J. Phys. Chem. B 2001, 105, 217-219
Acid Sites on SiO2-Al2O3 Monolayer Catalysts: Accessibility
31P
217
NMR Probes of Strength and
Bing Hu and Ian D. Gay* Department of Chemistry, Simon Fraser UniVersity, Burnaby, B.C. V5A 1S6, Canada ReceiVed: August 10, 2000; In Final Form: October 23, 2000
We have examined the acidic properties of a SiO2-on-Al2O3 monolayer catalyst by 31P NMR of adsorbed tricyclohexylphosphine and triphenylphosphine. These large probe molecules permit the assessment of steric effects, by comparison with trimethylphosphine, and reveal differences in acid strengths of surface sites. We find that there are significant constraints to accessing Lewis sites, but not Brønsted sites. The concentration of strong Brønsted sites is slightly less than half of the total concentration of Brønsted sites.
Introduction The traditional method for preparation of acidic amorphous silica-alumina catalysts has been the coprecipitation of the two oxides.1 These catalysts contain both H+ (Brønsted) sites and Lewis acid sites; both types of acid may be catalytically active, depending on the reaction. An alternative method of preparation, which has more recently been explored, is the deposition of SiO2 on an Al2O3 substrate, in amounts of the order of one monolayer of SiO2.2-8 By adjusting the amount of SiO2 deposited, this method permits the preparation of catalysts with controlled amounts of surface acidity, and to some degree permits the ratio of Brønsted to Lewis sites to be adjusted. Compared with precipitated silica-alumina catalysts, relatively little is known about the details of the surface acidity, and catalytic properties of these monolayer catalysts. Pure alumina is well-known9 to have a high concentration of Lewis acid sites resulting from coordinatively unsaturated Al3+, and little or no Brønsted acidity. Clearly, Lewis sites will remain at low submonolayer amounts of added SiO2. Increasing amounts of silica will lead to the possibility of Brønsted acid formation through interaction of the two oxides. Finally, deposition of multiple layers of silica would be expected to lead to a nonacidic surface, similar to pure amorphous silica. The presence of Brønsted sites in the intermediate range of silica depositions has been clearly demonstrated by infrared spectroscopy of adsorbed ammonia and pyridine,2 and by 31P NMR spectroscopy of adsorbed trimethylphosphine.3,4,6 Characterization of the acidic properties of a surface requires measurement both of the total acid, and of its strength. It is of course possible that a range of strengths exists on any particular surface. Sheng and Gay6 made a quantitative NMR determination of Lewis and Brønsted acid amounts as a function of amount of deposited silica, using adsorbed PMe3. As might be expected, these measurements show a monotonic decline in Lewis concentration with increasing SiO2. Brønsted acidity was found to pass through a maximum at a silica level of 7-8 SiO2 per nm2 of substrate surface. The Brønsted concentration at the maximum was 0.30 µmol/m2, about 50% higher than found for a typical commercial precipitated catalyst. These measurements determine the total amount of Brønsted acid which is strong enough to protonate PMe3; they say nothing more about the strength of the surface acid. So far the only direct measurement related to strength is the infrared observation by Niwa et al.2
that NH3 is protonated to a greater extent than is pyridine, from which these authors infer that there is a small population of strong acid sites, and a larger population of weaker sites. Several authors2-5 have studied catalytic test reactions, which throw some indirect light on the question of strength. Niwa and co-workers2 studied cumene cracking and 1-butene isomerization. Compared to commercial precipitated catalysts, the silica monolayer catalysts were of similar activity for butene isomerization, but were much less active for cumene cracking. Since the latter reaction is known to require a stronger acid10 as catalyst, it was inferred that the majority of Brønsted sites were weaker on the monolayer catalysts. This is consistent with the infrared observations mentioned above. On the other hand, Sato et al.3 studied xylene isomerization, cumene cracking, and n-heptane cracking, and in all cases found similar activities for monolayer and precipitated catalysts. From the above, it can be seen that there is some uncertainty as to the strength of surface Brønsted acids, which may be the result of different preparative procedures. All authors find that the rates of H+-catalyzed reactions go through a maximum at a deposited SiO2 level of 10-12 SiO2/nm2. This is in reasonable accord with the maximum in surface H+ concentration at 7-8 SiO2/nm2 as measured by PMe3 adsorption.6 We have recently shown that tricyclohexylphosphine (PCy3)11 and triphenylphosphine (PPh3)12 can be used to detect acid sites on solid surfaces. These molecules are much larger than PMe3, and would be expected to be excluded from acid sites that have stringent steric requirements. As discussed in ref 12, the sequence of base strengths for various phosphines is different for aqueous vs gas-phase environments, and it is thus uncertain as to the appropriate sequence for use in surface studies. The main uncertainty arises with PMe3 which is unusually weak in the gas phase.13 However it appears to be the case that PCy3 is much stronger than PPh3 in all environments. Thus if there is a distribution of acid strengths on the surface, we may learn about it by studying these two large phosphines which have similar steric demands, but differing base strengths. On the other hand, observation of an acid site with PMe3 but not with PCy3 is a clear indication that access to such a site is only possible for small molecules. In this paper we study adsorption of PPh3 and PCy3 on a SiO2-Al2O3 monolayer catalyst, and compare the results with those obtained using PMe3 as a probe. We also make comparisons with results previously obtained using the
10.1021/jp002926z CCC: $20.00 © 2001 American Chemical Society Published on Web 12/14/2000
218 J. Phys. Chem. B, Vol. 105, No. 1, 2001
Hu and Gay
large phosphine molecules on a commercial precipitated silicaalumina catalyst, Davison type 980. Experimental Section A monolayer catalyst was prepared by deposition of SiO2 from Si(OCH3)4 on Harshaw grade Al-3945 γ-Al2O3, which has a surface area of 225 m2/g; the manufacturer’s measurements indicate 28% of the pore volume below 60 Å diameter, and 90% below 100 Å. We prepared a catalyst containing 6.2 SiO2 per nm2, using the fluidized-bed procedure described in ref 6. NMR measurements were done on a 3.5T instrument (60.5 MHz for 31P) using MAS and high power (45-55 kHz) proton decoupling. All NMR measurements were carried out at an ambient probe temperature of 24 ( 2 °C. Detailed NMR procedures are described in references.6,11,12 For the experiments described here, samples were spun at rates between 1.5 and 2.5 kHz, using a spinner14 that permits samples to be sealed in a vacuum, and observed by NMR without exposure to air. Quantitative measurements were made from Bloch decay spectra, using a 90° pulse length of 4.5 to 5.5 µs. Relaxation delays were in the range of 1 to 8 s; for each sample experiments were done at increasing delays, until no change in peak ratios was observed for a 2-fold increase in delay time. Qualitative observations were also made using matched Hartmann-Hahn cross polarization with a contact time of 1 ms. The spectra presented here result from the signal averaging of between 103 and 105 scans. Sealed NMR samples of adsorbed phosphines were prepared by mixing solid phosphine with degassed adsorbent under vacuum, followed by heating, as described in refs 11 and 12. We previously found that heating at 200 °C was required to fully disperse these phosphines on the Davison 980 precipitated catalyst. On the present catalyst 150 °C was sufficient, and indeed there was only a slight change between the spectra of samples heated to 100 °C and 150 °C. This is likely due to easier diffusion because of the larger pore size of the present catalyst. Since we make comparison with samples in which PMe3 was adsorbed from the gas phase at room temperature, we have investigated the effect of heating the latter samples. For PMe3 on the present catalyst, there are no spectral changes and no differences in measured acid amount between samples which have been heated to 100 °C, and samples which have not been heated. Our measurements are directed to determining the total amount of surface Brønsted acid capable of protonating the various phosphines. Lewis sites are also present on these surfaces, so it is necessary to ensure that these are saturated with base, as they might fill preferentially when only small amounts of phosphine are added. Thus we carried out measurements on different samples with increasing amounts of phosphine, until a physisorbed line appeared, and we calculate the amount of surface H+ from the intensity of the PR3H+ line observed at the highest coverages. In the case of PPh3, the physisorbed and Lewis lines cannot be resolved (see below) and we have used a maximum phosphine coverage of 1.0 µmol/ m2. With PMe3 this coverage is more than sufficient to saturate all acid sites, and it is highly improbable that there are Lewis sites accessible to PPh3 but not to PMe3. Results and Discussion Adsorption of PMe3 on this catalyst gave spectra similar to those previously reported,6 and using the methods described there, we measure the amount of surface acid capable of protonating PMe3 as 0.32 µmol/m2. This figure is in good
Figure 1. A. 31P NMR spectrum of PCy3 on monolayer catalyst at a coverage of 0.94 µmol/m2, excitation by 90° pulses. B. Same sample as A, excitation by cross-polarization with 70 µs decoupler delay. C. PPh3 on monolayer catalyst at coverage of 0.40 µmol/m2, 90° pulse excitation. D. Same sample as C, cross-polarization with 70 µs decoupler delay.
Figure 2. A. 31P NMR spectrum of PCy3 at coverage of 0.49 µmol/ m2, 90° pulse excitation. B. 31P NMR spectrum of PCy3 at coverage of 1.68 µmol/m2, excited by cross-polarization.
agreement with the results of 6, where we found 0.27 µmol/m2 for a catalyst containing 4.2 SiO2/nm2, and 0.30 µmol/m2 for one with 8.2 SiO2/nm2. Figure 1 shows typical spectra for PCy3 and PPh3 adsorbed on the monolayer catalyst. Spectra A and B are for PCy3, C and D for PPh3. A and C are spectra recorded by 90° pulse excitation; B and D were recorded by cross-polarization with a 70 µs decoupler delay,15 which causes destruction of signals from P directly bonded to H. In solution, it is found that PCy3H+ resonates at 31 ppm11 and PPh3H+ at 4 ppm12. Thus the peaks at 25 ppm for PCy3 and 5 ppm for PPh3 arise from the phosphonium ions PR3H+ formed by protonation of the phosphines at Brønsted acid sites on the surface. The peaks at 8 ppm for PCy3 and at -5 ppm for PPh3 arise from unreacted physisorbed phosphine. In addition, PCy3 shows a small peak at -7 ppm, due to phosphine coordinated to Lewis acid sites on the surface.11 This peak is more evident in samples of lower phosphine coverage, where the line from physisorbed phosphine is less prominent, see Figure 2A. A Lewis peak cannot be seen with PPh3, because its shift is close to that of the physisorbed phosphine,12 and the peaks cannot be resolved with lines as wide as the present ones. The PCy3H+ peak is somewhat asymmetrical, and curve fitting (see below) shows it to contain two sub-peaks, with shifts of about 24 and 30 ppm. The composite character of this peak is more evident in cross-polarization spectra of samples with a high PCy3 coverage, as shown in Figure 2B. As can be seen in this figure, there are small shifts in the position of the physically adsorbed line, depending on coverage. Relative amounts of the various surface species were determined by fitting Lorentzian lines to the spectra generated by 90° pulses using a relaxation delay (typically 1 to 4 s) that was
Acid Sites on SiO2-Al2O3 Monolayer Catalysts
J. Phys. Chem. B, Vol. 105, No. 1, 2001 219
Figure 3. Lorentzian fit to the spectrum shown in Figure 1A. Upper: crosses are experimental points, curve is sum of fitting Lorentzians. Lower: the four Lorentzians used in the fit.
TABLE 1: Detected Brønsted Acid vs Total Adsorbed Phosphine. Quantities in µmol/m2 tricyclohexylphosphine
triphenylphosphine
total adsorbed phosphine
H+
total adsorbed phosphine
H+
0.41 0.49 0.79 0.94 1.68
0.30 0.36 0.42 0.40 0.36
0.40 0.65 0.74 1.04
0.16 0.13 0.15 0.15
determined experimentally to be long enough to not affect the relative intensity ratios of the peaks. An example of such a fit is shown in Figure 3. As noted above, the PCy3H+ peak is composite, and requires two superposed Lorentzians for accurate fitting. This indicates that two different phosphonium ion species are present, presumably associated with two different surface sites. The relative amounts of these two species are not well defined by the fitting procedure, and we report only the sum of the two, as a measure of the total Brønsted acidity present. We determine the total amount of each species as the fractional area of the Lorentzian peaks used to fit that species, times the total amount of phosphine in the sample. Table 1 shows the total amount of Brønsted acid detected with each of PCy3 and PPh3 as a function of the total amount of adsorbed phosphine. These measured amounts of H+ are reproducible to (0.02 µmol/m2 for repeated measurements on the same sample. Apart from the lowest PCy3 point, the variation with adsorbed coverage scarcely exceeds this, so we adopt average values of 0.39 ( 0.03 µmol/m2 and 0.15 ( 0.02 µmol/m2 as our measured values for the amounts of Brønsted acid detected by PCy3 and PPh3, respectively. The amount of Lewis acid detected with PCy3 is 0.09 ( 0.02 µmol/m2; as noted above, it is not possible to measure this with PPh3. This is less Lewis acid than found with PMe3; from the results of ref 6 we would expect about 0.3 µmol/m2 at the present SiO2 loading. This indicates that only a fraction of the Lewis sites are accessible to the larger PCy3 molecule. As can be seen, PCy3 detects slightly more H+ on this catalyst than does PMe3, whereas PPh3 detects substantially less. This shows that there is a range of Brønsted sites of differing acid strengths. The strongest sites, which are capable of protonating PPh3 comprise somewhat less than half of the total number of Brønsted sites. The fact that PCy3 detects more H+ than does
PMe3 suggests that steric hindrance on this surface is not important for access to H+. It is interesting to compare the above results for a monolayer catalyst with those previously obtained for the Davison 980 catalyst. On the latter catalyst we obtained H+ concentrations of 0.21,6 0.26,11 and 0.1812 µmol/m2 using PMe3, PCy3, and PPh3, respectively. Thus it can be seen that the present catalyst has slightly less of the strongest Brønsted sites (although the difference is close to the experimental uncertainty) and appreciably more of the weaker sites. There is a dramatic difference in behavior of PPh3 on this catalyst, compared with the Davison catalyst. As shown in ref 12, the amount of PPh3H+ formed on the latter passes through a maximum as the total amount of adsorbed phosphine is increased; there is a sharp decline in phosphonium ions for total coverages above 0.5 µmol/m2, which is also observed for other triaryl phosphines on the Davison catalyst. As shown in Table 1, no such phenomenon occurs here. We suggested12 that the disappearance of phosphonium ions on the Davison catalyst was due to their reaction with surface basic sites, this being facilitated by a high coverage of physisorbed species. If this explanation is correct, the present results indicate the absence of such sites on the monolayer catalyst. Another difference is the observation of Lewis sites using PCy3. On the commercial catalyst such sites were seen with PMe3,6 but not with PCy3.11 On the monolayer catalyst, Lewis sites are seen with PCy3, albeit in lower concentration than with PMe3. For both catalysts fewer Lewis sites are observable with PCy3 than with PMe3, suggesting that these sites may be generally less accessible on silica-alumina surfaces than are Brønsted sites. In summary, we find that SiO2 on Al2O3 monolayer catalysts contain at least two populations of Brønsted acids, of appreciably different strength. The concentration of the stronger sites is similar to that on commercial precipitated silica-alumina, while the concentration of the weaker sites is higher. Acknowledgment. This work was supported by a research grant from the Natural Sciences and Engineering Research Council of Canada. References and Notes (1) Ryland, Lewis. B.; Tamele, M. W.; Wilson, J. N. Catalysis VII; Emmett, P. H., Ed.; Reinhold: New York, 1960; Chapter 1. (2) Niwa, M.; Katada, N.; Murakami, Y. J. Phys. Chem. 1990, 94, 6441. (3) Sato, S.; Toita, M.; Sodesawa, T.; Nozaki, F. Appl. Catal. 1990, 62, 73. (4) Sato, S.; Sodesawa, T.; Nozaki, F.; Shoji, H. J. Mol. Catal. 1991, 66, 343. (5) Katada, N.; Toyama, T.; Niwa, M. J. Phys. Chem. 1994, 98, 7647. (6) Sheng, T.-C.; Gay, I. D. J. Catal. 1994, 145, 10. (7) Sheng, T.-C.; Lang, S.; Morrow, B. A.; Gay, I. D. J. Catal. 1994, 148, 341. (8) Katada, N.; Toyama, T.; Niwa, M.; Tsubouchi, T.; Murakami, Y. Res. Chem. Intermed. 1995, 9, 137. (9) Kno¨zinger, H.; Ratnasamy, P. Catal. ReV. Sci. Eng. 1978, 17, 31. (10) Tanabe, K. Solid Acids and Bases; Academic Press: New York, 1970; Chapter 5. (11) Hu, B.; Gay, I. D. Langmuir 1995, 11, 3845. (12) Hu, B.; Gay, I. D. Langmuir 1999, 15, 477. (13) Lias, S. G.; Liebman, J. F.; Levin, R. D. J. Phys. Chem. Ref. Data 1984, 13, 695. (14) Gay, I. D. J. Magn. Reson. 1984, 58, 413. (15) Opella, S. J.; Frey, M. H. J. Am. Chem. Soc. 1979, 101, 5854.
