Oxygen adsorption on palladium(100) surface ... - ACS Publications

Kamil Klier, Gary W. Simmons, Kenneth T. Park, Yarw-Nan Wang, James S. Hess, and Richard G. Herman. Langmuir 1998 14 (6), 1384-1391. Abstract | Full T...
14 downloads 0 Views 2MB Size
4522

J. Phys. Chem. 1991,95,4522-4528

Oxygen Adsorption

on Pd(100) Surface: Phase Transformatlons and Surface

Gary W.Simmons,* Yam-Nan Wang, Juan Marcos, and Kamil Klier Zettlemoyer Center for Surface Studies and Department of Chemistry, Lehigh University, Bethlehem, Pennsylvania 1801 5 (Received: December 4, 1990)

The various adsorbed oxygen phases p(2X2), 0.25 ML (ML = monolayer); c(2X2), 0.50 ML; p(5X5), 0.68 ML; and (V'5Xd5)R27°, 0.80 ML were produced under conditions of oxygen pressure and surface temperatures that resulted in the refinement of the phase diagram for this system. In addition, novel interpretationsof the surface structures and surface vibrational spectra are offered. We found for surface coverages of 0.35 ML 5 0 I0.80 ML that a mixture of the dense (V'5Xd5)R27° and less dense y(2X2)" is formed when the surface is heated to a temperature just below that required for oxygen desorption, irrespective of the phase(s) originally present. The phase transformation can be accounted for by an oxygen-stabilized surface reconstruction of the palladium surface. Proposed lateral shifts of palladium atoms in the top layer give rise to distorted fourfold sites that are stabilized by an increase in the palladium-oxygen bond strength. Evidence for oxygen adsorbed in distorted fourfold sites is afforded by the appearance of a HREELS vibration at 450 cm-' that is "silent" in fully symmetric fourfold sites. The implications of these results for oxidation reactions are discussed.

Introduction Certain transition metals have been reported to be active as methane oxidation catalystdf and in the case of palladium to give rise to partial oxidation products when promoted with chloromethane~.',' Because of interest in the mechanism of partial oxidation of methane as well as in finding the most suitable metallic catalyst(s) and selective reaction conditions, we conducted surface studies on single crystals of palladium. These studies have included reactions of palladium with oxygen, chlorine, dichloromethane, methane, and various combinations of these. In this paper, we present and discuss the results of our surface science studies of the reaction of oxygen with Pd( 100) surfaces. A number of studies of the Pd( 1OO)/oxygen system have been reported for a wide rangc of temperatures and surface coverage ~ For. coverages ~ ~of less ~ than 0.25 ML (ML = monolayer), oxygen dissociatively adsorbs with oxygen atoms occupying the fourfold hollow sites5characterized by a HREELS loss peak at 350 cm-l.- Adsorption of atomic oxygen occurs with a sticking coefficient of nearly unity? suggesting that oxygen dissociation is nonactivated and takes place via a molecularly adsorbed precursor state. For temperatures between 180 and 600 K the atomic oxygen exhibits an ordered ~ ( 2 x 2 phases-12 ) indicating N N and 2NN repulsive interactions and attractive 3NN interactionsss1*l2 (NN = nearest neighbor, 2NN = second nearest neighbor, 3NN = third nearest neighbor). After the ~ ( 2 x 2 sites ) have been filled, oxygen continues to dissociatively adsorb, but at a 2 orden of magnitude slower rate.g For temperatures 300-470 K,the atomic oxygen fills the fourfold sites in an ordered ~ ( 2 x 2 phase ) corresponding to a coverage of 0.50 ML.&12 The rate of formation of the ~ ( 2 x 2 phase ) is still sufficiently fast to indicate that the dissociation takes place from a molecular precursor state. The decrease in rate has been suggested by Imbihl and DemuthI3 to be caused by a decrease of electron donation by Pd as a result of the interaction between ( 1 ) Margotis, L. Ya. Adv. Catal. 1963, 14, 429. ( 2 ) Pitchai, R.; Klier, K. Catal. Rev.-Sci. Eng. 1986, 28, 13. ( 3 ) Cullu, C. F.; Kccnc, D. E.; Trimm, D. L. J . Catol. 1970, 19, 378. (4) Mann, R. S.;h i . M. K. J. Chem. Techno/.Btorechnol. 1979,29, 467. (5) Reider. K. H.; Stocker, W. Surf Sci. 1985, 150, L66. (6) Nykrg, C.; TenpthI, C. G. Solid Srore Commun. 1982, 44, 251. (7) Nykrg, C.;TenptBI, C. G. Surf Sci. 1982, 126, 163. (8) Stuve, E. M.; Madix, R. 3.; Btundle, C. R. Surf. Sci. 1984, 146, 144. (9) Orent, T. W.; Badcr, S. D. Surf. Sci. 1982, /IS,323. (10) Chang, S.-L.;Thiel, P. A. Phys. Rev. Lcrt. 1987, 59, 296. ( 1 1 ) Chang, S.-L.; Thiel, P. A. J . Chem. Phys. 1988,88, 2071. (12) Chang, S.-L.;Thiel, P. A.; Evans, J. W. Surf Sci. 1988. 205. 117. (13) Imbihl, R.; Demuth, J. E. Surf Sci. 1986, 173, 395.

0022-365419112095-4522302.50/0

oxygen and palladium at the sites in the ~ ( 2 x 2 structure. ) Studies of oxygen interactions with Pd( 100) that have been conducted at low temperatures, l e 1 2 0 K, also provide insight into the adsorption mechanism.-JO When a p(2X2) oxygen phase which formed at room temperature was exposed to oxygen at 80 K, a HREELS loss peak at 725 cm-I attributed to molecular oxygen was observed along with the 350-cm-I loss peak from atomic oxygen.&* Reaction of oxygen with a clean Pd(100) at 80 K,however, did not yield equivalent results. In this case, the HREELS spectrum showed not only the 350- and 725-cm-l loss peaks but also a third loss peak at 450 cm-I attributed by Nyberg and TengstHP' to oxygen in a non-fourfold site or to a "dense" oxygen phase. We find this observation particularly relevant to our work, since we observe a 430-cm-I loss peak for dense oxygen structures formed at high temperatures. Since atomic oxygen in the fourfold site is clearly the most stable form of adsorbed oxygen on Pd( 100) at low coverages, the formation of a second atomic oxygen state at low temperatures suggests that kinetic limitations owing to slow surface diffusion of dissociated species can give rise to metastable or nonequilibriumadsorbed phases. Indeed, Chang and ThiellO have observed the formation of a ~ ( 2 x 2 phase ) at 150-180 K and high impingement rates, while at room temperature the ~ ( 2 x 2 phase ) does not form until the ~ ( 2 x 2 phase ) is completely formed at 0.25 ML as noted earlier. Evans" has shown that these results are consistent with a model in which the formation of ~ ( 2 x 2 domains ) in addition to ~ ( 2 x 2 is ) a statistical consequence of random adsorption with NN blocking, where there is insufficient time for expansion to ~ ( 2 x 2 )domains before surrounding sites are filled. An alternative model for the apparent coexistence of the ~ ( 2 x 2 and ) ~ ( 2 x 2 phases ) has been offered by Bartelt, Roelofs, and E i n ~ t e i nin ' ~ which ~ ( 2 x 2 diffraction ) spots arise from disorder in the ~ ( 2 x 2 phase. ) While coverages as high as 0.50 ML can be obtained for oxygen exposures at room temperature, any further increases in coverage require higher temperatures. Ordered oxygen phases have also been observed at higher coverages with a ~ ( 5 x 5 phase ) corresponding to 0.70 ML forming at temperatures above 473 K and a (d5Xd5)R27° phase corresponding to 0.80 ML at temperatures above 573 K.9 Although Orient and Bader9have attributed the high-temperature, high-density surface structures to "surface oxide reconstruction", the structures and state of oxygen at the higher coverages are not well established. Furthermore, as indicated by recent studies by Chang and Thiel,"J2 details of the (14) Evans, J. W. J . Chem. Phys. 1987,87, 3038. (15) Bartclt, N. C.; Roelofs, L. D.; Einstein, T. L. Sur/. Sci. &It. 1989, 221, L750. Q - 1991 American Chemical Societv

Oxygen Adsorption on Pd( 100) Surface

The Journal of Physical Chemistry, Vol. 95, No. 11, 1991 4523 1.0

growth of these higher density oxygen phases and the transformations between them are more complex than found for the less dense phases. The characterization of oxygen in the dense phases and the transformations occurring in these phases as a function of coverage and temperature were a major focus of the present study.

Experimental Section All experimental work was conducted in two bakeable, stainless steel, ultra-high-vacuum systems. One comprised a three-grid 'display-type" low-energy electron diffraction (LEED) optics, an Auger electron spectrometer (AES) with cylindrical mirror analyzer (CMA), and a mass spectrometer (MS) for temperature-programmed desorption (TPD), while the other housed a high-resolution electron energy loss spectrometer (HREELS) and a mass spectrometer. Each unit was pumped by 200 L/s ion pumps boosted by a titanium sublimator, and each operated at a base pressure of low lO-'O Torr. The Pd( 100) crystals were spotwelded between two 0.25 mm diameter Ta wires. These wires allowed rapid resistive heating of the crystals to 1200 K as measured by a chromel-alumel thermocouple spot welded to the crystal edge. The unit equipped with HREELS was also capable of cooling the crystal to about 90 K with liquid N1. LEED patterns were taken with the incident electron beam energy between 50 and 150 eV. All LEED patterns were recorded at room temperature. In the AES operation, the primary electron beam energy was 3 keV, the beam current was 15-50 PA, and the modulation voltage on the CMA was 3 V peak-to-peak. During the thermal desorption experiments, the crystal was positioned about 1 cm in front of the quadmpole mass spectrometer and heated at a rate of ca. 40 K/s. The mass spectrometer was controlled by a microcomputer, and several mass peaks could be monitored during thermal desorption. HREELS measurements were taken at a specular angle of 60" from the surface normal, an incident beam energy of about 7 eV, and the crystal at room temperature. Elastic peak count rates were on the order of 2 X lo5 counts/s with a typical resolution of 50 cm-'. Sulfur, carbon, nitrogen, and oxygen were encountered at different stages of cleaning the Pd( 100) surface. Sulfur, nitrogen, and oxygen were readily removed by alternately argon ion etching and heating the crystal to 1000 K. The removal of carbon impurities proved to be more difficult, and more difficult to monitor, since the carbon Auger electron signal at 273 eV overlaps one of the palladium peaks at 278 eV. A technique similar to that described in ref 16 was used to monitor the carbon, viz., measuring the ratio of the Auger peak-to-peak intensities of the palladium 278 eV and the principal palladium 323-eV signals. The crystal was heated for 30 min in 5.0 X lo-* Torr of oxygen at 1000 K. The oxygen was pumped out, and then the specimen was cooled to room temperature and ion etched to remove oxygen from the surface. After several of these treatments, further carbon removal was achieved by dosing the crystal with oxygen at room temperature and then flashing the crystal to loo0 K. This led to the desorption of C02, CO, and O2when carbon was present. These procedures were repeated until a constant limiting value of 0.19 for the Pd(278 eV)/Pd(323 eV) ratio was achieved and neither CO nor C 0 2 was observed in the thermal desorption spectrum of an oxygen covered surface. The clean surface produced a sharp bright p(!Xl) LEED pattern. Research-grade gas bulbs (02and Ar, Matheson) were attached to the vacuum chambers through leak valves. To enhance the oxygen dosing pressure at the specimen surface, each unit was equipped with a microcapillary array beam doser. The exposures obtained with the doser were calibrated by comparing the time required to reach a specific surface coverage for the system (back-filled at a known pressure) with the time required to reach the same coverage with the doser controlled to produce a specific background pressure. The pressures at the specimen surface using (16) Behm, R. J.; Christmann, K.;Ertl, G.; Van Hove, M.A. J . Chem. Phys. 1980, 73. 2984.

0.8 v)

w

a 0.6

; 0

0.4 m W 2.

6

0.2

0.0 0.0

0.2

0.4 0.6 0.8 Oxygen Coverage, TDS

1.0

Figure 1. Comparison of the O(507 ev)/Pd(323 eV) Auger peak-to-peak height ratios and the areas under the desorption peaks for different oxygen coverages. The AES and TPD methods were calibrated on the assumption that the ~(2x2)and ~(2x2) structures correspond to 0.25 and 0.50 ML, respectively. 1.oo

0.80 2

I 6 0.80 0

6 0.40 e 0

b 0.20 0.oc

(470)

300

400

(650)

(550)

500 Temperature, K

800

Figure 2. Summary of the oxygen phases observed as a function of surface temperature and oxygen exposures. LEED patterns of the phases enclosed by a solid line were observed at room temperature. The dotted ) indicates that evidence for this phase was line around the ~ ( 2 x 2 phase observed at the temperature indicated.

the doser were determined to be 100 times higher than background. The ratio of the peak-to-peak Auger electron signals for oxygen (507 eV) and palladium (323 eV) was used as a measure of oxygen coverage along with the peak areas in the thermal desorption spectra. Both of these methods were calibrated assuming that the ~ ( 2 x 2 and ) ~ ( 2 x 2 LEED ) structures corresponded to coverages of 0.25 and 0.50 ML, respectively. Excellent agreement was found between the coverages determined by AES and TPD as indicated in Figure 1. To minimize the damage of oxygen overlayers caused by the AES or LEED electron beam, it was necessary to reduce the time of analysis and to reestablish the surface adlayer for each measurement. ReSult.9 The surface structures observed for Pd( 100) as functions of oxygen exposures and surface temperatures are summarized in Figure 2, which shows the conditions required to effect an increase in surface coverage and the concomitant change in surface phase. The LEED structures reported in Figure 2 enclosed by a solid line were recorded at room temperature after achieving the indicated coverages. Visual inspection of the LEED patterns at the higher temperatures showed variations in relative intensities of the (f1/2, *1/2), ~ ( 2 x 2 )and ~ ( 2 x 2 )phases and (0,f1/2), ~ ( 2 x 2 phase ) reflections. The ~ ( 2 x 2phase ) enclosed by a dotted line in Figure 2 indicates that the ~ ( 2 x 2 reflections ) were observed only at the temperatures shown. This observation indicates that ~ ( 2 x 2 and ) ~ ( 2 x 2 phases ) may coexist at elevated temperatures as has been described by Chang, Thiel, and Evans."J2 In addition

4524 The Journal of Physical Chemistry, Vol. 95, No. 1I , I991

r

I

I

...... I

?

.

.

.

.

.

. e . . . .

(b) (0,O) 0

0



0

0

4

0

e . 0

(1,l) (C) Figure 3. (a) LEED pattern of the ~ ( 5 x 5 phase ) and (b) LEED pattern of the (t/5Xt/S)R27° phase. Primary beam energy = 66 eV.

we find the appearance of the ~ ( 2 x 2 reflections ) for the ~ ( 2 x 2 ) phase at elevated temperatures even though the coverage is only 0.25 ML. The low coverage ~ ( 2 x 2 )phase (i.e. 8 < 0.25 ML) has been shown by Bartelt, Roelofs, and EinsteinIs to arise from the disordering of a ~ ( 2 x 2 phase ) through an intermediate ~ ( 2 x 2 ) phase. The appearance of diffraction spots for the ~ ( 2 x 2 ) structure is proposed to be a long-range ~ ( 2 x 2order ) arising from preferential occupation of two of the four possible ~ ( 2 x 2 sub) lattices lying on the same ~ ( 2 x 2 sublattice. ) The loss of long-range order in the ~ ( 2 x 2 )phase makes the ~ ( 2 x 2 diffraction ) spots diffuse and weak. Even at a coverage of 0.25 ML a transition through a ~ ( 2 x 2 )phase is predicted as the temperature is increased.Is The high-temperature p(2X2)/c(2X2) reflections should perhaps be referred to as a disordered ~ ( 2 x 2 )phase or "(2x2)" phase rather than as a coexistence of ~ ( 2 x 2 and ) ~(2x2) phases. For adsorption at room temperature, a well-ordered ~ ( 2 x 2 ) phase corresponding to 0.25 ML is formed after an oxygen exposure of 2 langmuirs, and a ~ ( 2 x 2 )phase corresponding to a saturation coverage of 0.50 is formed after 180-langmuir exposure (1 langmuir = 1Od Torr s). To produce any further increase in oxygen adsorption, a surface temperature of 470 K was required at an oxygen pressure of 8.0 X Torr to form the ~ ( 5 x 5 ) structure, whereas the temperature had to be increased to 550 K to form the (d5Xd5)R27' structure at this pressure. Extended heating of the ~ ( 5 x 5 )and (d5Xd5)R27' structures under 8.0 X Torr oxygen at 470 and 550 K,respectively, did not produce any changes in the structures or any increases in oxygen coverage, indicating that the fifth-order structures are equilibrium phases under these conditions of temperature and pressure. The coverages for the ~ ( 5 x 5 )and (d5Xd5)R27° structures were determined by AES and TPD17 to be 0.68 and 0.80 ML, respectively. LEED patterns of the ~ ( 5 x 5 ) and (d5Xd5)R27', are shown in Figure 3. In each case, a sketch of a section of the pattern is provided for clarity. These results are in general agreement with those reported earlier? As shown in Figure 2, the ~ ( 2 x 2 phase, ) which had been formed at room temperature, undergoes an irreversible surface phase (17) TPD and Kinetic Model Paper, to be published.

Simmons et al. transformation at about 470 K in vacuo to a mixed ~ ( 5 x 5 and ) "(2x2)" without loss of oxygen to the gas phase. A further irreversible phase transformation at 550 K in vacuo results in the formation of mixed "(2x2)" and (1/5Xd5)R27' phases. In fact, these transformations occur for any initial oxygen coverage greater than about 0.35 ML. A well-ordered ~ ( 5 x 5 phase ) also changes irreversibly to a mixture of the "(2x2)" and (d5Xd5)R27' phases when heated to 550 K in vacuo while a well-ordered (d5XdS)R27' structure does not undergo any phase transformation upon heating to 550 K. Heating the palladium crystal with initial coverages greater than 0.35 ML in vacuo at temperatures between 550 and 650 K resulted in partial desorption of oxygen and the formation of mixed (d5Xd5)R27' and "(2x2)" phases followed by the formation of a single "(2x2)" phase upon further loss of oxygen. It is particularly noteworthy that, during the decomposition of the (d5Xd5)R27' phase to "(2x2)" with partial desorption of oxygen, neither the ~ ( 5 x 5 ) nor the ~ ( 2 x 2 patterns ) were observed at room temperature for any of the intermediate coverages. Loss of oxygen from the "(2x2)" phase subsequently occurred at temperatures above 650 K and gave a clean p( 1X 1) surface. Most of our observations of phase transitions are in agreement with studies by Chang et al.11J2who reported the results for the coverage range of 0.5 14.60 ML and for temperatures between 400 and ca. 650 K. We did not observe the partial reversibility of the ~ ( 5 x 5 to ) (d5Xd5)R27' + "(2x2)" transition reported by Chang et a1.12 This discrepancy can be attributed to greater sensitivity for measuring the diffraction spot intensities at elevated temperatures in the LEED system used by these investigators. Neither we nor Orent and Baderg encountered the difficulty reported by Chang et a1.12 in preparing a single (d5Xd5)R27' phase. There is an apparent property of the surface such as the presence of defects that affects the activation barrier for the formation of ( 4 5 X d5)R27'. All of the oxygen phases described above were readily reduced at room temperature by residual CO or H2 after the gas-phase oxygen was pumped out. The ~ ( 5 x 5 phase ) was reduced directly to ~ ( 2 x 2 without ) the formation of an intermediate ~ ( 2 x 2 phase, ) and the (d5Xd5)R27' phase Was reduced directly to the ~ ( 2 x 2 ) phase without the formation of either the ~ ( 5 x 5 or ) the ~ ( 2 x 2 ) phase. The ~ ( 2 x 2 )phase was reduced directly to the ~ ( 2 x 2 ) structure. The ~ ( 2 x 2 phase ) was in all cases subsequently reduced to a clean surface before any adsorption of residual CO or H2 was observed, indicating that oxygen was not displaced by either of these gases but was removed as C 0 2or H 2 0 that readily desorbed at room temperature. The rate of reduction of adsorbed oxygen was found to be a function of the nature of the initial oxygen phase. For the single ~ ( 2 x 2 phase, ) reduction time at a total CO and H2 residual pressure of ca. Torr was found to be 5-10 min, corresponding to nearly unit efficiency, whereas, for the single (d5Xd5)R27' phase, the reduction time was on the order of 8 h. Even though extended time was required for the reduction of the more dense phases, the subsequent ~ ( 2 x 2 )phase was consistently reduced within 5-10 min. The fact that oxygen in the dense phase is more difficult to reduce than in the less dense suggests that the oxygen is more tightly bound to palladium sites in the dense phase. Alternatively, CO and H2 may not readily adsorb on the dense oxygen surface phase whereas there are available adsorption sites in the less dense phase. The HREELS spectra for each of the oxygen phases are r e p resented in Figure 4. For both the ~ ( 2 x 2 )and the ~ ( 2 x 2 ) structures, a single loss peak at 350 cm-' corresponding to atomic oxygen in the fourfold hollow sites was observed, in agreement with the results reported by Nyberg and Teng~tAI~.~ and Stuve, Madix, and Brundle.* For the ~ ( 5 x 5 structure, ) the loss peak at 350 cm-l became broader and a shoulder was observed near 400 cm-'. Upon the formation of the ( d W d 5 ) R 2 7 ' phase, this shoulder developed into a well-resolved peak at 430 cm-' in addition to the 350-cm-' transition. As pointed out earlier, the 430-cm-' loss found here for the dense oxygen phases coincided with that reported by Nyberg and Teng~tAl~.~ and Stuve et aL8 for oxygen adsorption on Pd(lO0) at 80 K. The fact that the

Oxygen Adsorption on Pd( 100) Surface

I '

I

I

0

I

I

I

I

1

645 Energy

The Journal of Physical Chemistry, Vol. 95, No. 11, 1991 4525

I

1290 LOSS

2

I

b

I

1935

(cm-')

Figure 4. Electron energy loss spectra (HREELS) of (a) p(2X2), (b)

~ ( 5 x 9and , (c) (d5Xd5)R27°. The spectrum from ~(2x2)is identical with that from ~(2x2) and is not shown.

characteristic 430-cm-' transition is not found for intermediate temperatures between 80 and 470 K will be discussed later in the context of a model proposed for the high-density oxygen phases on Pd( 100) surfaces.

Discussion In the discussion that follows, we develop a model that consistently accounts for (i) the splitting of the HREELS 350-cm-I loss of adsorbed oxygen in the fourfold sites at low coverages (10.5 ML) into two peaks at 350 and 430 cm-l for oxygen in the high density (>0.50 ML) phases, and (ii) the irreversible phase transformation from the ~ ( 2 x 2 and ) ~ ( 5 x 5 phases ) at ca. 550 K to the (d5Xd5)R27O phase and the "(2x2)" phase. We start our discussion by using the HREELS results to propose a model for the oxygen site in the high-density ~ ( 5 x 5 ) and (d5Xd5)R27O structures. Orent and Baderghave proposed that these structures may be oriented layers of PdO on Pd( 100). A PdO(1 lO)//Pd(l00) registry was suggested for the ~ ( 5 x 5 ) structure and Pd0(001)//Pd( 100) for the (dSXd5)Z?27O structure. We believe that an alternative model can be proposed for several reasons. First, the vibrational spectrum of PdO shows two bands in the 600-cm-I regionk8J9which are significantly higher than the 3 5 M 5 0 cm-' observed here for the high-density phases. Second, the (d5Xd5)R27O is not unique to palladium as implied by coincidence by the PdO overlayer with the palladium substrate. A (d/SXd5)R27O structure has also been reported for the Mo(100)/oxygen system for which oxygen coverage was 0.80 ML.wa Alkali metal ion scattering from this surface has shown that oxygen occupies only the fourfold sites:) although the authors indicated that they would not be able to resolve small distortions in these sites. Since, for oxygen coverages greater than 0.50 ML, some of the nearat-neighbor fourfold sites must be occupied, a change in surface structure is therefore necessary to overcome the NN repulsive interactions that are apparent by the formation of the ~ ( 2 x 2 and ) ~ ( 2 x 2 structures ) at coverages 50.50 ML. We propose that the stabilization of the high-density oxygen phases arise from a surface reconstruction of the top layer of palladium atoms that results in a distorted potential in the fourfold site. The HREELS spectra of surface oxygen in the dense structure provides evidence for this reconstruction. First, the (18) Kliche, K. Infrared Phys. 1986, 25, 381. (19) Goncharenko, G. I.; Lazarev, V. B.; Shaplygin, 1. S . Russ. J . Inorg. Chem. 1985, 30, 1273. (20) Miles, S.L.; Bernasek, S.L.; Gland, J. L. J. Phys. Chem. 1983,87, 1626. ( 2 1 ) Miles, S.L.; Bernasek, S. L.;Gland, J. L. Surf: Sei. 1983, 127, 271. (22) Zhang, C.; Van Hove, M. A.; Somorjai, G. A. Surf: Sei. 1985, 149, 326. (23) Overburg, S.H.; Stair, P.c. J . Vac. Sci. Techno/. 1983, Al, 1055.

Figure 5. The geometries of the Pd40 clusters used in the GF-Wilson matrix calculations: (a) the oxygen adsorbed onto a fourfold CdYsite and (b) the oxygen in a distorted C,,site. A, and E represent the eigenvectors of the vibrational motion. In (b) the E mode of (a) is projected onto the z axis and thereby becomes HREELS active.

possible assignment of the 350- and 430-cm-' losses to oxygen on bridge sites and to oxygen on top sites, or to any peroxo oxygen species, is ruled out because all of these species have different vibrational frequencies. For the adatom frequencies, a general rule has been established" that, for well-defined adsorption sites of a given adatom-metal configuration, the frequency of the AI mode (i.e., vibration mode perpendicular to the surface) decreases with increased ligancy, and vice versa for higher ligancy. Gates and Kesmode12sand Imbihl and Demuthk3reported a 480-cm-I loss for the threefold site of oxygen on Pd(ll1). Since the losses for the bridge and top sites would be expected to be even higher in energy than 480 cm-', neither the 350- or the 430-cm-' loss can be assigned to any of these sites. The presence of peroxo oxygen species in the dense oxygen structures on Pd( 100) can be ruled out since loss peaks at 700-850 cm-' previously attributed to these speciesk3were not observed. The basis for interpreting the HREELS spectra as evidence for oxygen adsorbed on distorted fourfold sites in the dense phases is given as follows. The undistorted fourfold site Pd40 has the symmetry C, and nine normal modes (2AI @ 2E @ 2BI @ B2), three of which are not infrared active (2B1 and B2),26 Of the remaining six modes (2A1 and 2E), only the AI modes ( u I and u2) satisfy the specular HREELS selection rules. The E modes (v6 and v,) are silent modes because the surface dipole selection rules that govern specular HREELS predict that only modes with a transition dipole moment perpendicular to the surface plane are excited.27" If the oxygen atom is moved out of the center position in the fourfold site (e.g., by lateral displacement that lowers the symmetry to C l h ) at , least one of the components of the C, E modes will acquire a nonzero value of the displacement eigenvector in the direction perpendicular to the surface plane and the mode will become active in specular HREELS as illustrated in Figure 5. Although the foregoing discussion presents a strong argument for a distorted oxygen site in the dense structure, a unique model for the reconstructed surface is more speculative. Calculations (24) Ibach, H.; Mills, D. L. EIectron Energy Loss Spectroscopy and Surface Vibrations;Academic Press: New York, 1982. (25) Gates, J. A.; Kesmodel, L. L. J . Caral. 1984, 83, 437. (26) Herzberg, G. Infrared and Raman Spectra: Van Nostrand Co. Inc: Princeton. NJ, 1945. (27) Ibach, H. Sur/. Sei. 1977, 66, 56. (28) Baro, A. M.: Ibach, H.; Bruchmann, H. D. Surf: Sci. 1979,88,384.

4526 The Journal of Physical Chemistry, Vol. 95, No. 11, 1991

Simmons et al.

a o o o o 8

c

0 O o o

0 0 - , 0 0 Q

O

O

O

~

~

@

O @ @ * O O ( 1 ,O)

(a)

(C)

F i 7. Clusters used in the calculation of vibrational frequencies by the GF-Wilson method: (a) Pd.0, C&; (b) PddO, CIA;and (c) PdlbOp

(b) Figure 6. (a) Proposed structure and simulated LEED pattern for the ~(5x5)phase, and (b) proposed structure and simulated LEED pattern for the (d5Xd5)R27° phase. The unit cells are marked by dashed lines.

of the expected vibrational frequencies and simulated kinematic diffraction patterns of several models were made. We present here the model that gave the best agreement between the calculated results and experiment. The important features of the model for the dense phase include a surface reconstruction that leads to an increase in the surfaceoxygen bond strength and a distortion of the oxygen bonding site. We propose that these features should be included in any proposed structure that differs from the one presented below. We suggest that the distorted fourfold site arises from a reconstruction of the top layer of palladium atoms that this reconstruction is sensitive to both oxygen coverage and temperature. The simplest reconstruction that is consistent with both the HREELS and LEED observations is a lateral shift of palladium atoms in the top layer. Surface reconstruction of this type has been proposed by Estrup et al.29who reported LEED evidence for lateral shifts of tungsten atoms in the top layer for the W(1 10)/H2 system. At hydrogen coverages above a critical value, the reconstruction took place over a temperature range of 85-300 K and did not require long-range order in the adsorbed layer. The shift in the top tungsten layer reduced the symmetry of the adsorption site and thus deepened the potential well in that site. The reconstruction was stabilized by the resulting increase in binding energy of hydrogen. Proposed shifts in the top layer of Pd( 100) surface atoms and positions of atomic oxygen that account for the ~ ( 5 x 5 )and (d5Xd5)R27° LEED patterns are shown in Figure 6. The coordinates of the palladium and oxygen atoms in each of these structures are given in Appendix I. Note that for oxygen in the dense structure, Figure 6b, the distance between the nearest neighbors is 1.94 A which is greater than the 1.21 A in molecular oxygen. The diameter of oxygen atoms in the unreconstructed surface, as measured by He diffra~tion,~ is 1.55 A so that there is sufficient room for pairs of oxygen atoms in the proposed structure. It is noted that Pd-0 "bonds", i.e., lines connecting the 0 adatom with its nearest Pd substrate atom, are at 90° with each other. Therefore, the dipoltdipole repulsion of the nearest Pd-0 dipoles vanishes, allowing the close approach of the two 0 adatoms. In contrast, the dipole moment of the oxygen adsorbed into the fourfold sites at low oxygen coverage is perpendicular to the surface, and the resulting dipoltdipole repulsion gives rise to the rare phase surface structures. (29) Chung, J. W.; Ying, s. C.; Estrup, P.J. Phys. Reo. Len. 1986, 56,

149.

Mode: A,(vJ Frequency, cm-l: 350 Eigenvector, A: 0.240

0.054

Mode: E ( v e l Frequency, cm-1: 430 Eigenvector, A: 0.000

E (v,) 136 0.000

Ai(vd 1eo

Figure8. Calculated vibrational frequencies and eigenvectors for various modes for the Pd40, C, sites. The eigenvectors represent the I component of the oxygen displacement.

In Figure 6 the shifts of the palladium atoms are exaggerated for purposes of illustration. Shifts of 5% from the ideal positions were used for simulating the LEED patterns and for the calculation of the oxygen-palladium vibrational frequencies. To illustrate the relationship between the proposed structure and the diffraction patterns, kinematic modeling of the LEED patterns via a Fourier transform of each structure was performed as described in Appendix 11. The fact that the LEED patterns represent the superposition of two orthogonal domains was taken into account. Comparison of the transforms shown in Figure 6 with the corresponding LEED patterns in Figure 3 demonstrates reasonable agreement within the limits of the kinematic calculations used. Both the symmetric and distorted fourfold sites depicted in Figure 7 were subjected to calculations of vibrational frequencies using the Wilson GF-matrix m e t h ~ d ~detailed * ~ I in Appendix I. The calculation for the symmetric C, site shows that the E mode ( v 6 ) has an eigenvalue higher than the A, mode ( u t ) and that the other two modes (Al or u2 and the degenerate E or Y,) have frequencies below the minimum frequency that can be resolved in the tail of the elastic peak (