J. Phys. Chem. 1991,95,4483-4489 number of typical MI sites is three. This MI adsorption requires relatively low activation energy (10.7 kcal mol-') and easily proceeds even below room temperature. Two more types of CH, chemisorption (type M2 and IM) have been also observed experimentally, Although they are very similar in nature to the MI adsorption, they need considerable activation in the initial stage of the adsorptions since their adsorption sites consist of MgLc2+ and OwE isolated each other. Activation by thermal energy and
4483
excitation by UV light lead to the M2 and the IM adsorptions, respectively.
Acknowledgment. We thank the Data Processing Center of Kyoto University for use of the FACOM M782 computer and the Computer Center O f the InstitUte for Molecular Science for permission to use the HITAC M680H "Puter. Registry NO. CHI, 74-82-8; MgO, 1309-48-4.
High-Resolution "Ai and 2aSi MAS NMR Investigation of SlO2-Ai2O9 Glasses R. K.Sato, P.F. McMillan,* Department of Chemistry, Arizona State University, Tempe, Arizona 85287- 1604
P.Dedson, and R. Dupree Department of Physics, University of Warwick, Coventry, CV4 7AI England (Received: September IO, 1990)
"Al and %i MAS NMR spectra have been obtained for SiO2-AI2O3glasses quenched from the melt at two rates: 105-l@ O C s-' (SQ) and @ l -103 OC s-l (NQ). TEM experiments detected phase separation in the S7& SQ and S52A48 SQ samples but no phase separation for the S4,A59 SQ composition. The NQ glasses were phase separated and, in the diffraction TEM experiments, showed diffuse spots indicating microcrystallinity. 27AIspectra for the SQ glasses show three peaks due to aluminum in 4-,5-, and 6-coordination.whereas NQ glasses show only peaks associated with aluminum in 4- and 6-coordination. Quantitative results for the 27Alspectra are discussed. The effect of spin speed on the 27AlMAS NMR spectra for SQ and NQ samples is demonstrated. Magnitudes of mean quadrupolar coupling constants and mean isotropic chemical shifts have been estimated. %i spectra of SQ glassa reveal a distribution of 4-cmrdinate Si sites, whereas NQ glasses show two inequivalent 4-coordinate sites.
Introduction There is a long-standing interest in understanding the structure of aluminosilicate glasses in relation to their material properties. One central problem has been the nature of the aluminum coordination in these glasses. In the present study, we have used Z7Aland %i magic angle spinning (MAS) NMR spectroscopy to investigate the local structures about Si and A1 in a series of binary Si02-A1203glasses. The results of many structural studies on aluminosilicate glasses have been extrapolated to gain an understanding of the corresponding high-temperature liquids, on the assumption that the glass is a good approximation of the melt.lJ In the present work, we have examined samples prepared with extremely different quench rates (10'103 and 105-106 OC s - I ) . ~ ~Comparison ~ of the spectra of one glass composition prepared with both quench techniques places limits on the validity of the extrapolation of glass structural models to the liquid, at least for the SiO2-AI2O3system. Most previouS structural studies on aluminosilicate glass systems have focused on ternary joins M20-Si02-A1203 or MO-Si02A1203,where M is an alkali or alkaline-earth metal, with mole ratio M20/A1203or MO/A1203 I1. For these glasses, it is generally concluded that both A1 and Si are in tetrahedral coordination to oxygen and form aluminosilicate frameworks with varying degrees of polymerization by corner sharing between tetrahedra.Is5d For compositions with MO/A1203 2 1, structural models involving aluminum in higher coordination have been proposed, consistent with observed crystal structures and physical properties of peraluminous melts and glasses.'** McMillan and Piriou' obtained Raman spectra for several glasses along the join SiO2-AI2O3and observed a loss of Raman band structure with increasing A1203content, which was interpreted as due to extreme disorder in local coordination. Risbud and co-workers9 obtained 27Aland 29si MAS NMR spectra for a series of roller quenched To whom cmcapondence should be addtesscd at Department of Geology, University of Illinois, Urbana, IL 61801.
SiO2-Al2O3glasses and ohserved peaks for 4-,5-, and 6-coordinate aluminum in the glasses. These workers found evidence for unmixing in their glass samples and also observed a weak peak near 15 ppm in their 27AlNMR spectra due to crystallization of corundum. In the present work, we have used 29Siand 27AlMAS NMR to study the same series of samples studied by McMillan and P i r i ~ uto , ~investigate the effects of quench rate on the glass structure, and to quantify the numbers of A1 sites observed in the NMR experiments.
Experimental Section Sample Characterization. The glass samples were synthesized for a previous Raman spectroscopic study,3 and details of the (1) Seifert, F. A.; Mysen, B. 0.;Virgo, D. Geochim. Cosmochim. Acra 1981,45, 1879-1884. (2) Sweet, J. R.; White, W. B. Phys. Chem. Glasses 1969, 10, 246-251. McMillan, P. F. Am. Mineral. 1984,69,622-644. Seifert, F. A,; Mysen, B. 0.;Virgo, D. Am. Mineral. 1982, 67, 696-717. Mysen, B. 0.;Virgo, D.; Seifert, F. A. Am. Mineral. 1985, 70, 88-105. Mysen. B. 0.Structure and Properties of Silicate Melts; Elsevier Science: New York, 1988. (3) McMillan, P. F.; Piriou, B. J . Non-Cryst. Solids 1982,53, 279-298. (4) Coutures, J. P.; Bcrjoan, R.; Benczech, G.;Granicr, B. Rev. IN. Hautes Temps. Refract. 1978, IS, 103. (5) McMillan, P. F.; Piriou, B.; Navrotsky, A. Geochim. Cosmochim.Aero 1982,46,2021-2037. Matson, D. W.; Sharma, S.K.; Philpotts, J. A. Amer. Mineral. 1986, 71, 694-704. (6) Oestrike, R.; Wang-Wong, Y.; Kirkpatrick, R. J.; Hervig, R. L.; Navrotsky, A. and Montez, B. Geochim. Cosmochim. Acra 1987, 51, 2199-2209. (7) Aksay, I. A.; Pask, J. A,; Davis, R. F. J . Am. Ceram. Soc. 1979.62, 332-336. Riebling, E. F. Rev. Int. Haures Temps. Refract. 1967,4,65-76. Mehta, S.; Risbud, S. H. J . Am. Ceram. Soc. 1979,62,641-642. Leonard. A.; Suzuki, S.;Fripiat, J. J.; De Kimpe, C. J . Phys. Chem. 1964, 68, 2608-2617. Hanada, T.; Soga, N. J. Am. Ceram. Soc. 1982,65, (284x86. (8) Morikawa, H.; Miwa, S A ; Miyake, M.; Marumo, F.; Sata, T. J . Am. Ceram. Soc. 1982,65, 78-8 1. (9) Risbud, S. H.; Kirkpatrick, R. J.; Taglialavore, A. G.; Montez, B. J . Am. Ceram. Soc. 1987, 70, ClOC12.
0022-365419112095-4483502.50/0 0 1991 American Chemical Society
4484 The Journal of Physical Chemistry, Vol. 95, No. I I , 1991
I
I
II I
I I I
B I Figure 1. TEM micrographs showing phase separation of SQ glasses S16A24and SS2Aw Glass droplets are dispersed in a glass matrix.
sample preparation have been described elsewhere? Two quench methods were used to prepare the samples. One series of glasses was quenched from the liquid by simply removing the sample from the focal point of the solar beam and allowing the liquid to cool in air on the water-cooled sample stage. This resulted in a quench rate timed at 1 0 t 1 0 3 OC s-', termed "normal quench" (NQ). A second series of samples was obtained by hammer quench techniques, giving a quench rate estimated at 105-106 OC s-'. These samples are denoted "super quench" (SQ). Sample compositions were determined by electron microprobe and X-ray fluorescence analyses as part of the previous study.3 Five samples with bulk compositions S76A24 (NQ and SQ; the subscripts refer to the mole per cent Si02(S)and A1203 (A) components in the glass), S71A29(NQ), S52A48 (SQ), and S41A59 (SQ) were examined in this study. The samples were further characterized by transmission electron microscopy in this study using a JEOL JEM-2000 FX instrument in the Center for High Resolution Electron Microscopy (Center for Solid State Science) at Arizona State University. The samples were crushed and mounted on carbon-coated copper grids by using an acetone sample slurry. A 120-kV accelerating voltage was used to perform selected area diffraction and high-contrast imaging at magnifications up to 500kX to detect phase separation and/or microcrystallinity within the samples. The normal quenched sample S71A29(bulk composition) was phase-separated, with glassy droplets approximately 10 pm across of composition S4,AS9dispersed in a matrix of approximate composition &a2+ Both s7&24 NQ and SQ samples were found to be homogeneous on the scale of the electron microprobe analyses (on the order of 2-5 pm), but TEM studies showed evidence for phase separation at the 20-100-nm level (Figure 1). The SQ sample S&8 (bulk and matrix composition) showed occasional large (5-10 pm) glassy inclusions with composition S4'AS9,but
Sat0 et al. TEM imaging showed no evidence for phase separation on the 10-20-nm scale throughout the sample. No evidence for phase separation in the S4,AS9S Q sample was observed with magnifications up to 500kX in the TEM experiments. From these observations, it is likely that the low-silica pole of one of the suggested metastable miscibility gaps in the Si02-A12.03 glass systemlo lies near S41A59 This composition for the low-silica pole agrees well with the thermodynamic calculations of Aksay and Pask" and Risbud and Pask.12 Risbud and Pask suggested that the high-silica pole extended to at least 90 mol 96 Si02, and showed direct evidence for unmixing in a glass with 85 mol 5% S O 2 , in agreement with our observations. Extremely diffuse diffraction spots were observed for the S7&a and S71A29NQ glass samples, indicating the presence of poorly crystalline regions within the glass. Mullite is the liquidus phase in this region of the composition diagram,12 so these diffuse spots are probably due to nuclei of crystalline mullite within the glass samples. No crystals were observed by imaging at magnifications up to 500kX, and no crystal peaks were observed in the Raman spectra for these glasses,3 so it is unlikely that this small amount of crystallinity affected the NMR results. NMR Spectroscopy. Solid-state magic angle spinning nuclear magnetic resonance (MAS NMR) spectra were obtained at spin speeds up to 15 kHz by using a Doty MAS probe in a Bruker MSL-360 (8.45 T) at the Department of Physics, University of Warwick. Higher field spectra were obtained on a Bruker AMX-500 (1 1.7 T) at the Bruker Regional Center in Coventry, England, using a MAS probe built with Doty dual air bearing drive unit at the University of Warwick. Dual-bearing 5-mmdiameter rotors made from Si3N4(for spectra) and alumina (for 29Sispectra) were used with the above'instrumentation and contained up to -0.1 g of crushed sample. Additional spectra were obtained at the Department of Chemistry, Arizona State University, on a Bruker AM-400 instrument (9.4 with a Bruker MAS probe capable of spin speeds up to 6 kHz using Delrin single-bearing rotors with up to -0.3 g of finely crushed sample. The magic angle was set by maximizing the intensity of the spinning sidebands for 79Br.13 27Al MAS NMR runs on the MSL-360 utilized 1-ps pulse lengths corresponding to amlution pulse angle of -30° and pulse delay of 0.5 s. Instrument dead time between the pulse and acquisition of the free induction decay was 6.8 ps. AMX-500 spectra were collected with a 2-ps pulse length, 18-ps dead time, and 0 5 s pulse delay. Spectra accumulated on the AM-400 utilized a 1-ps pulse length, 25-ps dead time, and pulse delay of 1 s. 27Alspectra were generally limited to 2000 scans and referenced to external 1 M [A1(H20)6l3+. Due to aluminum-containing components in the NMR probes used in the MSL-360 and AMX-500 instruments, free induction decays (FIDs) were accumulated for the background and subtracted from the FIDs obtained for the samples, which were then processed with 100-Hz exponential smoothing. For experiments conducted on the AM400, no signal from the probe could be detected for the duration of the "Al experimental run times, and background subtraction was not carried out. For 27AlMAS NMR experiments, quantitative analyses were carried out to determine the amount of alumina detected in the sample. A reference free induction decay (FID) was obtained for a known quantity of corundum (a-A1203)with a predetermined number of scans. Weighed samples were run under identical conditions with the same number of scans to obtain FIDs. Points were removed from the start of the FID before being Fourier transformed, and integrations carried out to determine area under the peaks of the spectra. For each successive point removed from (10) Aramaki, S.; Roy, R. J. Am. Cerum. Soc. 1962, 45, 229-242. MacDowell, J. F.; Beall, G. H. J. Am. Cerum. Soc. 1969,52,17-25. Jantzen, C. M.; Herman, H. J . Am. Cerum. Soc. 1979,62,212-213. Risbud, S. H.; Pask, J. A. J. Am. Cerum. SOC.1979.62.214-215. Risbud, S. H. J . NonCryst. Solids 1982, 49, 241-252. ( 1 1) Aksay, I. A.; Pask, J. A. J . Am. Cerum. Soc. 1975, 58, 507-512. (12) Risbud, S. H.; Pask, J. A. J. Am. Cerum. SOC.1977,60,418-424. (13) Frye, J. S.; Maciel, G. E. J . M u p . Reson. 1982,48, 125-131.
27Al and 29SiMAS N M R Spectra of Si02-A1203 Glasses log (area) vs polnts removed
The Journal of Physical Chemistry, Vol. 95, NO. 11, 1991 4485 27~1 V
1 '
2.3
0
5
10
pints removed
Figure 2. Plot of log area verses the number of points removed from the FID for the 27A1spectrum obtained at a spinning speed of 14 kHz for sample S 7 d Z 4SQ. The instrument dead time was 6.8 ps, and time between points was 1.2 ps. Areas represent integrations carried out for regions between 400 and -400 ppm. Locations of the first two points are affected by probe ringing.
the FID, integrated peak areas of the spectrum were determined and correlated with the instrument dead time (time between the end of the pulse and the start of the FID acquisition) represented by the first point of the modified FID. Areas representing aluminum detected by the NMR experiment were found to vary logarithmically with the time associated with the instrument dead time. Extrapolation back to zero time then allowed the amount of aluminum relative to the alumina standard observed for a 6.8-ps dead time to be estimated (Figure 2). 29si MAS NMR spactra gathered on the MSL360 used a pulse length of 1 ps corresponding to a 10-15O pulse angle, dead time of 24 ps, and pulse delay of 120 s. 29Sispectra were acquired on the AM-400 with a pulse length of 4.5 ps (30° pulse angle), 24-ps dead time, and pulse delay of 120 s. For 29Si runs on both instruments, the number of scans was chosen to maximize signal-to-noise but was generally in the range of 800-1200 scans. %i spectra were referenced to external T M S (tetramethylsilane) and processed with 100-Hz exponential smoothing.
Results and Discussion 27AI MAS NMR Spectroscopy. 27Al MAS N M R spectra obtained at 8.45 T and spin speeds near 15 kHz for the Si02A1203(SA) SQ glasses (Figure 3) show three peak maxima. The S76A24 SQ glass (indicating 76 mol 9% SiO2and 24 mol 76 A1203) shows a peak at 29.9 ppm with shoulders at 12.3 and 51.3 ppm. For the S52A48 SQ sample, the peak and shoulders occur at 4.6, 28.1, and 58.4 ppm. These features are located at 6.7,26.0, and 50.0 ppm for the S4IASg SQ sample. Because of the rapid spin rate, none of these maxima can be attributed to spinning sidebands from the central transition. The first spinning sidebands for this transition occur at approximately 200 and -125 ppm. The three peaks are in fact better resolved at lower spinning speeds, although the spectra are complicated by spinning sidebands and a remnant broad central static line shape. The effect of spin rate on the N M R spectra is discussed in more detail below. Changes in relative height of the three bands are apparent as composition is varied for the three SQ samples. The shoulders occurring near 7 and 50 ppm increase in intensity relative to the middle peak as the AI203 content is increased, consistent with either a change in relative populations of AI sites, or a shift in peak maximum of the central feature. These spectra are similar to those obtained by Risbud and co-workers9 for SiO2-AI2O3
,
800
'
160
'
A
PPM
"
'
-100
o
-200
Figure 3. 27AlMAS NMR spectra of SQ samples showing peaks corresponding to 4, 5-, and 6-coordinated A1 sites, and NQ spectra with only peaks corresponding to 4- and 6-coordinate sites. Sample compositions in mol 5% SiOl (S)and A1203(A) are given for each spectrum. Spectra were obtained at 8.45 T (93.83 MHz) at spin speeds near 15 kHz.
glasses, except that we see no evidence for crystalline A 4 0 3 in our spectra. The 27Alchemical shifts for aluminum in various oxygen coordinations fall into distinct characteristic range^.'^,'^ This allows assignment of maxima in the SQ glass spectra near 50 ppm to A104 tetrahedral units and those near 7 ppm to A106 octahedral units. The peak near 28 ppm can be assigned to aluminum 5-coordinated to oxygen following Risbud and c o - ~ o r k e r sby ,~ its intermediate position between the 4- and 6-coordinate peak positions?.l6 and by the similarity in peak position to the 5-coordinated aluminum peak positions for crystalline a n d a l ~ s i t e , ~ ~ , ' ~ barium glycoaluminate,18and pyrophyllite dehydro~y1ate.I~ The 27Al MAS N M R spectra for the two N Q glasses S76A24 and S71A29 are also shown in Figure 3. For the S71A29 N Q glass, a narrow peak near -0.6 ppm is well resolved from a broad peak centered near 50.5 ppm. These are assigned to 6- and 4-coordinated sites, respectively. The spectrum for the s7&24 NQ sample is noisier due to a limited amount of sample. The peak position of the narrow peak for the S76A24 N Q sample occurs near -4.3 ppm, and the broad peak is centered near 43.4 ppm. The peak positions for the S76A24N Q spectrum are shifted to lower frequency relative to the S71A29 N Q sample. There is no evidence (14) Kinsey, R. A.; Kirkpatrick, R. J.; Hower, J.; Smith, K. A.; Oldfield, E. Am. Minerd. 1985, 70, 537-548. Muller, D.; Gtssner, W.; Behrens, H. J.; Scheler, G. Chem. Phys. Len. 1981, 79, 59-62. (15) Kirkpatrick, R. J.; Smith, K. A,; Schramm,S.;Turner, G.; Yang, W. H. Annu. Rev. Plonet. Sei. 1985, 13, 29-47. (16) Dupree, R.;Farnan, I.; Forty, A. J.; El-Mashri, S.; Bottyan, L. J .
Phys. (Poris) 1985, 46, 113-1 17. (17) Alemany, L. B.; Kirker.
G. W. J. Am. Chem. Soc. 1986, 108, 6158-6162. (18) Cruickshank, M. C.; Dent Glasser, L. S.;Barni, S.A. I.; Popplett, 1. J. F. J. Chem. Soc., Chem. Commun. 1986, I , 23-24. (19) Fitzgerald, J. J.; Dec, S. F.; Hamza, 1. A. Am. Minerol. 1W9,74, 1405-1 408. (20) Nakajima, Y.; Morimoto, N.;Watanabe, E. Proc. Jpn. Acod. 1975, 51, 173-178. Sadaga, R.; Tokonami, M.; Takeuchi, Y.Act4 Crystollogr. 1962, IS,65-68. Burnham, C. W.Cornegie Inst. Wush. Teur Book 1963, 62, 223-227.
4486 The Journal of Physical Chemistry, Vol. 95, No. 11, 1991
for 5-coordinated aluminum sites in either N Q sample. The broad peak observed in the N Q glass spectra near 50 ppm associated with 4-coordinated aluminum could not be further resolved with extended runs at higher magnetic field strengths (a single broad smooth peak was obtained at 9.4 and 11.7 T for runs greater than 2000 scans). Relative peak areas for the 4- and 6-coordinate sites are different for the two samples, indicating different abundances of aluminum in these coordinations. 27AlQuantirarion. Quantitation experiments are difficult to perform since signal from fast relaxing sites can be lost during the pulse breakthrough period inherent in pulse methods.21-22 Further, incomplete excitation of sites with large electric field gradients results in incomplete detection of aluminum nuclei.8919921-23Aluminum nuclei in sites with large electric field gradients also suffer extreme quadrupolar line broadening, which can result in no observable peak in the high-resolution N M R e ~ p e r i m e n t . ~In ’ , ~addition, ~ Massiot and c o - ~ o r k e r have s ~ ~ recently shown that spinning sidebands from outer transitions can affect intensity in the region of the central transition. Quantitative experiments were carried out with the Bruker MSL-360 instrument (8.45 T). As described above, a free induction decay (FID) was obtained for a known quantity of a-A1203 (corundum) by using the standard set of run conditions described in the previous section. Glass samples were weighed and run under identical conditions. Points were removed from the start of the FID, which was then Fourier transformed. Each spectrum was integrated from 400 to -400 ppm, and a percentage of A1 seen in the sample relative to the standard was calculated. For each successive point of the FID that was removed, the relative percentage of aluminum detected was plotted against the time represented by the first point of the modified FID. The relation was observed to be logarithmic. Extrapolation to zero time allowed determination of the relative amount of aluminum observed for the actual 6.8-1s dead time, with an estimated error of 4~15%. For the NQ samples, the results of the quantitative experiments indicate that all (100 f 10%) of the aluminum was detected by the NMR experiment, whereas for the SQ samples, 70-858 of the total aluminum relative to the standard was observed. A similar loss of NMR signal has been observed for A1 in mixed AI(O,N), tetrahedral sites, with a large electric field gradient at the AI nucleus.a For the oxide glasses in this study, the undetected aluminum must be in sites with large electric field gradients resulting from distorted local A10, coordinations. In their Raman study of the super quenched samples in the SO2-A1203 glass series, McMillan and Piriou3 suggested that the absence of Raman band structure in the high alumina glasses was consistent with a continuous range of “nonintegral” A1 coordinations, rather than discrete populations of A104 and A106 polyhedra. The loss of N M R signal for the SQ glasses suggests that a range of highly distorted AlO, polyhedra is in fact present in these glasses, in addition to the 4-, 5-, and 6-coordinate sites observed. From the quantitative NMR experiments, these sites must represent 15-30% of the total aluminate polyhedra present, in the SQ Si02-A1203 glasses. Effect of Spin Speed. 27Alstatic and MAS NMR spectra at several spin speeds were obtained for samples on the MSL 360 (8.45 T ) instrument. The static spectrum as well as spectra obtained at different spin speeds for the S71A29 NQ glass sample are shown in Figure 4. The static spectrum shows a single broad band between approximately 400 and -300 ppm with a maximum near 33 ppm. Spinning the sample at a spinning speed of 5 kHz a t 54.7O to the magnetic field results in the appearance of two peaks a t 48 and 0 ppm corresponding to 4- and 6-coordinate AI sites in the glass, but there remains a broad baseline feature that resembles the static spectrum. As the spin speed is increased to (21)Kirkpatrick, R. J.; Oestrike, R.; Weiss Jr., C. A.; Smith, K. A,; Oldfield, E. Am. Mineral. 1986,71, 705-71 1. (22)Oestrike, R.; Kirkpatrick, R. J. Am. Mineral. 1988,73, 534-546. (23)Dupree, R.; Lewis, M. H.; Smith, M. E. J. Appl. Crystallogr. 1988, 21, 109-116. (24) Massiot, D.; Bessada, C.; Coutures, J. P.;Taulelle, F. J . Magn. Reson.,
in
press.
Sat0 et al. 27AI ‘%lA29 NQ
100
500
400
300
100
100
0
-100
-200
-300
-400
-100
-600
*P*
Figure 4. 27AINMR spectra for the S7!AZ9NQ sample showing the effect of spin speed. Bottom spectrum is for the static case showing the characteristic broad static lineshape. Initial increase in spin spccd to 5 kHz reveals two peaks. Further increased spin speed shows decreased contribution of the static pattern as well as decreased resolution. Spectra were obtained at 8.45 T (93.83 MHz).
9 kHz, the area under the broad baseline decreases relative to the 4- and 6-coordinate A1 peaks. At the same time, the 4- and 6-coordinate peaks become less well-resolved. Finally, spinning the sample a t 12 kHz results in a relatively flat baseline, but resolution of the 4- and 6-coordinate peaks is further decreased. The effect of spin speed on the spectra can be understood by considering the relative band-broadening effects of chemical shift dispersion and anisotropic interactions. These effects contribute to the broadened band of the static spectrum. Spinning a t the magic angle removes dipolar broadening, broadening due to chemical shift anisotropy and reduces quadrupolar broadening for the least distorted ~ i t e s . ~ ’ - For ~ ~ half-integer quadrupolar nuclei, first-order quadrupolar interactions do not contribute to the l i n e ~ i d t h . ~ ~In- *practice, ~ quadrupolar broadening of the central to transition can be reduced by a factor of 3.5-4.0 by spinning a t the magic angle, once the spin speed becomes greater than the width of the static broadening contribution^.^^ Therefore, a t a given spinning speed, the peaks for 4- and 6-120ordinate sites observed in Figure 4 are a subset of the A1 sites present in the sample with more regular geometries and a smaller contribution from quadrupolar broadening. The remainder of the sites contribute to the unresolved background signal. Spinning at faster speeds results in further reduction of quadrupolar broadening for more distorted sites with larger quadrupolar contributions to the line width. However, these more distorted (25) Andrew, E. R. Rev. Phys. Chem. 1981,l. 195-224. Harris, R. K. Nuclear Magnetic Resonance Spectroscopy; Pitman Books: London, 1986. (26) Engelhardt, E.; Michel, D. High-Resolution Solid-State NMR of Silicates and Zeolites; John Wiley and Sons: New York, 1987. Ganapathy, S.;Schramm, S.;Oldfield, E. J. Chem. Phys. 1982,77,4360-4365.Meadows, M. D.;Smith, K.A,; Kinsey, R. A,; Rothgeb, T.M.; Skarjune, R. P.;Oldfield, E. Proe. Nail. Acad.Sci. U S A . 1982,79,1351-1355.Samoson, A.; Lipmaa, E. Chem. Phys. Lett. 1983, 100, 205-208. Behrens, H.-J.; Schnabel, B. Physica 1982,1148,185-190. Samoson, A.; Lipmaa, E.; Alma-Zeestraten, C. Bruker Rep. 1984,I , 14-15. Smith, M. E. A High Resolution Multinuclear Magnetic Resonance Study of Ceramic Phascs. Ph.D. Thesis, University of Warwick, Coventry, England, 1987;pp 34-36.
27A1and 29SiMAS NMR Spectra of Si02-AIz03 Glasses
The Journal of Physical Chemistry, Vol. 95, No. 11, 1991 4487
27AI
s71A29 NQ
274
9-
11.7 T18.45 T
-
1.39
4-coordinuc = 1.33 6-coordinrte = 0.81
elr
4Ao
260
i
PPY
.
8
-200
'
-400
q
V
-600
Figure 5. 27Al MAS N M R static spectrum for the SS2AUIS Q sample obtained at 8.45 T (93.83 kHz) shown at bottom. Spinning the sample at 5 kHz resolves peaks corresponding to 4-,5-,and 6-coordinated sites. As the spin speed is increased, the broad static feature decreases but resolution of peaks is lost, as discussed in the text.
4- and 6-coordinate sites cover a wider range of chemical shifts than the less distorted sites, so that peaks for particular aluminum coordination environments are broadened as spin speed is increased, and peak resolution is diminished. In addition to these arguments for the loss in peak resolution with increased spin speed, Massiot and co-workersZ4have recently shown that there is an increased contribution from spinning sidebands of outer transitions in the region of the central transition as the spin rate is increased. This would also tend to reduce peak resolution. For the three SQ samples which show peaks near 0, 30, and 50 ppm associated with 6-, 5-, and 4-coordinated Al sites, increased spinning speed similarly distributes the broad baseline feature into the respective peaks. However, due to the small initial separations of the three peaks, resolution of the peaks is completely lost with faster spinning. The resulting spectrum obtained at 14-15 kHz appears as a central peak with two shoulders, compared to spectra obtained for the same sample at slower spinning speeds with three distinct peaks (Figure 5). Spectra were obtained for each sample at different spinning speeds from 5 to 15 kHz to confirm the presence of peaks corresponding to aluminum in 4-, 5-, and 6coordination in these fast-quenched samples. 27AIMAS N M R Spectroscopy at Different Field Strengths. By utilization of peak positions obtained at two magnetic field strengths (8.45 and I 1.7 T) it was possible to estimate mean values for quadrupolar coupling constants for the aluminum sites.27 Calculations of the mean quadrupolar coupling constants for the Scoordinate sites of the SQ samples fell in the range 3.0-3.5 MHz, smaller than the quadrupolar coupling constants determined for the 5-coordinate sites in andalusite (5.9 MHz)17 and pyrophyllite dehydroxylate ( I 0.5 MHz).I9 The calculated mean quadrupolar coupling constant for the Ccoordinate site of the S41AS9SQ sample was 3.5 MHz, comparable to values determined for the 4-coor-
40000
20000
'
1___
0
-20000
-40000
HERTZ
Figure 6. 27AlM A S N M R spectra for the S7IA29 N Q sample obtained at 8.45 T, 93.83 MHz (bottom) and 11.7 T,130.35MHz (top) for a spin speed of 9 kHz.
dinate sites in albite (3.29 MHz) and microcline (3.22 MHZ).~* As noted above, peak widths in 27Al MAS N M R spectra discussed above are determined by a convolution of chemical shift dispersion and quadrupolar coupling effects. Extremely distorted sites with large quadrupolar coupling constants have broadened peaks, but peak breadth can also be attributed to chemical shift dispersion where a distribution of similar sites exists in the sample. In an effort to resolve the relative contribution of these broadening mechanisms, spectra were obtained at two different field strengths for the S71A29NQ sample (Figure 6), which show two resolved peaks for 4- and 6-coordinate A1 sites. Spectra were obtained a t 8.45 and 11.7 T, spinning the sample a t -9 kHz. Comparison of the ratio of magnetic field strengths (1.39) to the ratios of the 4- and 6-coordinate peak widths (fwhm) obtained at the two magnetic field strengths, 1.33 and 0.81, respectively, shows that nearly all of the width of the Ccoordinate peak is due to chemical shift dispersion but that most of the width of the narrow peak is due to quadrupolar broadening. This implies that the range of chemically distinct 4-coordinate AI environments in the glass is larger than that for 6-coordinate sites. Calculations utilizing peak positions obtained a t the two magnetic field strengths also allow approximate determination of isotropic chemical shift values.27 For the 6-coordinate site, the isotropic chemical shift values are shifted 1-2 ppm from the peak positions obtained at 8.45 T. The isotropic chemical shift for the 5-coordinate site is shifted approximately 3 ppm from the position determined at 8.45 T, and for the 4-coordinate site, the isotropic chemical shift is deshielded 3-4 ppm. 29SiMAS NMR Spectroscopy. Because 29Siis a dipolar nucleus, these spectra were not subject to the complications associated with the interpretations of the 27AIqaudrupolar nucleus. However, the signal-to-noise ratio of the spectra is much poorer due to the low natural abundance of 29Si,its low NMR sensitivity, and long relaxation times.29 ~~~~~
(27) Kohn, S. C.; Dupree, R.;Smith, M.E.Geochim. Cosmochim. Acra 1989, 53, 2925-2939.
(28) Kirkpatrick, R.J.; Kinsey, R.A,; Smith, K. A.; Henderson, D. M.; Oldfield, E. Am. Mineral. 1W5,70, 106-123.
4488 The Journal of Physical Chemistry, Vol. 95, No. 1 1 , 1991 29si
(29) Kirkpatrick, R. J. MAS NMR Sptctroecopy of Minerals and Glasscs. In Specrroscopic Methods in Mineralogy and Geology; Rev. Min. 18; Hawthorne, F. c., Ed.; Mineralogical Society of America: Washington, D.C., 1988; pp 341-404. Stebbins, J. F. NMR Spectroscopy and Dynamic Proasses in Mineralogy and Geochemistry. In Specrroscopic Merhods in Mineralogy und Geology; Rev. Min. 18; Hawthorne, F. C., Ed.;Mineralogical Society of America: Washington, D.C., 1988; pp 405-430. (30) Stebbins, J. F.; McMillan, P. F. Am. Mineral. 1989, 74, 965-968. Xue, X. Y.;Stebbins, J. F.; Kanzaki, M.; Tronnw, R. G. Science 1989, 245, 962-964. (31) Smith, J. V.; Blackwell, C. S. Narure 1983, 303, 223-225. Smith, J. V.; Blackwell, C. S.; Hovis, G. L. Narure 1984,309, 140-142. Smith, K. A.; Kirkpatrick, R. J.; Oldfield, E.; Henderson, D. M. Am. Minerul. 1983, 68, 12061215. Ramdas. S.; Klinowski. J. Nurure 1984, 308, 521-523. Engelhardt, 0.;Radeglia, R. Chem. Phys. Lctr. 1984, 108,271-274. (32) Fyfe, C. A.; Gobbi, G. C.; Hartman, J. S.;Klinowski, J.; Thomas, J . M. J . Phys. Chem. 1982,86. 1247-1250. (33) Pettifer, R. F.; Duprec, R.; Farnan, 1.; Sternberg, U. J . Non-Crysr. Solids 1988. 106.408-412. (34) Murdoch; J. B.; Stebbins, J. F.; Carmichael, I. S.E. Am. Mineral. 1985, 70, 332-343. (35) Selvaray. U.; Rao, K. J.; Rao, C. N. R.; Klinowski, J.; Thomas, J. M. Chem. Phys. Lett. 1984, 114,24-27. (36) Lipmaa, E.; Magi. M.; Samoson, A.; Tarmak, M.; Engelhardt, G. J . Am. Chem. Soc. 1980, 102, 4889-4893.
Sat0 et al. position also has been observed with depolymerization of the glass network,6~’s.2Z~34~36‘8 which could suggest the presence of some Si-0- groups with nonbridging oxygens in the glass network. On increasing alumina content to the composition of the SS2& SQ sample, the peak position remains near -109 ppm, but the unresolved shoulder near -90 ppm increases in intensity to become equal in intensity to the maximum near -109 ppm, reflecting the relative increase in alumina content of the glass. The signal to noise for the S4’AS9SQ spectrum is poor due to the small quantity of sample available (-20 mg). Relative to the S52A48 spectrum, the entire band appears to shift to a more shielded position, in the direction opposite from that expected from known compositional trends. In addition, the fwhm of the band appears to have decreased for the S4’AS9S Q sample indicating a decreased distribution of Si sites types. Further discussion must await better spectra obtained for a larger quantity of sample. However, we note that the S4,AS9SQ sample was the only glass sample in this study that was not phase separated and the 29Sispectrum could reflect this difference. The %i spectrum of the S79%SQ sample is similar to spectra obtained by Risbud and co-workers9 for roller quenched Si02A1203glasses with silica contents of 72 and 80 mol %. Reported peak positions for the roller quenched samples coincide with the -109 ppm peak position determined for the dominant peak of the SQ glass. The spectrum obtained for the Ss2& SQ sample of the present study with the deshielded shoulder at equal height with the -109 ppm peak resembles the spectrum of their roller quenched S63A37 sample. Effect of Quench Rate on the S 7 d 2 4 Composition. Glass samples for composition S76A24 were prepared by using both quench rates. The 27Al N M R spectra for both N Q and SQ samples are presented in Figure 3. Three peaks are present in the 27AI spectrum for the S76A24 SQ sample, indicating the presence of aluminum in 4-, 5-, and 6-coordination. From quantitative work, 15-30% of the aluminum in the sample was not detected in the N M R experiment, suggesting that highly distorted AI sites are present in addition to the 4-, 5-,and 6-coordinated sites represented by the peaks. For the S76A24 N Q sample, aluminum is present in 4- and 6-coordinated sites ( 5 coordination is not present). The quantitative work indicated that 100% of the aluminum in the sample was detected, so that no additional sites are present in this sample. The loss of the 5coordinated site as well as complete detection of the aluminum for the NQ sample suggests that the A1 coordination environments are much better ordered in the N Q sample compared with the SQ glass sample. The quench rate difference between these two samples results in a higher fictive temperature for the S Q glass, although this is complicated by the Occurrence of phase separation in these glasses. The large variations in A1 and Si coordination environment for this glass composition with quench rate indicate that extrapolation of the structural information for the normal quenched glass and probably also the super-quenched glass to the high-temperature liquid would not be appropriate. The fact that the 4- and 6-coordinate A1 sites are retained in the NQ glass sample but the 5-coordinate site (and highly distorted sites representing the “missing” NMR signal) is not present indicates that either the structural relaxation rate of the 5-coordinate A1 (and other distorted) sites is more rapid than the 4- and 6-coordinate sites or that the 5-coordinate sites are stable only at higher temperature. Conclusions For the hammer-quenched Si02-A1203 glasses examined with A1203contents ranging from 25 to 60 mol %, aluminum is present in 4-, 5-, and 6-coordination, confirming the results of Risbud and (37) Engelhardt, G.; Noh, M.; Forkel, K.; Wihsmann, F. G.; Magi, M.; Samoson, A.; Lipmaa, E. Phys. Chem. Glasses 1985, 26, 157-165. (38) Fyfe, C. A. Solid State NMR for Chemists; CRC Press: Guelph, Ontario, Canada, 1984. Lipmaa, E.; Magi, M.; Samoson, A.; Tarmak, M.; Engelhardt,G. J . Am. Chem. Soc. 1981,103,4992-4996. Magi, M.; Lipmaa, E.; Samoson, A.; Engelhardt, G.; Grimmer, A. R. J . Phys. Chem. 1984,88, 1518-1522.
J. Phys. Chem. 1991, 95,4489-4495 c o - ~ o r k e r s .Quantitation ~ of the Z7AlNMR data showed that all of the Al signal was detected for the NQ samples but that only 7045% of the signal was detected by the N M R experiment for the SQ samples, indicating that 1 5 3 0 % of the aluminum is in other highly distorted sites, in addition to the 4-, 5-, and 6-coordinate sites observed in the spectra. Samples quenched with slower quench rate showed only 4- and 6-coordination for AI, reflecting differences in the local structure of the glass with quench rate, due to different structural relaxation rates for different AI sites or to differences in thermodynamic stability for 4-, 5-, and 6-coordinate sites. Only 4-coordinate Si was detected in the samples. Contrary to the situation for spin I = systems or for quadrupolar spectra of crystalline materials, increased spinning speeds resulted in a loss of resolution, even though this results in more effective line narrowing due to more effective removal of second-order quadrupolar broadening at higher spin speeds. The
4489
observed loss of resolution can be understood from a consideration of the combined contributions from quadrupolar interactions and chemical shift dispersion. Though quadrupolar broadening is removed from more distorted sites as the spinning speed is increased, these more distorted sites contribute to a wider distribution of chemical shifts, resulting in diminished peak resolution.
Acknowledgment. R.K.S. and P.F.M. were supported by National Science Foundation Grants EAR 8616990 and EAR 8916004 to P.F.M. R.D. was supported by SERC. The N M R instrument at ASU was purchased with funds from ASU and the National Science Foundation (CHE 8409644). The TEM instrument was purchased with funds from the Department of Energy (DOE 300860053). We thank Dr.A. Curton for allowing use of the AMX-500 at Bruker Spectrospin, Coventry, England. We are appreciative of careful reviews and useful comments. Registry No. A1203, 1344-28-1.
Molecular Transport in Porous Silica. Quenching of Fluorescence from Surface- Immobilized Pyrene A. L. Wong, M. L. Hunnicutt, and J. M. Harris* Department of Chemistry, University of Utah, Salt Lake City, Utah 841 I 2 (Received: September 21, 1990; In Final Form: January 8, 1991)
Quenching of fluorescence from a surface-immobilized probe is developed as a method for measuring molecular transport in porous silica. In these experiments, 3 4 I-pyrenyl)propyldimethylchlorosilane is covalently bonded to the surface of porous silica, and the rates of fluorescence quenching of this fluor by iodine were measured in a series of primary alcohols. Interfacial quenching rates were compared to those measured for 1-methylpyrene in free solution in the same solvents. Differences in the measured rates of quenching at the liquidlsolid interface could be attributed to the volume excluded from quencher by the solid surface. A simple hemispherical diffusion model provided reasonable predictions of the rates of quenching based on the diffusion coefficients of iodine measured in free solution. The results indicate that the viscosity of alcohols within the pores of silica does not differ significantly from bulk solution values.
Introduction
Considerable effort has been recently directed toward the characterization of liquid/solid interfaces and the role played by the surfaces of porous solids in interfacial chemical reactions. It is known that the rates and products of chemical reactions in porous media can be modified by differences in molecular transport in porous structures and by the large surface area of the interface. A number of solid substrates, including zeolites, porous membranes and polymers, porous glasses and minerals, possess unique structural and interfacial properties that can impact chemical reactions. The widespread use of these materials in chemical separations and heterogeneous reactions has allowed the application of chemical processes at liquid/solid interfaces to grow faster than the fundamental understanding of the physical nature of the interface and the dynamics of interfacial chemistry. This study focuses specifically on limiting rates of diffusioncontrolled reactions taking place at the surface of porous silica gel. We attempt to asses the role of surface geometry in transport near the interface and detect any changes in viscosity solvent within a porous matrix. Porous silica is used extensively in chemical separations as a substrate for chromatography and as an adsorbing medium, and as a support for immobilized reagents and catalysts. In addition, researchers have used porous silica as a model substrate to investigate chemistry in confined geometries. Since porous silica is optically transparent, generally pure, and inert to a number of different molecules,it is ideally suited to these studies. There are a broad range of porous silica materials which possess 0022-365419 1/2095-4489%02.50/0
different pore sizes and surface areas.' The average pore diameters can be as small as 20 A or as large as 2000 A. The specific surface areas also vary over three decades ranging from 1 to lo00 m2/g.Z There are two issues involved in the question of molecular transport in a porous matrix: the geometry of the substrate and the hydrodynamic properties of the interfacial solvent. The first issue, geometry of the solid, has generated controversy over what approach should be used to describe the surface structure: classical geometry (e.g., pore size, particle diameter) versus fractal geometry (e.& spectral and fractal d i m e n ~ i o n ) . ~The - ~ classical approach begins with a primary picture of a local pore (a cylinder or sphere, for example) and develops a spatial distribution of those primary pore element^.^ The fractal approach, however, assumes the same structural features exist for a given surface on all length A purely fractal surface is one where the morphology of surface irregularities is independent of the magnification at which the surface is observed. Some of the controversy over this issue may (1) Iler, R. K. The Chemistry of Silica: Solubility, Polymerization, Colloid and Surface Properties, and Biochemistry; Wiley: New York, 1979. (2) Gregg, S. J.; Sing, K. S . W. Adsorption, Surface Area and Porosity; Academic Press: New York, 1982. (3) Avnir, D., Ed. The Fractal Approach to Heterogeneous Chemistry; Wiley: Chischester, U.K., 1988. (4) Klafter, J., Drake, J. M., Eds. Molecular Dynamics in Restricted Geometries; Wiley: New York, 1989. (5) Avnir, D.; Farin, D.; Pfeifer, P. Nature 1984, 308, 261.
0 1991 American Chemical Society
