Journal of Non-Crystalline Solids 104 (1988) 95-106 95

North-Holland, A m s t e r d a m

PREPARATION OF MONODISPERSE SILICA PARTICLES: C O N T R O L O F S I Z E AND M A S S

FRACTION

G.H. B O G U S H , M.A. T R A C Y and C.F. Z U K O S K I IV

Department of Chemical Engineering, University of lllinois, Urbana, Illinois 61801, USA

Received 13 July 1987

Revised manuscript received 26 February 1988

We extend the work of Stober, Fink and Bohn (J. Colloid lnterfac Sci. 26 (1968) 62) to establish the ranges of reagent
concentrations which result in the precipitation of monodisperse silica particles from ethanol solutions containing ammonia,
water, and tetraethyl orthosilica. A correlation is presented which can be used to predict final particle sizes over concentra-
tions of 0.1-0.5 M TEOS, 0.5-17.0 M H 2 0 and 0.5-3 M N H 3. The m a x i m u m particle size achievable for these conditions at
25 o C is = 800 n m and near the maximum, monodispersity is difficult to maintain. The effects of reaction temperature are
studied and particle size distributions are analyzed. A seeded growth technique for preparing larger particles a n d / o r
increasing solids mass fraction up to a theoretical limit of = 24 wt% of silica is described.

I. Introduction detail the set of reagent concentrations and reac-

tion temperatures which result in the precipitation
The work of Stober et al. [1] showing that the of particles with narrow size distributions. Our
hydrolysis of silicon alkoxides leads to the precipi- results are presented as a correlation to aid others
tation of essentially monodisperse SiO 2 particles wishing to tailor particle size for particular needs.
has led to use the alkoxides of m a n y metals in the In addition, we describe a seeded growth technique
preparation of a variety of ceramic precursor which can be used to prepare suspensions of par-
powders [2,3,7]. These studies have shown that ticles with narrow size distributions at mass frac-
while solids are easily precipitated from a wide tions of 17%.
range of starting alkoxides, the narrow size distri-
butions achieved with silica are difficult to repro-
duce. Often when conditions are found which 2. Experimental
result in mondispersity the suspension mass frac-
tions are so low as to preclude their use in the TEOS (Fisher Scientific Co., Reagent Grade)
fabrication of ceramic bodies large enough for was vacuum distilled before use for particle
even model studies. As part of an effort to over- synthesis. Dehydrated ethyl alcohol (US Industrial
come these difficulties, we present here an exten- Chemicals Co.) and a m m o n i u m hydroxide (J.T.
sive study on the preparation of powders by the Baker Chemical Co., Reagent Grade) were used as
hydrolyses of tetraethyl orthosilicate (TEOS) in received. Titrations indicated an a m m o n i a con-
aqueous ethanol solutions containing ammonia. centration of 14 M. A m m o n i a saturated ethanol
In previous work on this system by Stober et al. solutions were prepared by passing the a m m o n i a
[1] and Van Helden et al. [4], final particle sizes through a drying column and then bubbling it
are presented in a graphical form which are dif- through ethanol. The resulting a m m o n i a con-
ficult to read and cover limited concentration centration, determined just prior to use, was about
ranges. We have expanded on these studies to 3 M at room temperature.
0022-3093/88/$03.50 © Elsevier Science Publishers B.V.
(North-Holland Physics Publishing Division)
96 G.H. Bogush et al. / Preparation of monodisperse silica particles

For each experiment the necessary amounts of 3. Results and discussion

ethanol, ammonium hydroxide, ethanol saturated
with ammonia and deionized water (18 MI2 cm In their early work Stober et al. [1] and later
prepared with a Barnstead Nanopure Deioniza- Van Helden et al. [4] and Bridger et al. [6] pre-
tion Apparatus) were mixed in the reaction vessel sented results which established that the hydroly-
and placed in a constant temperture bath. After sis of TEOS in ethanol solutions containing water
the solution had come to temperature, the TEOS and ammonia can result in particles with narrow
was added quickly. However, extremely rapid ad- size distributions. This has stimulated a great deal
ditions, as required, for example, when using of work in using the resulting amorphous S i O 2
titanium alkoxides as a starting material [5], were particles as model colloids and in studies of
not needed. ceramic fabrication processes. In addition,
A series of experiments demonstrated that the titanium, zinc, aluminium, and zirconium al-
final particle size and size distribution were not koxides, to name a few, have been used in at-
functions of the degree of agitation during the tempts to prepare cermaic precursor powders of
reaction. In all subsequent experiments low to similar morphologies [2,7]. However, it is com-
moderate mixing speeds were used. monly reported that narrow size distributions are
Both round bottom flasks equipped with pad- difficult to achieve. Indeed, even in the silica
dle stirrers and polyethylene bottles mixed with system, which is by far the most extensively
magnetic stirbars were used as reactors with no studied, when attempting to prepare larger par-
alteration in final product. The majority of the ticles or suspensions with high mass fractions,
results reported here were carried out with reactor extrapolations of the results of Stober et al. and
volumes of 100-150 ml. However, vessels with Van Helden et al. can lead to heterodisperse sus-
volumes up to 4 1 have been used with no observa- pensions.
ble effects on final particle size or size distribu- In this paper we wish to establish in a more
tion. All vessels were washed with a 2% H F solu- quantitative fashion the concentration ranges over
tion then rinsed with deioized water before use. which monodispersity is achieved and develop a
For the growth experiments, seed suspensions correlation which can be used to predict final
were prepared as described above. After the reac- particle sizes for a given set of ammonia, water
tion had come to completion (3-8 h after the first and TEOS concentrations. To achieve this, a wide
addition of TEOS), TEOS and water were added range of initial reaction conditions were studied.
to the suspension in a 1 : 2 mole ratio. This mole Primarily, the effects of initial concentrations
ratio is required as the production of one mole of on final particle size were investigated at a reac-
SiO 2 from one mole of TEOS consumes two moles tion temperature of 25 ° C. Our approach was to
of water and produces four moles of ethanol. hold the TEOS concentration fixed and then vary
Additions of up to twice the number of moles of ammonia and water concentrations in a systematic
TEOS in the original solution were repeated at 8 h manner. Typical results of particle morphology
intervals for up to 10 additions. are shown in fig. 1. The dependence of final
Particle sizes were determined from electron particle size on ammonia and water concentra-
micrographs taken on a Phillips 400 transmission tions are displayed in figs. 2 and 3, for fixed
electron microscope. The diameters of over a TEOS concentrations of 0.17 M and 0.3 M, re-
hundred particles were used in calculations of the spectively. It will be noted that the size goes
average size and standard deviation of each sam- through a maximum as both ammonia and water
ple. A carbon replica of a diffraction grating (SPI concentrations are raised. Similar behavior was
Supplies) was used as a magnification standard. observed for all TEOS concentrations studied
The solids density was detrmined on samples (0.1-0.5 M) where a maximum particle size was
dried at 1 0 0 ° C for 24 h with a Quantachrome achieved at approximately 7 M water and 2 M
Micropycnometer, in which helium was used to ammonia. This maximum moves to higher am-
measure the volume displaced by the particles. monia and water concentrations as TEOS con-
 G.H. Bogush et aL / Preparation of monodisperse silica particles 97

Fig. 1. (a) SiO 2 particles of average size 79.9 nm+7.3% from a solution containing 0.2 M TEOS, 1.0 M NH 3 and 2.0 M H 2 0 in
ethanol and (b) SiO 2 particles of average size 392 nm_+ 2.6% synthesized at 25 ° C from a solution containing 0.25 M TEOS, 0.5 M
NH3, and 7.5 M H 2 0 in ethanol.
 98 G.H. Bogush et aL / Preparation of monodisperse silica particles

1000
i
i o 0.5M NH 3
o 1.0M NH
× 2.0M NH ~,
<, ,3.0M NH
800

E × --~ × E
c
/ x ¢,
600

0 / ," t2._ ~ c¢,, - -

5 / ' u~ / ~ ~ e- ~ ~,:... C3
/
~,)-
o., 400 / / ~ .o × ~ 4OO
0

'~"

0
t,,Jo
-J

O0
I "

,
/o

I
0

O0 5.0 10.0 15.0 20.0 0.0 5,0 I0,0 15,0 20,0 25.0

Water Concentration (M) Water Concentration (M)

Fig. 2. Average particle size as function of a m m o n i a and water Fig. 4. Average size of particles precipitated at 25 ° C from
concentrations at a constant T E a S concentration of 0.17 M at solutions containing 0.17 M TEAS, 1.0 M N H 3 over an ex-
25 o C. See text for synthetic details. Curves calculated from eq. tended range of H 2 0 concentration. Curve is calculated from
(1) for constant T E a S and N H 3 concentrations as indicated. eq. (1).

centration is increased. However, the maximum

achievable size remained at approximately 800 nm Limited data was obtained for water concentra-
for all T E a S concentrations studied. tions above 17 M. In fig. 4, final particle sizes are
presented for 0.17 M T E a S and 1 M ammonia
1000
and H 2 0 concentrations up to 25 M. N o phase
_ o 0.5M NH
separation was observed at these water concentra-
/" ~ [] 1.0M NH tions and the particle size appears to approach a
/ × ",, × 2.0M NH
/
\ o 3.0M NH plateau size at high water concentration.
800 -

x -.\
x .x.
This result is in contrast to the work of Bridger
\. et al. [6] who observed a phase separation at water
C
¢, .,'. concentrations greater than 15 M at a T E a S con-
600 " / O\".
/ centration of 0.28 M. Phase separation is expected
× \ ".
E / to be a complex function of TEAS, N H 3 and H 2 0
o
i5 0
a
"" concentration as well as temperature. Conse-
~ 400-
0 t ,' quently, as Bridger et al. carried out their reaction
, J'
at water concentrations well above those typically

200-
i
'~ /

/o
O
studied in our work and at 30 ° C, this result is not
surprising. However, in agreement with the trends
reported in fig. 4, they report that at a N H 3
concentration near 1.7 M, the final particle diame-
0 i i i i / i i i t l r i i i I i i I i
0.0 5. 10,0 15.0 20,0 ter decreased from 290 to 67 nm as water con-
Water Concentration (M) centration increased from 15 to 26 M.
Fig. 3. Average particle size as functions of a m m o n i a and
Within minutes of the time the reaction had
water concentrations at a constant T E a S concentration of 0.3
M at 25 o C. See text for synthetic details. Curves calculated
been initiated, we observed that the reacting solu-
from eq. (1) for constant T E a S and N H 3 concentrations in tion became a turbid white for water concentra-
indicated. tions greater than approximately 2 M. At very low
 G.H. Bogush et al. / Preparation of monodisperse silica particles 99

water concentrations, a few hours were required to

observe opacity in the reaction mixture. This in-
creased apparent induction time can be accounted 8OO (

for by the small final particle size where it was not

until late in the reaction that the particles achieved
a size capable of scattering visible light. The over-
2 ©
all reaction time determined by electron micro-
graphs of growing particles was on the order of 3
h for intermediate reagent concentrations (0.17 M
TEOS, 1.3 M NH3, 3.3 M H 2 0 ) and decreased as
the water concentration was raised reaching a
600

400
o//S
reaction time of approximately 40 rain at 24 M
H 2 0 [81. 200 0 0 i
Results gathered on over 100 sample prepara- l

tions synthesized at 2 5 ° C and concentration

ranges of 0.1-0.5 M TEOS, 0.5-17.0 M H 2 0 and o , , , , , , i , l ! , , : F !
0 200 400 600 800 1000
0.5-3.0 M N H 3 were used in the development of a Meosured Diemeter (qm)
correlation relating final particle size to intial
Fig. 5. Diameters calculated from eq. (l) as a function of
reagent concentrations. The resulting expression, experimentally determined diameters. A perfect correlation
which has been developed to fit the experimental would correspond to the center line. The two outer lines
observations and thus contains no mechanistic represent a + 20% error in the correlation.
information, is written,
in limited extrapolations outside the range of val-
d= A[H20] 2 exp(-B[HzO]'/=), (1) ues for which it was designed. It can be seen that
with the correlation deviates most strongly at small
particle sizes and at higher TEOS concentrations.
A = [ T E O S 1 ' / 2 ( 8 2 - 151[NH31 Fitting the data at the low water concentrations
where small particles are generated is difficult
+ 1200[NH3 ]2 _ 366[NH 3 ]3),
without the introduction of a more complex fitting
and function. At elevated TEOS concentrations, the
particle sizes were less reproducible and broader
B = 1.05 + 0.523[NH3] - 0.128[NH3] 2,
size distributions were observed. The correlation
where d is the average diameter in nanometers as presented in eq. (1) has been chosen to act as a
and the reagent concentrations are given in tool/1. practical guide and thus will be useful when desig-
Fig. 5 shows the predicted diameter from the ning reaction conditions required to achieve a
correlation plotted versus the experimentally desired size.
observed diameters. The majority of the data falls As a general rule, spherical particles with a
within the lines representing a deviation of 20%. narrow size distribution as shown in fig. 1 were
Due to the complexity of the functional depen- obtained over the range of concentrations used in
dence of particle size on reagent concentrations, a the development of eq. (1). The use of distilled
better correlation would be difficult to achieve TEOS was found to produce a more uniform
without using an even less straightforward fitting particle morphology than using the TEOS as sup-
function. The curves drawn in figs. 2 - 4 are calcu- plied by the vendor. Near the m a x i m u m achieva-
lated using eq. (1). It should be noted that the ble size for any given TEOS concentration, mono-
data used in developing the correlation did not dispersity was often not achieved and bimodal
include results for water concentrations greater final particle size distributions were c o m m o n (fig.
than 17 M. Thus the predictions up to 25 M water 6). Stober et al. [1] reported a similar type of
as shown in fig. 4 suggest that eq. (1) can be used behavior for the larger particle sizes which they
100 G.H. Bogush et aL / Preparation of monodisperse silica particles

--,,iF V

Fig. 6. Bimodal particle size distribution precipitated at 25 ° C from a solution contmning 0.4 M TEOS, 0.5 M NH3, and 5.5 M H20.

prepared. We also observed heterodisperse particle

size distributions at high TEOS and ammonia
concentrations, mostly at 0.5 M TEOS and 2.0-3.0 5o
M ammonia. These samples showed signs of being
flocculated. 25
In the range where monodispersity was
achieved, we found, in agreement with Van Helden
et al. [4], that the standard deviation expressed as ~2o.
a percentage of average size decreased with in-
creasing particle size (fig. 7). Surprisingly, these ~ 15
O o
results appear to follow a function which is inde- 0
pendent of initial reagent concentrations. For O 0

O
example, particles synthesized with 0.17 M TEOS ~ ~0- 0 o O o O 00
and 1 M N H 3 and with 3.0 and 19.0 M HzO had 0 O O0 O
0 o o 0~0 0 o %
final particle diameters of 279 and 270 nm respec- ~ o 0%
O 0O ~ O0 O
tively. The standard deviations in therespective
particle size distributions were 5 and 8%. This 0 0
similarity in the breadth of final particle sizes 0
' ' '28o' ' '460' ' '660' ' '~60' ' '1o0o
occurred even though the water concentrations Average DTemeter (nrn)
were substantially different and the total reaction
Fig. 7. Standard deviation in particle size distribution ex-
times were 180 min for 3.0 M H 2 0 but only 45 pressed as a percentage of average size as a function of average
rain for reaction containing 19.0 M H 2 0 . size.
 G.H. Bogush et al. / Preparation of monodisperse silica particles 101

1880 -- - g / c m 3 for particles in the 80-900 nm diameter

o .IOM TEOS
[] .25M TEO5
range. It is felt that the helium samples micro-
× .35M TEOS pores inside the particles. Consequently, the mea-
80C sured density does not represent the particle's
bulk density (as defined by the mass of a single
E
J
particle divided by its volume which has been
600 - calculated on the basis of the electron microscope
©
i
E i diameter). The density determined by pycnometry
does, however, fall in the range of densities re-
x
8, a2,C
I O S
ported by Van Helden et al. [4] and Vrij et al. [9]
L t o x as determined by measuring suspension densities
4 as a function of solids mass fraction, a method
28,S
O
expected to give densities very similar to those
! °
J determined via pyconometry. Particle bulk densi-
ties were estimated by following the esterification
procedure of Van Helden et al. [4] which replaces
10 20 30 40 5Q 60
Tem~erctdre (dec) C)
all silanol groups on the internal and external
surfaces with steryl alcohol groups. After rigor-
Fig. 8. Particle size as a function of reaction t e m p e r a t u r e for
s o l u t i o n s c o n t a i n i n g 1.0 M N H 3, 7.0 M H 2 0 a n d T E O S ously washing the resulting hydrophobic particles
c o n c e n t r a t i o n s of 0.10, 0.25 a n d 0.35 M. free of steryl alcohol and drying them at 90 ° C for
48 h, the particle densities were again determined
by helium pycnometry. These paticles had a den-
sity of 1.9-1.95 g / c m 3. Based on this density and
the assumption that the C~8 chains from the steryl
alcohol (O = 0.77 g / c m 3) plug all pores effectively
The effect of reaction temperature in final par- excluding helium, we calculate a bulk unesterified
ticle properties was explored over a temperature particle density of 1.78-1.86 g / c m 3. These values
range from 9 to 55 o C. The reactions were carried are in line with the calculations of previous inves-
out in 1.0 M ammonia, 7.0 M water and TEOS tigators [4,6,9] and indicate a particle porosity of
concentrations ranging from 0.1 to 0.35 M. The 11-15%.
results (fig. 8) indicate that final particle size From these results we conclude that SiO 2 par-
decreases monotonically as temperature increases. ticles of narrow size distribution can be reproduci-
At the lowest temperature studied monodispersity bly synthesized over a size range of 20-800 nm
was not always maintained, where as at elevated diameter. While the " b e s t " conditions will depend
temperatures narrow size distributions were found on what size particles are desired, the particles
for all TEOS concentrations. with the smallest percentage standard deviation
Elemental analysis carried out on the particles are achieved for average sizes greater than about
washed in deionized water and dried at 90 ° C for 300 nm. Near the maximum achievable size for
48 h yielded 0.15 wt% carbon, 0.67 wt% hydrogen any TEOS concentration, broad or multimodal
and 43.32% silicon. These results compare favora- particle size distributions are often obtained. For
bly with those of Van Helden et al. [4]. Assuming the TEOS system, particles of size greater than
all of the carbon originates from unhydrolyzed approximately 800 nm are difficult to synthesize
ethoxy groups which remain after the reaction has without resorting to the seeded growth technique
come to completion, we calculate that the final described below, and we recommend that a set of
particles contain 92.7% SiO2, 0.25% O C E H 5 groups experiments be carried out in the region of the
and 5.75% H20. maximum to determine the exact conditions which
Density determinations made using a helium produce large monodisperse particles if this growth
pycnometer yielded a solids density of 2.04-2.10 technique cannot be followed.
102 G.H. Bogush et al. / Preparation of monodisperse silica particles

!
Fig. 9. (a) Seed particle synthesized at 5 5 ° C from a solution containing 0.2 M TEOS, 1.0 M NH3, and 6.0 M H 2 0 . (b) Particles
growtn from seed particles shown in (a) by 10 single additions of TEOS and corresponding water. See text for explanation of single
addition nomenclature.
 G,H. Bogush et al. / Preparation of monodisperse silica particles 103

4. Seeded growth that the use of as supplied TEOS in subsequent

growth steps was not detrimental to the size distri-
While very narrow particle size distributions bution or morphology of the final product.
can be obtained with a one step growth process, In a typical experiment, two suspensions with
the solids content in the resulting suspension average diameter of 170 nm and mass fractions of
achieves a m a x i m u m values of 3% at a TEOS 1.5% were prepared. To the first sample, an amount
concentration of 0.5 M. We have found that at- of TEOS equal to that originally present i n the
tempts to increase the solids weight fraction above solution (4.0 ml in a total reaction volume of 90.0
this value by simply increasing the initial TEOS ml) and the corresponding volume of water (re-
concentration results in heterodisperse particle size ferred to below as a single addition) were added at
distributions. In order to achieve both larger par- 8 h intervals. To the second sample, twice the
ticles a n d / o r larger final mass fractions, we have original volume of TEOS and the corresponding
developed a seeded growth technique which re- volume of water were added at each interval (re-
suits in monodisperse particles at solids contents ferred to below as a double addition). After 10
with a theoretical limit of 24%. additions each, the first sample reached a mass
In this technique, a seed suspension is prepared fraction of 10.6%, and the second a mass fraction
using the techniques described above. The size of of 14.3%. The particles in each case remained
the seed does not influence the mondispersity of spherical and increased in size (fig. 9). In line with
the final product and thus should be selected on the results presented for conventionally grown
the basis of the desired final size and mass frac- particles, it was found that the percentage stan-
tion. TEOS and H 2 0 are added to the seed sus- dard deviation of the seeded growth particles
pension in a 1 : 2 mole ratio after the seed suspen- decreases as their size increases (fig. 10).
sion has stopped reacting. During ensuing hydrol- It was found that during seeded growth the size
ysis and condensation, the number of colloidally of the particles can be related to the total volume
stable particles remains constant but their size of TEOS added to the solution by a simple equa-
increases. While the use of distilled TEOS is nec- tion.
essary to achieve uniform seed particles, we found
d=do(V/Vo) 1/3, (2)
where d is the average diameter, d o is the average
100 ]
diameter of the seed particles, V0 is the moles (or
J equivalently, volume) of TEOS used to produce
I the seed particles and V is the total moles of
80 !
TEOS added to the solution (including V0). This
equation fits the results of the two experiments
very well (fig. 11).
eo J
The form of eq. (2) suggests that the volume of
the growing particles increases in proportion to
d~
q3
the number of moles of TEOS added to the solu-
L
u 40 i
tion. As a result, the number and the density of
i o
o i
,
,~
×
o the particles must remain constant while they are
o
20
I ~ X
growing. As would be expected, however, for a
seeded growth process, if the added amount of
TEOS exceeds a critical value, nucleation will
result in the growth of a second population of
00 ~
o
v-- i
~DO 2~o sSo 46o s6o 600 colloidally stable particles. The critical TEOS value
Average Diameter (nm) depends on initial particle size and number den-
Fig. 10. Standard deviation after each growth step of particles, sity, and reaction temperture. For example, at
shown in fig. 11 o, single additions, x double additions. 5 5 ° C seed particles (prepared with 1 M N H s, 5
104 G.H. Bogush et al. / Preparation of monodisperse silica particles

600 M H 2 0 and 0.2 M TEOS) were grown successfully

into stable particles in suspensions having volume
fractions of greater than 10% by using both single
500
and double additions of TEOS for 10 growth
steps. However, at 25 o C, double additions to seed
E
particles prepared under the same conditions
resulted in a heterodisperse particle size distribu-
@
tion (fig. 12).
E .300
h3 We routinely use this seeded growth technique
to prepare 4 1 of suspensions with mass fractions
b~ 200 in the 11-15% ( w / w ) range. The technique works
well for particles with final diameters of 50-900
nm. When small final particles are desired, even
100
smaller seed particles must synthesized. For exam-
ple, to prepare 50 nm particles at a final mass
fraction of 11%, seeds with a diameter of ap-
0 5 10 15 20 25 proximately 18.5 nm must be synthesized. Condi-
V/Vo
tions resulting in particles in particles in this size
Fig. 11. Average particle size after each growth step during a
range can be easily determined for eq. (1) or by
seeded growth experiment o, single additions, × double ad- looking at the data presented in figs. 2-4. Growth
ditions. Curve calculated from eq. (2). of particles of any size appears to present no

Fig. 12. Heterodisperse particle size distribution produced at 25 o C with 10 double additions of TEOS to a seed suspension prepared
from 0.2 M TEOS, 1.0 M NH 3 and 5.0 M H 2 0 in ethanol.
 G.H. Bogush et al. / Preparation of monodisperse silica particles 105

particular difficulties as long as the critical TEOS the data presented in this paper and predicts the
concentration described above is not exceeded. correct kinetic trends, further work is required to
unravel the dependence of the rates of hydrolysis,
condensation, nucleation, and aggregation on re-
5. Mechanism of formation agent concentrations and reaction temperature.
Work along these lines suggests that the nuclea-
While the primary purpose of this paper is to tion and aggregation mechanism can be developed
establish experimental conditions which result in into a quantitatively predictive model [8].
silica particle with narrow size distributions, these
results also begin to lay a solid foundation on
which to develop mechanistic models of particle
growth. Relevant observations drawn from our 6. Conclusions
studies include the data presented in figs. 7 and 10
which indicate that as the particles grow their size
distribution becomes narrower. This suggests the In this paper we present conditions which re-
growth mechanism is self sharpening (i.e., small sult in the precipitation of SiO 2 particles from
particles grow faster than large particles). Sec- aqueous ethanol solutions containing ammonia
ondly, because the density (porosity) of the final and TEOS. This work extends earlier studies to
particles and their external granular morphology broader reagent concentration ranges and sum-
are independent of particle size or whether pre- marizes the data in a more quantitative fashion
pared in a single or multistep process, it can be than has been available previously in the litera-
concluded that the growth mechanism does not ture. While the resulting correlation contains no
alter dramatically during the course of the reac- mechanistic information, it can be used as an
tion. Finally, if precipitation is to result in engineering tool in the preparation of single sized
monodisperse particles, the growing particles must, S i O 2 powders. Surprisingly, we find that the pre-

at some point, achieve colloidal stability. That is, rcentage standard deviation of the resulting par-
as van der Waals attractive forces are ever present, ticles is independent of initial reagent concentra-
a substantial repulsive pair interaction potential tions and is solely a function of final particle size.
must develop to forestall flocculation and the The seeded growth technique described here has
resulting heterodispersity in size distribution. been used to prepare large quantities of S i O 2
These observations and kinetic studies reported particles at mass fractions up to 17 wt%. It is
elsewhere [8,10,11] have led us to postulate that found that the number and density of the particles
particle growth occurs primarily through an aggre- in the suspension remains constant during the
gation mechanism. Monodispersity is achieved growth steps thus allowing for accurate prediction
through size dependent aggregation rates where of increases in particle size and mass fraction for
the probability of aggregation between two par- each addition of TEOS.
ticles of the same size decreases as the particles
grow. Such aggregation laws are predicted by the
DLVO theory of colloidal stability [12] where the The authors would like to thank J. Nemeth and
repulsive interparticle potential is due to charges T. McCarthy for their elemental analysis and J.W.
bound to particle surfaces. This model suggets Goodwin for his helpful suggestions in the devel-
that silica nuclei formed in a supersaturated solu- opment of the seeded growth procedure. The
tion are colloidally unstable and aggregate. As the authors would also like to thank J.C. Chang, B.
particle size distribution moves to larger average Schwartz, and R.Y. Sugahara for their help in
sizes, the bigger particles stop interacting and determining particle densities.
grow primarily through aggregation with smaller This work was supported by the US Depart-
particles and freshly formed nuclei. While this ment of Energy, Division of Material Sciences
model fits the qualitative observations drawn from under contract DOE DEA AC02 76ERl198.
106 G.H. Bogush et al. / Preparation of monodisperse silica particles

References [7] E.A. Barringer and H.K. Bowen, J. Am. Ceram. Soc. 69
(1982) C-199; B. Fegley, E.A. Barringer and H.K. Bowen,
[1] W. Stober, A. Fink and E. Bohn, J. Coll. Interf. Sci. 26 J. Am. Ceram. Soc. 67 (1984) C-113; K.S. Mazdyasni,
C.T. Lynch and J.S. Smith II, J. Am. Ceram. SOc. 48
(1968) 62.
[2] B. Barringer, N. Jubb, B. Fegley, R.L. Pober and H.K. (1965) 372.
[8] G.H. Bogush, Masters Thesis, Department of Chemical
Bowen, in: Ultrastructural Processing of Ceramics, Glasses
and Composites, eds., L.L. Hench and D.R. Ulrich (John Engineering, University of Illinois at Urbana-Champaign,
1987.
Wiley and Sons, New York, 1984) ch. 26 and the contents
[9] A. Vrij, J.W. Jensen, J.K.G. Dhont, C. Path-
of the Better Ceramics through Chemistry Symposium
mamanoliaran, M.M. Kops-Werkhoven and H.M. Fijnaut,
Series published by the Materials Research Society, Bos-
Faraday Discuss. Chem. Soc. 76 (1983) 19.
ton, MRS Symp. Proc. Series 32 (1984) and 73 (1986).
[10] G.H. Bogush and C.F. Zukoski IV, to be published in the
[3] M.D. Sacks and T-Y Tseng, J. Am. Ceram. Soc. 67 (1984)
Proc. of Microstructures '86, held in Berkeley, CA in
526, 532; T. Shimohira, A. Makishima, K. Kotoni and M.
August 1986.
Wakakuwa, Proc. Int. Symp. of Factors in Densification
[11] G.H. Bogush and C.F. Zukoski IV, to be published in the
and Sintering of Oxide and Non-Oxide Ceramics (1978) p.
Proc. of the Third Int. Conf. on Ultrastructure Processing
1109.
[4] A.K. Van Helden, J.W. Jansen and A. Vrij, J. Coll. Interf. of Ceramics, Glasses and Composites, held in San Diego,
CA in February 1987.
Sci. 81 (1981) 354.
[5] J.H. Jean and T.A. Ring, Langmuir 2 (1986) 251. [12] E.J.W. Verwey and J. Th. G. Overbeek, Theory of the
[6] K. Bridger, D. Fairhurst and B. Vincent, J. Coll. Interf. Stability of Hydrophobic Colloids (Elsevier, New York,
Sci. 68 (1979) 190. 1948).