Ethene Adsorption and Decomposition on the Cu(410) Surface - The

Nov 11, 2009 - Corresponding author. E-mail: [email protected]., †. CNISM and Dipartimento di Fisica, Università di Genova. , ‡. Universit...
0 downloads 0 Views 4MB Size
J. Phys. Chem. C 2009, 113, 20881–20889

20881

Ethene Adsorption and Decomposition on the Cu(410) Surface Tatyana Kravchuk,†,| Vinay Venugopal,† Luca Vattuone,*,† Luke Burkholder,‡ Wilfred T. Tysoe,‡ Marco Smerieri,† and Mario Rocca§ CNISM, IMEM-CNR, and Dipartimento di Fisica, UniVersita` di GenoVa, 16146 GenoVa, Italy, and Department of Chemistry and Biochemistry, UniVersity of WisconsinsMilwaukee, Milwaukee, Wisconsin 53211 ReceiVed: May 22, 2009; ReVised Manuscript ReceiVed: October 6, 2009

The influence of open steps on the surface properties is shown by investigating the interaction of molecular ethene with Cu(410). We find a surprisingly low-temperature, site-selective chemistry at the strongly undercoordinated step sites. Ethene bonds either in a π-bonded or in a di-σ-bonded state or undergoes complete dehydrogenation. All pathways involve the low-coordination sites at the step, since the first species is partially stabilized with respect to low-Miller-index surfaces, while the other two are observed only on Cu(410). When annealing the surface, dehydrogenation and transformation into the di-σ-bonded moiety proceed, both processes being favored by faster heating rates. The so-generated carbon (presumably C2 admolecules) decorates the step edges, thereby blocking the active sites for subsequent dissociation and permitting only π-bonding of ethene. The dipole loss of carbon disappears in high-resolution electron energy loss spectroscopy when annealing to room temperature, indicating that carbon moves to more coplanar or even to subsurface sites where it still influences the surface chemistry. The surface reactivity is recovered when heating the crystal to 900 K since C dissolves then deep enough into the bulk. 1. Introduction The improvement of industrial catalysts for heterogeneous reactions is strongly motivated by economical reasons. Knowledge-based designs to improve efficiency and/or selectivity may be achieved by a thorough fundamental understanding of the action of the catalyst. Using surface science methods and techniques, we can investigate the elementary steps, identify the active sites, and unravel the reaction mechanisms at the nanoscale. Although the importance of special surface active sites for catalysis has been known for many years, only a few investigations deal with the influence of under-coordinated sites on reaction selectivity.1,2 Controlling such a parameter is a central aspect for hydrocarbon chemistry on metal surfaces, which is important for both heterogeneous catalysis and nanotechnology. Carbon nanotube synthesis is based on the formation of C-C bonds and C-H scission,3 whereas the inhibition of carbon growth (C-C bond breaking) is important for preventing the poisoning of industrial catalysts.4 In the steam re-forming process, typically on Ni-based catalysts, hydrocarbons dissociate on the surface to form molecular hydrogen, and the remaining carbon reacts to produce carbon monoxide and additional hydrogen.4 However, a severe problem with nickel catalysts is that carbon can accumulate to form graphitic deposits, which ultimately deactivate the catalyst.1 For an improved catalyst, carbon removal should be facilitated, a process which would be thermodynamically favored by weaker metal-carbon bonding. Copper-based catalysts may be good candidates for achieving this goal, since the reactivity of copper exceeds that of silver and gold but is far lower than that of main-group transition metals. We focused our interest on the * Corresponding author. E-mail: [email protected]. † CNISM and Dipartimento di Fisica, Universita` di Genova. ‡ University of WisconsinsMilwaukee. § IMEM-CNR and Dipartimento di Fisica, Universita` di Genova. | Current address: Schulich Faculty of Chemistry, Technion - Israel Institute of Technology, Haifa, 3200 Israel.

adsorption of a small hydrocarbon molecule, ethene, on Cu(410), a surface characterized by well-defined, open steps, organized in an ordered staircase. It is known from previous work that ethene adsorbs reversibly on low-Miller-index Cu surfaces in a π-bonded state.5-7 However, recent studies showed that adsorption8 and decomposition9 can occur even on gold in the presence of defects (steps, adatoms, rough single crystals). Similarly, ethene was demonstrated to be stabilized at the open step edges of Ag(410),10 its adsorption energy being 3 times larger than on the (100) terraces. Detailed cluster calculations on the C2H4-Cu(111) interaction potential revealed two potential energy minima at different adsorbate-substrate distances, separated by an activation barrier.11 The first corresponds to the π-bonded species observed experimentally. The other is due to a strongly distorted, chemisorbed C2H4 moiety whose existence was excluded experimentally, at least for pristine Cu(111).5 The coexistence of such species was, however, suggested to occur for real systems11 and was experimentally observed to be possible for other sample preparation conditions.12 Di-σ-bonded ethene was indeed detected on Cu(100) as a result of photoinduced dissociation of CH2I2 in temperatureprogrammed desorption (TPD) as evidenced by its higher (240 K) desorption temperature12 relative to that for the π-bonded form (120 K).5 In spite of the intensive studies of ethene adsorption on metal surfaces, this remains a complex system because a number of rather subtle factors determine the processes that actually occur, surface defects being just one of them. In this paper, we investigate adsorption of ethene on Cu(410) at low temperatures by high-resolution electron energy loss spectroscopy (HREELS), X-ray photoelectron spectroscopy (XPS), and TPD. The ethene bonding and reaction were analyzed as a function of temperature and coverage. While ethene is reversibly π-bonded on the (100)-like terraces, the step sites induce not only di-σ-bonding but also complete ethene

10.1021/jp904794n CCC: $40.75  2009 American Chemical Society Published on Web 11/11/2009

20882

J. Phys. Chem. C, Vol. 113, No. 49, 2009

Kravchuk et al.

dehydrogenation even at low temperatures. The resulting carbon decorates the steps edges. A short report on these results was published in ref 13. 2. Experimental Details The results presented here were obtained in two different laboratories using three different experimental setups. HREELS experiments were performed at the University of Genoa, Italy. The ultrahigh vacuum (UHV) system (base pressure ∼5 × 10-11 mbar) is equipped with a home-built HREEL spectrometer, a low-energy electron diffractometer (LEED), a cylindrical-mirror analyzer for Auger spectroscopy, an ion gun for in situ sample cleaning, and a supersonic molecular beam.14 HREEL spectra were recorded using specular detection at an incident angle of ∼67° and with a low incident electron energy of 2.9 eV. Ethene was dosed either by backfilling the chamber or by a supersonic molecular beam in the temperature range from 125 K to room temperature. The beam allows the direction of the gas-phase molecules as well as their kinetic energy to be controlled.14 The rotational population in the beam is also very different for beam and backfilling experiments, since this degree of freedom is efficiently cooled by the adiabatic expansion occurring in the former case. The temperature dependence of the adsorbed layer was also monitored by HREELS, heating the crystal to the indicated temperature, and recording the spectrum after cooling the sample to the lowest temperature attainable in each experiment (from 125 to 150 K). XPS and TPD investigations were performed at the University of WisconsinsMilwaukee, using two different chambers. The sample cooling system typically allowed the sample to reach ∼110 K. Exposures were performed at this temperature by backfilling only. The XPS chamber was operated at a base pressure of 7 × 10-10 mbar and equipped with a SPECS X-ray source and double-pass cylindrical mirror analyzer. Spectra were typically collected using a Mg KR X-ray source, with 250 W power and an analyzer pass energy of 50 eV. Temperaturedependent XPS data were collected after increasing the sample temperature to 300 K and keeping it constant during the measurement. The TPD data were collected with a linear heating rate in another UHV chamber with base pressure of 7 × 10-10 mbar. The desorbing species were detected using a Dycor quadrupole mass spectrometer placed in line of sight of the sample and shielded from the rest of the chamber so that only molecules leaving the sample surface are detected. The chamber was also equipped with a double-pass, cylindrical-mirror analyzer for Auger electron spectroscopy and an ion gun for surface preparation. The Cu(410) single crystal is a 10 mm diameter disk, oriented to within 0.1° of the (410) plane. The surface was prepared by cycles of sputtering with Ne+ or Ar+ ions followed by annealing for 3 min to ∼900 K, until spectra indicative of the clean surface were observed by HREELS or by Auger spectroscopy. The surface geometry of Cu(410) is shown in Figure 1 and consists of three-atom-row wide (100) terraces alternating with monatomic steps running in the 〈001〉 direction and thus showing an open geometry. 3. Results 3.1. TPD Data. TPD spectra of deuterated ethene (m/e ) 32) from Cu(410) are presented in Figure 2 for sequential doses of 0.25 L performed by backfilling the chamber at a sample temperature of 110 K (1 L ) 1 × 10-6 Torr s, equivalent to 0.15 ML; 1 ML corresponds on Cu(410) to 1.53 × 1015 molecules cm-2). The heating rate was 4.3 K/s up to a maximum

Figure 1. Geometry of the Cu(410) surface.

temperature of 440 K. The spectra reported in panel a were collected by dosing the surface with ethene by backfilling, collecting the TPD spectrum, and then dosing the surface again without cleaning it. For the first dose, two well-defined desorption states are detected at ∼143 and ∼236 K. Note that the sharp onset of the low-temperature (∼143 K) state suggests that the peak desorption temperature is less than 143 K. The well-separated peaks suggest that two distinct surface species are present. On the basis of previously published results, the low-temperature peak is associated with π-bonded ethene,5 while the feature at 236 K assigned to di-σ-bonded molecules.12 The latter corresponds reasonably to ethene located at the step, since this species is not observed on low-Miller-index surfaces. The broad desorption peak at lower temperatures arises both from ethene at terraces (since a value of 140 K was reported for ethene desorption from Cu(100) at heating rates of 2.5 K/s15,16) and at steps (since for sputtered Cu(111) stabilization of the π-bonded molecules from 9017 to 190 K was found). Subsequent ethene doses result in a decrease in the intensity of the high-temperature peak and in the appearance of a desorption feature at lower temperature. No further changes were observed after six cycles. The ethene desorption states are then at ∼143 and ∼169 K, respectively. The sample was finally annealed to 900 K in vacuo, cooled to 110 K, and exposed again to ethene. The resulting TPD spectrum then recovered the shape observed for desorption from the clean surface. The disappearance of the high-temperature feature indicates that ethene adsorption and desorption cycles modify the surface, deactivating the sites responsible for di-σ-bonding. It is evident from Figure 2b that heating to higher temperatures gradually restores the initial situation. Thus, the surface modification survives when annealing to 535 K, but it is eliminated when heating to 900 K. From the desorption temperature, the species leaving the surface at 169 K is reasonably ascribed to π-bonded ethene, but in another, more stable configuration. Most probably it consists of ethene located close to the deactivated step sites. As shown in the inset to Figure 2a, the TPD area is nearly constant for subsequent cycles. This indicates that the initial sticking probability is little affected by the modification of the surface, showing only a limited increase after several cycles. A detailed investigation of sticking probability by retarded reflector measurements is reported in the accompanying paper.18 The results presented later in the paper point out that this π-bonded ethene is stabilized by carbon which decorates the step edges after ethene dissociation. Such carbon inhibits diσ-bonding of the ethene but stabilizes π-bonded molecules adsorbed next to it. This conclusion is supported by our HREELS results.

Ethene Adsorption on Cu(410)

J. Phys. Chem. C, Vol. 113, No. 49, 2009 20883

Figure 3. TPD spectra of ethene (m/e ) 27) from Cu(410) for different exposure at 110 K. Each spectrum was collected using a heating rate of 3.6 K/s after dosing ethene on a sample that had been precovered with ethene and annealed to 900 K. Inset: the total area of TPD spectra as a function of exposure.

Figure 2. Sequential TPD spectra of 0.25 L exposure (corresponding to 0.063 ML) of ethene (m/e ) 32) on Cu(410). The heating rate was 4.3 K/s up to a maximum temperature of 440 K. (a) Spectra for sequential experiments in which the surface was exposed to ethene at 110 K, flashed to 440 K during TPD, cooled again to 110 K, and exposed for the next measurement without cleaning the sample. (b) The crystal was pre-exposed to 0.25 L of ethene, heated to the reported temperature, cooled again to 110 K, and exposed to a new dose of ethene for which we show the desorption spectrum. Inset of the top panel shows the area of the TPD spectra as a function of dose.

Figure 3 displays TPD spectra for different exposures of ethene (C2H4, m/e ) 27) on Cu(410) at 110 K by backfilling the chamber. The sample was cleaned by heating to 900 K and cooled again to 110 K before each dose. For the first dose (0.025 L), two desorption states are evident at 143 and 248 K. These peaks shift with ethene exposure to 148 and 223 K at the largest dose (1 L). The low-temperature peak, associated with π-bonded ethene, shows only a relatively small shift as a function of exposure, in accord with previous measurements on the ethene-Cu(111) system.5 The high-temperature feature appearing between 223 and 248 K is due to di-σ-bonded ethene. The shift could be induced by carbon accumulation on the surface as a result of dehydrogenation at larger ethene exposures, since the data in Figure 2 show that the former has a strong influence on the second desorption peak.

For larger exposures, a gradual growth of both peaks was detected. Saturation of the high-temperature feature occurs at ∼0.5 L. The total TPD area is plotted in the inset of Figure 3 as a function of exposure. It reveals a linear increase in coverage with exposure, i.e., a constant sticking probability, in agreement with the results presented in the accompanying paper.18 3.2. HREELS Data. To further investigate the nature of the ethene species on the surface, HREELS measurements were performed for ethene/Cu(410). The resulting spectra, presented in Figure 4, were recorded after dosing ethene by backfilling the chamber at various sample temperatures. The energy loss peaks at 935 cm-1 [ν7(CH2)], 1290 cm-1 [ν3(CH2)], 1569 cm-1 [ν(CC)], 3018 cm-1 [ν1(CH2)], and 3134 cm-1 [ν9(CH2)] at 145 K are the signatures for π-bonded ethene,5 and the frequencies correspond to a σ-π-parameter of ∼0.1-0.2.19 Since these experiments were performed by dosing at 145 K, these results are consistent with the presence of π-bonded ethene (Figure 2). Essentially no intensity is present at ∼1150 cm-1 [ν(CC)]20 and ∼2900 cm-1 [ν1(CH2)]20 (Figure 4 spectrum a), which would be typical of di-σ-bonded ethene. This suggests that either the di-σ-bonded species is initially absent on the surface and its formation is thermally activated or its intensity is too low to be detected. In order to check whether it forms at higher dosing temperatures, the surface was exposed to a larger dose of ethene at 193 K (Figure 4 spectrum b), a temperature at which most of the π-bonded moiety has already desorbed, while the di-σbonded one should still be stable (see Figure 2). The resulting HREELS spectrum, recorded after cooling the sample back to 150 K, shows a broad feature at ∼935 cm-1, intensity at 1123 cm-1, and a peak at ∼1330 cm-1. The feature at ∼935 cm-1 is relatively insensitive to the hybridization state of ethene,19 but this mode is dominant in the case of π-bonding. As we will show later on and in the accompanying paper,18 two different π-bonded states, π0 and π1, characterized by different binding energies, must be present at the surface. Only π1 is, however,

20884

J. Phys. Chem. C, Vol. 113, No. 49, 2009

Figure 4. Normalized HREELS spectra of ethene adsorbed by backfilling on clean Cu(410) at different temperatures. Spectrum a corresponds to a dosing temperature of 145 K. The losses correspond to π-bonded ethene (see the text for the assignment of the vibrational frequencies). Spectra b and c correspond to a dose of 51 L of C2H4 at 193 K and of 116 L dose at room temperature, respectively. At 193 K the shift of the losses, from 1290 and 1569 cm-1 to 1123 and 1330 cm-1, indicates an increased σ-π-parameter. Some π-bonded ethene must be present to explain the intensity at 935 cm-1. The weaker relative intensity of the CH stretch around 3000 cm-1 indicates, however, that these admolecules are bonded in a different configuration than when dosing at lower T. Moreover, at 193 K dosing temperature, the loss at 388 cm-1, witnessing ethene dissociation, dominates the spectrum. No adsorption takes place at 300 K. The weak loss at 1876 cm-1 in the 145 K spectrum is due to traces of CO.

significantly populated at the temperatures at which ethene is dosed in HREELS experiments. However, an increase in the value of the σ-π parameter significantly decreases the vibrational frequencies of the 1569 and 1290 cm-1 modes, shifting them to the observed values of ∼1330 and 1123 cm-1.19 Due to the hybridization of the two modes, the lower frequency (1123 cm-1) now corresponds mainly to the CC stretching mode. The measured spectrum is thus compatible with the presence of di-σ-bonded ethene, although the intensity of the 935 cm-1 feature indicates that π-bonded molecules are present, too. Since the dipole moments of the energy losses associated with the π- and σ-bonded moieties are comparable,21 the weakness of the latter in spectrum b is indicative of a lower coverage. The absence of all of the energy loss features on dosing at 300 K (Figure 4c) is consistent with the TPD data (Figure 2). The formation of di-σ-bonded ethene must therefore be mediated by a precursor that becomes short-lived at higher temperatures. A loss also appears at 388 cm-1, which now dominates the spectrum (Figure 4b). Interestingly, such a loss is absent when dosing ethene at 300 K (HREELS recorded at 300 K). Thus, the species responsible for this feature must also form via a precursor mechanism. To understand the origin of the loss at 388 cm-1, we performed HREELS measurements as a function of the annealing temperature. In this experiment, the results of which are displayed in Figure 5, the sample was exposed to a supersonic

Kravchuk et al. molecular beam. This has the advantage of preserving a better vacuum in the sample chamber and decreases the readsorption probability. Since ethene was seeded in Ne, the molecules in the gas phase have the relatively low kinetic energy of 0.10 eV. The beam was incident at a grazing angle along the step down direction. The spectrum obtained immediately after adsorption at 145 K shows the vibrational signatures of π-bonded ethene. As shown in Figure 5a on flashing to 250 K this moiety disappears while a loss at 1151 cm-1, indicative of di-σbonded admolecules, is now apparent. These molecules leave the surface after heating above 250 K, in agreement with the TPD results of Figure 2. The 388 cm-1 loss persists up to 500 K and thus cannot be due to ethene. Since no CHrelated losses are observed upon annealing and since it is unlikely that the CC bond will cleave, we ascribe this mode to the C2-substrate vibration, in accord with previous work.20,22 The higher kinetic energy of the impinging ethene, compared to backfilling, favors the dehydrogenation reaction when dosing at low temperatures. An additional feature is observed in Figure 5 at 597 cm-1. This energy coincides with that of H-Cu(100)23,24 and is therefore assigned to adsorbed hydrogen produced by the dehydrogenation reaction. This loss is strongly attenuated when heating to ∼190 K and disappears when heating to 270 K, indicating that either hydrogen desorbs in this temperature range or it moves subsurface. From this analysis we may thus conclude that ethene dehydrogenation occurs and that the surface modification affecting the subsequent exposures of Figure 2 is due to carbon accumulating at the surface. The ethene losses in spectrum b of Figure 4 thus correspond to a mixed layer composed of both di-σ-bonded ethene and of π-bonded molecules stabilized by carbon, the latter corresponding to the desorption peak at ∼170 K in Figure 2. Figure 5b also shows the effect of subsequent ethene doses, by using the beam at temperatures slightly higher than in the experiments reported in Figure 5a. Evidently, at this temperature the π-bonded species has already saturated after the first dose while the dehydrogenation process persists with exposure. After heating the sample, the vibrational signature for the di-σ-bonded loss is not as evident as in the experiment of Figure 5a. The reason for this apparent disagreement is, as will be shown below, that the conversion probability from π to di-σ depends both on dosing temperature and on heating rate, i.e., on the initial coverage of the π-bonded species and on the metastable population reached at high temperature when heating. This effect is emphasized by the data displayed in Figure 6, where ethene is again dosed using the molecular beam. As can be seen, after annealing to 200 K, no ethene associated losses are observed when using a slow heating rate (0.3 K/s) and only the peak corresponding to C2 survives. On the contrary, for the higher heating rate (1.5 K/s) the signatures of di-σ-bonded ethene, albeit weak, are present at 1151, 1360, and 2903 cm-1. The larger intensity of the CH stretch, compared to the spectra in Figure 5, is due to the larger coverage of ethene achieved in this experiment, which may also explain the slightly different frequencies of ν(CC) and ν3(CH2) with respect to the experiment of Figure 4. 3.3. Dehydrogenation Kinetics. The picture that emerges is that ethene adsorbs at low temperatures predominantly in π-bonded states at terraces and at steps and that these admolecules either desorb or transform into more strongly

Ethene Adsorption on Cu(410)

J. Phys. Chem. C, Vol. 113, No. 49, 2009 20885

Figure 5. Normalized HREELS spectra recorded after exposing Cu(410) to the supersonic molecular beam at 0.10 eV kinetic energy (obtained by seeding C2H4 in Ne) and -60° incidence (corresponding to step down). (a) 0.12 ML of ethene, dosed at 145 K, and subsequently annealed to the indicated temperatures. The spectrum of the clean surface is shown for reference. Initially, ethene is mostly in the π-bonded state but additional losses are present at 388 and 597 cm-1. The loss at 388 cm-1 survives annealing and corresponds therefore to a fragment of the ethene, which we identify with adsorbed C2. The loss at 1151 cm-1, corresponding to di-σ-bonded ethene, can be seen best after annealing to 250 K when the π-bonded molecules have desorbed. The loss at 597 cm-1 is compatible with the H resulting from ethene dissociation. (b) The same as spectra a but for a slightly higher dosing T and for different subsequent doses. The vibrational signatures for dissociation grow with dose. After annealing no clear signature for di-σ-bonded molecules is present. The presence of C2 may be connected either with the slightly higher kinetic energy of the incoming molecules with respect to dosing by backfilling or with the higher temperature at which the dosing occurred in the data shown in the figure. [1 L of ethene on Cu(410) ) 0.25 ML.]

Figure 6. Effect of the heating rate on the conversion probability from the π-bonded to the di-σ-bonded state of ethene. The HREELS spectrum in part b corresponds to a 5 times larger heating rate than part a. The loss at 597 cm-1 is indicative that more dissociation has taken place at the higher heating rate. [1 L of ethene on Cu(410) ) 0.25 ML.]

bound species upon heating: π0 converts into di-σ while π1 dehydrogenates, leading to hydrogen and C2 species. Dehydrogenation may take place also for the di-σ-bonded molecules. Since the formation of the di-σ-bond is not expected to favor

dehydrogenation, the pathway from di-σ to CC is most probably indirect, involving step sites. This branching reaction can be explored by performing TPD experiments at different heating rates, and the results are displayed in Figure 7.

20886

J. Phys. Chem. C, Vol. 113, No. 49, 2009

Kravchuk et al.

Figure 7. (a) 4 amu (D2) TPD profiles at various heating rates, including the results of fitting for D2 due to fragmentation of molecularly desorbed C2D4 (- -), surface decomposition of C2D4 ( · · · ), and the total D2 yield (- · -), i.e., the sum of the fragmentation and decomposition profiles. (b) 32 amu TPD profiles of C2D4 on Cu(410) at various heating rates, including the analytical fits to first-order desorption described in the text.

TABLE 1: Parameters Involved in the Fit of C2D4 TPD Data Shown in Figure 7aa heating rate/K s-1

C2D4 peak temp/K

Adesorb/s-1

Eact (desorb.)/kJ mol-1

Adecomp/s-1

Eact (decomp)/kJ mol-1

3.6 4.3 8.6 10.4

238 ( 2 237 ( 2 237 ( 2 240 ( 2

7 × 109(1 7 × 109(1 7 × 109(1 7 × 109(1

46.0 ( 0.4 45.4 ( 0.4 44.1 ( 0.4 44.3 ( 0.4

5 ( 4 × 1011 5 ( 2 × 1011 5 ( 1 × 1011 5 ( 0.5 × 1011

67 ( 3 64 ( 2 62.5 ( 0.5 60.2 ( 0.2

a

For comparison with ref 8, be reminded that 1 eV/atom ) 96.485 kJ/mol.

The apparent increase in desorption yield at higher heating rate arises because the time over which the data are collected is correspondingly shorter at higher heating rates. For the analysis of the spectra, the 32 amu (C2D4) desorption profile was fitted to the following analytical equation for the coverage during the desorption sweep appropriate for a firstorder process:25

( ( (

Θ ) Θ0 exp

TP2 2

T

exp

Eact 1 1 R T TP

)))

where TP is the peak temperature, Eact is the activation energy of desorption, and Θ0 is the initial coverage. The desorption rate is then given by

( )

rate ) k1 exp -

Eact RT

where k1 is the first-order desorption rate constant. The first-order desorption profile is shown in Figure 7b, where the fitting variables are the desorption activation energy Eact(desorption) and the pre-exponential factor A, using the experimental heating rate β. These parameters were constrained to yield the experimental C2D4 desorption peak temperature (∼238 K). The pre-exponential factor was allowed to vary to provide the best overall fit to the desorption profile to yield a value of ∼7 × 109 s-1 and a desorption activation energy of 45 ( 1 kJ/mol (see Table 1).

This enabled the calculation of the relative coverage of C2D4 as a function of temperature during the desorption sweep.25 The corresponding D2 (4 amu) profile is displayed in Figure 7b. Ethene exhibited a small 4 amu signal in our mass spectrometer (∼1% of the 32 amu signal) and the contribution due to C2D4 fragmentation is indicated. The difference between these curves and the experimental D2 desorption profiles is due to hydrogen desorption from C2D4 decomposition. The D2 desorption occurring between 220 and 260 K corresponds from the recombination of D formed during the ethene dehydrogenation process at that temperature, since hydrogen is then unstable, as demonstrated by HREELS (see Figure 5; the loss associated with hydrogen adatoms at 597 cm-1 formed by lower T dehydrogenation disappears already below 190 K). Since this deuterium is generated from di-σ-bonded ethene, the rate of its formation due to C2D4 decomposition, is given by Θ(C2D4)Adecomp exp[-Eact(decomp)/RT], where the C2D4 coverage during the desorption sweep is obtained for the C2D4 desorption profile as described above. The 4 amu profile was fitted by varying the pre-exponential factor, and the values of activation energy for decomposition to yield the values for each of the four different heating rates are listed in Table 1, and the resulting fit to the experimental desorption profile, including the sum of the fragmentation and decomposition, is shown in Figure 7a (- · -). The agreement with the experimental data is satisfying. The relative 4 amu contribution due to decomposition at the surface (Figure 7a) increases with increasing heating rate, being nearly zero at a rate of 3.6 K/s and becoming eventually larger than the fragmentation peak at 10.4 K/s. This yields an activation

Ethene Adsorption on Cu(410)

J. Phys. Chem. C, Vol. 113, No. 49, 2009 20887

Figure 8. Amount of accumulated carbon (%) after dosing and after subsequent annealing to 300 K as inferred from XPS measurements. No cleaning or high-temperature annealing was performed in between the additional exposures. Inset: XPS C 1s peak measured after three subsequent doses of 1, 3, and 3 L, followed by heating to 300 K after each dose (the relative amount of carbon can be inferred from the open square of the last measurement).

energy for dehydrogenation of 63 ( 3 kJ/mol and a preexponential factor of ∼5 × 1011 s-1. This analysis implies that the dissociation efficiency increases with heating rate, i.e., the conversion and dissociation processes are linked, and that hydrogen is released during dissociation. Indeed, as shown in Figure 5, hydrogen desorbs from Cu(410) below 189 K, since the relevant loss at 597 cm-1 is then no longer present. This desorption T is in accord with literature on H/(Cu(111)26 and H/Cu(100).27 Additional evidence for ethene dissociation and carbon accumulation was obtained by XPS. In Figure 8, the carbon coverage (in percent of monolayer of Cu atoms) is displayed after some ethene adsorption-desorption cycles, where the coverage is calculated from the relative integrated areas of C 1s and Cu 2p peaks and corrected for the relevant sensitivity factors. An initial, very small amount of bulk carbon signal was subtracted from all spectra. It can be seen that approximately half of the admolecules dissociate during the first exposure-annealing cycle, while much smaller additional carbon coverages accumulate for the subsequent doses, suggesting occupation of the reactive sites by the dissociation products. The C 1s peak is at 284.5 eV (see inset of Figure 8), which is consistent with the presence of C2 species. This binding energy could also be assigned to ethene, while graphite and carbidic C are expected at 284 and 283 eV binding energies, respectively. However, the presence of CH species is excluded by the HREELS results. 3.4. Nature of the Ethene Desorption State on CarbonContaminated Cu(410). To obtain additional insights into the origin of 169 K desorption peak formed on a carboncontaminated surface (Figure 2), HREELS spectra were collected for ethene adsorbed (dosed by backfilling) on carbonprecovered Cu(410) surfaces. The result is displayed in Figure 9. Carbon contamination was achieved by repeated ethene adsorption-desorption cycles as for the TPD experiments (Figure 2a). The carbon-related loss is apparent at 367 cm-1. However, frequencies are evident at 942 and 1313 cm-1, and the frequency shift from 1290 to ∼1313 cm-1 suggests a slightly higher σ-π parameter for ethene on a carbon-contaminated surface.19 This experiment allows us to conclude that the 169 K desorption peak in TPD spectra is associated with π-bonded ethene. This is in accord with previous literature22 for ethene on carbon (or oxygen) precovered metals for which only π-bonded configurations were reported.

Figure 9. Normalized HREELS spectra measured on carbon precovered Cu(410) after an exposure of 0.15 L of ethene at 145 K dosed by backfilling. Carbon contamination was achieved by repeated ethene adsorption-desorption cycles as for the TPD experiments in Figure 2.

Finally, we note that, at such a small dose by backfilling, little or no dissociation takes place. The loss at 296 cm-1 corresponds most probably to the ethene-surface vibration, expected at this frequency for ethene at close-packed steps.28 Dissociation is therefore favored at low coverage by the relaxation of the step edge induced by carbon resulting from previous uptakes. A similar phenomenon was reported for O2 dissociation on Ag(210).29 3.5. Identification of Dehydrogenation Sites. To further investigate the path leading to ethene decomposition, we performed HREELS investigations of ethene adsorption on bare (after sputtering and annealing) and on CO-precovered (0.3 L) Cu(410) surfaces. In both cases, identical exposures (1 L) of ethene were used by backfilling the chamber. The crystal was then heated to desorb both CO and ethene, which leave the surface at almost identical temperatures. The normalized intensities of the low-frequency HREELS features, due to the remaining carbon fragments, are presented in Figure 10. CO initially adsorbs at the steps, where its binding energy is higher than on the terraces, and the saturation CO coverage corresponds to 50% of the step sites. If dissociation of ethene were to occur at terraces, similar loss intensities would be expected in both experiments. This is evidently not the case, since the loss due to carbon fragments is at least three times more intense for ethene adsorbed on the bare Cu (410) surface. It can then be concluded that ethene dehydrogenates and leads to carbon decorating the open step edges. Such carbon inhibits further dissociative poisoning of the active sites and permits adsorption only in the π-bonded state. 4. Discussion The picture that emerges from the experiments is that ethene adsorbs predominantly in a π-bonded configuration on Cu(410).

20888

J. Phys. Chem. C, Vol. 113, No. 49, 2009

Figure 10. Normalized HREELS intensity measured on Cu(410) after dosing 1 L of ethene on bare Cu(410), followed by heating to 233 K (solid line), and after dosing 1 L of ethene on Cu(410) predosed with 0.3 L of CO and heating to 204 K (dashed line). The slight difference in the C-Cu loss around 365 cm-1 is connected to the different carbon coverage.

It desorbs from the terraces at ∼143 K, a value which compares well with that reported for Cu(100).16 More strongly bound π-bonded molecules sit at the undercoordinated Cu atoms at the steps, most probably in atop configurations by analogy to the case of ethene/Ag(410).10 Such molecules contribute to the high temperature tail of the desorption peak. Di-σ-bonded ethene is present, too, and desorbs30 at ∼240 K with an activation energy of ∼45 kJ/mol. Its site cannot be determined from the present data. Decomposition to C2 species occurs, too, with an activation energy of ∼63 kJ/mol. The remaining carbon decorates the steps. It is difficult to gauge the coverage of di-σ-bonded ethene, since it is evident in all spectra that the losses associated with it are weak compared to those of π-bonded ethene. At least at low exposure, the two populations should be comparable according to the TPD results (Figures 2 and 3). The different HREELS intensities may thus be due to different HREELS cross sections for the two forms of ethene, but this is not very likely. More probably, the difference corresponds to a smaller formation efficiency of the di-σ-bonded species at 145 K, the dosing temperature corresponding to the HREELS experiments, compared to 110 K, corresponding to the TPD results. To check this hypothesis, we exposed the sample to a large ethene dose (51 L) at 193 K, where mostly di-σ-bonded species would be expected from the TPD result. The outcome, reported in spectrum b of Figure 4, shows indeed only a very small loss associated to di-σ-bonded ethene. This result can be rationalized only assuming that the latter state is accessed through a precursor that becomes short-lived at the higher temperature. If formation of di-σ-bonds implies a precursor on the terraces (rather than at the steps), which is only slightly populated at 145 K, then only a very small amount of the former species will be produced in the HREELS experiments. The active site should, however, still be close to the step edge, since no σ-bonded ethene is

Kravchuk et al. reported for low-Miller-index Cu surfaces. A scheme of the complex potential energy surface resulting from our investigations is reported in the accompanying paper.18 A decrease in the ethene desorption yield is expected if carbon blocks the adsorption sites at the steps. This effect is not evident from the inset in Figure 2, in which the area under the TPD peaks is nearly identical for subsequent spectra, but it does occur when measuring subsequent ethene uptakes, as is shown in the accompanying paper.18 Ethene adsorption is thus hindered and di-σ-bonding is inhibited on the carbon-contaminated surface. Most probably, less stable adsorption sites close to the step become populated once the step itself is saturated by carbon atoms. This bond must be more stable than the π-bond in absence of carbon, since the desorption temperature of this moiety (169 K) is slightly higher. As the heating rate of the TPD experiments was increased, the fraction of 4 amu (D2) relative to 32 amu (C2D4) was found to increase, suggesting a pathway to surface decomposition of C2D4 in competition with desorption. These data were then fitted using a procedure to differentiate between D2 due to the fragmentation of molecularly desorbed C2D4 in the mass spectrometer and D2 due to the decomposition of C2D4 on the surface. The results are listed in Table 1 and shown in Figure 7. The variation in selectivity as a function of heating rate arises because of the widely varying kinetic parameters for desorption and decomposition. Transition-state theory indicates that the preexponential factor for desorption should equal kT/h, where k is the Boltzmann constant and h is Planck’s constant, if the partition functions of the adsorbate and transition state are identical. At the desorption temperature of ∼240 K, this yields a value of ∼5 × 1012 s-1, significantly larger than the experimentally derived value of 7 × 109(1 s-1, implying that the transition state to desorption is significantly less mobile than the adsorbate and may suggest that desorption (as well as decomposition) is mediated by the steps. A simple thermodynamic calculation, reported in the Supporting Information of ref 13, allows the magnitude of the C-Cu bond strength to be estimated. It shows that to render ethene dissociation on copper thermodynamically possible, the C-Cu binding energy should be larger than 170 kJ/mol. For comparison, the bond energy of C on Ni(100) was estimated from calorimetric experiments to be ∼203 kJ/mol.31 On the other hand, the calculated dissociation energies of monocarbides32 of Ni and Cu are D0(NiC) ) ∼381 kJ/mol and D0(CuC) ) ∼230 kJ/mol. The average binding energy of carbon adsorbed on copper as a result of ethene dissociation should thus be larger than 170 kJ/mol but smaller than 230 kJ/mol. These limits should be compared to the C-Ni binding energy, which lies between 203 and 380 kJ/mol. Since Ni is used as catalyst for the steam re-forming process, our study indicates that copperbased catalysts could work as well if open step edges are present. As demonstrated elsewhere, the under-coordination induced by roughening Cu(111) surfaces by ion bombardment at 145 K and by Cu atom evaporation at 90 K is still insufficient for activating ethene dehydrogenation28 and also to allow for diσ-bond formation. It is, however, sufficient to stabilize π-bonded ethene. The steps generated by the latter procedures must, therefore, be mostly compact and thus different from the open ones of Cu(410). On the basis of the presented results, we can summarize the processes occurring for ethene on Cu(410). The molecule may adsorb either as π-bonded or as di-σ-bonded or undergo complete dehydrogenation. The final state depends on adsorption

Ethene Adsorption on Cu(410) site, dosing temperature, and surface treatment. The π-bonded molecules are at the terraces as well as at the steps, while diσ-bond formation and dehydrogenation occur only close to or at the open steps. When dosing at 110 K, the ratio between πand di-σ-bonded molecules is close to unity until the steps become saturated. Further exposure at 110 K results only in molecules in their π-bonded state. Dosing at higher temperatures (145 and 193 K) on clean Cu(410) results in π- and di-σ-bonded molecules, too. However, the amount of the latter species is smaller than when dosing at 110 K. This result indicates that the formation of di-σ-bonds is mediated by a weakly bound precursor. Dehydrogenation of ethene is less temperature dependent and occurs even at 193 K, however, not at room temperature (Figure 4), indicating that it is precursor-mediated, too. The process takes place at the step edges and this reaction is poisoned by carbon accumulation at such sites. The dehydrogenation reaction can also be inhibited by preadsorbing molecules such as CO that compete for the step sites. Upon annealing the crystal to 900 K, carbon can diffuse into the subsurface region recovering the initial reactivity of clean Cu(410). The possibility of dehydrogenation occurring directly during the collision of ethene molecules with the surface is ruled out by experiments performed at higher surface temperature (see Figure 4 and experiments reported in ref 18) showing absence of carbon formation. The process is therefore precursor-mediated. The additional dissociation process taking place during the TPD heating during the desorption of the di-σ-bonded layer, demonstrated by the dependence of the D2 desorption yield on the heating rate, is less easy to rationalize. The breaking of a CC bond occurring in di-σ-bond formation is indeed not expected to favor dehydrogenation. As discussed in the accompanying paper,18 dehydrogenation occurs most probably through a pathway involving the conversion of the di-σ-bonded molecules into the more strongly bound, though π-bonded, molecules at the step edge. As soon as the surface becomes covered by carbon, the adsorption of ethene results in smaller and smaller amounts of di-σ-bonded molecules. One might expect a slight decrease in the ethene desorption yield with carbon precoverage, if it blocks the adsorption sites. However, this is not the case, at least at low exposures. Ethene adsorption is still possible, albeit only in the π-bonded state. Most probably adsorption sites close to the step become populated once the step itself is saturated by carbon. This bond is more stable than the π-bond on the terrace and results thus in a slightly higher desorption temperature. 5. Conclusions Our data demonstrate low-temperature site-selective chemistry of ethene on strongly under-coordinated sites on copper surfaces. On Cu(410) they induce not only di-σ-bonding but also promote complete dehydrogenation at low temperature. No partially dehydrogenated radicals could be detected. The so-generated carbon decorates the step edges, thereby blocking the active sites for subsequent dissociation and/or adsorption in the di-σbonded state. However, the dehydrogenation activity can be restored merely by heating the sample to 900 K. This fact and the smaller C-Cu bonding energy compared to C-Ni make Cu-based catalysts interesting for steam re-forming processes.

J. Phys. Chem. C, Vol. 113, No. 49, 2009 20889 Acknowledgment. We thank Compagnia S. Paolo for financial support. We thank M. Okada for providing the Cu(410) sample. V.V. undertook this work with the support of the “ICTP Programme for TRIL, Trieste, Italy”. M.R. finished this paper during his stay in Rio and thanks INMETRO and CNPq for their hospitality. Discussion with A. Kokalj is acknowledged. References and Notes (1) Vang, R. T.; Honkala, K.; Dahl, S.; Vestergaard, E. K.; Schnadt, J.; Lægsgaard, E.; Clausen, B. S.; Nørskov, J. K.; Besenbacher, F. Nat. Mater. 2005, 4, 160. (2) Vattuone, L.; Savio, L.; Rocca, M. Surf. Sci. Rep. 2008, 63, 101. (3) Ajayan, P. M. Chem. ReV. 1999, 99, 1787. (4) Bengaard, H. S.; Norskov, J. K.; Sehested, J.; Clausen, B. S.; Nielsen, L. P.; Molenbroek, A. M.; Rostrup-Nielsen, J. R. J. Catal. 2002, 209, 365. (5) Linke, R.; Becker, C.; Pelster, Th.; Tanemura, M.; Wandelt, K. Surf. Sci. 1997, 377-379, 655. (6) Nyberg, C.; Tengstal, C. G.; Andersson, S. Chem. Phys. Lett. 1982, 87, 87. (7) Kubota, J.; Kondo, J. N.; Domen, K.; Hirose, C. J. Phys. Chem. 1994, 98, 7653. (8) Yim, W.-L.; Nowitzki, T.; Necke, M.; Schnars, H.; Nickut, P.; Biener, J.; Biener, M. M.; Zielasek, V.; Al-Shamery, K.; Kluner, Th.; Ba1umer, M. J. Phys. Chem. C 2007, 111, 445. (9) Vinod, C. P.; Hans, J. W. N.; Nieuwenhuvs, B. E. App. Catal. A 2005, 291 (1-2), 93. (10) Kokalj, A.; Dal Corso, A.; de Gironcoli, S.; Baroni, S. Surf. Sci. 2004, 566-568, 1018. (11) Witko, M.; Hermann, K. Appl. Catal., A 1998, 172, 85. (12) Kovacs, I.; Solomosi, F. J. Phys. Chem. B 1997, 101, 5397. (13) Kravchuk, T.; Vattuone, L.; Burkholder, L.; Tysoe, W. T.; Rocca, M. J. Am. Chem. Soc. 2008, 130, 12552. (14) Rocca, M.; Valbusa, U.; Gussoni, A.; Maloberti, C.; Racca, L. ReV. Sci. Instrum. 1991, 62, 2172. (15) Jenks, C. J.; Xi, M.; Yang, M. X.; Bent, B. E. J. Phys. Chem. 1994, 98, 2152. (16) Arvanitis, D.; Baberschke, K.; Wenzel, L.; Doebler, U. Phys. ReV. Lett. 1986, 57, 3175. This paper reports ethene desorption at 190 K for a nonspecified heating rate. To be compatible with our data and with ref 15, such temperature implies either an unphysically high heating rate or that the surface was carbon contaminated due to ethene dissociation induced by synchrotron radiation. (17) Skibbe, O.; Binder, M.; Otto, A.; Pucci, A. J. Chem. Phys. 2008, 128, 194703. (18) Venugopal, V.; Vattuone, L.; Kravchuk, T.; Smerieri, M.; Savio, L.; Jupille, J.; Rocca, M. J. Phys. Chem. C, accompanying paper in this issue (DOI: 10.1021/jp9047924). (19) Stacchiola, D.; Burkholder, L.; Tysoe, W. T. Surf. Sci. 2002, 511, 215–228. (20) Sock, M.; Eichler, A.; Surnev, S.; Andersen, J. N.; Klotzer, B.; Hayek, K.; Ramsey, M. G.; Netzer, F. P. Surf. Sci. 2003, 545, 122. (21) Sheppard, N. J. Electron Spectrosc. Relat. Phenom. 1986, 38, 175. (22) Chesters, M. A.; Sheppard, N. Spectroscopy of Surfaces; John Willey & Sons Ltd: New York, 1988; Chapter 7. (23) Astaldi, C.; Bianco, A.; Modesti, S.; Tosatti, E. Phys. ReV. Lett. 1992, 68 (1), 90. (24) Lai, W.; Xie, D.; Yang, J.; Zhang, D. H. J. Chem. Phys. 1992, 121 (15), 7434. (25) Redhead, P. A. Vacuum 1962, 12, 203. (26) Lee, G.; Poker, D. B.; Zehner, D. M.; Plummer, E. W. Surf. Sci. 1996, 357-358, 717. (27) Kolovos-Vellianitis, D.; Kammler, Th.; Kuppers, J. Surf. Sci. 2000, 454-456, 31. (28) Skibbe, O.; Binder, M.; Pucci, A.; Kravchuk, T.; Kokalj, A.; Vattuone, L.; Venugopal, V.; Rocca, M. J. Chem. Phys. 2009, 131, 024701. (29) Vattuone, L.; Savio, L.; Rocca, M. Phys. ReV. Lett. 2003, 90, 228302. (30) Michalak, A.; Witko, M.; Hermann, K. J. Mol. Catal. A 1997, 119, 213. (31) Vattuone, L.; Yeo, Y. Y.; Kose, R.; King, D. A. Surf. Sci. 2000, 447, 1. (32) Gutsev, G. L.; Andrews, L.; Bauschlicher, C. L. Theor. Chem. Acc. 2003, 109, 298.

JP904794N