Langmuir 1989,5, 316-318
316
Infrared Spectroscopic Study of C12/02 Coadsorption on Al( 11 1) V. M. Bermudez* and A. S. Glass*>+ Naval Research Laboratory, Washington, D.C. 20375-5000 Received August 8, 1985. I n Final Form: September 28, 1988 Fourier transform infrared reflection-absorption spectroscopy (FT-IRRAS) has been used to study the adsorption of chlorine on clean and on O,-exposed Al(111) surfaces. On clean Al, C1 at surface sites gives rise to a stretching mode at 745-760 cm-' (depending on coverage), and a second band at 820 cm-' is tentatively assigned to subsurface C1. Exposure to 0, either before or after Cl, leads ultimately to a single broad band at a frequency intermediate to those for exposure of the clean surface to either Cl, or 0,. The results indicate strong coupling between A1-0 and A1-Cl modes and suggest that the species are intimately mixed during coadsorption. Infrared reflection-absorption spectroscopy (IRRAS) is becoming an important technique for obtaining structural information concerning adsorbates on surfaces. The high resolution ( 4 5 cm-') attained in many IRRAS experiments, together with the nondestructiveness and freedom from the requirement for ultrahigh vacuum, make IRRAS useful in studies of complex reactions on surfaces. One class of surface chemical phenomena in which IRRAS has been applied'" is "coadsorption", wherein a preadsorbed species influences the subsequent adsorption of a second and vice versa. This paper reports work aimed at obtaining insight into the mechanism for chlorination of A1 and the structure of the 0-C1 coadsorbed phase. IRRAS has already been shown6 capable of distinguishing between isolated subsurface 0 and 0 in oxide islands, on the basis of the frequency (830 vs 865 cm-') for vibration normal to the surface, and of observing the conversion of isolated to island species with increasing coverage or with annealing. C1 adsorption on Al(111) has been found7-using Auger spectroscopy, work function measurement, and temperature-programmed desorption-to occur at surface sites at low exposure and also a t subsurface sites at higher exposure. C1 at surface sites gives rise to a broad thermal desorption peak at about 200 "C. Preadsorbed 0 inhibits formation of the subsurface phase but does not greatly affect C1 adsorption a t surface sites. However, the techniques used in ref 7 do not permit direct observation of distinct C1 and 0 species. IRRAS, on the other hand, provides a bond-specific probe with which different species can be distinguished during the chemisorption process. Most of the experimental details (sample preparation, etc.) have been described previ~usly.~J The chamber was equipped with KBr windows according to the design of Hollins and Pritchard.8 Radiation striking the sample was p-polarized and incident at -86" with an angular spread of f4". A Mattson Sirius 100 FTIR system was used with a liquid-nitrogen cooled wide-band MCT (HgCd, Tel-,) photoconductive detector providing a usable signal-to-noise ratio down to 600 cm-l. Bidirectional scans (2000,20-min total scan time) were averaged at 8-cm-' resolution, and fourfold zero-filling of the double-sided interferograms was used to give a point spacing of 1 cm-'. Data obtained at 4-cm-' resolution showed no additional structure. Spectra were recorded for the clean surface and after each successive gas exposure, and the quantity 6R/R (Rex RcJ/Rcl,where Rex ( R c J ,the reflectance of the exposed (clean) surface, was computed. The noise level correN R L / O N T Postdoctoral Research Associate.
0743-7463/89/2405-0316$01.50/0
sponded to 6RIR = 1 X except toward the end of the detector range, where it increased several fold. All spectra are shown "as recorded (no smoothing). The optical path was purged with dry N, to reduce interference in the 640-690-cm-' range from atmospheric COz. Since cooling of the sample to room temperature (following cleaning by sputtering and subsequent annealing6) was slow, all experiments were done at a slightly elevated temperature (30-38 "C). Exposures are in langmuirs, where 1langmuir = lo4 Toms = 3.59 X 1014Oz/cm2or 2.41 X 1014Clz/cm2. Auger analyses were performed before and after each series of IRRAS experiments. Following large C12 exposures on the clean surface a small amount of 0 was the only impurity d e t e ~ t e d . ~ FT-IRRAS results for O/Al(111) for O2 exposures of 5-50 langmuir showed only the subsurface 0 band, with a maximum at 815-860 cm-l depending on coverage, in agreement with previous6 IRRAS data obtained using tunable diode lasers. No additional structure was observed at the present higher resolution (8 vs 20 cm-l). As before! the surface 0 mode seen in EELS (electron energy-loss spectroscopy) at 550-690 cm-I was not detected. For a very large 0, exposure (several thousand langmuirs) the p-polarized 6RIR peak occurs at about 930 cm-', close to the bulk A1,0, LO phonon frequencyg of 950 cm-l. The only available vibrational data for a halogen on clean Al are the EELS work of Chen et for dissociative adsorption of CHJ on Al(111) which indicates an A1-I mode a t 335 cm-'. Presumably, because of its large size, I occupies only surface sites. Assuming a purely quadratic potential, equal AI-Cl and AI-I force constants, and a rigid surface plane, one would expect the corresponding AI-C1 mode at 640 cm-l, Le., an increase by a factor of (MI/Mc1)1/2 = 1.91. Surface-enhanced Raman spectra (SERS) of halogen ions on Ag and Au electrodesll show an M-Cl/M-I (1) Hayden, B. E.;Kretzschmar, K.; Bradshaw, A. M. Surf. Sci. 1983, 125, 366. (2) Tornquist, W. J.; Griffin, G. L. J . Vac. Sci. Technol. A 1986, 4,
1437. ( 3 ) Ryberg, R. J. Chem. Phys. 1985,82, 567. (4) Yates, J. T., Jr.; Trenary, M.; Uram, K. J.; Metiu, H.; Bozso, F.; Martin, R. M.; Hanrahan, C.: Arias, J. Philos. Trans. R. SOC.London, Ser A 1986, 318, 101. (5) Uram, K. J.; Ng, L.; Yates, J. T.,Jr. Surf.Sci. 1986, 177, 253. (6) Bermudez. V. M.: Rubinovitz. R. L.: Butler. J. E. J . Vac. Sci. Technol. A 1988, 6, 717. (7) Bermudez, V. M.; Glass, A. S. J . Vac. Sci. Technol. A. 1989, 7, in press. (8) Hollins, P.; Pritchard, J. J. Vac. Sci. Technol. 1980, 17, 665. (9) Bruesch, P.; Kotz, R.; Neff, H.; Pietronero, L. Phys. Reu. B 1984, 29, 4691. (IO) Chen, J. G.; Beebe, T. P., Jr.; Crowell, J. E.; Yates, J. T., Jr. J. Am. Chem. SOC.1987, 109, 1726.
0 1989 American Chemical Society
Langmuir, Vol. 5, No. 2, 1989 317
IR Study of C12/02Coadsorption
I
I
I
I
,
I
I“
1000
800
,
I
I
,
,
.I
600
cm-1
Figure 1. FT-IRRAS spectra for Cl/Al(111) vs C12 exposure. (6R/R),refers to the fractional change in p-polarized reflectance induced by Clz exposure. The sharp feature at 660 cm-’ arises from atmospheric C02,the concentration of which varies slowly with time as a result of imperfect N2 purging of the optical path.
1000
solo
cn r
600
(M = Ag, Au) frequency ratio of 2.1-2.2, suggesting a larger force constant for the M-C1 bond. This leads to a rough estimate of 700-740 cm-’ for the surface A1-C1 vibration. Recent EELS work by Ng et a1.12 on the electron-stimulated decomposition of (CF,H),O on oxidized Al( 111) suggests that the AI-F mode lies at 870 cm-’. Scaling this value by ( M C ~ / M ~ =) ’1.36 / ~ gives an A1-C1 mode at 640 cm-’, identical with the uncorrected estimate based on the AI-I energy. However, the effect of adsorbed C and 0 on the reported A1-F frequency is not known. The results for exposure of the clean surface to C1, are shown in Figure 1. At the lowest exposure a band is observed at about 745 cm-’ which shifts to about 760 cm-l at 20 langmuirs and reaches a nearly constant intensity. The position of this band agrees reasonably well with the estimate for the surface AI-C1 mode noted above. Briefly annealing the sample a t 200 “C causes nearly complete elimination of this band. These observations are all consistent with C1 adsorption at surface sites. At the lowest exposures a shoulder is seen on the high-energy side of the 745-cm-’ band. This feature develops into a peak at about 820 cm-l and continues to grow slowly in intensity (but not shift significantly) up to 80 langmuirs (the highest exposure used here). This mode energy is considerably higher than that estimated for the surface AI-C1 stretch, and its dependence on Clz exposure is consistent with adsorption a t subsurface sites.7 However, assignment of this band must be done with caution because of the sensitivity of the 780-840-cm-’ region of the Cl/Al(111) spectrum to small amounts of 0 (see below). For a mode frequency of 745 cm-I an Al-35Cl - Al-37Cl isotopic shift of 20 cm-’ with a 3/1 relative intensity (relative abundance) for the higher frequency (lighter isotope) would be expected. The spectra show no clear indication of such a splitting. This might result from dipole coupling13between the Al-35Cl and Al-37Cl modes, the uncoupled (“singleton”) frequencies of which are sep-
Figure 2. FT-IRRAS spectra for a sequence of Clz exposures for Al(111) preexposed to (a) 50 and (b) 10 langmuirs of 02.Note compression of (6R/R),scale relative to that of Figure 1. The vertical lines show the shift, or lack thereof, in the energies of various features in the spectra.
(11) Ping, G.; Weaver, M. J. J. Phys. Chen. 1986,90, 4057. (12) Ng, L.; Chen, J. G.; Basu, P.; Yates, J. T., Jr. Langmuir 1987,3, 1161.
(13) Hollins, P.; Pritchard, J. F’rogr. Surf. Sci. 1985, 19, 275. (14) Hollins, P.; Davies, K. J.; Pritchard, J. Surf. Sci. 1984, 138, 75. (15) Shaw, K. D.; Garrell, R. L.; Krimm, S. Surf. Sci. 1983, 124, 625.
arated by less than the line width. Such interaction is k n o ~ n ’ to ~ Jtransfer ~ intensity from the lower to the higher energy mode and in this case might reduce the Al-37Cl intensity below the detection limit. Verification of this effect would entail recording spectra for a series of 35c12/37c12mixtures of varying relative concentration. Isotopically pure C1, was not available during the course of these experiments. At very low C1 coverage, for which dipole coupling is at a minimum, the signal-to-noise ratio is too low to permit observation of a weaker Al-37Cl component. Mixed isotope experiments would also aid in determining the relative contributions of chemical and dipole coupling effects to the 745-760-cm-’ coverage-dependent shift.13 It is worth noting that the coverage-dependent shift observed15in the SERS of Cl- on Ag has been explained quantitatively in terms of dipole coupling. Figures 2 and 3 show data obtained for exposure to Clz (0,) following exposure to 0, ((31,). The results are not simply sums of O/Al and Cl/Al spectra, indicating that the surface cannot be described as a “phase separated” composite of chlorinated and oxygenated patches and that a model based on a mixture of interacting A1-C1 and A1-0 bonds is more appropriate. The effects of the interaction are seen clearly when a surface pretreated with one species (even at high exposure) is given a small exposure to the other. In all cases significant changes in band intensity and position occur. The simplest situation to discuss is that of C1, exposure following a 50-langmuir exposure to 0, in Figure 2a. The O2preexposed surface (ref 6 and work cited) is “saturated in the sense that all energetically favorable 0 sites are
318 Langmuir, Vol. 5, No. 2, 1989
crrr
Figure 3. Same as Figure 2 but for a sequence of O2exposures for Al(111) preexposed to (a) 37 and (b) 10 langmuirs of Clz.
occupied, and further 0 uptake is very slow. The surface layer is not, however, equivalent to A1,03 since the subsurface 0 vibrational mode, at 850 cm-', is well below the bulk oxide LO frequency of 950 cm-', and the A1 L2,3VV Auger line shape6 shows that not all surface A1 sites are fully 0-coordinated. Preadsorbed 0 blocks C1 adsorption at subsurface sites7but not at surface sites, which saturate a t low C12 exposure. The IRRAS data show that the surface C1 has a significant effect on the subsurface A1-0 mode, shifting the peak 10 cm-' to lower energy and causing an approximately twofold increase in intensity. Further Cl, exposure has little or no effect, consistent with Auger data7 showing that C1 uptake for this O2 preexposed surface ceases when the surface sites are filled. We envision the resulting system as consisting of a layer of subsurface 0 underlying a mixed A1-C1 and A1-0 surface layer. A qualitative interpretation of the data in Figure 2a can be formulated with reference to slab model calculations by Strong et a1.16 for O/Al(lll). They used fixed force constants and considered only lattice dynamical effects associated with different 0 binding sites while ignoring any dependence of valence chemistry on 0 coverage. They find that the presence of subsurface 0 causes a shift of about 60 cm-' to higher energy in the mode frequency for surface 0. Similarly, for certain subsurface 0 binding sites, the presence of surface 0 causes a shift of about 30 cm-I to lower energy for the subsurface mode. These results are consistent with the -10-cm-' shift in the subsurface 0 mode as surface C1 sites are filled. Note also that the 10-langmuir C1, exposure causes a significant increase in band intensity, whereas the band resulting from the same C1, exposure on the clean surface (Figure 3b) is barely visible on the scale of Figure 2a. Since the C1 Auger intensities are similar in (IS)Strong, R. L.; Firey, B.; deWette, F. W.; Erskine, J . L. Phys. Rev. B 1982, 26 3483; J . Electron Spectrosc. Relnt. Phenom. 1983, 29, 187.
Bermudez and Glass both cases, this implies that, in addition to the lattice dynamical effect discussed above, there is a chemical interaction between adsorbed 0 and C1 leading to an enhancement of the A1-0 dynamic dipole moment. The situation is more complicated in other exposure regimes (e.g., Figures 2b and 3a,b) for which both 0 and C1 occupy both surface and subsurface sites. In this case the possible combinations of chemical and lattice dynamical effects are complex; however, the data all show evidence for interaction of different species. For example, for a low O2 preexposure (Figure 2b), insufficient for "saturation" as described above, the surface A1-C1 mode is observed a t about 735 cm-'. However, there is no obvious shift in the position of this band to higher energy with increasing C1 coverage, as is seen for C1 on the clean A1 surface. This suggests that the interaction between adsorbed C1 atoms producing this shift may be impeded by the presence of 0. It is also interesting to note that a C12 exposure of only 5 langmuirs causes both a significant shift and intensity increase in the 815-cm-' 0 band. For a C1-covered surface showing clear 740- and 820-cm-' bands (Figure 3a), addition of 0 leads eventually to a single broad band at 800 cm-'. Further insight into the structure of the 0 + C1 coadsorbed phase can be obtained from the data in Figures 2 and 3. Addition of a small amount of 0 when most of the surface C1 sites are filled (Figure 3b) yields a band at 780 cm-', intermediate between the low-coverage subsurface 0 (815 cm-', Figure 2b) and surface Cl(745 cm-') energies. Further O2exposure, up to 50 langmuirs, yields a band very similar in position and intensity to that for 10 langmuirs of C1, following 50 langmuirs of O2 (Figure 2a). This shows that the distribution of C1 and 0 between surface and subsurface sites is the same in either case. Were this not the case, a 50-langmuir O2 preexposure would prevent occupation by C1 of those sites populated by a small C12 exposure on the clean surface. This conclusion is also consistent with Auger data7 showing that the initial C1 uptake is independent of O2 preexposure and that O2 exposure does not lead to desorption of Cl preadsorbed at surface sites. To summarize, we have, first, obtained high-resolution vibrational spectroscopic data identifying a surface C1 mode at 745-760 cm-' on Cl,-exposed Al(111). The assignment has been supported by correlation with previous7 Auger, work function, and temperature-programmed desorption results and with vibrational datal0 for I/Al(lll). Another band, tentatively assigned to C1 at subsurface sites, has also been observed at 820 cm-l. This latter assignment is complicated by the sensitivity of this part of the spectrum to small amounts of 0. Second, coadsorption of 0 and C1 has been shown to yield an intimately mixed phase rather than distinct chlorinated and oxygenated patches. The coadsorption data have been discussed qualitatively, in terms of strongly coupled A1-0 and A1-C1 motions, with reference to the results of slab model calculations'6 for O/Al(lll).
Acknowledgment. We are most grateful to J. E. Butler for the generous loan of the FTIR apparatus, the infrared detector, and other equipment and for encouragement during the course of this work. Preliminary FT-IRRAS experiments for 02/Al(I l l ) , including the configuring of the optics, were performed by R. L. Rubinovitz. We thank K. B. Koller for help in learning the use of the FTIR hardware and software. This work was supported by the Office of Naval Research. Registry No. AI, 7429-90-5; CI,, 7782-50-5; Oz, 7782-44-7.
