Langmuir 1986, 2, 588-593
588
Surface Interaction in the Pt/y-A1203System. 5. Effects of Atmosphere and Fractal Topology on the Sintering of Pt Walter G. Rothschild," H. C. Yao, and H. K. Plummer, Jr. Scientific Laboratory, Ford Motor Company, Dearborn, Michigan 48121 Received December 20, 1985. I n Final Form: M a y 23, 1986 The high-temperature, thermally induced sintering of Pt particles on y-alumina in air and hydrogen was studied by infrared transmission spectroscopy (IR) of chemisorbed CO and stereotransmissionelectron microscopy (TEM) of the sintered Pt. The IR absorption band of CO chemisorbed on Pt/y-A1203presintered at 750 " C in air exhibited a bimodal shape: an intense narrow band at 1092 cm-l of 12 cm-l width, representing the CO molecules chemisorbed on large crystallites, superposed on a broad asymmetric background peaked at 2060 cm-l, representing the CO chemisorbed on dispersed Pt/A1203regions. The narrow band is absent when Pt/A1203was presintered in H P . The results suggest two different modes of P t coarsening: (i) In H z , Pt sinters through migration of Pt atoms (Ostwald ripening). (ii) In air or 02, Pt02is formed and vaporized. The Pt particles grow through vapor-phase transport as well as by surface migration of PtO,, the latter process being preponderant. Sintering in H2 leads to thin Pt hexagons of a narrow size dmtribution (average TEM diameter 150 A). Sintering in air or O2yields large three-dimensional Pt crystallites (300 A) that exhibit several low-index facets. The broad IR profiles of the chemisorbed CO show that sintering of Pt in a reducing as well as oxidizing atmosphere leaves a certain portion of Pt that effectively resists high-temperature-inducedparticle growth and remains in well-dmpersed configurations. We propose that this phenomenon is caused by regions of fractal topologies of Pt adsorption sites on the alumina.
I. Introduction Sintering or particle growth of Pt on an alumina support has been the subject of extensive investigations. Three principal mechanisms have been proposed to explain the thermally induced increase of the particle size and the corresponding decrease of Pt surface area on the alumina. The first is the mechanism of atomic diffusion; it pertains to the Ostwald ripening process and regards the differences in the interfacial energies of the particles as the driving force for the transport of Pt from smaller to larger particles.l-s The second mechanism is crystallite migration. It suggests thermally induced migration of smaller crystallites, followed by collisions and coalescence into larger particles.s-12 The third mechanism involves the presence of two phases of Pt (or PtO,) on alumina, namely, a dispersed and a particulate phase, which can reach an equilibrium state after sintering in oxygen a t a high temperature. The higher the sintering temperature, the larger the resulting Pt particle size, and the lower the PtO, concentration in the dispersed phase.13J4 Three possible pathways for these interparticle transport mechanisms were ~uggested:~vapor-phase transport, (1)Flynn, P. C.;Wanke, S. E. J . Catal. 1974,34,390; 400. (2)Wanke, S. E.; Flynn, P. C. &tal. Reu. (Sci. Eng.) 1974,12,93. (3)Wanke, S.E.In Sintering and Heterogeneous Ca&dj&; Kuczynski, G. C., Miller, A. E., Sargent, G. A., Eds.; Plenum: New York, 1984; p 223. (4)Lee, H.H. J. Catal. 1980,63,129. (5)Huang, F.H.; Li, C.-Y. Scr. Metall. 1973,1, 1239. (6)Wynblatt, P.; Gjostein, N. A. Progr. Solid State Chem. 1975,9,21. (7)Wynblatt, P.; Gjostein, N. A. Acta Metall. 1976,24,1165. Wynblatt, P. Ibid. 1976,24,1175. (8) Wynblatt, P.; Ahn, T.-M. In Sintering and Catalysis; Kuczynski, G. C., Ed.; Plenum: New York, 1975;p 83. (9)Ruckenstein, E.; Pulvermacher, B. AICHE J. 1973,19,356. (10)Ruckenstein, E.; Pulvermacher, B. J . Catal. 1973,29,224;1974, 35,115. (11)Ruckenstein, E.;Dadyburjyr, D. B. J . Catal. 1977,48,73. (12)Chu, Y. F.; Ruckenstein, E. J. Catal. 1978,55,281. (13)Yao, H.C.; Sieg, M.; Plummer, H. K., Jr. J.C a t d 1979,59,365. (14)Yao, H. C.; Wynblatt, P.; Sieg, M.; Plummer, H. K., Jr. In Sintering Processes; Kuczynski, G. C., Ed.; Plenum: New York, 1980;pp 561-571. See also paper 4 in this series: Yao, H. C. Appl. Surf. Sci. 1984, 19,398.
diffusion on the transport surface, and volume diffusion. While the latter is not considered to be important for Pt under the present experimental conditions, the other two mechanisms can proceed in parallel. Evidence presented in previous s t u d i e ~ l ~indicated -'~ that sintering of Pt proceeds more rapidly in an oxidizing than in a reducing atmosphere (or in vacuo), the difference beginning to increase a t temperatures above 500 OC.14 Recently, renewed emphasis has been made that sintering of Pt in H2involves some chemical reaction between noble metal and support material, appreciating that metal particle growth at elevated temperatures is not merely a physical agglomeration process but that the distribution of Pt particle growth depends also on the nature of the support.1"21 In this study we have further probed the factors that affect sintering of Pt on y-alumina a t 750 " C using transmission electron microscopy of the deposited Pt and transmission infrared spectroscopy of the CO stretching fundamental of chemisorbed carbon monoxide. In addition, we have drawn on some previously published chemisorption data to support the present conclusions. The results are discussed in terms of Pt surface migration of Pt or Pt02under reducing or oxidizing conditions, with a small contribution of vapor-phase transport for PtO,. Furthermore, we have taken account of the recently established fractal surface topology of y-alumina as it conceivably affects the results and conclusions. In particular, we are advancing some ideas how such fractal effects may generate the peculiar shape of the infrared profile of chemisorbed CO. (15)Maat, H. J.;Moscou, L. Proc. Int. Congr. Catal. 3rd, 1964 1965, 2. 1277. -, -- - -
(16)Herrmann, R. A.;Adler, S.F.; Goldstein, M. S.; Debaun, R. M. J. Phys. Chem. 1961,65,2189. (17)Gruber, H. J. Phys. Chem. 1962,66,48. (18)Hughes, T. R.; Houston, R. T.; Sieg, R. P. I n d . Eng. Chem. Process Des. Deu. 1962,1, 96. (19)Arai, M.; Ishikawa, T.; Nakayama, T.; Nishiyama, Y. J. Colloid Interface Sci. 1984,97,254. (20)Lagarde, P.; Murata, T.; Vlaic, G.; Freund, E.; Dexpert, H.; Bournonville. J. P.J. Catal. 1983,84,333. (21)Den Otter, G. J.; Dautzenberg, F. H. J. Catal. 1978,53, 116.
0 1986 American Chemical Society
Surface Interaction in the Pt/y-Al,O,
System
Langmuir, Vol. 2, No. 5, 1986 589
0.8
8 0.6 z F 0.4
z a uz + 0.2
t3
3"" a
2100
2020 cm"
1940
1360
Figure 2. Room temperature infrared spectra of the C-0 stretching fundamental of CO linearly chemisorbed ('on-top" p i t i o n ) on a 4.6 pmol/m2 (BET) Pt/y-A120zcatalyst: Spectrum A, after outgassing in vacuo a t 25 'C of the original 25-torr-CO charge; spectrum B, after further in vacuo outgassing a t 300 O C ; spectrum C = spectrum A - B. Duration of outgassing steps = 30 min. Half-width: spectrum C, 12 cm-'; spectrum B, 36 cm-'. 8.0-,
Figure 1. Stereo transmission electron micrographs of the hahit of Pt of a 12.3 pmol/mz (BET) Pt/y-A1203sample: (A) heated in air at 750 "C for 5 h; (B) heated in H2 a t 750 OC for 5 h.
11. Experimental Section 1. Materials. A granular y-A1203support was prepared by
agglomerating Dispal-M y-Al,O, powder (Conoco Chemicals) by wetting with water, drying, and calcining a t 600 'C. The solid mass was crushed and sieved, and the 0.5-1.0-mm-diameter fraction was chosen for the support. The support for each sample was immersed in distilled water to which aqueous H,PtCI, of desired concentration was added. The samples were evaporated and dried in air a t 120 "C, rewetted with water, and redried in order to distribute the Pt ions uniformly. Each sample was calcined in air at 400 OC for 2 h and finally reduced in H, a t 300 'C and thoroughly outgassed in vacuo at the same temperature. 2. Adsorption Measurements. The apparatus and p d u r e for measuring BET area and CO uptake of supported Pt samples as well as their results have been extensively described in previous papers'3,"and are referenced here, whenever needed, for purpases of sample characterization. 3. Transmission Electron Microscopy (TEM) a n d Infrared (IR) Spectroscopy. TEM micrographs were obtained using a Siemens Elmiskop I TEM and employing an extraction-replication procedure similar to those described in the lite r a t ~ r eas ~ well ~ . ~as ~ in our previous studies.13J4 The catalyst specimen contained 12.3,mol of Pt/m2 (BET), or 23.1 wt 90Pt, by X-ray analysis. This high-load Pt/AI2O3sample was specially selected as it was expected to lead to large Pt crystals upon sintering. The IR spectra of the CO stretching fundamental of chemisorbed CO were obtained hy using 4.6 pmol of Pt/m2 (BET-A1203 samples (13.8 wt 70 Pt). This particular Pt load was expected to yield relatively large Pt crystals upon sintering but was not considered high enough to mask any effects of the alumina support: The sample still falls on a linear relation linking amount of Pt(wt %) with its spread [rrmolof R/m2 (BET)]; the 12.3 #mol of Pt/mz (BET) sample does not." The spectroscopy as well as the reduction/CO-charging/outgassing procedures was performed with a vacuum-tight dualpurpose Pyrex/Vicor cell equipped with KBr windows. This (22) Reneai. H. A.; Curtis. H.M ;Studer, H. P.J . C o r d 1968.10,328. (23) Dalmai-lmelik.G.; I a e l e q . C.:Mutin, 1. J . Mtcmor. (Po&) 1574,
20. 123.
yy,,io ,
I
0. I
2100 2060 2020 1980 1940 1900 C"'
Figure 3. Room temperature infrared spectra of the C-0 stretching fundamental of CO linearly chemisorbed on a 4.6 rmol/m2 (BET) Pt/y-Al,O, catalyst presintered in H, a t 750 "C for 5 h. (Lower spectrum) Saturation coverage of CO, fwhm = 42 cm-l. (Upper spectrum) Spectrum taken after outgassing in vacuo at 280 "C for 30 min; fwhm = 36 cm-'. construction permitted the movement of the sample wafer out of its position in the optical path to a cell volume for treatment at high temperatures in static or flowing gases.= Whatever the pretreatment of the 4.6 @molof Pt/m2 (BET-A1203 sample, prior to spectroscopy it was reduced in its cell in H, at 300 OC, then outgassed, and cooled to room temperature in vacuo before being charged with CO a t 25 torr. Excess CO was removed by evacuation; the CO coverage was therefore, initially, saturated. Spectra were recorded a t room temperature with a PerkinElmer 180 under normal settings. Slit widths were kept a t about '/,o of the bandwidth, making corrections unnecessary. To facilitate the spectral evaluations, vertical marker pips were injected electronically each 2 cm-' by a special attachement.
111. Results 1. TEM of the Sintered Pt/y-Al,O,. T h e stereo TEM pictures of high-load Pt/AI,03 samples [12.3 Nmol/m2 (BET)], sintered at 750 "C for 5 h in air a n d in H2, are shown in Figure 1, parts A a n d B, respectively. After sintering in air, some Pt has grown to three-dimensional piles of extraordinarily large crystallites (average diameter 300 A), some as large as -2000 A, as can be seen from Figure 1A. These large crystallites appear to possess (24) Dalls Betta, R.J . Phys. Chem. 1975, 79,2519.
Rothschild et al.
590 Langmuir, Vol. 2, No. 5, 1986 several different crystal facets. In contrast, Pt sintering in Hz (Figure 1B) forms thin clusters of distorted hexagons but with a relatively narrow size distribution of average diameter 150 A. It should be noted that the extraction-replication procedure (to wash away the alumina support) may cause layers of Pt clusters to touch, thus increasing the apparent thickness. 2. IR of CO Chemisorbed on the Sintered Pt/yAlz03. Room-temperature spectra of the carbon-oxygen stretching fundamental of CO chemisorbed on Pt/y-Al,03 presintered in the two atmospheres are shown in Figures 2 and 3. Figure 2: Spectrum A was obtained with a Pt/y-A1203sample presintered in air a t 750 "C; CO is a t saturation coverage (see section 11.3). It shows a sharp CO band a t 2092 cm-' with a full width a t half-maximum (fwhm) of only 12 cm-l, superposed on a much larger and broader absorption region. Spectrum B was taken after the CO-charged sample had been degassed in vacuo at 300 "C for 30 min: The narrow 2092-cm-' band has disappeared but the broad asymmetric absorption region, with maximum absorption near 2060 cm-l and fwhm = 36 cm-', has remained. It is only removed by extensive outgassing a t elevated temperatures. The narrow CO band (spectrum C) is astonishingly similar, in fwhm as well as peak frequency, to that of CO chemisorbed on well-defined Pt planes such as on Pt(ll1) (9.0 and 2101 cm-I), r e s p e c t i ~ e l y . We ~ ~ ~have ~ ~ not succeeded in finding literature sources which also report such a narrow absorption band on supported platinum. Figure 3: The spectra were obtained from the P t / y Alz03sample presintered in Hz at 750 "C. They show a CO stretching vibration peaked a t 2078 cm-' (saturation coverage) and 2058 cm-' (after outgassing for 1 h a t 280 "C), respective fwhm = 42 and 36 cm-l, and a CO-coverage-dependent red-shift of the profile. Evidently, no isolated, sharp CO absorption peak is observed after hightemperature presintering of Pt/ y-A1203 in hydrogen.
IV. Discussion 1. Sintering of Pt/y-A1203in an Oxidizing Atmosphere. a. TEM Data. The above TEM results, in combination with our data of Pt-sintering rates,l3,l4demonstrate that sintering in oxygen and air leads to significantly larger Pt particles (and a t a considerably faster rate) than sintering in hydrogen. A mechanistic model of the sintering process in oxygen assumes that Pt sinters on y-Al,03 via PtO, molecules that migrate by surface diffusion and vapor-phase transport through cycles of decomposition, reoxidation, and vaporization of the metal ~xide.~ Previously, .~ we had observed a rapid drop of CO uptake after high-load Pt/alumina catalysts had been sintered in air a t increasing temperatures between 400 to 600 "C.13 This is now understood to signify that at these temperatures, where decomposition of PtOz is not significant under air of atmospheric pressure, the compound sinters until the equilibrium between Pt02in the dispersed and the particulate phases on y-alumina is attained.14 For instance, at 500 "C and 760 torr of O,, the vapor pressure of PtOz is approximately 7 X torr,7 the equilibrium concentration of Pt02 in the dispersed phase is 2.2 pmol/m2 (BET) with average Pt particle diameter of 15 A.13J4 This means that there remains a large fraction of nondispersed (particulate) Pt02 for a high-Pt catalyst. Above 600 "C, however, there is significant decomposition of PtO, in air a t normal pressures. Hence, the (25) Shigeishi, R. A.; King, V. A. Surf. Sci. 1976,58, 379. (26) Severson, M. W.; Tornquist, W. J.; Overend, J. J . Phys. Chem. 1985, 88, 469.
saturation concentration of surface Pt02decreases. In a low-Pt specimen, this leads to sintering of the highly dispersed Pt phase to form small Pt cluster^.'^^^^ For the high-Pt catalyst employed in these experiments, it causes the observed growth to the extraordinarily large, threedimensional Pt crystallites shown in Figure 1A. It has recently become clearer that the mechanism of this agglomeration process is rather more complex than previously suggested3i8J4because of important contributions from reaction steps involving platinum oxychlorides.28~29 At 750 "C, the vapor pressure of Pt02 in air is of the order of lo4 t ~ r r . ~ A ,rough ~ estimate of ideal-gas kinetic effusion rates dn/dt = '/gw, with n = molecules/cm3, u = cm/s (average), predicts that 7 X 1013PtOz molecules/s impinge on unit area. For a typical 5-h sintering, this deposits ideally 13 X lo1' molecules PtOz/cm2,a factor of about of the total Pt. Hence, contributions of PtO, vapor transport to sintering in air are probably small.3 b. IR Spectra. We propose that the narrow spectral component (Figure 2, spectrum C) is from CO molecules that are chemisorbed on surface elements of the large Pt clusters. This assignment is straightforward since the CO profile closely resembles the corresponding Pt(ll1) spectrum in frequency and ~ i d t h . This ~ ~ corroborates * ~ ~ that sintering in air must have generated clusters possessing a considerable fraction of low-index crystal facets. Some basic theories on the surface and vapor transport of PtO, suggest that such results reflect "classical ripening", "nucleation-inhibited growth", and "abnormal g r ~ w t h " . ~ As has also been demonstrated, particle growth by diffusion within micropores of a support may show rate relations that fit any theoretical model8 because of pore size and shape effects. Since y-alumina has since then been identified as a fractal object,3l~~~ these theories merit fresh attention. An assignment of the broad 2060-cm-l spectral profile, which remains after partial desorption of the saturation CO layer (spectrum B, Figure 2), is not obvious. Current thinking would tend to rationalize that a CO frequency distribution near 2060 cm-' represents the range of CO oscillators chemisorbed on the various low-coordination sites of Pt crystals such as kinks, terraces, or other faults33 generated during the rapid sintering of the Pt particles. However, if the literature evidence is scrutinized, it appears that this suggestion explains neither the asymmetry, nor the width, nor the frequency range of the observed CO band. Indeed, it seems, a t the high CO coverage of our experiments, that CO frequencies of CO chemisorbed on a large variety of high-index Pt sites fall, nevertheless, rather close into the 2080-2100-~m-~ range of the observed narrow feature (spectrum C ) . For instance, (i) CO on Pt(321), with 60% 6-fold, 20% 5-fold, and 20% 3-fold terrace coordination, gives a sharp infrared feature at 2095 cm-l at full coverage.34 (ii) CO on Pt(533), leading to a surface with (111)terraces and (100) steps, shows a sharp spectral feature peaked at 2080 cm-' (40% coverage);this feature diminishes with increasing coverage (70-100%) in favor of a narrow band peaked between 2090 and 2100 cm-'. The initial adsorption is assigned to the step atoms.% (27) Rothschild, W. G.; Yao, H. C. J. Chem. Phys. 1981, 74, 4186. (28) Lietz, G.; Lieske, H.; Spindler, H.; Hanke, W.; Volter, J. J.Catal. 1983, 81, 17. (29) Foger, K.; Jaeger, H. J. Catal. 1985,92, 64. (30) Alcock, C. B.; Hooper, G. W. Proc. Roy. Soc. (London) 1960,254, 551.
(31) Avnir, D.; Farin, D.; Pfeifer, P. Nature (London) 1984,308, 261. (32) Pfeifer, P.; Avnir, D. J. Chem. Phys. 1983, 79, 3550. (33) Haal, D. M.; Williams, F. L. J. Catal. 1982, 76, 450. (34) McClellan, M. R.; Gland, J. L.; McFeeley, F. R. Surf. Sci. 1981, 112, 63.
Surface Interaction in the Ptly-Al,OJ System (iii) CO chemisorption on @io2)supported Pt particles of essentially identical dimension (truncated octahedra) but different and known average diameter (11-105 A) exhibits three discernible infrared spectral features, about 10 cm-’ apart and located in the range 2070-2087 cm-’ a t full coverage, within 2062-2082 cm-’ at 10% coverage.36 Interestingly, these band positions are not particle-size-dependent; they are assigned to faces (high-frequency band), to edges (lowest frequency band), and to corners of the Pt octahedra.36 Clearly then, a t or near CO saturation coverage there is no reason to assume that the broad profile exhibited by spectrum B (Figure 2) is caused predominantly, let along solely, by CO adsorption on low-coordination sites of the large Pt crystallites; if a t all, any such contributions should be reflected by the “tail” of the narrow spectral feature (C). Notice that we are only concerned here with linearly adsorbed CO (“on top” position); the bridged CO structure, a weak band peaked a t a considerably lower frequency (5 1850 cm-’), was not observed in our work, nor does it enter into the discussion above. We propose that the broad, asymmetric spectral band of chemisorbed CO (spectrum B, Figure 2) represents adsorption sites on well-dispersed Pt. By “dispersed” we understand not, only a distribution of particle size but also, equally important, of particle shape. The reasoning is as follows: (i) Low Pt load Pt/y-A1,03 catalysts, especially prepared to consist essentially of truly dispersed Pt (CO/Pt = 0.83; no Pt particles observed by TEM),14show an infrared CO profile at saturation coverage that is nearly identical with spectrum B of Figure 2 in peak frequency (-2060 cm“), in fwhm (-50 cm), and in the asymmetry toward long wavelength^.,^ Notice also that sintering of the catalyst (in H2 at temperatures not exceeding 500 “C), which leads to formation of a particulate phase of small Pt clusters (CO/Pt = 0.64, 10-35-A particle diameter), does not muterially change the shape of this CO profile.27 (ii) The ready disappearance of the sharp CO spectral feature (spectrum C, Figure 2) upon outgassing in vacuo and the contrasting persistance of the broad spectral feature (spectrum B) during this operation point to two Pt regions of widely different CO adsorption strength each. Since we are quite convinced that the narrow spectral CO feature describes the CO adsorption site distribution on the large Pt clusters (see above), the broad CO spectral feature should, by elimination, represent the distribution of adsorption sites of the dispersed (as defined above) Pt/y-A1203regions. (iii) Temperature-programmed CO desorption studies on similarly dispersed Pt/alumina catalysts have shown a broad range of desorption energies of (linearly) chemisorbed CO, assigned to the many different, available adsorption sites on the small Pt particles.37 2. Sintering of Pt/r-A1203in a Reducing Atmosphere. a. TEM Data. The formation of large Pt crystals during sintering of Pt/y-Alz03at 750 OC in H2can only take place by agglomeration of atoms or clusters that diffuse through the pore structure of the alumina support. Since the larger of these crystals are, predominantly, flat and hexagonal or hexagonally distorted (see Figure lB), we propose that (i) sintering takes place by atomic migration over inside surfaces of the A1203, (ii) the Pt crystals (35) Greenler, R. G.: Leibsle, F. M.; Sorbello, R. S. Phys. Rev. B 1985, 32, 8341. (36) Greenler, R. G.; Burch, K. D.; Kretzschmar, K.; Klauser, R.; Bradshaw, A. M. Surf. Sci. 1985,1521153, 338. (37) Foger, K.; Anderson, J . R. Appl. Surf. Sci. 1979, 2, 335.
Langmuir, Vol. 2, No. 5, 1986 591 grow predominantly in the larger pores of the support, and (iii) the Pt seeds grow at their edges. Whereas point ii is rather obvious, point i is supported by energy considerations: The Pt-Pt binding energy amounts to 20 kcal/mol, the activation energy for Pt atomic migration on the alumina is of the order of 5 k ~ a l / m o l . ’ ~ b. IR Spectra. The vibrational profile of the CO stretch of chemisorbed CO associated with the H2-sintered Pt/yAl,03 (Figure 3) is also revealing. Although one would expect that CO molecules adsorbed on the nearregular hexagonal Pt crystallites would yield a relatively narrow and symmetric band profile,%the actually observed band shows again considerable asymmetry to lower frequencies. From this we conclude that a considerable fraction of Pt has remained in dispersed configurations (see section 1V.l.b). It has been suggested that dispersed Pt forms bonds with the 0 atoms of the s u p p ~ r tor , ~even an alloy,2l when heated in H2 at temperatures above 500 O C . Such bonding effects cannot, a priori, be predicted merely from the spectral profile of an adsorbate. However, with the hindsight of such strong metal-support interactions, the infrared band of CO chemisorbed on Pt/y-A1203sintered in H2 a t temperatures above or a t 500 OC shows indeed singular features, previously glossed over and not understood, which are obviously absent in the corresponding CO spectra of Pt/alumina sintered in air (or in vacuo) at high temperatures. These features are the following: (i) CO desorption leads to band narrowing (see Figure 3). The same operation always yields broadening for catalysts presintered in air or in vacuo.ni38 (ii) A significant fraction of CO remains relatively strongly chemisorbed on a catalyst presintered a t higher temperatures in H2,.as compared to the same specimen presintered in air.27 (in) We reported previously that the same (high-load) Pt/y-A1203catalyst sintered less rapidly in vacuo than in H2, the sintering effect having been monitored by subsequent CO uptake.14 At that time, this result was not understood: Why should sintering of Pt occur a t a faster rate in H2 than in vacuo when, as very likely, Pt agglomeration takes place by atomic migration in both situations. Apart from the unlikely event of a platinum hydride as an active diffusing species, we now support the notion that care must be taken in relating sintering-induced adsorbate uptake data to adsorbent particle size in the presence of possible chemical modifications of the metal-support structure by the sintering atmosphere.20,21 3. Effect of Pt Crystal Size on CO Adsorption Strength. Some useful conclusions on the adsorption strength of CO on the various phases of air-sintered Pt/ y-Al,03 can be drawn immediately from the effect of outgassing on the appearance of the IR profile of the Chemisorbed CO molecules (see Figure 2): Evidently, the change of spectrum A to spectrum B indicates that CO is Chemisorbed significantly weaker on Pt adsorption sites on the larger Pt crystallites (spectrum C) than on Pt adsorption sites within the dispersed regions (spectrum B). We propose that catalytic reactions which proceed preferentially when the CO bond is weakened are twofold disadvantaged by the sintering of the metal (in air). First, the active metal surfaces of the catalyst is decreased by sintering. Therefore, fewer CO molecules adsorb per number of Pt atoms. Second, the overall efficiency of “activating the C-0 bond” of a chemisorbed carbon monoxide molecule is decreased since a portion of CO molecules now chemisorbs (relatively weaker) on the large (38) Rothschild, W. G. J. Chem. Phys. 1976, 65, 455.
Rothschild et al.
592 Langmuir, Vol. 2, No. 5, 1986 crystallites. Contrarily, reactions that are rate-limited by surface diffusion/desorption steps of chemisorbed CO, such as the oxygen deposition during the Langmuir-Hinshelwood sequence of COz formation on Pt, should in principle be favored when the metal has been sintered. By combination of these two points, it seems that sintering of A1203-supported Pt in air may lead to entirely different product yields or mixes, compared to a fresher catalyst, for catalytic reactions of CO that contain such diverse activation processes.39 4. Effects of the Fractal Surface Topology of yAlzO,. It has recently been established that activated alumina of the type used in our work has a fractal surface dimension of d = 2.8.31332 This signifies that the distribution of pore sizes is scale-independent; within a wide range of linear size (between the diameter of a nitrogen molecule and the radius of gyration of polystyrene molecules), all sizes and shapes of (inside) pores are present, following a distribution of N of (linear) size r of the form3,
N(r) = C = constant (1) with d = 2.8. In other words, (i) the number of smaller pores greatly exceeds the number of larger-size pores, (ii) the outside surface area of the alumina is negligible compared to the inside surface area (the “walls” of the pores), and (iii) the shape of the pores is highly irregular. Although direct data on the dimension of the dispersed Pt within the Al,03 support are not available, it appears reasonable to assume that the fractal surface dimension of the alumina causes, a t least to a degree, some of the deposited metal to be in a dispersed state that has a fractal (noninteger) dimension and, therefore, exhibits the hyperbolic size distribution of the form of eq 1 (with, a priori, a different value of d ) . Since finer dispersed Pt undergoes stronger interactions with the alumina s ~ p p o r t it , ~is~therefore ,~~ not surprising that a certain fraction of the Pt resists sintering (see sections N . l b and IV.2b). This must be, according to the above suggestions, the Pt clusters that find themselves within the smaller size ranges of the alumina pores.40 This idea also explains, in a qualitative way, the unusual appearance (long-wavelength absorption tail) of the IR profile of the CO stretch of CO chemisorbed on regions of dispersed Pt (see Figures 2 and 3): The smaller the size of the Pt clusters (the weaker their metallic properties), the stronger the bond between a Pt site and the C atom of a chemisorbed CO molecule, thus the weaker its CO bond and, consequently, the lower the CO resonance frequency. The latter, well-known effect can be explained with the help of “back-bonding”theory.41 It has also been indirectly observed by adsorption of (isoelectronic) N, on Pt that sufficiently small Pt clusters (15-70-A diameter) cause strong adsorption of the nitrogen molecules, under considerable distortion of the electronic orbitals of the adsorbate. (The infrared-forbidden N, fundamental transition is induced). Interestingly, the N2-peakfrequency is red-shifted by 100 cm-’ with respect to its (Raman) gas-phase value.42 Furthermore, the exceedingly broad appearance of the CO profile is readily rationalized by a fractal topology of the dispersed Pt since, according to eq 1,there are many more smaller than larger pores: Consequently, there should exist a significant excess of small Pt clusters over larger ones. Clearly, once a “critically large” Pt cluster has
1.
1k.Q pi a m
Figure 4. Cross section (Sierpinskicarpet, fifth approximant) of a 2.89...-dimensional self-similar filament structure serving as a prototype, idealized Pt/alumina surface after sintering. The filament is generated by translation of the carpet normal to and out of the paper plane. Dark regions represent Pt agglomeration serving as chemisorption sites for CO. Distance a indicates an average separation. For separations larger than a, Pt is in its pure metallic state. For separations smaller than a, Pt is surface-bonded (dispersed);the perturbing effect of the surrounding aluminasurface elements on the electronic levels of Pt increases with decreasing r / a . The carpet is only developed in the lower left quadrants of successive approximants. Outside surface area of
the filament is negligible.
been generated (by sintering, for instance), its electronic metallic properties are no longer modified upon further A CO molecule adsorbed on a given Pt site of such a critically large cluster will have the same CO resonance frequency as if it were adsorbed on the equivalent site of an even larger Pt cluster. On the other hand, the resonance frequency of a CO molecule that is chemisorbed on a “subcritical” Pt cluster should be sensitive to the particular size and shape of the Pt cluster since this determines the electronic properties of the Pt (and therefore the bonding characteristics between this Pt site and the C atom of its chemisorbed CO molecule). It is noteworthy here that CO desorption from welldispersed Pt/y-Alz03has been observed to yield not only a red-shift of the IR profile but also an increase in its bandwidtheZ7In our model, this is caused by preferential desorption of weaker bonded CO molecules (which absorb at higher spectral frequencies). Notice that moleculemolecule interactions which are coverage-dependent, such as dipole-dipole effects, could not lead to band broadening with decreasing CO c o ~ e r a g e . ~ ’ * ~ ~ , ~ ~ Figure 4 shows these ideas40by the example of a simulation of alumina in terms of a filament structure that is constructed through the out-of-plane translation of a Sierpinski carpet, a fractal object of dimension4j d = log 8/log 3. This leads to d = 1+ log 8/log 3 = 2.89 ... for the inside surface dimension of the filament.32 Figure 5 shows a simulation of the frequency distribution of the tail of the infrared spectral intensity of the fundamental of CO molecules chemisorbed on fractal surface elements of dispersed Pt/y-Alz03. We introduce a scale factor f to relate the self-similarity of the fractal surface directly to the average CO oscillator frequency; this avoids the difficulty of distinguishing between “fractal” and (43) Huizinga, T.; van’t Blik, H. F. J.; Vis, J. C.; Prins, R. Surf.Sci.
(39) Erdohelyi, A.; Pasztor, M.; Solymosi, F. J . Catal. 1986, 98, 66. (40) Rothschild, W. G. Bull. Am. Phys. SOC.1985, 30,413. (41) Byholder, G. J . Phys. Chem. 1976, 79,756. (42) Van Harveled, R.; von Montfoort, A. Surf. Sci. 1966, 4,396.
1983, 135, 580.
(44) Primet, M. J . Catal. 1984, 88, 273. (45) Mandelbrot, B. B. The Fractal Geometry of N a t u r e ; Freeman: New York, 1983; p 142.
Surface Interaction in the PtlyA1208System
0
w [red- shif t]
Figure 5. Simulated infrared absorption spectrum (low-frequency portion) of the CO stretch of CO chemisorbed on Pt dispersed on a fractal support of (arbitrary) scale factor f = 7/6. Each rectangle represents a distribution (spectral intensity) of CO adsorbed on fractal surface sites constructed as follows: The smaller the linear size r of the sites, the wider their distribution (length of rectangle) and the greater the red-shift of the average CO transition frequency of the chemisorbed CO (middle points on the frequency axis). The equal height of the rectangles represents the (assumed) frequency- and coverage-independent CO extinction coefficient. Notice that CO desorption is simulated by successive removal of rectangles starting from the top (weaker-adsorbed CO molecules have higher infrared transition frequencies), leading to a decrease in intensity of the profile with an accompanying red-shift and increase in its half-width.
“fracton” dimension^.^^ We also assume that the CO extinction coefficient is independent of frequency and coverage. As the caption to the figure explains, the two observed spectral effects caused by CO desorption from well-dispersed Pt/y-A1z03,27namely, a red-shift of the band and an increase of its width (see above), are correctly simulated; unless Pt is dispersed in a perfect (mono)layer on the support, we expect that 1 C f C 2.47
V. Summary We summarize here our ideas on the possible fractaltopological effects of the alumina support and their manifestations on the spectra of the chemisorbed CO. (i) The degree of Pt dispersion that remains after sintering is greater the smaller the size of pores within the A1203support; there are many more smaller than larger pores. (46) Alexander, S.; Orbach, R. J. Phys., Lett. 1982, 43, L625. (47) Pfeifer, P.; Avnir, D.; Farin, D. J. Stat. Phys. 1984, 5 / 6 , 699.
Langmuir, Vol. 2, No. 5, 1986 593 (ii) The energy distribution of CO adsorption sites on the Pt, and hence the frequency range of the resonance frequencies of the CO oscillators adsorbed on these sites, is wider the greater the Pt dispersion. (iii) The center-of-gravity frequency of this frequency range is lower (the CO is the stronger adsorbed) the greater the degree of Pt dispersion (the greater the perturbation of the electronic levels of Pt involved in CO bonding). (iv) The characteristics of the CO adsorption sites on the large Pt crystallites, generated by sintering of Pt/ alumina a t high temperatures in air or oxygen, resemble those of low-index planes of pure Pt in their perturbing effects on the vibrational energy levels (and relaxation process)48of the adsorbed CO molecules. These ideas furnish a different way of looking a t sintering and at dispersion of Pt on high-surface, irregularly shaped supports, in other words, supports where the notion of “flat surface” breaks down even on a very small scale of size and where support-metal interaction may generate concordingly irregularly shaped Pt. In essence, the spectral profile of adsorbed CO acts as a probe of the position of the electronic Fermi levels of the metal p a r t i ~ l e .If~ ~ ~ ~ ~ our ideas are correct, CO chemisorption on Pt dispersed within a narrow distribution of sites, such as the regularly shaped pores of zeolites, should give CO IR-absorption profiles that are relatively narrow and lack significant asymmetry (unless strong chemical metal-surface interactions overpower the topological aspects of the catalyst). Infrared spectral profiles and desorption data of CO chemisorbed on Pt/alumina that was presintered in Hz at high temperatures are now better understood; they agree with proposed strong metal-support interaction effects induced by H2.
Acknowledgment. We are grateful to R. M. Pick, P. Pfeifer, and E. C. Su for helpful advice and interest in our work. Registry No. A1203,1344-28-1;CO, 630-08-0; Pt, 7440-06-4. (48) Rothschild, W. G. Abstracts of Papers, 188th National Meeting of the American Chemical Society, Philadelphia, PA; American Chemical Society: Washington, DC, 1984; COLL 87. (49) Cyrot-Lackmann, F. Growth and Properties of Metal Clusters; Bourdon, J., Ed.; Elsevier: The Netherlands, 1980; p 241. (50) Backelett, G. B.; Bassani, F.; Bourg, M.; Julg, A. J.Phys. C: Solid State 1983,16,4305.
