- Email: [email protected]
Carbon monoxide adsorption on the kinked Pt(321) surface
Surface Science I 12 ( 198 I) 63-77 North-Holland Publishing Company
CARBON SURFACE Michael
MONOXIDE
63
ADSORPTION
R. MCCLELLAN
ON THE KINKED
Pt(321)
*
Depurtnwnt of Chemisttyv und Center /or Muteriols Science und Engineering, of Technologv, Cumhndge, Mussuchusetts O_‘I39, USA
Mussuchusetts
Imtrtute
John L. GLAND Ph.vsiwI USA
Chemistq
Deportment,
Generul Motors
Reseurch
L&oratories,
Wurreu. Michigun
48090,
and
F.R. McFEELEY Depurtment of Chemrstryv ond Center/or Muteriuls Science und Engineering, of Technologv, Cumhridge. Mussuchusetts 0.?139, USA Received
23 June
1981; accepted
for publication
5 August
Mussuchusetts
Institute
1981
Carbon monoxide adsorption and desorption were studied on the kinked Pt(321) surface using high resolution electron energy loss spectroscopy, thermal desorption spectroscopy, Auger electron spectroscopy, and low energy electron diffraction. Two chemisorbed states are observed for CO adsorption on the Pt(321) surface. Carbon monoxide adsorbed along the rough steps (which have a high density of kinks) has an adsorption energy of 151 kJ/mol (36 kcal/mol). Carbon monoxide adsorbed on the terraces has an adsorption energy of 96 kJ/mol(23 kcal/mol). Both terminal and bridge bonded CO adsorb on the atomically rough steps although terminal binding is the dominant adsorption mode. Adsorption along the rough steps and terraces does not occur sequentially at 100 K indicating that the CO mobility is restricted on the rough Pt(321) surface. However, sequential desorption does occur since CO desorbs from the terraces before it desorbs from steps. Molecular adsorption of CO on the Pt(321) kinked surface clearly dominates. In fact, weakening of the carbon-oxygen bond does not occur contrary to several recent suggestions that surface defects on platinum weaken the carbon-oxygen bond.
1. Introduction Recently, unique adsorption states of CO have been proposed to exist in conjunction with kink sites on platinum [l] single crystals. In order to systematically investigate this possibility, we report here the results of high * Present
address:
0039-6028/8
Sandia
National
l/0000-0000/$02.75
Laboratories,
Livermore,
California
0 198 1 North-Holland
94550, USA
resolution electron energy loss and thermal desorption measurements on CO adsorbed on the kinked Pt(321) surface. These experiments have also served as a prelude to a complete study of carbon monoxide oxidation on this surface, which will be reported separately [2,3]. The Pt(321) surface is an interesting surface from several viewpoints. First, it is extremely rough on an atomic scale since 40% of the surface platinum atoms are coordinatively unsaturated (see fig. 1). Yet the surface depicted in fig. 1 is stable in vacuum even at high temperature and does not reconstruct in the presence of adsorbates [4]. This type of stable, rough single crystal surface allows us to characterize in detail the effect of local surface morphology on chemisorption and to investigate the possibility of unique surface reaction pathways on platinum kinked surfaces [4,5]. Carbon monoxide adsorption on close packed Pt(ll1) surfaces (which comprise the terraces on the (321) surface) has been well characterized by a number of techniques [ 1,669]. At low coverage the CO adsorbs in a terminal (atop) configuration. As coverage increases a bridged configuration becomes populated, leading to roughly equal numbers of terminal and bridge bonded molecules at saturation. The existence of two bonding configurations is clearly indicated by infra-red reflection spectroscopy [6], high resolution electron
Fig. I. Photograph
of a ball model of the Pt(321) surface.
M. R. McClellm
et ul. /
Curhm
momride
udsorptiort
on kirtkrd
Pt(321)
65
energy loss spectroscopy [1,7], dynamic work function change measurements [8], and low energy electron diffraction [9]. A single broad CO desorption peak develops as coverage increases, indicating that the bonding energies of these two configurations are essentially equal [9- 111. Carbon monoxide adsorption on platinum surfaces with linear or rough steps has not been thoroughly characterized. The coverage dependence of the CO thermal desorption spectra has been studied for two platinum surfaces with linear steps [lo]. On both of these surfaces, two desorption peaks were observed for large CO coverages. The low temperature peak was identified as desorption from the (111) terraces, since its characteristic desorption temperature was identical to that observed in desorption from the ideal (111) surface, while the high temperature peak was identified as desorption from the steps. These two binding states were found to populate sequentially; the higher energy sites along the linear steps populate first. The relative peak areas further indicate that carbon monoxide packs more densely along the steps than on the terraces [lo]. Carbon monoxide adsorption on a stepped platinum surface has also been characterized using high resolution electron energy loss spectroscopy [l]. In contrast to the behavior observed on Pt( 11 l), several CO stretching frequencies were observed at low coverage. The results suggested that the dominant adsorption configuration for CO at low coverages involved terminal adsorption along the top of the linear steps [l], while a weak v(C-0) transition observed to adsorption along the at 1690 cm ~ ’ for low CO coverages is attributed bottom of the linear step [l]. Two additional weak Y(C-0) transitions were observed at 1560 and 1410 cm-‘. The authors argued that these bands are due to carbon monoxide with a substantially weakened carbon-oxygen bond resulting from adsorption at kink sites occurring as imperfections on the stepped surface [l]. At high coverage the usual terminal (2100 cm-‘) and bridge bonded (1860 cm-‘) configurations dominated as the terraces became populated. In addition to this argument for anomalous bond weakening at kink sites, dissociative carbon monoxide adsorption at kink sites on platinum surface has been proposed recently on the basis of X-ray photoemission results [5]. In contrast to these proposals, no evidence has been found in the experiments reported here either for dissociative chemisorption or for molecular CO with an anomalously weakened CO bond, despite the high density of kink sites (20% of the surface) on the crystal face.
2. Experimental The apparatus employed for these experiments is schematically illustrated in fig. 2. The system consists of an UHV stainless steel vacuum chamber equipped with a single pass cylindrical mirror analyzer for Auger electron spectroscopy
66
4 QUADRUPOLE
RESOLUTION ELECTRON ENERGY LOSS SPECTROMETER
Fig. 2. A schematic
+%LT,C,
ARRAY
VIEWPORT
diagram
of the apparatus
with the experimental
stations
labeled
(AES), low energy electron diffraction (LEED) optics, a multiplexed quadrupole mass spectrometer and a 127” cylindrical sector high resolution electron energy loss spectrometer (HREELS). The crystal was spark cut from a single crystal rod and oriented using back-reflection Laue diffraction to within “0.5”. The sample was then spark planed and mechanically polished. In order to remove calcium from the bulk, the sample was treated in flowing atmospheric pressure oxygen at 1300 K for 24 h. Subsequent to this treatment, the crystal was mounted in the UHV system by spot-welding two 0.05 cm diameter platinum wires which served as leads for resistive heating and as thermal links between the crystal and a liquid nitrogen reservoir. The crystal temperature could be varied from 100 to 1300 K as monitored by a 0.008 cm diameter chromel-alumel thermocouple spotwelded directly to the crystal. The surface was cleaned by Argon Ion sputtering at low temperature (- 100 K) to remove small amounts of residual calcium from the surface. Carbon was then removed by heating the crystal in oxygen (= 900 K). prior to carbon monoxide adsorption the surface was characterized using LEED and AES in order to ensure that it was clean and well-ordered. A typical AES of the clean surface produced in this manner is shown in fig. 3. LEED observations of this surface revealed diffraction patterns characteristic of a well-ordered ideal (321) surface. No evidence of surface faceting was observed during these expsriments. After obtaining a clean well-ordered surface, the crystal was cooled to 100 K and dosed with carbon monoxide using a multichannel molecular beam doser. A multichannel molecular beam doser is an array of microscopic
M. R. McClellan
et al. /
Curhor~ motloxide
&sorption
on kinked
Pt(_l.?I)
61
Pt (321) CLEAN
--dN(E dE
I
'0
I
200
I 300
ENERGY
I
I 400
I
I 500
I
6C
(eV)
Fig. 3. An Auger spectrum of the clean Pt(321) surface. The primary beam energy modulation was 2V peak-to-peak; and a sweep rate of 5 eV/s was used.
was 3 keV; the
capillary tubes that produce a directional flow of the adsorbate. The purpose of using a doser is to preferentially expose the front face of the crystal which increases the signal to background in the thermal desorption experiment. To ensure reproducible surface coverages, the crystal was returned to the same position in front of the doser for subsequent doses. All adsorption was done with the crystal at 100 K. The exposures were estimated using an uncalibrated nude ion gauge. Since the actual exposures were made with the aid of a multichannel molecular beam doser, separate thermal desorption experiments were performed to estimate the pressure differential between the crystal and the ion gauge. This was done by comparing total desorption yields for doser exposures and background exposures to CO. The CO flux in front of the doser was found to be enhanced by a factor of - 15. The relative concentration of CO molecules adsorbed for various exposures was determined using integrated thermal desorption spectra (the heating rate was 10 K/s for these experiments). The relative coverages reported in this paper are references to the integrated area of the thermal desorption spectrum shown in fig. 5g (called &,,,). For exposures greater (- 2 X ) than the exposure required to produce 0,, the coverage is called saturation. An estimate of the absolute coverage of adsorbate CO was obtained during CO oxidation studies on this same surface [3]. Carbon dioxide yields were estimated from tempera-
EXPOSURE
(L
1
Fig. 4. The total amount of CO adsorbed on the Pt(321) surface at 100 K as a function exposure. Calibration methods are discussed in the text. One langmwr exposure of CO is 3.X X 10’4/cm2 for a gas at T= 300 K.
of CO
equal to
ture programmed reaction experiments where a known initial amount of adsorbed oxygen was completely consumed. The oxygen concentrations were determined by comparing oxygen Auger results for the oxygen saturated Pt( 111) surface with Auger results for oxygen on the Pt(321) surface. The saturation coverage estimated in this way is approximately 8 X 10’4/cm2. The complete exposure versus coverage curve is shown in fig. 4. The initial slope is constant (consistent with precursor state kinetics), and yields a value between 0.5 and 1.0 for the sticking coefficient. This value is typical for both flat and stepped Pt surfaces [9,1 l- 131. Energy loss spectra were obtained in the specular direction using a - 5 eV primary beam. The relative intensities of individual energy loss features depends on spectrometer tuning and crystal position, therefore relative peak intensities were used only qualitatively. Thermal desorption spectra (TDS) were obtained subsequent to the energy loss spectra (typically a 15-20 min time delay between exposure and desorption). The reproducibility of the data presented here was verified by removing the Pt(321) sample from the system and preparing a new Pt(321) surface. Satisfactory consistency was invariably obtained during spot checks of the data.
3. Results and discussion 3.1. The Pt(321) surface A ball model of the ideal Pt(321) surface is shown in fig. 1 (cf. ref. [14]). In - the usual step notation this surface can be described as Pt(321)
+ Pt(S) - [3( 111) x (3li)])
indicating that the surface is comprised separated by (311) steps of monatomic decomposed as (3li)
-2(loo)
of (111) terraces, three atoms wide, and height. The (311) steps can be further
x (iii).
Thus, the step itself exposes two atom wide sections of open (100) and close packed (111) planes. A more useful representation of this surface is the microfacet notation of Van Hove and Somorjai [17]. In this scheme the surface is described as Pt(321)
+ Pt(S) -[(3,‘2),(lll)
+ (1/2),(11i)
+ (l),(lOO)].
Here the Miller index of the surface has been vectorally decomposed in a basis set made up of low index planes [the (11 l), the (1 Ii), and the (loo)]. These vector decomposition coefficients are given in the parentheses before the low index planes. The subscript on the coefficient is the number of unit cells of that low index plane contained in unit cell of the surface. Using this representation the characteristic parameters of the surface are easily determined. These parameters are summarized in table 1. A close examination of fig. 1 reveals that the platinum surface is quite rough and that the atoms reside in three distinct local environments. These different environments may be characterized by the number of surface nearest neighbors possessed by the atom in question. Using this classification, the surface consists of: (1) terrace atoms with the maximum 6-fold coordination, comprising 60% of the surface; (2) step atoms, 5-fold coordinated, comprising 20% of the surface; and (3) kink atoms, which have only 3-fold coordination and comprise the remaining 20% of the surface. Note that during this discussion we use rough step site as a general term to describe adsorption on or near either step or kink atoms. LEED observations of the crystal revealed a diffraction pattern characteristic of the ideal Pt(321) structure. While the background intensity was somewhat higher than is typically observed on Pt(l1 l), the diffraction features themselves were sharp. Occasionally, extremely small splittings of the LEED features were observed. In view of our ability to reproduce all of our results on two Pt(321) surfaces, we are confident that the crystal consists overwhelmingly of ideal surface and that the results presented in this paper are due to carbon monoxide adsorbing onto this ideal surface. Table I Characteristic
parameters
A. Characteristic
of the Pt(321) surface (i) Kink to kinkz4.8
distances:
(ii) Terrace B. Densities
of different
platinum
types:
width=5.5
A A
(i) Kink density=3.5X lOI cm-’ (ii) Step density=3.5 X lOI cm --z (iii) Terrace density= 1.0X IO” cm-*
No ordered CO overlayer structures were observed using a conventional four grid LEED system during these experiments despite the wide range of coverages and annealing temperatures studied. This observation is not surprising in light of the rough surface used and the lack of mobility for adsorbed CO discussed later in this paper. 3.2. Thermal desorption spectroscopy Thermal desorption of CO from the Pt(321) surface is for a series of CO exposures. For low initial CO coverages /
I
I
I
COVERAGE
Pt (321) co/co TA=
I
I
illustrated in fig. 5 a single first order
IOOK
010028,,, b) 0.060 c) 0.088 d) 0. I 78 elO.498
max mox mox max
I 0
400
600 TEMPERATURE
600 (K)
Fig. 5. A series of carbon monoxide desorption surface at 100 K. The heating rate was IO K/s.
spectra
for various
CO coverages
on the Pt(321)
h4. R. McCleIlun er al. / Curhon monoxide udsorptron on kinked P1(321)
71
desorption peak is observed at 556 K. As the coverage increases the high temperature peak moves down to 546 K and increases in size. For coverages above 0.28,,,,, a second low temperature peak is observed at 436 K. This low temperature peak at 436 K can be identified as desorption of CO bound to the (11 I) terraces by analogy with published results for CO desorption from the Pt( 111) surface [ 1,101. the higher temperature peak at 556 K can therefore be identified as desorption of CO bound to the atomically rough step sites. The desorption temperature, > 500 K, is typical for CO desorption from smooth steps [1,10,11,13]. Further evidence that these two desorption peaks are caused by desorption from two distinct chemisorbed states is furnished by the partial desorption experiment illustrated in fig. 9. All the carbon monoxide in the low temperature peak was desorbed by annealing the sample in vacuum at 400 K until desorption ceased (see arrow in the upper desorption spectrum of fig. 9). Only the high temperature desorption peak is observed in the subsequent desorption experiment (lower desorption spectrum of fig. 9) indicating that no redistribution of CO occurs with cooling. Therefore, the two TDS peaks correspond to two chemisorbed states which desorb sequentially. The desorption parameters for the high temperature peak have been estimated using several methods. The heat of desorption obtained is about 15 1 kJ/mol (36 kcal/mol) regardless of the method used. Fig. 6 illustrated CO desorption from a saturated surface using a series of heating rates. As illustrated in fig. 7 the shift in the high temperature peak with heating rate indicates a heat of desorption of 151 kJ/mol (36.2 kcal/mol) and a preexponential factor of 2 X 1014/s [16]. Alternatively, if we assume that the interactions in the adsorbed layer of CO are negligible at the lowest coverage illustrated in fig. 5 (0.028,,,,), the high temperature peak width and position alone are used to calculate a desorption energy of 15 1 k 16 kJ/mol (36 k 4 kcal/mol) [ 171. Using this desorption heat [ 181 the pre-exponential factor was estimated to be 1 X 1014/s. Finally, using a desorption isostere method the desorption heat estimated for the high temperature peak ranges from about 134 kJ/mol (32 kcal/mol) below 0.28,, up to 151 kJ/mol (36 kcal/mol) between 0.3 and 0.458,,,, [ 191. Since the sticking coefficient is near one (fig. 4), adsorption is not an activated process. Therefore the heat of adsorption should be equal to the heat of desorption. In summary, the heat of adsorption on the Pt(321) surface for CO adsorbed on the rough step sites is approximately 151 kJ/mol (36 kcal/mol). Similar heats have been previously reported for CO adsorbed on the step sites of several platinum surfaces [lo,1 l] confirming that the high temperature peak is caused by desorption of CO from the rough step sites on the Pt(321) surface. Detailed analysis of the low temperature desorption peak is more difficult since the peak is fairly small and overlaps with the leading edge of the high temperature peak. The shift in the low temperature peak with heating rate
HEATING
RATE LK/sl a) 2.4
bl5.0 cl 7.5 ej23.0
270
430
590
TEMPERATURE
750
(K)
Fig. 6. A series of thermal desorption spectra heated at a series of linear heating rates.
,3_
Pt (321) co/co ( sot.) TA =IOOK HIGH TEMPERATURE
for CO desorption
from a saturated
Pt(321) surface
PEAK
IO-
91.75
1.80
I85 IOOO/Tm
Fig. 7. A plot using the temperature to estimate the heat of desorption (36.2 kcal/mol).
1.90
I .95
(K1 (T,,) of the maximum
desorption rate and the heating rate (j3) from the data in fig. 6. The heat of desorption is 151 kJ/mol
M. R. McClelluz
et (11./ Curhon monoxrde adsorpfion on kinked Pt(321)
73
indicates a heat of desorption of 96 k/mol(23 kcal/mol) and a pre-exponential factor of 2.5 X lO”/s. If we assume a pre-exponential factor of 10i3/s, the heat for the low temperature peak is estimated to be 109 kJ/mol (26 kcal/mol) using the peak temperature [18]. Similar heat have been previously observed for CO adsorption on the low index Pt(ll1) surface [9,12] and on the (111) terraces of linear stepped surfaces [9,1 l] again confirming that the low temperature peak is caused by desorption of CO from the terraces of the Pt(321) surface. In contrast to the results obtained on linearly stepped surfaces [ 10,111, CO does not adsorb sequentially first on the rough step sites and then on the terrace sites. For instance, at a CO coverage of 0.49(3,, (spectrum e, fig. 5) where the low temperature terrace peak is clearly visible the high temperature step peak has attained only about half of its saturation intensity. Above about half saturation coverage both chemisorbed states are populated at about the same rate. This observation indicates that CO mobility is quite low on this rough surface even near the desorption temperatures, as substantial mobility coupled with significant site to site variation in the binding energy would result in stepwise adsorption. As the surface approaches saturation coverage the high temperature peak becomes approximately twice as large as the low temperature peak. Since the rough step sites cover 40% of this surface, these desorption results imply that the carbon monoxide surface concentration is roughly three times larger near the rough step sites than on terrace sites. Similar density comparisons for CO adsorbed on linearly stepped surfaces have been reported previously [lo]. Because of the large concentration of step and kink sites on the Pt(321) surface the low temperature terrace desorption peak does not dominate the high temperature step desorption peak even at saturation coverage. McCabe and Schmidt [ 1 l] previously observed similar behavior on the rough Pt(210) surface. 3.3. Vibrational spectroscopy A series of high resolution electron energy loss spectra (HREELS) for increasing CO concentrations adsorbed on the Pt(321) surface are shown in fig. 8. Two features are observed at 2065 and 480 cm-’ for low CO coverages (spectrum a, fig. 8). These vibrational transitions are characteristic of CO bound to the platinum surface in a terminal (atop) configuration. The low frequency v(C-0) transitions previously assigned to adsorption at kink sites [l] were never observed for CO adsorption on this kinked surface. As CO coverage increases a new feature appears at 1865 cm-’ which is characteristic of bridge bonded CO. No new low frequency modes accompany the growth of the bridge bonded transition. As coverage increases the terminal band at 2065 cm-’ moves up to 2095 cm-. This increase in frequency is similar to the shift observed for CO adsorption on the Pt( 111) surface ascribed to dipole-dipole coupling [ 121.
COVERAGE al 0020mlJx
+ co TA = IOOK
‘t (321)
b) 0000max c) 0 17%,3x d) %ox
16
2075 \
jyLy$
1
0
I
I, 400
I 800
ENERGY
I 1200
LOSS
I, 1600
I 2000
(cm-‘)
Fig. 8. A series of high resolution electron energy loss spectra the Pt(321) surface at 100 K. The beam energy was -5 eV.
for carbon
monoxide
adsorbed
on
In order to establish the relationship between the two chemisorbed states observed by desorption and these vibrational results several partial desorption experiments were performed. Typical results are illustrated in fig. 9. A vibrational spectrum was first obtained from a CO saturated surface at 100 K (top spectrum, fig. 9). All the CO in the low temperature terrace peak was then desorbed by annealing the sample in vacuum at 400 K until desorption ceased. The vibrational spectrum of the remaining CO bound to the rough step sites was then obtained (middle spectrum, fig. 9). Desorption of CO from the rough step sites is illustrated by the lower desorption spectrum in fig. 9 which was
M.R.
McCiellun
et ul. /
ENERGY LOSS
Curbon
ICI-II-‘)
monoxide
adsorption
OIT kinked
Pt(321)
TEMPERATURE
(K )
Fig. ‘9.A series of high resolution
electron energy loss spectra taken after carbon monoxide is (a) adsorbed at 100 K, (b) desorbed from the terraces, and (c) desorbed from the surface. The pretreatment temperature used is indicated by an arrow in the upper desorption spectrum. The lower desorption spectrum was taken following pretreatment at 400 K and indicates that desorption from the terraces is complete. Note that the vibrational results indicate that CO can be bridge bonded or terminally bonded on the terraces.
taken after the vibrational data were obtained. Note that a single high temperature desorption peak is observed confirming that all the CO adsorbed on the terraces has been desorbed and that no redistribution of CO occurs with cooling (all HREELS spectra were taken at 100 K). After the sample was heated to 900 K a HREELS spectrum (bottom spectrum, fig. 9) and an Auger spectrum were taken to verify that all the adsorbed CO desorbed without dissociation..The vibrational spectrum of CO adsorbed on the rough step sites (middle spectrum, fig. 9) clearly indicates that CO interacts with these sites by forming both bridge bonded and terminally bound configurations. The single desorption peak observed indicates that these two configurations result in similar interaction energies with the surface (about 15 1 kJ/mol(36 kcal/mol)). Detailed comparison of the two energy loss spectra for CO adsorbed only on the rough step sites (middle spectrum, fig. 9) and for high coverage CO adsorbed on both terraces and steps (top spectrum, fig. 9) reveals a significant change in peak width. The metal-carbon stretching peak at 480 cm-’ is about 20 cm-’ broader at high coverage while the carbon-oxygen stretching peak at 2095 cm-’ is about 10 cm- ’ broader at high coverage. We interpret this result
as indicating that the vibrational frequencies of terminally bound carbon monoxide molecules are only slightly different for binding to step or terrace atoms. Thus the spectra resulting from step and terrace features together are somewhat broader than those resulting from the step associated molecules alone, as the frequency differences are too small to result in separate resolvable features.
4. Conclusion From the experimental results presented here, the following picture of carbon monoxide adsorption on the atomically rough Pt(321) surface emerges. Carbon monoxide adsorbs on the atomically rough step sites primarily in a terminal atop configuration; however, some carbon monoxide adsorbs at these sites in a bridge bonded configuration. Two chemisorbed states are observed using thermal desorption spectroscopy. Carbon monoxide adsorbed on atomically rough step sites desorbs in a first order peak at 556 K and has a heat of adsorption of 151 kJ/mol (36 kcal/mol). Carbon monoxide desorbs from the terraces at 436 K and has a heat of adsorption of about 96 kJ/mol (23 kcal/mol). Adsorption into these states is not sequential indicating that CO mobility is limited on this rough surface. Finally, under all the conditions employed in these experiments, there was no evidence whatsoever for dissociative chemisorption or any unusual weakening of the carbon monoxide bond associated with adsorption at the step or kink sites.
Acknowledgements It is a pleasure to acknowledge many stimulating discussions with and the expert assistance of Gary E. Mitchell and Edward B. Kollin. One of us (M.R.M.) would like to thank General Motors Research Laboratories for the opportunity to perform these experiments. Partial support for this research was furnished by the MIT Center for Materials Science and Engineering under NSF Grant DMR 78-24185 and by the Alfred P. Sloan Foundation.
References [I] [2] [3] [4] [5] [6] [7]
H. Hopster and H. Ibach, Surface Sci. 77 (197X) 109. M.R. McClellan, J.L. Gland and F.R. McFeeley. in preparation. J.L. Gland, M.R. McClellan and F.R. McFeeley, in preparation. SM. Davis and G.A. Somor~ai. Surface Sci. 92 (1980) 4X9. Y. Iwasawa. R. Mason, M. Textor and G.A. Somorjai, Chem. Phys. Letters 44 (1976) 468 H.J. Krebs and H. Luth. Appl. Phys. 14 (1977) 337. H. Froitzheim, H. Hopster, H. Ibach and S. Lehwald. Appl. Phys. 13 (1977) 147.
[X] P.R. Norton, J.W. Goodale and E.B. Selkirk. Surface Sci. X3 (1979) 189. [9) G. Ertl. M. Neumann and K.M. Streit. Surface Sci. 64 (1977) 393. [IO] D.M. Collins and W.E. Spicer, surface Sci. 69 (1977) 85. [I I] R.W. McCabe and L.D. Schmidt. Surface Sci. 66 (1977) 101. [ 121 R.A. Shigeishi and D.A. King, Surface Sci. 58 (1976) 379. [ 131 I. Fair and R.J. Madix, J. Chem. Phys. 73 (I 980) 3480. [I41 J.F. Nicholas, An Atlas of Models of Crystal Surfaces (Gordon and Breach, New York, 1965) pp. 42-43. [ 151M.A. Van Hove and G.A. Somorjai, Surface Sci. 92 (1980) 4X9. [ 161 M.W. Roberts and C.S. McKee, Chemistry of the Metal-Gas Interface (Oxford Univ. Press, 1978) p. 273. [ 171 D. Edwards, Surface Sci. 54 ( 1976) I. [ 181P.A. Redhead, Vacuum I2 (1962) 203. [ 191 J.L. Gland, B.A. Sexton and G.B. Fisher, Surface Sci. 95 (19X0) 587.
CARBON SURFACE Michael
MONOXIDE
63
ADSORPTION
R. MCCLELLAN
ON THE KINKED
Pt(321)
*
Depurtnwnt of Chemisttyv und Center /or Muteriols Science und Engineering, of Technologv, Cumhndge, Mussuchusetts O_‘I39, USA
Mussuchusetts
Imtrtute
John L. GLAND Ph.vsiwI USA
Chemistq
Deportment,
Generul Motors
Reseurch
L&oratories,
Wurreu. Michigun
48090,
and
F.R. McFEELEY Depurtment of Chemrstryv ond Center/or Muteriuls Science und Engineering, of Technologv, Cumhridge. Mussuchusetts 0.?139, USA Received
23 June
1981; accepted
for publication
5 August
Mussuchusetts
Institute
1981
Carbon monoxide adsorption and desorption were studied on the kinked Pt(321) surface using high resolution electron energy loss spectroscopy, thermal desorption spectroscopy, Auger electron spectroscopy, and low energy electron diffraction. Two chemisorbed states are observed for CO adsorption on the Pt(321) surface. Carbon monoxide adsorbed along the rough steps (which have a high density of kinks) has an adsorption energy of 151 kJ/mol (36 kcal/mol). Carbon monoxide adsorbed on the terraces has an adsorption energy of 96 kJ/mol(23 kcal/mol). Both terminal and bridge bonded CO adsorb on the atomically rough steps although terminal binding is the dominant adsorption mode. Adsorption along the rough steps and terraces does not occur sequentially at 100 K indicating that the CO mobility is restricted on the rough Pt(321) surface. However, sequential desorption does occur since CO desorbs from the terraces before it desorbs from steps. Molecular adsorption of CO on the Pt(321) kinked surface clearly dominates. In fact, weakening of the carbon-oxygen bond does not occur contrary to several recent suggestions that surface defects on platinum weaken the carbon-oxygen bond.
1. Introduction Recently, unique adsorption states of CO have been proposed to exist in conjunction with kink sites on platinum [l] single crystals. In order to systematically investigate this possibility, we report here the results of high * Present
address:
0039-6028/8
Sandia
National
l/0000-0000/$02.75
Laboratories,
Livermore,
California
0 198 1 North-Holland
94550, USA
resolution electron energy loss and thermal desorption measurements on CO adsorbed on the kinked Pt(321) surface. These experiments have also served as a prelude to a complete study of carbon monoxide oxidation on this surface, which will be reported separately [2,3]. The Pt(321) surface is an interesting surface from several viewpoints. First, it is extremely rough on an atomic scale since 40% of the surface platinum atoms are coordinatively unsaturated (see fig. 1). Yet the surface depicted in fig. 1 is stable in vacuum even at high temperature and does not reconstruct in the presence of adsorbates [4]. This type of stable, rough single crystal surface allows us to characterize in detail the effect of local surface morphology on chemisorption and to investigate the possibility of unique surface reaction pathways on platinum kinked surfaces [4,5]. Carbon monoxide adsorption on close packed Pt(ll1) surfaces (which comprise the terraces on the (321) surface) has been well characterized by a number of techniques [ 1,669]. At low coverage the CO adsorbs in a terminal (atop) configuration. As coverage increases a bridged configuration becomes populated, leading to roughly equal numbers of terminal and bridge bonded molecules at saturation. The existence of two bonding configurations is clearly indicated by infra-red reflection spectroscopy [6], high resolution electron
Fig. I. Photograph
of a ball model of the Pt(321) surface.
M. R. McClellm
et ul. /
Curhm
momride
udsorptiort
on kirtkrd
Pt(321)
65
energy loss spectroscopy [1,7], dynamic work function change measurements [8], and low energy electron diffraction [9]. A single broad CO desorption peak develops as coverage increases, indicating that the bonding energies of these two configurations are essentially equal [9- 111. Carbon monoxide adsorption on platinum surfaces with linear or rough steps has not been thoroughly characterized. The coverage dependence of the CO thermal desorption spectra has been studied for two platinum surfaces with linear steps [lo]. On both of these surfaces, two desorption peaks were observed for large CO coverages. The low temperature peak was identified as desorption from the (111) terraces, since its characteristic desorption temperature was identical to that observed in desorption from the ideal (111) surface, while the high temperature peak was identified as desorption from the steps. These two binding states were found to populate sequentially; the higher energy sites along the linear steps populate first. The relative peak areas further indicate that carbon monoxide packs more densely along the steps than on the terraces [lo]. Carbon monoxide adsorption on a stepped platinum surface has also been characterized using high resolution electron energy loss spectroscopy [l]. In contrast to the behavior observed on Pt( 11 l), several CO stretching frequencies were observed at low coverage. The results suggested that the dominant adsorption configuration for CO at low coverages involved terminal adsorption along the top of the linear steps [l], while a weak v(C-0) transition observed to adsorption along the at 1690 cm ~ ’ for low CO coverages is attributed bottom of the linear step [l]. Two additional weak Y(C-0) transitions were observed at 1560 and 1410 cm-‘. The authors argued that these bands are due to carbon monoxide with a substantially weakened carbon-oxygen bond resulting from adsorption at kink sites occurring as imperfections on the stepped surface [l]. At high coverage the usual terminal (2100 cm-‘) and bridge bonded (1860 cm-‘) configurations dominated as the terraces became populated. In addition to this argument for anomalous bond weakening at kink sites, dissociative carbon monoxide adsorption at kink sites on platinum surface has been proposed recently on the basis of X-ray photoemission results [5]. In contrast to these proposals, no evidence has been found in the experiments reported here either for dissociative chemisorption or for molecular CO with an anomalously weakened CO bond, despite the high density of kink sites (20% of the surface) on the crystal face.
2. Experimental The apparatus employed for these experiments is schematically illustrated in fig. 2. The system consists of an UHV stainless steel vacuum chamber equipped with a single pass cylindrical mirror analyzer for Auger electron spectroscopy
66
4 QUADRUPOLE
RESOLUTION ELECTRON ENERGY LOSS SPECTROMETER
Fig. 2. A schematic
+%LT,C,
ARRAY
VIEWPORT
diagram
of the apparatus
with the experimental
stations
labeled
(AES), low energy electron diffraction (LEED) optics, a multiplexed quadrupole mass spectrometer and a 127” cylindrical sector high resolution electron energy loss spectrometer (HREELS). The crystal was spark cut from a single crystal rod and oriented using back-reflection Laue diffraction to within “0.5”. The sample was then spark planed and mechanically polished. In order to remove calcium from the bulk, the sample was treated in flowing atmospheric pressure oxygen at 1300 K for 24 h. Subsequent to this treatment, the crystal was mounted in the UHV system by spot-welding two 0.05 cm diameter platinum wires which served as leads for resistive heating and as thermal links between the crystal and a liquid nitrogen reservoir. The crystal temperature could be varied from 100 to 1300 K as monitored by a 0.008 cm diameter chromel-alumel thermocouple spotwelded directly to the crystal. The surface was cleaned by Argon Ion sputtering at low temperature (- 100 K) to remove small amounts of residual calcium from the surface. Carbon was then removed by heating the crystal in oxygen (= 900 K). prior to carbon monoxide adsorption the surface was characterized using LEED and AES in order to ensure that it was clean and well-ordered. A typical AES of the clean surface produced in this manner is shown in fig. 3. LEED observations of this surface revealed diffraction patterns characteristic of a well-ordered ideal (321) surface. No evidence of surface faceting was observed during these expsriments. After obtaining a clean well-ordered surface, the crystal was cooled to 100 K and dosed with carbon monoxide using a multichannel molecular beam doser. A multichannel molecular beam doser is an array of microscopic
M. R. McClellan
et al. /
Curhor~ motloxide
&sorption
on kinked
Pt(_l.?I)
61
Pt (321) CLEAN
--dN(E dE
I
'0
I
200
I 300
ENERGY
I
I 400
I
I 500
I
6C
(eV)
Fig. 3. An Auger spectrum of the clean Pt(321) surface. The primary beam energy modulation was 2V peak-to-peak; and a sweep rate of 5 eV/s was used.
was 3 keV; the
capillary tubes that produce a directional flow of the adsorbate. The purpose of using a doser is to preferentially expose the front face of the crystal which increases the signal to background in the thermal desorption experiment. To ensure reproducible surface coverages, the crystal was returned to the same position in front of the doser for subsequent doses. All adsorption was done with the crystal at 100 K. The exposures were estimated using an uncalibrated nude ion gauge. Since the actual exposures were made with the aid of a multichannel molecular beam doser, separate thermal desorption experiments were performed to estimate the pressure differential between the crystal and the ion gauge. This was done by comparing total desorption yields for doser exposures and background exposures to CO. The CO flux in front of the doser was found to be enhanced by a factor of - 15. The relative concentration of CO molecules adsorbed for various exposures was determined using integrated thermal desorption spectra (the heating rate was 10 K/s for these experiments). The relative coverages reported in this paper are references to the integrated area of the thermal desorption spectrum shown in fig. 5g (called &,,,). For exposures greater (- 2 X ) than the exposure required to produce 0,, the coverage is called saturation. An estimate of the absolute coverage of adsorbate CO was obtained during CO oxidation studies on this same surface [3]. Carbon dioxide yields were estimated from tempera-
EXPOSURE
(L
1
Fig. 4. The total amount of CO adsorbed on the Pt(321) surface at 100 K as a function exposure. Calibration methods are discussed in the text. One langmwr exposure of CO is 3.X X 10’4/cm2 for a gas at T= 300 K.
of CO
equal to
ture programmed reaction experiments where a known initial amount of adsorbed oxygen was completely consumed. The oxygen concentrations were determined by comparing oxygen Auger results for the oxygen saturated Pt( 111) surface with Auger results for oxygen on the Pt(321) surface. The saturation coverage estimated in this way is approximately 8 X 10’4/cm2. The complete exposure versus coverage curve is shown in fig. 4. The initial slope is constant (consistent with precursor state kinetics), and yields a value between 0.5 and 1.0 for the sticking coefficient. This value is typical for both flat and stepped Pt surfaces [9,1 l- 131. Energy loss spectra were obtained in the specular direction using a - 5 eV primary beam. The relative intensities of individual energy loss features depends on spectrometer tuning and crystal position, therefore relative peak intensities were used only qualitatively. Thermal desorption spectra (TDS) were obtained subsequent to the energy loss spectra (typically a 15-20 min time delay between exposure and desorption). The reproducibility of the data presented here was verified by removing the Pt(321) sample from the system and preparing a new Pt(321) surface. Satisfactory consistency was invariably obtained during spot checks of the data.
3. Results and discussion 3.1. The Pt(321) surface A ball model of the ideal Pt(321) surface is shown in fig. 1 (cf. ref. [14]). In - the usual step notation this surface can be described as Pt(321)
+ Pt(S) - [3( 111) x (3li)])
indicating that the surface is comprised separated by (311) steps of monatomic decomposed as (3li)
-2(loo)
of (111) terraces, three atoms wide, and height. The (311) steps can be further
x (iii).
Thus, the step itself exposes two atom wide sections of open (100) and close packed (111) planes. A more useful representation of this surface is the microfacet notation of Van Hove and Somorjai [17]. In this scheme the surface is described as Pt(321)
+ Pt(S) -[(3,‘2),(lll)
+ (1/2),(11i)
+ (l),(lOO)].
Here the Miller index of the surface has been vectorally decomposed in a basis set made up of low index planes [the (11 l), the (1 Ii), and the (loo)]. These vector decomposition coefficients are given in the parentheses before the low index planes. The subscript on the coefficient is the number of unit cells of that low index plane contained in unit cell of the surface. Using this representation the characteristic parameters of the surface are easily determined. These parameters are summarized in table 1. A close examination of fig. 1 reveals that the platinum surface is quite rough and that the atoms reside in three distinct local environments. These different environments may be characterized by the number of surface nearest neighbors possessed by the atom in question. Using this classification, the surface consists of: (1) terrace atoms with the maximum 6-fold coordination, comprising 60% of the surface; (2) step atoms, 5-fold coordinated, comprising 20% of the surface; and (3) kink atoms, which have only 3-fold coordination and comprise the remaining 20% of the surface. Note that during this discussion we use rough step site as a general term to describe adsorption on or near either step or kink atoms. LEED observations of the crystal revealed a diffraction pattern characteristic of the ideal Pt(321) structure. While the background intensity was somewhat higher than is typically observed on Pt(l1 l), the diffraction features themselves were sharp. Occasionally, extremely small splittings of the LEED features were observed. In view of our ability to reproduce all of our results on two Pt(321) surfaces, we are confident that the crystal consists overwhelmingly of ideal surface and that the results presented in this paper are due to carbon monoxide adsorbing onto this ideal surface. Table I Characteristic
parameters
A. Characteristic
of the Pt(321) surface (i) Kink to kinkz4.8
distances:
(ii) Terrace B. Densities
of different
platinum
types:
width=5.5
A A
(i) Kink density=3.5X lOI cm-’ (ii) Step density=3.5 X lOI cm --z (iii) Terrace density= 1.0X IO” cm-*
No ordered CO overlayer structures were observed using a conventional four grid LEED system during these experiments despite the wide range of coverages and annealing temperatures studied. This observation is not surprising in light of the rough surface used and the lack of mobility for adsorbed CO discussed later in this paper. 3.2. Thermal desorption spectroscopy Thermal desorption of CO from the Pt(321) surface is for a series of CO exposures. For low initial CO coverages /
I
I
I
COVERAGE
Pt (321) co/co TA=
I
I
illustrated in fig. 5 a single first order
IOOK
010028,,, b) 0.060 c) 0.088 d) 0. I 78 elO.498
max mox mox max
I 0
400
600 TEMPERATURE
600 (K)
Fig. 5. A series of carbon monoxide desorption surface at 100 K. The heating rate was IO K/s.
spectra
for various
CO coverages
on the Pt(321)
h4. R. McCleIlun er al. / Curhon monoxide udsorptron on kinked P1(321)
71
desorption peak is observed at 556 K. As the coverage increases the high temperature peak moves down to 546 K and increases in size. For coverages above 0.28,,,,, a second low temperature peak is observed at 436 K. This low temperature peak at 436 K can be identified as desorption of CO bound to the (11 I) terraces by analogy with published results for CO desorption from the Pt( 111) surface [ 1,101. the higher temperature peak at 556 K can therefore be identified as desorption of CO bound to the atomically rough step sites. The desorption temperature, > 500 K, is typical for CO desorption from smooth steps [1,10,11,13]. Further evidence that these two desorption peaks are caused by desorption from two distinct chemisorbed states is furnished by the partial desorption experiment illustrated in fig. 9. All the carbon monoxide in the low temperature peak was desorbed by annealing the sample in vacuum at 400 K until desorption ceased (see arrow in the upper desorption spectrum of fig. 9). Only the high temperature desorption peak is observed in the subsequent desorption experiment (lower desorption spectrum of fig. 9) indicating that no redistribution of CO occurs with cooling. Therefore, the two TDS peaks correspond to two chemisorbed states which desorb sequentially. The desorption parameters for the high temperature peak have been estimated using several methods. The heat of desorption obtained is about 15 1 kJ/mol (36 kcal/mol) regardless of the method used. Fig. 6 illustrated CO desorption from a saturated surface using a series of heating rates. As illustrated in fig. 7 the shift in the high temperature peak with heating rate indicates a heat of desorption of 151 kJ/mol (36.2 kcal/mol) and a preexponential factor of 2 X 1014/s [16]. Alternatively, if we assume that the interactions in the adsorbed layer of CO are negligible at the lowest coverage illustrated in fig. 5 (0.028,,,,), the high temperature peak width and position alone are used to calculate a desorption energy of 15 1 k 16 kJ/mol (36 k 4 kcal/mol) [ 171. Using this desorption heat [ 181 the pre-exponential factor was estimated to be 1 X 1014/s. Finally, using a desorption isostere method the desorption heat estimated for the high temperature peak ranges from about 134 kJ/mol (32 kcal/mol) below 0.28,, up to 151 kJ/mol (36 kcal/mol) between 0.3 and 0.458,,,, [ 191. Since the sticking coefficient is near one (fig. 4), adsorption is not an activated process. Therefore the heat of adsorption should be equal to the heat of desorption. In summary, the heat of adsorption on the Pt(321) surface for CO adsorbed on the rough step sites is approximately 151 kJ/mol (36 kcal/mol). Similar heats have been previously reported for CO adsorbed on the step sites of several platinum surfaces [lo,1 l] confirming that the high temperature peak is caused by desorption of CO from the rough step sites on the Pt(321) surface. Detailed analysis of the low temperature desorption peak is more difficult since the peak is fairly small and overlaps with the leading edge of the high temperature peak. The shift in the low temperature peak with heating rate
HEATING
RATE LK/sl a) 2.4
bl5.0 cl 7.5 ej23.0
270
430
590
TEMPERATURE
750
(K)
Fig. 6. A series of thermal desorption spectra heated at a series of linear heating rates.
,3_
Pt (321) co/co ( sot.) TA =IOOK HIGH TEMPERATURE
for CO desorption
from a saturated
Pt(321) surface
PEAK
IO-
91.75
1.80
I85 IOOO/Tm
Fig. 7. A plot using the temperature to estimate the heat of desorption (36.2 kcal/mol).
1.90
I .95
(K1 (T,,) of the maximum
desorption rate and the heating rate (j3) from the data in fig. 6. The heat of desorption is 151 kJ/mol
M. R. McClelluz
et (11./ Curhon monoxrde adsorpfion on kinked Pt(321)
73
indicates a heat of desorption of 96 k/mol(23 kcal/mol) and a pre-exponential factor of 2.5 X lO”/s. If we assume a pre-exponential factor of 10i3/s, the heat for the low temperature peak is estimated to be 109 kJ/mol (26 kcal/mol) using the peak temperature [18]. Similar heat have been previously observed for CO adsorption on the low index Pt(ll1) surface [9,12] and on the (111) terraces of linear stepped surfaces [9,1 l] again confirming that the low temperature peak is caused by desorption of CO from the terraces of the Pt(321) surface. In contrast to the results obtained on linearly stepped surfaces [ 10,111, CO does not adsorb sequentially first on the rough step sites and then on the terrace sites. For instance, at a CO coverage of 0.49(3,, (spectrum e, fig. 5) where the low temperature terrace peak is clearly visible the high temperature step peak has attained only about half of its saturation intensity. Above about half saturation coverage both chemisorbed states are populated at about the same rate. This observation indicates that CO mobility is quite low on this rough surface even near the desorption temperatures, as substantial mobility coupled with significant site to site variation in the binding energy would result in stepwise adsorption. As the surface approaches saturation coverage the high temperature peak becomes approximately twice as large as the low temperature peak. Since the rough step sites cover 40% of this surface, these desorption results imply that the carbon monoxide surface concentration is roughly three times larger near the rough step sites than on terrace sites. Similar density comparisons for CO adsorbed on linearly stepped surfaces have been reported previously [lo]. Because of the large concentration of step and kink sites on the Pt(321) surface the low temperature terrace desorption peak does not dominate the high temperature step desorption peak even at saturation coverage. McCabe and Schmidt [ 1 l] previously observed similar behavior on the rough Pt(210) surface. 3.3. Vibrational spectroscopy A series of high resolution electron energy loss spectra (HREELS) for increasing CO concentrations adsorbed on the Pt(321) surface are shown in fig. 8. Two features are observed at 2065 and 480 cm-’ for low CO coverages (spectrum a, fig. 8). These vibrational transitions are characteristic of CO bound to the platinum surface in a terminal (atop) configuration. The low frequency v(C-0) transitions previously assigned to adsorption at kink sites [l] were never observed for CO adsorption on this kinked surface. As CO coverage increases a new feature appears at 1865 cm-’ which is characteristic of bridge bonded CO. No new low frequency modes accompany the growth of the bridge bonded transition. As coverage increases the terminal band at 2065 cm-’ moves up to 2095 cm-. This increase in frequency is similar to the shift observed for CO adsorption on the Pt( 111) surface ascribed to dipole-dipole coupling [ 121.
COVERAGE al 0020mlJx
+ co TA = IOOK
‘t (321)
b) 0000max c) 0 17%,3x d) %ox
16
2075 \
jyLy$
1
0
I
I, 400
I 800
ENERGY
I 1200
LOSS
I, 1600
I 2000
(cm-‘)
Fig. 8. A series of high resolution electron energy loss spectra the Pt(321) surface at 100 K. The beam energy was -5 eV.
for carbon
monoxide
adsorbed
on
In order to establish the relationship between the two chemisorbed states observed by desorption and these vibrational results several partial desorption experiments were performed. Typical results are illustrated in fig. 9. A vibrational spectrum was first obtained from a CO saturated surface at 100 K (top spectrum, fig. 9). All the CO in the low temperature terrace peak was then desorbed by annealing the sample in vacuum at 400 K until desorption ceased. The vibrational spectrum of the remaining CO bound to the rough step sites was then obtained (middle spectrum, fig. 9). Desorption of CO from the rough step sites is illustrated by the lower desorption spectrum in fig. 9 which was
M.R.
McCiellun
et ul. /
ENERGY LOSS
Curbon
ICI-II-‘)
monoxide
adsorption
OIT kinked
Pt(321)
TEMPERATURE
(K )
Fig. ‘9.A series of high resolution
electron energy loss spectra taken after carbon monoxide is (a) adsorbed at 100 K, (b) desorbed from the terraces, and (c) desorbed from the surface. The pretreatment temperature used is indicated by an arrow in the upper desorption spectrum. The lower desorption spectrum was taken following pretreatment at 400 K and indicates that desorption from the terraces is complete. Note that the vibrational results indicate that CO can be bridge bonded or terminally bonded on the terraces.
taken after the vibrational data were obtained. Note that a single high temperature desorption peak is observed confirming that all the CO adsorbed on the terraces has been desorbed and that no redistribution of CO occurs with cooling (all HREELS spectra were taken at 100 K). After the sample was heated to 900 K a HREELS spectrum (bottom spectrum, fig. 9) and an Auger spectrum were taken to verify that all the adsorbed CO desorbed without dissociation..The vibrational spectrum of CO adsorbed on the rough step sites (middle spectrum, fig. 9) clearly indicates that CO interacts with these sites by forming both bridge bonded and terminally bound configurations. The single desorption peak observed indicates that these two configurations result in similar interaction energies with the surface (about 15 1 kJ/mol(36 kcal/mol)). Detailed comparison of the two energy loss spectra for CO adsorbed only on the rough step sites (middle spectrum, fig. 9) and for high coverage CO adsorbed on both terraces and steps (top spectrum, fig. 9) reveals a significant change in peak width. The metal-carbon stretching peak at 480 cm-’ is about 20 cm-’ broader at high coverage while the carbon-oxygen stretching peak at 2095 cm-’ is about 10 cm- ’ broader at high coverage. We interpret this result
as indicating that the vibrational frequencies of terminally bound carbon monoxide molecules are only slightly different for binding to step or terrace atoms. Thus the spectra resulting from step and terrace features together are somewhat broader than those resulting from the step associated molecules alone, as the frequency differences are too small to result in separate resolvable features.
4. Conclusion From the experimental results presented here, the following picture of carbon monoxide adsorption on the atomically rough Pt(321) surface emerges. Carbon monoxide adsorbs on the atomically rough step sites primarily in a terminal atop configuration; however, some carbon monoxide adsorbs at these sites in a bridge bonded configuration. Two chemisorbed states are observed using thermal desorption spectroscopy. Carbon monoxide adsorbed on atomically rough step sites desorbs in a first order peak at 556 K and has a heat of adsorption of 151 kJ/mol (36 kcal/mol). Carbon monoxide desorbs from the terraces at 436 K and has a heat of adsorption of about 96 kJ/mol (23 kcal/mol). Adsorption into these states is not sequential indicating that CO mobility is limited on this rough surface. Finally, under all the conditions employed in these experiments, there was no evidence whatsoever for dissociative chemisorption or any unusual weakening of the carbon monoxide bond associated with adsorption at the step or kink sites.
Acknowledgements It is a pleasure to acknowledge many stimulating discussions with and the expert assistance of Gary E. Mitchell and Edward B. Kollin. One of us (M.R.M.) would like to thank General Motors Research Laboratories for the opportunity to perform these experiments. Partial support for this research was furnished by the MIT Center for Materials Science and Engineering under NSF Grant DMR 78-24185 and by the Alfred P. Sloan Foundation.
References [I] [2] [3] [4] [5] [6] [7]
H. Hopster and H. Ibach, Surface Sci. 77 (197X) 109. M.R. McClellan, J.L. Gland and F.R. McFeeley. in preparation. J.L. Gland, M.R. McClellan and F.R. McFeeley, in preparation. SM. Davis and G.A. Somor~ai. Surface Sci. 92 (1980) 4X9. Y. Iwasawa. R. Mason, M. Textor and G.A. Somorjai, Chem. Phys. Letters 44 (1976) 468 H.J. Krebs and H. Luth. Appl. Phys. 14 (1977) 337. H. Froitzheim, H. Hopster, H. Ibach and S. Lehwald. Appl. Phys. 13 (1977) 147.
[X] P.R. Norton, J.W. Goodale and E.B. Selkirk. Surface Sci. X3 (1979) 189. [9) G. Ertl. M. Neumann and K.M. Streit. Surface Sci. 64 (1977) 393. [IO] D.M. Collins and W.E. Spicer, surface Sci. 69 (1977) 85. [I I] R.W. McCabe and L.D. Schmidt. Surface Sci. 66 (1977) 101. [ 121 R.A. Shigeishi and D.A. King, Surface Sci. 58 (1976) 379. [ 131 I. Fair and R.J. Madix, J. Chem. Phys. 73 (I 980) 3480. [I41 J.F. Nicholas, An Atlas of Models of Crystal Surfaces (Gordon and Breach, New York, 1965) pp. 42-43. [ 151M.A. Van Hove and G.A. Somorjai, Surface Sci. 92 (1980) 4X9. [ 161 M.W. Roberts and C.S. McKee, Chemistry of the Metal-Gas Interface (Oxford Univ. Press, 1978) p. 273. [ 171 D. Edwards, Surface Sci. 54 ( 1976) I. [ 181P.A. Redhead, Vacuum I2 (1962) 203. [ 191 J.L. Gland, B.A. Sexton and G.B. Fisher, Surface Sci. 95 (19X0) 587.