Formation of colloidal silica particles from alkoxides

Formation of colloidal silica particles from alkoxides

Colloids and Surfaces, 63 (1992) 151-161 Elsevier Science Publishers B.V., Amsterdam 151 Formation of colloidal silica particles from alkoxides J.K...

3MB Sizes 2 Downloads 58 Views

Recommend Documents

No documents
Colloids and Surfaces, 63 (1992) 151-161 Elsevier Science Publishers B.V., Amsterdam


Formation of colloidal silica particles from alkoxides J.K. B a i l e y a n d M . L . M e c a r t n e y

Department of Chemical Engineering and Materials Science, University of Minnesota, Minneapolis, Minnesota 55455, USA (Received 19 September 1990; accepted 22 October 1990) Abstract The growth of solution-derived monodisperse silica particles was studied by cryogenic transmission electron microscopy (cryo-TEM) in order to provide direct observation of the growth mechanism in colloidal particle formation from alkoxides. The silica particles were grown from tetraethoxysilane solutions using n-propanol and ethanol as solvents with an ammonia catalyst. A range of water concentrations was used. The reacting solutions were cryogenically fastfrozen at various times in the reaction and the samples were directly observed in the TEM. The results indicate that low density particles are seen initially, and these undergo a collapse to form the colloidally stable seed for further growth. Nuclear magnetic resonance was also used to follow the particle formation in order to test this growth model. Continued growth is postulated to occur by addition of small, low density particles to the growing seed, surmised from a lack of dense small particles and no evidence for aggregation of larger particles in the cryo-TEM results.

Keywords: Colloidal silica; particle growth; transmission electron microscopy.



High quality advanced ceramic materials can be manufactured by controlled sintering of compacts of unagglomerated, monodisperse powders. Thus, there is considerable interest in understanding the growth mechanisms which lead to formation of monodisperse powders. Chemical synthesis of colloidal powders is an attractive production method since it offers the potential advantages of high purity and control over particle size and distribution. St6ber et al. [1] first showed that monodisperse silica particles can be derived from silicon alkoxides, and Bogush et al. [2] demonstrated that control over particle size and mass fraction can be obtained. In this process (the SFB process) a solution of silicon alkoxide monomers undergoes

- S i - O R + H 2 0 ~ - S i - O H + ROH

Correspondence to: J.K. Bailey, Sandia National Laboratories, Inorganic Materials Chemistry, Division 1846, P.O. Box 5800, Albuquerque, NM 87185, USA. 0166-6622/92/$05.00


and condensation - S i - O H + - S i - O H ~ - S i - O - S i - - + H 20


to form porous particles of amorphous silica [3]. To date, only empirical relations are known for controlling the size and mass fraction of the colloids produced [2]. To model and control more accurately the process conditions which influence growth, an accurate understanding of the growth mechanism of the colloids is needed. In addition, to apply the knowledge learned from understanding the silica system to other technologically interesting systems, it is vital that the growth mechanism be understood well enough to verify if the same mechanism operates in other systems. To understand the growth mechanism, Bogush and Zukoski [4,5] performed kinetic and structural investigations on particles formed by the SFB

© 1992 - - Elsevier Science Publishers B.V. All rights reserved.


process. They inferred that nucleation of particles was occurring continuously throughout the reaction, and discredited the idea of a traditional nucleation and growth mechanism (LaMer mechanism [6]) where relief of supersaturation creates a fixed number of particles which grow uniformly. They proposed a nucleation and aggregation mechanism whereby small particles nucleate and then aggregate until reaching a colloidally stable size. These larger, colloidally stable particles would then diffuse through the solution aggregating vaith small, colloidally unstable particles which are continuing to nucleate. Hence, only those particles which become colloidally stable early in the reaction should be able to grow. Later in the reaction, small particles, which are constantly forming, must aggregate onto the growing particles before they can reach a colloidally stable size. Harris and co-workers [7,8] have also investigated the growth of particles produced by the SFB process. They assumed that the first hydrolysis of an alkoxide group is the rate limiting step in the formation of nuclei. They claim that once the first alkoxide group is hydrolyzed, the remaining groups hydrolyze rapidly, and small nuclei are created from fully hydrolyzed species. In the early stages of the reaction, these small nuclei aggregate, forming seed particles which are colloidally stable, similar to the mechanism of Bogush and Zukoski [4,5]. However, in contrast to these authors, they propose that in the later stages of the reaction, particle growth occurs mainly by addition of monomer to the surface, not by aggregation of the small nuclei particles. They suggest that the monomer growth mechanism provides the explanation for the smooth particle surface. The differing views, presented above, on the mechanism of the particle formation need to be reconciled. A technique which can differentiate between the proposed growth mechanisms is cryogenic transmission electron microscopy (cryoTEM). This technique allows one to observe directly structures in the liquid state and thus to observe directly the changes in structure between different stages of the reaction. For the nucleation

J.K. Bailey, M.L. Mecartney/Colloids Surfaces 63 (1992) 151-161

and aggregation mechanism of Bogush and Zukoski [4,5], one would expect to see small nuclei forming throughout the reaction. For the mechanism of Harris and co-workers [7,8], one would expect to see small nuclei form and aggregate at the beginning of the reaction. As the reaction proceeded, one would expect the number of nuclei to decrease as monomer addition to the particle surface became the dominant growth mechanism. For the LaMer mechanism [6], or Iler's model [9], which is based on a similar nucleation and growth mechanism but includes Ostwald ripening, one would expect to see small particles only at the start of the reaction which would grow throughout the reaction. The cryo-TEM technique has previously been used to follow the gelation of acidand base-catalyzed TEOS [10]. For base-catalyzed gelation of TEOS, the cryo-TEM work has shown the formation of 4 nm silica particles which subsequently aggregate into a gel network. For the formation of colloidal particles from alkoxides, this technique will allow observation of particles larger than 2-3 nm in diameter and therefore it should provide unambiguous information for determining the actual growth mechanism.

Experimental Materials

Silica colloids were prepared from a solution of 0.17 M tetraethoxysilane (TEOS), 1.0-1.3 M NH3, and 1.0-3.8 M H 2 0 in n-propanol solvent [1,2]. The TEOS (Fischer) was distilled before use. For a 1.0 M water concentration system, ammonia was added by bubbling anhydrous ammonia through the alcohol solvent for several hours and titration was used to find the concentration immediately before use. For the higher water concentrations, concentrated ammonium hydroxide was used. Normal propanol was used in the TEM studies since frozen n-propanol films are more resistant to electron beam damage in the TEM than are ethanol films. Ethanol was used in the N M R studies to reduce the number of species present in order

J.K. Bailey, M.L. Mecartney/Colloids Surfaces 63 (1992) 151-161

to increase resolution. There is no difference in the morphology of the particles produced by using the different alcohols, and it is presumed that the growth mechanism is also similar.


standard silicon NMR terminology, where Qi represents a silicon nucleus surrounded by i siloxane bonds, so that Qo represents uncondensed monomer and Q4 represents a fully condensed silicon nucleus.

Techniques Results Cryo-TEM samples were prepared at intervals throughout the reaction. At a given time in the reaction, a drop of solution was taken from the reaction vessel and was placed on a perforated carbon grid suspended in a controlled environmental chamber. The grid was blotted to produce thin liquid films in the grid holes. The grid was then rapidly plunged into liquid ethane which vitrified the solvent. The frozen specimens were transferred to the TEM via a liquid nitrogen cold stage and transfer module. The structures present in the fast-frozen liquid films were then directly imaged. The electron microscopes and cold stages used for this work were a JEOL 100CX electron microscope with a JEOL cold stage and a Philips CM-30 electron microscope with a Gatan cold stage. All microscopy was done at 100 kV accelerating voltage and low dose imaging with Kodak SO-163 film was used to reduce electron beam damage to the sample. This technique is described more fully elsewhere [11,12]. Dried TEM samples were prepared by taking a drop of the reacting sol at a given time in the reaction and placing the drop onto a perforated carbon grid resting on blotter paper. These samples were then air dried. The NMR results were obtained with a 500 MHZ General Electric spectrometer. A 10 mm sample tube was used and the spectra were referenced to a TMS standard. The solution was prepared in bulk and an aliquot transferred to the NMR tube which was rapidly inserted into the spectrometer. Chromium acetylacetonate (l%wt.) was used as a relaxation agent and 500 scans were taken per spectrum. When one set of scans was completed a second set was started immediately. Peaks were identified according to Kelts and Armstrong [13]. Throughout the text, we use

The growth of silica colloids as followed using the cryo-TEM technique for a 1.0 M water concentration solution is shown in Figs l a - l d . Figures 2a-2c show corresponding images from dried TEM samples for comparison. This solution took 36 min to first show turbidity. Figure l a shows the cryofixed sol at 6 min reaction time. This solution contains no visible particles and is featureless. In comparison, a dried sample taken at the same reaction time is shown in Fig. 2a: the dried sample shows particles of average size 14 nm. Figure lb shows a cryo-fixed solution at 16 min of reaction time, and one can see low contrast particles visible in the solution with an average size of 26 nm. The low contrast is due to the low density of the particles. At 24 min of reaction, Fig. lc, the cryoTEM sample shows high density particles in the micrograph. The high density particles have a rough texture and an average size of 20 nm. In this micrograph, some low density particles are also present. If a solution is dried at this time, as shown in Fig. 2b, there are high density particles which have an average size of 30 nm, and many smaller particles, also of high density are seen; some appear attached to the larger particles in contrast to the cryo-fixed sample (Fig. lc) which shows no agglomeration. Figure l d shows particles at 66 min into the reaction which were cryo-fixed. There are only high density particles at this time. Figure 2c shows the particles close to their final size, 24 h after mixing. The particles have an average size of 200 nm and the size distribution is narrower. Figure 3 shows a cryo-TEM sample of a silica solution with a 2.0 M water concentration at 75 s into the reaction. This solution first showed turbidity at 4 min. Low density particles are seen, similar to Fig. lb. This solution also contained


J.K. Bailey, M.L. Mecartney/Colloids Surfaces 63 (1992) 151-161

Fig. 1. Growth sequence observed by cryo-TEM for particles prepared from a solution of 0.17 M TEOS, 1.0 M H 2 0 , and 1.0 M NH 3 in n-propanol. (a) Sample frozen 6 min into reaction, no particles visible; (b) sample frozen 16 min into reaction, arrow indicates low density particle; average particle size, 26 nm; (c) sample frozen 24 min into reaction, arrow indicates high density particle; average particle size, 20 nm; (d) sample frozen 66 min into the reaction; average particle size, 48 nm.

J.K. Bailey, M.L. Mecartney/Colloids Surfaces 63 (1992) 151-161


Fig. 2. Growth sequence observed using dried samples prepared from the same solution as Fig. I. (a) Sample dried 6 min into the reaction; average particle size, 14 nm; (b) sample dried after 25 min of reaction; average large particle size, 30 nm; (c) sample dried after 24 h, near end of reaction; average particle size, 200 nm.


J.K. Bailey, M.L.

Mecartney/ColloidsSurfaces 63 (1992) 151-161 TEOS




21 rain










Fig. 4. 298i NMR spectra from solution containing 0.17M TEOS, 1.0M H 2 0 and 1.0M NH3 in ethanol. Spectra are shown in perspective and are from 21, 33, and 92 min into the reaction.

Fig. 3. Cryo-TEM micrograph of particle growing in solution containing 2.0 M water concentration. Sample frozen at 75 s into reaction which took 240 s to show turbidity, arrow indicates low density particle.

tubes were used, the Q4 peak, which is off the left side of the spectrum in Fig. 4, is obscured and cannot be resolved. The peak intensities for successive times in the reaction are shown in Table 1. Discussion

higher density particles later in the reaction; however, the high density colloids had a smoother texture than the 1.0 M water solutions. Cryo-TEM revealed that samples with water concentrations in the range 1.0-3.8 M, made at less than twothirds of the time to turbidity, showed no distinct colloids. Furthermore, the smallest high density colloid observed, later than two-thirds of the time to turbidity, had a diameter of 12 nm. N M R results for the solutions are shown in Fig. 4, where spectra from three times in the reaction, (21, 33, and 92 min) are compared. The reaction time indicated is the time at the midpoint of data acquisition, which took approximately 10 min per spectra. Two peaks can be identified in the Qo region of the spectra: the TEOS peak, which remains large throughout the reaction, at - 8 2 p p m , and the first hydrolysis product, Si(OCH2CH3)aOH, at - 7 8 ppm. It can be seen that aside from these peaks, there are no other major peaks in the spectra. Since glass sample

The most striking feature in the cryo-TEM micrographs is that early in the reaction, there are no dense colloidal particles in solution. In the liquid phase, one sees that there are low density, i.e. ramified, particles present. Previous work using the cryo-TEM technique has been able to resolve TABLE 1 TEOS and H1 concentrations from NMR spectra Time (rain)

[TEOS] (M)

[H1] (M)

0 9 21 33 44 56 68 79 92

0.170 0.161 0.156 0.145 0.136 0.128 0.122 0.111 0.112

0 0.0086 0.0083 0.0149 0.0155 0.0185 0.0188 0.0158 0.0197

J . K . Bailey, M . L . Mecartney/Colloids Surfaces 63 (1992) 151-161

silica particles of 4 nm diameter in solution in a base-catalyzed TEOS gel [10] so if there were small dense particles of this size or larger they would have been visible. The cryo-TEM pictures indicate that the dense particles seen upon drying are artifacts of the dried specimen preparation procedure, and are not present in the liquid state. There are many processes which can take place upon drying as has been described by Brinker et al. [14] including a rise in reaction rate which could lead to particles forming during the drying process or to low density particles collapsing upon drying. The earliest time that dense particles are present in the micrographs is 24 min, and prior to this time, no dense silica particles were seen. It is also evident from Fig. lc that solutions which have formed dense particles in the later stages may also contain low density particles, indicating that not all particles are at the same stage in their growth. From the cryo-TEM work, no dense particles in solution between the sizes of 2-12 nm were seen at any significant concentration. Therefore, the mechanism for formation of stable particles cannot involve aggregation of dense particles in this size range. The N M R data presented in Table 1 allow the calculation of rate constants for the first and second hydrolysis reactions of TEOS under these reaction conditions. The disappearance of TEOS monomer from solution by hydrolysis (Eqn(1)) can be described by the rate equation d[TEOS]/dt = - khl [ H 2 0 ] [TEOS]


where kh~ is the rate constant for the first hydrolysis. Since the water concentration is six times greater than the TEOS concentration and since less than one half of the TEOS is reacting during the N M R experiment, we shall assume that the water concentration is constant throughout the reaction. This assumption allows us to integrate the rate equation directly. Solving the integrated rate equation gives: In [TEOS] = In [TEOS]o - k~x t


where k~l is the product of kh~ and the water


concentration and [TEOS]o is the initial TEOS concentration. Plotting In ITEOS] vs. time (Fig. 5) gives a straight line of slope k~l = 4 . 9 . 1 0 -3. We can then use this value for k~l to solve for the rate of the second hydrolysis from the rate equation for the first hydrolysis product, Si(OCH2CH3)3OH, abbreviated H1. d [H 1]/dt = kh ~[H2 O] [TEOS] - kh2 [H 2 O] [H ~]

k¢1[Hi] [sion]



where kh2 is the rate constant for the second hydrolysis, and kc~ is an averaged rate constant for condensation with all other SiOH groups in the system. If we use the assumption of Harris et al. [7] that condensation is negligible for species which are not fully hydrolyzed, and again assume relatively constant water concentration, we can solve for Hi analytically as: [H~ ] = {[TEOS]o k'~/(k'h2


k'ht )}

x {exp (-- k~l t) - exp ( - k~,2t)}


where k~2 is analogous to k~,l. Performing a nonlinear regression using this equation gives a value of k~2 = 7.32.10 -3, i.e. a value of kh2 which is 1.5 times greater than kh~. Figure 6 shows the fit of [H~] obtained using Eqn(6). Although there is scatter in the experimental measurements, Fig. 6 shows that there is an underprediction at the start of the reaction and an overprediction later in the

















-1.8 -1.9










• 60



Twae (m)

Fig. 5. TEOS monomer concentration as a function of time for solution described in Fig. 4. Slope of the straight line is the rate constant for the first hydrolysis constant.

J.K. Bailey, M.L. Mecartney/Colloids Surfaces 63 (1992) 151-161

158 0.025


I -7,





0.02 ~ 0.015 0.01 0.005 0


I .





Fig. 6. Concentration




40 60 Tun= (m) first







Si(OCH2CH3)a(OH)] as a function of time for the same solution as Fig.4. The line is a best fit using values of rate constants given in text.

reaction. One source of deviation is error in the assumption that the water concentration was constant; however, the good fit of the [TEOS] vs. time plot (Fig. 5) argues against this explanation. The difference between experiment and prediction is probably due to the assumption that condensation is negligible, since condensation would become increasingly important as the reaction proceeds. Full treatment of the kinetic rate equations requires numerical integration and fitting of a large number of rate constants and is beyond the scope of this paper [15-17]. This first approximation suggests two findings: the second hydrolysis reaction is approximately 50% faster than the first, so while the first hydrolysis is slower, it is not sufficiently slower to be considered rate limiting; and condensation of only partially hydrolyzed species cannot be neglected. The cryo-TEM work shows that there are low density particles in solution which are presumed to be the precursors to the colloidally stable seed particles. Our proposed growth mechanism is that the alkoxide monomers undergo polymerization to form expanded polymeric species in solution (microgel particles or clusters). These microgel molecules continue to grow by addition of monomer, meanwhile they are also cross-linking internally by intramolecular reactions leading to their densification. They continue growing until

they are no longer soluble, at which point they collapse. It is known from polymer science that the solubility of polymers decreases with increases in size and degree of cross-linking [18]. At a given size, the intramolecular enthalpic attraction overcomes the entropic solvation forces, and the molecule collapses. The process of microgel particle collapse is driven by the same forces that drive syneresis of gels [3]. The solubility may also be changing throughout the reaction since the solvent is changing as water is consumed and alcohol is produced. Once collapsed, hydrolysis and condensation continue, bonding the particle together. If there are expanded polymers in solution, there must be Q2 species in solution which are not being detected in the NMR spectra (Fig. 4). As calculated below, one can have large visible particles in solution while maintaining a low concentration of Q2 species. Assuming that a random coil polymer is formed, the size of a coil is given as rg = nl 2 where n is the number of links, l is the link length and r~ is the radius of gyration [18]. Assuming the link length is about twice the Si-O bond distance (0.18 nm), for an rg of 13 nm there are approximately 5000 links in this size chain. There may be twice as many silicon nuclei as links and so a conservative estimate is that there are approximately 104 silicon nuclei incorporated in a low density particle this size. Although this estimate fails to include effects for non-random coils and cross-linking, it has overestimated the relationship of links to nuclei to compensate and should produce an order of magnitude estimate. To calculate the number of particles per area observed in cryo-TEM, one 26 nm particle in an area of 1 lxm square is assumed, the highest particle density seen in the cryo-TEM. The TEM films are approximately 0.1 I.tmthick and, therefore, the total volume per particle is 10-19m 3. For a 0.17M concentration, this volume contains approximately l07 silicon nuclei. If 10* of these nuclei are Q2 silicon atoms involved in a polymer, the ratio of peak intensities of Qo to Q2 on the NMR spectra would be 999 to 1, i.e. the Q2 peak will be three orders of magnitude smaller than the monomer

J.K. Bailey, M.L. Mecartney/Colloids Surfaces 63 (1992) 151-161

peak and, therefore, it is not surprising that it is not observed in Fig. 4. There are also at least three different Q2 peaks depending on the terminal groups and since some nuclei will also be Q3 silicon nuclei as well, the ratio of peak areas will be even smaller than estimated. Therefore the Q2 and Q3 peaks will not be seen in the NMR spectra if the concentration is this low. It is thus possible that there are polymers in solution which can be seen by cryo-TEM but which contain too few nuclei, compared to the number of monomers, to be detected by NMR. One stipulation in the above calculations is that the distribution of polymer size is bimodal. We have assumed that there are monomers and high molecular weight species and only few intermediate mass species, precisely the type of distribution which has been observed by Klemperer et al. in a mass spectroscopy study of base-catalyzed TEOS reactions 1-19]. They found a distribution in which there are low molecular weight species and high molecular weight species with essentially no intermediate weight species. This type of distribution can be obtained in cases where the condensation is much faster than hydrolysis. Once formed, the dense, colloidally stable particles grow by the addition of hydrolyzed monomer and polymeric species. The rate of mass addition to the surface of the particles depends on the probability of contacting a particle. While smaller particles diffuse faster than larger particles, they have a smaller surface area. Since both of these factors depend on the square of the particle radius, it has been demonstrated that the probability of random collision with a particle does not depend on the particle radius l-4]. As described by Bogush and Zukoski, as the particles grow, the size distribution narrows because the radius of spheres increases at a decreasing rate with constant mass addition to the surface, i.e. it takes less mass for a small radius particle to increase its radius by an incremental amount than for a large particle I-4]. Therefore, a requirement for growing monodisperse particles is a fixed window of stable particle formation. No new particles can be formed


after a certain time in the reaction because the growing polymers are swept up by the dense seed particles before they can grow to a large enough size to collapse and form a stable particle by themselves. The distinguishing difference between our mechanism and that of Bogush and Zukoski is our proposal of the mechanism of colloidally stable (seed) particle formation and the nature of the species attaching to the surface of the particles. In summary, hydrolyzed TEOS monomer polymerizes to form large microgel clusters which reach a size and degree of cross-linking where they become insoluble and collapse. After, or during, collapse, the microgel undergoes additional condensation to form high density, colloidally stable particles (seed particles). Once the particles are formed, they are prevented from aggregating with other seed particles due to double layer effects as described by Bogush and Zukoski [4,5]. However, monomers are still being hydrolyzed in solution and these react until they either form polymers which reach a large enough size to collapse or until they encounter a particle and attach to its surface. When polymers react with a surface, they can collapse onto the surface, enabling particles to maintain a spherical shape. This postulated growth mechanism is schematically illustrated in Fig. 7. From this work it appears that in 1.0 M solutions, when the polymers reach greater than 26 nm in size, they are no longer soluble. In solutions with different TEOS, ammonia, and water concentrations, the distribution of hydrolyzed monomer will change, and, therefore, it would be expected that the particles will have different solubilities. If the concentrations are changed such that the rate of hydrolysis increases relative to the rate of condensation, as may be expected for a higher water and ammonia concentration, the distribution of alkoxide monomers would include more hydrolyzed species and thus the degree of cross-linking in a microgel particle would be higher and these particles would not be as expanded in solution. The limit of this behavior is that hydrolysis goes to completion before particle growth begins, which may lead to the formation of dense particles (fully

J.K. Bailey, M.L. Mecartney/Colloids Surfaces 63 (1992) 151-161




Collapse of Clusters to Form Particles

Additional Clusters Coat Particles

Stable, Growing Colloids

Fig. 7. Growth mechanism for the formation of monodisperse colloidal silica particles from alkoxides (see text for details).

cross-linked) from the beginning of the reaction. In non-limiting cases, competition between hydrolysis and condensation leads to particles which incorporate partially condensed species and can be soluble. These continue to grow and react until they are no longer soluble and collapse, forming dense particles. For the solutions studied in this work, the latter model is appropriate. Further study to determine reaction rates and solubilities as functions of reactant and catalyst concentration would be required to map out the entire range of behavior for the systems. If the reaction rates, mechanisms and solubility behavior were better known, a model for predicting the degree of hydrolysis and condensation, the expected size of microgel particles before collapse, and the particle size distribution could be developed.

Conclusions The growth process for the formation of colloidal silica particles from alkoxides has been directly observed using cryo-TEM. The cryo-TEM results indicate that small, low density particles are forming early in the reaction, and these collapse upon reaching a certain size. The collapsed, denser, particles are colloidally stable towards aggregation with each other. Comparison of the dried TEM samples to the cryo-TEM results indicates that cryo-TEM is a useful tool for eliminating artifacts that occur during the drying of samples and gives

additional information about the growth mechanism. The dried samples show small particles present throughout the reaction but cannot indicate that they are actually solvated in solution. Additionally, the size of dried particles is larger than that of particles in cryo-TEM samples at the same time in the reaction, which indicates that growth is occurring during the drying process. Using the results from the cryo-TEM experiments, a growth mechanism has been postulated for the formation of colloidal particles by the SFB process. Hydrolyzed monomer reacts to form microgel polymers which collapse upon reaching a certain size and cross-link density which makes them insoluble. Collapsed particles densify by condensation and are colloidally stable with respect to each other. The denser seed particles grow by addition of hydrolyzed monomer or polymer addition to their surface. The rate of growth of the polymers must be slow enough that after a sufficient number of seeds has formed, the polymers attach to a particle surface before they grow to a large enough size to collapse and form a seed particle themselves. N M R experiments were also conducted to test the growth mechanism. The data are consistent with this growth mechanism.

Acknowledgments Support for this work was provided by Sandia National Laboratories, contract number 05-3974.

J.K. Bailey, M.L. Mecartney/Colloids Surfaces 63 (1992) 151-161 Professors







t h a n k e d for the use o f t h e c o n t r o l l e d e n v i r o n m e n t v i t r i f i c a t i o n system.


A.V. M c C o r m i c k

a n d J. S a n c h e z a r e t h a n k e d for a s s i s t a n c e w i t h t h e NMR Brinker

studies. of










References 1 W. Stfber, A. Fink and E. Bohn, J. Colloid Interface Sci., 26 (1968) 62. 2 G.H. Bogush, M.A. Tracy and C.F. Zukoski, J. Non-Cryst. Solids, 104 (1988) 95. 3 C.J. Brinker and G.W. Scherer, Sol-Gel Science, Academic Press, Boston, 1990. 4 G.H. Bogush and C.F. Zukoski, in J.D. Mackenzie and D.R. Ulrich (Eds), Ultrastructure Processing of Advanced Ceramics, Wiley, New York, 1988, p. 477. 5 G.H. Bogush and C.F. Zukoski, in J.A. Pask and A.G. Evans (Eds), Ceramic Microstructures '86, Plenum, New York, 1987, p. 475. 6 V.K. LaMer and R.H. Dinegar, J. Am. Chem. Soc., 72 (1950) 4847. 7 M.T. Harris, O.A. Basaran and C.H. Byers, in J.D. Mackenzie and D.R. Ulrich (Eds), Ultrastructure Processing of Advanced Ceramics, Wiley, New York, 1988, p. 843.

161 8 M.T. Harris, R.R. Brunson and C.H. Byers, J. Non-Cryst. Solids, 121 (1990) 397. 9 R.K. Iler, The Chemistry of Silica, Wiley, New York, 1979. 10 J.K. Bailey, T. Nagase, S.M. Broberg and M.L. Mecartney, J. Non-Cryst. Solids, 109 (1989) 198. 11 J.K. Bailey, J.R. Bellare and M.L. Mecartney, in J.C. Bravman, R.M. Anderson and M.L. McDonald (Eds), Specimen Preparation for Transmission Electron Microscopy of Materials, Mat. Res. Soc. Symp. Proc., Vol. 115, Material Research Society, Pittsburgh, 1988, p. 69. 12 J.R. Bellare, H.T. Davis, L.E. Scriven and Y. Talmon, J. Electron Microscopy Technique, 10 (1988) 87. 13 L.W. Kelts and N.J. Armstrong, J. Mater. Res., 4 (1989) 423. 14 C.J. Brinker, A.J. Hurd and K.J. Ward, in J.D. Mackenzie and D.R. Ulrich (Eds), Ultrastructure Processing of Advanced Ceramics, Wiley, New York, 1988, p. 223. 15 R.A. Assink and B.D. Kay, J. Non-Cryst. Solids, 104 (1988) 112. 16 J.C. Pouxveil and J.P. Boilot, J. Non-Cryst. Solids, 94 (1987) 374. 17 J.K. Bailey, C.W. Macosko and M.L. Mecartney, J. NonCryst. Solids, 125 (1990) 208. 18 P.C. Heimenz, Polymer Chemistry, Marcel Dekker, New York, 1984. 19 W. Klemperer, V.V. Mainz, S.D. Ramamurthi and F.S. Rosenberg, in C.J. Brinker, D.E. Clark and D.R. Ulrich (Eds), Better Ceramics Through Chemistry III, Mat. Res. Soc. Symp. Proc., Vol. 121, Material Research Society, Pittsburgh, 1988, p. 15.