most general case for a Langmuir–Hinshelwood reaction and using a simple first-order dependence on θO and θCO , assuming an isotropic distribution of molecules [26], we arrive at dθO/ dt = −k ∗ × θO θCO .
(1)
With k ∗ assumed to be independent of the coverages, (1) should hold for different surface coverages during the reaction, i.e., for different CO pressures. However, k ∗ is not a constant but varies also significantly as a function of reaction time within the saturation-pressure experiment [8], as was also pointed out by Wintterlin et al. [26]. One interpretation could be that k ∗ is strongly coverage dependent, as it was suggested by Zaera et al. [25]. Alternatively, the simple assumption of an isotropic distribution of adsorbates could be wrong. Based on the already mentioned observation by Wintterlin et al. that the reaction is restricted to oxygen island edges [26, 27], (1) can
be written as dθO/ dt = −k × c × θOα g (θCO ) ,
(2)
where c × θOα = θO,edge is the number of oxygen atoms at island edges per substrate atom, with α = 0.5 for smooth (nonfractal) edges, and g a still unknown function depending on θCO . c depends on the total number of islands, their shape and size distribution [8]. Thus, without knowledge of the island morphology, it is only possible to determine the product k = k × c from our measurements. The function g(θCO) in (2) accounts for the density of CO molecules on reactive sites, i.e., at the edge of oxygen islands. Fortunately, in the saturationpressure limit, it becomes greatly simplified: then all reactive sites are always fully populated and g reduces to unity. In this framework, the equation dθO/ dt = −k × θOα ,
(3)
should be suitable to evaluate our isothermal reaction data. In order to avoid a numerical calculation of the derivative and smoothing of the data curves, it seems better to use the integral form of (3), that is 1 θO = θO |t=t0 1−α − (1 − α) k (t − t0 ) 1−α ,
(4)
which can be fitted to the curves in Fig. 7. θO |t=t0 is chosen to be 0.16 ML and t0 is the time at which this coverage is observed, as mentioned above. A fit of (4) with three variable parameters, namely α, k and t0 yields a mean value of α = 0.63 ± 0.15. Within the denoted statistical error margin this result is consistent with the hypothesis of a one-dimensional reaction front, i.e., reaction at the edges of oxygen islands. The large uncertainty indicates, that the shape of the curve shows a relatively weak dependence on α. This derived value is comparable with the one of 0.55 given by Völkening et al. [27] obtained from the STM observations. In a second fit procedure α is held fixed at 0.63 for all temperatures, resulting in the curves displayed in Fig. 7 together with the data points. Now k can be used to extract the activation energy E a of the reaction by assuming an Arrhenius-type behaviour of the reaction rate, i.e., k = cν exp (−E a /kB T ) ,
(5)
where ν is the preexponential factor and c the parameter defined in (2). The inset in Fig. 7 shows k as a function of the inverse temperature as well as a fit according to (5), that results in the parameters E a = (0.53 ± 0.04)eV, cν = 4.7 × 106±0.7s−1 .
FIGURE 7 Time evolution of the atomic oxygen coverage for CO oxidation at a CO pressure of 1.3 × 10−6 mbar for varying reaction temperatures between 280 and 305 K. Solid lines are the results from a fit of the rate equation (4) to the data. The rate constants obtained are shown in the inset as Arrhenius plot, with the resulting exponential behavior given as the straight line
Using, instead, a fixed value of α = 0.5 for the fit, the results would stay within the given error bars (0.52 eV, 2.4 × 106 s−1 ). As c is generally not known (c 1 for extended islands), in any case ν alone cannot be determined. From the STM work of Völkening et al. (Fig. 1a of [27]) we estimate that c is of the order of 0.05, giving a preexponential factor of ∼ 108 s−1 , so
DENECKE
Surface chemistry studied by in situ X-ray photoelectron spectroscopy
983
by simply taking the result from the Arrhenius plot one underestimates ν by a factor of ∼ 20. In this STM study an activation energy of 0.49 V and a prefactor of 1.6 × 107 s−1 are obtained in a temperature region from 237 to 274 K [27], in good agreement with our observation. Within their quoted uncertainty, Zaera et al. [25] found comparable values of E a = 0.39 eV and ν = 2.5 × 105 s−1 in the limit of zero coverage in the range of 300 – 400 K, using, however, the incorrect assumption of an isotropic distribution of adsorbates (see discussion above and [8] for details). From density functional theory calculations, Eichler obtained an activation energy of 0.74 eV and a preexponential factor of 5 × 1012 s−1 [34]. Both values differ substantially from our and other experimental data. 3.3
Adsorption of methane on Pt(111) and Ni(111)
Methane is the main component of natural gas and thus represents a basic ingredient for the chemical industry; one important process is, e.g., the production of synthesis gas in the steam reforming process. The initial step for technologically relevant reactions is always the dissociation of CH4 , which has been studied in some detail on single crystal surfaces (for a review see [35]). Model studies on Pt(111) performed under UHV conditions show molecular adsorption at low surface temperatures [36], and completely reversible desorption with the rate maximum at 73 K. At higher surface temperatures methane adsorption is dissociative in nature, but with rather small initial sticking coefficients, S0 , e.g., of about 10−6 at 500 K [37]. By using the supersonic molecular beam for gas dosing, an initial sticking coefficient of about 3 × 10−2 on Pt(111) is found at a kinetic energy of 0.85 eV, almost independent of the surface temperature [38]. For lower kinetic energies, the sticking coefficient decreases remarkably; in addition, a further reduction is obtained for a surface temperature of 150 K as compared to 550 K [36]. By IR spectroscopy these authors identified methyl (CH3 ) as the dissociation product at 150 K. Similar findings are reported for CH4 adsorption on Ni(111). For dissociative adsorption, again a strong increase in S0 with increasing translational energy is observed [39]. The absolute values are smaller than for Pt(111). The sticking probability at 0.85 eV is 5 × 10−3 for a nozzle temperature of 550 K, but increases to 2 × 10−2 for 1050 K; this change with nozzle temperature is due to vibrational excitation of the impinging methane molecules (see [35] and references therein). Methyl is identified as the adsorbed species at 80 K [40]. Exposing the Pt(111) crystal surface at 120 K to a methane molecular beam with a kinetic energy of 0.71 eV leads to the appearance of two obvious peaks in the C 1s spectral range, as can be seen in Fig. 8a. Initially appearing at binding energies of 282.45 and 282.85 eV, both peaks shift slightly to higher binding energies with increasing coverage due to adsorbateadsorbate interactions. The binding energy of the main component at the highest coverages observed (282.59 eV) is in good agreement with values found in previous XPS studies, in which CH3 was formed by irradiating a low temperature molecular methane layer with X-rays [41, 42]. CH3 is formed on the surface independent of the kinetic energy of the incoming methane molecules in the range from 0.25 to 0.83 eV, as is evident from the identical C 1s spec-
C 1s spectra obtained after dissociative methane adsorption at 120 K, together with a deconvolution into the vibrational states. (a) CH3 on Pt(111), obtained for a kinetic energy of 0.83 eV (Tnozzle = 1273 K, 5% CH4 in He). (b) CD3 on Pt(111), obtained for a kinetic energy of 0.71 eV (Tnozzle = 1073 K, 5% CD4 in He). (c) CH3 on Ni(111), obtained for a kinetic energy of 0.67 eV (Tnozzle = 973 K, 2.5% CH4 in He) FIGURE 8
tral shape [9]. The observation that the sticking coefficient increases remarkably with kinetic energy [38], is reflected in our time-dependent XPS data in a faster growth of the total C 1s signal with higher kinetic energies [43]. The detailed fine structure observable in the C 1s spectrum is due to vibrational splitting, i.e., excitation of the C–H stretching mode in the photoemission process; similar effects have been observed in XPS of both gas phase and adsorbed hydrocarbons (see, e.g., [5, 44, 45]). By studying isotope effects, the vibrational origin of the fine structure can be convincingly shown, as obvious in the corresponding CD3 spectra in Fig. 8b. The most intense peak in both cases is the adiabatic transition (no vibrational excitation in ground and final state of the photoemission process, labelled “0–0”), while the peak at higher binding energies represents a transition to the first vibrationally excited final state (‘0–1’). The peak corresponding to the transition to the second vibrationally excited final state (“0–2”, at about 283.2 eV in the CH3 spectrum) can also be reproduced by the fits of suitable model functions to the data,
984
Applied Physics A – Materials Science & Processing
described in detail in [9]. The values obtained for the respective energy splittings are given in Table 1; within the given error bars, the values are compatible with equidistant vibrational levels. For CH3 , the “0–0” and “0–1” transitions exhibit an energy difference of 400 ± 4 meV, while a peak separation of 294 ± 5 meV is found for the CD3 species. This is in very √ good agreement with the expected decrease by roughly 1/ 2 when going from H to D [45]. A similar good agreement is found for the “0–2” states. The ratios of the peak areas, which also remain constant during the adsorption experiment, are included in Table 1. In the linear coupling model, the intensity ratio of the first vibrationally exited state and the adiabatic peak is called S factor [46]. It is proportional to the number of carbon-hydrogen bonds in the molecule, and thus is a good value for fingerprinting in the analysis of XP spectra [47]. Typical values are 0.33 for the CH3 group of gas phase methanol [45] and 0.44 to 0.49 for methane in gas phase [47]. The S factor of 0.50 obtained as an average of the experiments here, independent of the kinetic energy of the impinging molecules, is much higher than expected from gas phase results and is also larger than observed for CH3 groups in adsorbed molecules. A possible reason could be the strong bonding of methyl on Pt(111), changing the internal charge distribution. Using the linear coupling model, an intensity ratio between the second excited and the adiabatic peak of 0.13 is derived. The value of 0.09 ± 0.03 determined from our fitting results deviates slightly from this ratio, indicating some anharmonicity of the potential. When going √ from CH3 to CD3 the S factor increases accordingly by about 2, reaching a value of about 0.69 ± 0.03. The intensity ratio between second excited peak and the adiabatic transition is 0.22, in agreement with the expected value from the linear coupling model. Looking at the spectrum in Fig. 8c taken after exposing Ni(111) to CH4 , the smaller splitting between the adiabatic and excited states, as compared to the Pt(111) data, is quite obvious. A splitting of only 356 meV is found, with an S factor of 0.52, very similar to the value for Pt(111). The observed energy splitting on Pt(111) agrees very well with gas phase XPS data of methane [44, 47], where values of 400 meV are found. Thus, no C−H mode softening upon adsorption is observed for methyl on Pt(111) [48]. In contrast to that, the significantly lower value of 356 meV indicates C−H mode-softening for Ni(111), as was also proposed from vibrational spectroscopy [40]. DFT calculations suggest that CH3 is bound in an on-top site on Pt, while they predict a hollow
site for Ni and Cu [49]; at the on-top site, the H atoms have less possibility to interact with the substrate electron density, which could explain the absence of mode softening for Pt(111). Thus, both from the observed energy splitting and the intensity ratio, methyl on Pt(111) displays the most clearly resolved vibrational fine structure found in adsorbate systems, making it a model system. In addition, in larger hydrocarbons other vibrational modes might interfere with the C–H stretching mode, e.g., as in ethylene on Ni(100) [50]. To study the thermal evolution of the adsorbed CH3 layers, the crystals were heated to 500 K at a linear heating rate of 0.5 K/s, using a filament at the back of the sample. Simultaneously, XP spectra were acquired, in average every 10 K, to minimize radiation damage. The spectra for T > 500 K have been obtained after heating to the denoted temperatures by direct resistive heating. Significant changes occur upon increasing the temperature, as can be seen from the C 1s raw data in Figs. 9a, b and c for Pt(111) and Ni(111), respectively. While Fig. 9a shows two example spectra for the Pt(111) surface, the intensities of whole sets of C 1s spectra are displayed in Figs. 9b and c as grey scales values (low intensity = white, high intensity = black). The spectra have been analysed using a fitting procedure as described in [9], and the resulting thermal evolution of various species is presented in Figs. 10a and b for Pt(111) and Ni(111), respectively. For Pt(111) (Fig. 10a), upon heating up to 240 K, only small changes in the spectra are observed. At about 240 K
CH3 /Pt(111) CD3 /Pt(111) CH3 /Ni(111) Energy splitting (meV) (adiabatic-1st excited2nd excited) S factor (1st excited/adiabatic) Energy splitting (meV) (1st excited - 2nd) Intensity ratio (2nd excited/adiabatic)
400 ± 4
294 ± 5
356 ± 4
0.50 ± 0.04
0.69 ± 0.03
0.52 ± 0.04
386 ± 18
304 ± 5
351 ± 5
0.09 ± 0.03
0.22 ± 0.02
0.09 ± 0.03
TABLE 1 Energy splittings (in meV) and intensity ratios for the vibrationally resolved peaks of CH3 adsorbed on Pt(111) and Ni(111), as well as CD3 /Pt(111), as derived from peak fitting spectra as shown in Fig. 8
Two example spectra (a) and series of C 1s spectra taken during temperature-programmed XPS (heating rate = 0.5 K/s) after dissociative adsorption of methane at 120 K on Pt(111) (a) + (b) and Ni(111) (c). Intensities in (b) and (c) are displayed as grey scales (low intensity = white, high intensity = black), spectra are taken every 10 K
FIGURE 9
DENECKE
Surface chemistry studied by in situ X-ray photoelectron spectroscopy
985
Between 500 and 540 K, decomposition of CH to carbon occurs, as is indicated by the drop of the CH peaks in Figs. 9b and 10a, and the increase of new C 1s peaks at higher binding energies. Also, a prominent H2 peak is observed in TPD in this temperature range [55]. The observed C 1s binding energies are consistent with isolated C (283.8 eV) or so-called chain carbon (284.1 eV) [56]. In the temperature range studied here, the literature value for graphitic carbon (284.6 eV) has not been observed [56], possibly due to the low carbon coverage. For the Ni(111) surface, the methyl dehydrogenation reaction is more complex [40]. This is apparent in Figs. 9c and 10b. Around 170 K, CH is formed with its maximum coverage at 220 K, but already at 240 K a new C 1s binding energy position signals the appearance of a new species. This species is assigned to acetylene (C2 H2 ) formed in a C–C coupling reaction, from a comparison to the spectra of molecular acetylene adsorbed in a separate experiment at 110 K (data not shown here). This assignment confirms a previous study by HREELS [40]. For temperatures above 420 K complete dehydrogenation is observed. As can be seen from Fig. 10b, no significant desorption takes place, in contrast to the situation on Pt(111). FIGURE 10 Results of a quantitative analysis of the data shown in Figs. 9b and c, showing the relative intensities of the various species on a common scale. The drop in intensity for the Pt(111) system in panel (a) marks recombinative methane adsorption, while no obvious desorption is observed for the Ni(111) surface (panel (b))
a sudden decrease of the CH3 species and at 265 K an increase of the peaks at 283.61 and 284.02 eV, which have been assigned to CH, occurs. The same species can also be generated by radiation damage [9]. The total C 1s intensity decreases by around 75% and a CH4 desorption peak is observed in TPD spectra, in agreement with [55]. This indicates that in this temperature range recombinative desorption of CH3 and hydrogen takes place, emphasizing again the assignment of the initially formed species to CH3 . Note, that the amount of desorbing CH4 depends on the initial CH3 coverage [55]; it ranges from 0 to 75% for low and high initial coverages, respectively. This finding suggests that CH3 dissociation only occurs if a sufficient number of adsorption sites is empty. CH (methylidyne) formation has also been proposed for Pt(110) [51]. For Pt(111), however, the formation of ethylidyne (CCH3 ) from CH3 produced from hyperthermal methane was proposed from vibrational spectroscopy as the next step in thermal chemistry [52]. A similar conclusion is drawn for methyl produced directly on Pt(111) by a methyl source [53], but only for methyl coverages above 0.3 ML (the saturation coverage is reached at 0.45 ML); below 0.3 ML, dehydrogenation to CH is reported to be the preferred pathway. Furthermore, coadsorption of hydrogen, as in our case from the initial dissociation, enhances the dehydrogenation rate [53]. By comparison with CO adsorption, we estimate that our maximum methyl coverage is always below 0.25 ML and, therefore, no ethylidyne should be observed, in agreement with our observations. A separate study of the thermal evolution of C2 H4 [43], where it is well established that ethylidyne is formed on the surface [54], yields a different C 1s binding energy than observed here for CH.
4
Summary
Using three examples of simple surface reactions, the potential of a new experimental apparatus, combining high-resolution X-ray photoelectron spectroscopy at synchrotron sources with a supersonic molecular beam for gas dosing, have been demonstrated. Time-dependent and quantitative spectroscopic data not only allow to determine the kinetics of surface processes, but are also very useful in detecting and identifying intermediate species. Comparing the coadsorption of water and CO with the pure CO adsorption on Pt(111), interesting details could be revealed. The presence of water changes the preferred CO adsorption site from on-top to bridge and leads to a new feature in C 1s core-level spectra, which is suggested to be CO adsorbed in hollow sites. Water is replaced by adsorbing CO from the bilayer into multilayers, from where it desorbs upon heating, restoring the original CO site distribution known from clean Pt(111). The sticking coefficient of CO on a bilayer of water is about 13 times smaller than that found on clean Pt(111) [9]. In order to determine the kinetics of the central reaction step of CO oxidation on Pt(111), namely the reaction between adsorbed atomic oxygen and CO to CO2 , which desorbs immediately at the reaction temperature, other steps like the adsorption of CO or the dissociation of oxygen need to be taken into account. One way to be independent of these steps is by starting with a surface covered with atomic oxygen and working with a saturation supply of CO. The latter can be achieved by using the high local pressure of 1.3 × 10−6 mbar from a molecular beam, while keeping the UHV environment in the analysis chamber. Under these conditions, an activation energy of 0.53 eV was found for the oxidation step, using a rate equation derived from quantitative O 1s data, which only accounts for reactive sites at edges of oxygen islands, in agreement with previous STM results [26, 27]. The ap-
986
Applied Physics A – Materials Science & Processing
parent preexponential factor derived from the present data (4.7 × 106 s−1 ) includes a factor describing the oxygen island morphology and density. Taking this information from the STM work [27], yields a preexponential factor of ∼ 108 s−1 . Methane adsorption on Pt(111) or Ni(111) is either molecular and weak at low temperatures, or dissociative at high gas pressures and temperatures. Tuning the kinetic and vibrational energy of the methane molecules by using a supersonic molecular beam, dissociative adsorption can be observed under UHV conditions and at low surface temperatures. For this case, the chemical nature of the adsorbed species has been determined as CH3 for both surfaces, independent of kinetic energy. The C 1s spectra reveal a fine structure, which is due to vibrational excitation in the photoemission process. A detailed analysis of this vibrational fine structure gives confidence in the identification of the species. While the vibrational energy splitting for CH3 on Pt(111) is almost identical to that observed in gas phase XPS studies, the expected mode-softening is found on the Ni(111) surface. Temperature-programmed XPS investigations of the adsorbed species confirm different dehydrogenation pathways. While on Pt(111) only CH is found as intermediate before complete dehydrogenation, a C–C coupling reaction on Ni(111) yields adsorbed acetylene from the CH species; this acetylene dehydrogenates completely in one step for higher temperatures. The reported combination of techniques promises to be a very useful tool in the study of surface reactions. Further improvements on the time-resolution will increase the ranges of accessible parameters. Combined with simultaneous gas phase analysis a complete picture of surface processes can be obtained. ACKNOWLEDGEMENTS I would like to thank M. Kinne for careful assembly and testing of the complicated apparatus. M. Kinne and T. Fuhrmann collaborated closely in performing the experiments and the data analysis presented here. The rest of the synchrotron group had a great share in the successful outcome of the various beamtimes. I thank Hans-Peter Steinr¨uck for his continuing support, many stimulating discussions, and for the close collaboration. Support by the BESSY staff during the beamtimes and by the workshop in Erlangen in assembly and maintenance is gratefully acknowledged. The work was supported by the DFG (Ste 620/4-2).
REFERENCES 1 A. Baraldi, G. Comelli, S. Lizzit, M. Kiskinova, G. Paolucci: Surf. Sci. Rep. 49, 169 (2003) (and references therein) 2 A. Baraldi, L. Rumiz, M. Moretuzzo, M. Barnaba, G. Comelli, S. Lizzit, G. Paolucci, R. Rosei, F. Buatier de Mongeot, U. Valbusa: J. Vac. Sci. Technol. A 20, 683 (2002) 3 Y. Teraoka, A. Yoshigoe: Appl. Surf. Sci. 216, 8 (2003) (and references therein) 4 R. Denecke, M. Kinne, C.M. Whelan, H.-P. Steinrück: Surf. Rev. Lett. 9, 797 (2002) 5 J.N. Andersen, A. Beutler, S.L. Sorensen, R. Nyholm, B. Setlik, D. Heskett: Chem. Phys. Lett. 269, 371 (1997) 6 M. Kinne, T. Fuhrmann, B. Tränkenschuh, J.F. Zhu, R. Denecke, H.-P. Steinrück: Langmuir 20, 1819 (2004) 7 M. Kinne, T. Fuhrmann, C.M. Whelan, J.F. Zhu, J. Pantförder, M. Probst, G. Held, R. Denecke, H.-P. Steinrück: J. Chem. Phys. 117, 10 852 (2002) 8 M. Kinne, T. Fuhrmann, J.F. Zhu, C.M. Whelan, R. Denecke, H.-P. Steinrück: J. Chem. Phys. 120, 7113 (2004)
9 T. Fuhrmann, M. Kinne, C.M. Whelan, J.F. Zhu, R. Denecke, H.-P. Steinrück: Chem. Phys. Lett. 390, 208 (2004) 10 C.T. Rettner, H.E. Pfnür, D.J. Auerbach: Phys. Rev. Lett. 54, 2716 (1985) 11 S. Kneitz, J. Gemeinhardt, H.-P. Steinrück: Surf. Sci. 440, 307 (1999) 12 H. Froitzheim, H. Hopster, H. Ibach, S. Lehwald: Appl. Phys. 13, 147 (1977) 13 H. Steininger, S. Lehwald, H. Ibach: Surf. Sci. 123, 264 (1982) 14 J.-S. McEwen, S.H. Payne, H.J. Kreuzer, M. Kinne, R. Denecke, H.-P. Steinrück: Surf. Sci. 545, 47 (2003) 15 P.J. Feibelman, B. Hammer, J.K. Nørskov, F. Wagner, M. Scheffler, R. Stumpf, R. Watwe, J. Dumesic: J. Phys. Chem. B 105, 4018 (2001) 16 M. Kinne: Ph.D. thesis, Universität Erlangen-Nürnberg, 2004 17 D.L. Doering, T.E. Madey: Surf. Sci. 316, 305 (1982) 18 J. Yoshinobu, M. Kawai: J. Chem. Phys. 103, 3220 (1995) 19 K.-H. Allers, H. Pfnür, P. Feulner, D. Menzel: J. Chem. Phys. 100, 3985 (1994) 20 J. Bergeld, B. Kasemo, D.V. Chakarov: Surf. Sci. 495, L815 (2001) 21 X.-Q. Gong, P. Hu, R. Raval: J. Chem. Phys. 119, 6324 (2003) 22 R.L. Palmer, J.N. Smith: J. Chem. Phys. 60, 1453 (1974) 23 T. Engel, G. Ertl: Adv. Catal. 28, 1 (1979) 24 M.D. Xu, J.Y. Liu, F. Zaera: J. Chem. Phys. 104, 8825 (1996) 25 F. Zaera, J.Y. Liu, M.D. Xu: J. Chem. Phys. 106, 4204 (1997) 26 J. Wintterlin, S. Völkening, T.V.W. Janssens, T. Zambelli, G. Ertl: Science 278, 1931 (1997) 27 S. Völkening, J. Wintterlin: J. Chem. Phys. 114, 6382 (2001) 28 J.L. Gland, E.B. Kollin: J. Chem. Phys. 78, 963 (1983) 29 J.L. Gland, E.B. Kollin: Surf. Sci. 151, 260 (1985) 30 K.L. Kostov, P. Jakob, D. Menzel: Surf. Sci. 377, 802 (1997) 31 J. Yoshinobu, M. Kawai: J. Chem. Phys. 103, 3220 (1995) 32 A. Baraldi, S. Lizzit, D. Cocco, G. Comelli, G. Paolucci, M. Kiskinova, R. Rosei: Surf. Sci. 385, 376 (1997) 33 I.Z. Jones, A. Bennet, M. Bowker: Surf. Sci. 439, 235 (1999) 34 A. Eichler: Surf. Sci. 498, 314 (2002) 35 J.F. Weaver, A.F. Carlsson, R.J. Madix: Surf. Sci. Rep. 50, 107 (2003) 36 A.F. Carlsson, R.J. Madix: Surf. Sci. 458, 91 (2000) 37 A.C. Luntz, H.F. Winters: J. Chem. Phys. 101, 10 980 (1994) 38 D.J. Oakes, M.R.S. McCoustra, M.A. Chesters: Faraday. Discuss. 96, 325 (1993) 39 M.B. Lee, Y.Y. Yang, S.L. Tang, S.T. Ceyer: J. Chem. Phys. 85, 1693 (1986) 40 Q.Y. Yang, K.J. Maynard, A.D. Johnson, S.T. Ceyer: J. Chem. Phys. 102, 7734 (1995) 41 Y. Matsumoto, Y.A. Gruzdkov, K. Watanabe, K. Sawabe: J. Chem. Phys. 105, 4775 (1996) 42 R. Larciprete, A. Goldoni, A. Groso, S. Lizzit, G. Paolucci: Surf. Sci. 482, 134 (2001) 43 T. Fuhrmann, M. Kinne, B. Tränkenschuh, J.F. Zhu, R. Denecke, H.-P. Steinrück: submitted to New J. Phys. 44 T.X. Carroll, N. Berrah, J. Bozek, J. Hane, E. Kukk, L.J. Saethre, T.D. Thomas: Phys. Rev. A 59, 3386 (1999) 45 M. Wiklund, A. Jaworowski, F. Strisland, A. Beutler, A. Sandell, R. Nyholm, S.L. Sorensen, J.N. Andersen: Surf. Sci. 418, 210 (1998) 46 L.S. Cederbaum, W. Domcke: J. Chem. Phys. 64, 603 (1976) 47 L.J. Saethre, O. Svaeren, S. Svensson, S. Osborne, T.D. Thomas, J. Jauhiainen, S. Aksela: Phys. Rev. A 55, 2748 (1997) 48 N. Sheppard, C. De la Cruz: Adv. Catal. 42, 181 (1998) 49 A. Michaelides, P. Hu: J. Chem. Phys. 114, 2523 (2001) 50 R. Neubauer, C.M. Whelan, R. Denecke, H.P. Steinrück: J. Chem. Phys. 119, 1710 (2003) 51 D.T.P. Watson, S. Titmuss, D.A. King: Surf. Sci. 505, 49 (2002) 52 D.J. Oakes, H.E. Newell, F.J.M. Rutten, M.R.S. McCoustra, M.A. Chesters: J. Vac. Sci. Technol. A 14, 1439 (1996) 53 D.H. Fairbrother, X.D. Peng, R. Viswanathan, P.C. Stair, M. Trenary, J. Fan: Surf. Sci. 285, L455 (1993) 54 N. Sheppard, C. De La Cruz: Adv. Catal. 41, 1 (1996) 55 D.J. Oakes, H.E. Newell, F.J.M. Rutten, M.R.S. McCoustra, M.A. Chesters: J. Vac. Sci. Technol. A 14, 1439 (1996) 56 N.M. Rodriguez, P.E. Anderson, A. Wootsch, U. Wild, R. Schlögl, Z. Paal: J. Catal. 197, 365 (2001)