J. Phys. Chem.

C 2007, 111, 18155-18158 18155

Modeling Aqueous Silica Chemistry in Alkali Media

Miguel J. Mora-Fonz, C. Richard A. Catlow, and Dewi W. Lewis*

Department of Chemistry, UniVersity College London, 20 Gordon Street, London WC1H 0AJ, United Kingdom
ReceiVed: September 6, 2007

A method for modeling the reactions of siliceous species is presented, which allows chemically accurate
deprotonation and dimerization energetics to be calculated, at conditions reflecting those of high pH
hydrothermal zeolite synthesis. The free energies of condensation reactions leading to silicate species up to
the linear tetramer are considered at room temperature and at 450 K.

The precise mechanism of zeolite formation is far from being
completely understood, a consequence of the complex reaction
space and the difficulties in probing this chemistry under
reaction conditions. Notwithstanding such problems, the initial
processes of nucleation and subsequent crystallization under
such hydrothermal conditions have been widely studied,1,2 and
key pre-nucleation species have been identified.3,4 We have also
recently demonstrated how computational methods can con-
tribute to an understanding of the reactivity of silicate oligomers
in the formation of specific “zeolitic” building units, with the
aim of establishing a more general picture of nucleation.5
However, key to progress in this area is establishing whether
computational models and methods are both accurate and also
reflect, as closely as possible, the conditions present during
synthetic and natural zeolite formation.
Silicic acid (Si(OH)4), the monomeric silicate molecule, is
the simplest but, perhaps, the most important species involved
in the formation of zeolite nucleation centers. Little zeolite
nucleation is noted under neutral conditions,2 and zeolite
syntheses are typified by high pH (ca. 10-14). Two processes,
in particular, are therefore vital in controlling the subsequent
chemistry: deprotonation of the monomer, Si(OH)4 + OH- f
Si(OH)3O- + H2O (and further deprotonations of the resulting
anion), and dimerization by condensation of both neutral and
anionic species, for example, 2Si(OH)4 f (OH)3-Si-O-Si-
(OH)3 + H2O.
Experimental measurements of monomer deprotonation and
dimerization energies have been reported and summarized by
Iler6 and Brinker,7 with McCormick8 reviewing more recent
measurements. Theoretical calculations have also been per-
formed using a variety of different models.9-14 The complexity
of silicate chemistry limits the experimental studies to only a
few species, specifically the monomer and dimer, as many
different reactions occur once polymerization is initiated, the
individual speciation of which proves almost impossible.
Similarly, computational methods have also been restricted by
their relative high cost and also lack, in some instances, the
incorporation of descriptions of solvent and pH effects.
Here, we describe how a dielectric solvent model combined
with explicit hydration allows DFT methods to describe
accurately the chemistry of small silicate oligomers, exemplified
here by Si(OH)4 and (OH)3-Si-O-Si-(OH)3, under conditions
typical of zeolite formation. Specifically, we report the energet-
* Corresponding author. Tel.: +44 20 7679 4779. Fax: +44 20 7679
7463. E-mail: d.w.lewis@ucl.ac.uk.
10.1021/jp077153u CCC: $37.00 © 2007 American Chemical Society
Published on Web 11/15/2007
ics of the deprotonation of both the monomer and the dimer,
together with those of the dimerization, trimerization, and
tetramerization reactions at 298 and 450 K. We also validate
our method with the well-characterized water autoprotolysis
reaction.
The calculations were performed using DMOL,3 version 2.2,15
using a double numeric basis set plus polarization (DNP) and
the BLYP functional. A description of the water solvent is
included via the COSMO16 approach, with initial gas-phase
optimized structures being reoptimized in the presence of the
solvation model. Molecular dynamics using DFT forces are also
used to ensure initial models do not become trapped in local
minima. The Gibbs free energy is calculated by combining the
electronic energy (BLYP/DNP), zero-point energy, together with
the rotational, vibrational, and translational contributions to the
energy, using standard statistical mechanics methods at 298 K
(the temperature of most of the experimental measurements)
and 450 K (characteristic of zeolite synthesis). Improvements
to this initial solvated model are made by the inclusion of
sodium counterions in the charged clusters, together with the
addition of some explicit (together with the continuum descrip-
tion of COSMO) water molecules in the first coordination sphere
of all of the species considered, to represent more completely
the most important short-range interactions. Details of the
number of explicit water molecules used for individual species
are reported in Table 1, which gives the free energies calculated
together with values derived from experimental pKa or equi-
librium constants reported in the cited literature. Examples of
the models used are shown in Figure 1.
The reliability of the methodology is established by consider-
ing the autoprotolysis of water. It is clear that describing this
reaction either (most simplistically) by “gas-phase” models or
even with the inclusion of a COSMO solvation model drastically
overestimates the free energy change. However, the inclusion
of some explicit water molecules improves the description,
giving the free energy within 4 kJ mol-1 of experiment. The
experimental value is obtained using what we consider the most
reliable current estimates of the entropies of the solution species
given by Aue et al.,17 specifically that there is no difference in
entropy between a solvated water and a hydronium ion.
Inclusion of a full explicit hydration sphere does not appear
necessary, as sufficient water is present to allow formation of
the stabilized complexes, with three and four water molecules
for the H+ and OH- ions, respectively. These complexes,
OH-(H2O)4 and H3O+(H2O)3, are found to be the most prevalent
18156 J. Phys. Chem. C, Vol. 111, No. 49, 2007 Mora-Fonz et al.

TABLE 1: Free Energies of Reaction at 298 and 450 K (kJ mol-1) in the Gas Phase and COSMO Solvation (COSMO) as
Compared to Silicates Species up to the Tetramer with Experimental Results (expt)a
free energy (∆G)/kJ mol-1
298 K 450 K 298 K
COSMO
reaction gas COSMO exptb COSMO (Sefcik radii)
water autoprotolysis H3O+ + OH- f H2O + H2O -1013 -258 -261
H3O+(H2O)3 + OH-(H2O)4 f (H2O)4 + (H2O)5 -509 -97 -10120,21 -91
monomer I M + OH- f M- + H2O -226 -61 -64 -42
M + NaOH f MNa + H2O -90 -49 -44
M + NaOH(H2O)3 f MNa(H2O)3 + H2O -27 -29 -336; -368 -30 -29
monomer II M- + OH- f M2- + H2O 285 2 2 0
MNa + NaOH f MNa2 + H2O -64 -43 -43
MNa(H2O)3 + NaOH(H2O)3 f MNa2(H2O)6 + H2O -47 -17 -188; -1722; -236 -14 -16
dimer I D + OH- f D- + H2O -296 -91 -92 -64
D + NaOH f DNa + H2O -100 -79 -76
D + NaOH(H2O)3 f DNa(H2O)3 + H2O -27 -35 -388 -30
dimer II D- + OH- f D2- + H2O 119 -25 -23 -14
DNa + NaOH f DNa2 + H2O -73 -43 -40
DNa(H2O)3 + NaOH(H2O)3 f DNa2(H2O)6 + H2O -28 -25 -298 -24
dimerization M + M(H2O) f D(H2O) + H2O -10 -4 -78 -3
M + M f D + H2O -13 -2 0
M + MNa(H2O)3 f DNa(H2O)3 + H2O -12 -7 0
M + MNa f DNa + H2O -23 -32 -31
M + M- f D- + H2O -83 -33 -28
trimerization M + DNa f TrNa + H2O -42 -16 -8
M + DNa(H2O)3 f TrNa(H2O)3 + H2O -43 -10 -4
tetramerization M + TrNa f TNa + H2O -16 9 16
M + TrNa(H2O)3 f TNa(H2O)3 + H2O -18 -10 -6
a Monomers I and II are the first and second deprotonation reactions for the monomer, respectively, with similar notation used for the dimer. For

ease of comparison, those results for the most comprehensive model are in bold. Results are presented with standard COSMO atomic radii at both
298 and 450 K and for selected species at 298 K with the optimized radii given by Sefcik and Goddard13 (see text for further details). Monomer,
dimer, trimer, and tetramer are symbolized by M, D, Tr, and T, respectively. b The experimental free energy for the proton transfer between two
water molecules in solution is calculated from a thermodynamic cycle. The experimental enthalpy of formation of the species in the gas phase is
taken from Dewar et al.,20 and the change in entropy for a proton transfer is considered zero, after Aue et al.17 The free energies of hydration are
taken from Pearson.21

Figure 1. Models of the singly deprotonated monomer and dimer, and that of the doubly deprotonated dimer. The models depicted are the most
comprehensive considered here, with three explicit water molecules and a Na+ ion coordinated to the deprotonated oxygen, together with a dielectric
continuum model of the remaining solvent. We also considered the same silicate cluster but only surrounded by the continuum water model and
also simply as gas-phase species (see Table 1). The structures shown are the energy minima for the models shown, and dotted lines indicate
hydrogen bonds.

species in such solutions.18,19 Little change in either geometry
or energetics is found when further explicit water molecules
are added.
The first deprotonation energy (as pKa) of the monomer is
the most reliably determined thermodynamic property of silicate
oligomers. Thus, this value serves as a key reference in
evaluating the reliability of any theoretical method whose aim
is to investigate the self-assembly of silicate oligomers. From
our results, it is evident that the gas-phase and the COSMO
models are again inadequate to model correctly the monomer’s
first deprotonation energy. Similarly, simply including the
counterion (Na+) together with COSMO results in poor energet-
ics as compared to experiment. It is only when both explicit
solvation and the Na+ counterion are included in our model,
making the system considered overall neutral, that the experi-
mental deprotonation energies are described accurately (within
4 kJ mol-1).
While there is general agreement that deprotonation energies
calculated with DFT-COSMO correlate well with experimental
pKa and that it is possible to estimate pKa values by linear
extrapolation,9,22,23 DFT-COSMO alone, using the standard
atomic radii to compute the solvation cavity, is not considered
able to reliably compute absolute pKa values.9 Here, however,
through the inclusion of some explicit solvation and counterions,
Modeling Aqueous Silica Chemistry in Alkali Media
good agreement with experiment is achieved. Thus, such an
approach, whereby some, but not necessarily all, of the inner
solvation sphere of a deprotonated species is considered, may
be more generally applicable, as discussed further below.
The presence of multiple species in solution makes accurate
experimental determination of the pKa of the dimer significantly
more difficult. Thus, values tend to be extrapolated from a
composite value following a fitting procedure to determine the
relative populations of the monomer and dimer.8 Again, there
is good agreement (within 4 kJ mol-1) between our calculated
deprotonation energies (once explicit hydration is considered)
and the available experimental results, giving both added
confidence in our method but also in the robustness of the
experimental fitting procedure, in this case.
Higher deprotonation energies have also been determined for
the monomer together with the first and second deprotonation
energies for the dimer.24 However, these rely increasingly on
fitting procedures, where the concentrations of the more strongly
deprotonated species and of the dimer species are very small
as compared to the dominant singly and doubly deprotonated
monomeric species. Hence, these values are likely to be less
reliable. Nevertheless, it is clear that the first and second
deprotonation energies of the dimer are well described with our
method (Table 1), giving us increased confidence in modeling
larger silicate species.
The dimerization reaction is the fundamental reaction of silica
chemistry. Again, the range of conditions (acidic, neutral, basic)
and the difficulties in attempting to restrict further polymeri-
zation occurring make it a challenging reaction to study
experimentally. The experimental estimate of the ∆G is = -7
kJ mol-1.8 In addition, there are a number of computed values:
using DFT/TNP, Catlow et al.12 reported a gas-phase dimer-
ization free energy of -9.2 kJ mol-1, while Tosell,11 using MP2/
6-311+G(2d,p), reported a value of +24.7 kJ mol-1 in the gas
phase but +8.8 kJ mol-1 when a G2/COSMO model is applied.
Most recently, Trinh et al. reported a value of +9 kJ mol-1 for
the neutral dimerization and -28 kJ mol-1 for the reaction of
a monomer with a deprotonated monomer, both using B3LYP/
6-31+G(d,p) with COSMO solvation.23 However, none of these
studies consider explicit hydration.
Our calculations reproduce well the experimental data for the
dimerization reaction, both under neutral and under basic
conditions. Note, that these calculations correctly reflect how
an increase in pH makes polymerization more favorable.
Moreover, we find the reaction of two deprotonated monomers
(with cations and explicit water) to be slightly less favorable
(-3.4 kJ mol-1), again reflecting experimental observation that
very high pH does not necessarily result in increased polym-
erization. Thus, we believe that the combination of the DFT
method used, together with a suitable explicit representation of
the inner solvation sphere of the key species and a continuum
representation of the remainder of the solvent, allows the
aqueous chemistry of silicate oligomers to be modeled reliably.
The effect of including explicit water molecules can be
considered two-fold. First, they shield the highly charged regions
from the continuum, allowing electrons to redistribute more
“naturally”. Second, they result in an improvement in the
description of the shape of the cavity in the dielectric continuum
that is constructed by the COSMO methodology. COSMO uses
atomic radii to construct the cavity,16 and the large difference
in atomic size between neutral atoms and these charged and
ionic species is significant and results in a poorer description
of the surface of the molecule. Thus, addition of explicit water
near such atoms, particularly bare O- and Na+, effectively
J. Phys. Chem. C, Vol. 111, No. 49, 2007 18157
places these atoms “inside” the molecule and reduces the errors
in the cavity shape and size. Similar conclusions have been
drawn before for silicates13 and for related systems, for example,
concentrated aluminum hydroxide solutions.24,25
An alternative to the inclusion of explicit solvent, to improve
quantitative agreement with experiment, is to refit the radii
used.13 However, such a fitting procedure rescales the radii of
all atoms of the same element. Here, the most significant change
in atomic size is due to the deprotonation of the oxygen atoms
in the silicate fragments. The semi-ionic nature of these species
means that the charge is fairly well localized on this particular
oxygen. Hence, rescaling the radius of all oxygen atoms is
unrealistic. Furthermore, the fitted radii are dependent on the
level of theory used,13 which will lead to some compromise in
transferability. To illustrate that our method of inclusion of
explicit water, which in essence ensures that changes in atomic
size do not result in a poorer representation of the cavity, is
both transferable and, indeed, gives better agreement with
experiment, we present in Table 1 the results obtained without
explicit water but with the Sefcik and Goddard13 COSMO radii.
These authors used B3LYP 6-31G**+//6-31G** for their
calculations, and they note that the fitted radii are therefore
specific to this level of theory. Hence, our deprotonation energies
using their radii but with the BLYP/DNP/COSMO method do
not exactly reproduce the experimental values. On the contrary,
when the Sefcik and Goddard radii are used together with
explicit water, the results are almost comparable to our original
results (and hence experiment). Thus, our hypothesis that the
explicit water provides an effective shielding of the ionic species
is supported.
We acknowledge that the level of theory considered here
should not, perhaps, be expected to give as good an agreement
as found. Nevertheless, the method performs consistently for
all of the species and conditions considered and at a relative
low computational cost. Indeed, the inclusion of the combination
of explicit and COSMO solvation appears to provide a good
compromise between expense and accuracy. Note that, at this
level of theory, the largest cluster presented here (the hydrated
tetramer) requires 25 000 min on a 2.8 GHz Xeon processor,
with 2 GB memory. Furthermore, little change is noted, for
example, when the number of explicit waters is increased further
here, nor when Car Parinello quantum molecular dynamics were
used (at considerable higher cost) to consider aluminate spe-
cies.25 Hence, the method, which allows us to consider condi-
tions that are chemically relevant to zeolite synthesis, can, we
believe, be applied with confidence to further silicate species.
We show, for example, in Table 1 the results for the reactions
discussed above at 450 K, a temperature akin to typical zeolite
synthesis, and also for the trimerization and tetramerization
reactions. Of note here is that there is little variation in the free
energy of adding a monomer to a hydrated monomer, dimer,
or trimer at 298 K. However, the formation of such small
fragments through monomer addition is less favorable at
elevated temperatures, suggesting that lower temperatures are
required to initiate formation of larger silicates. Indeed, we note
that room-temperature aging of synthesis gels is often critical
to the successful formation of zeolite crystals.
Further studies using this method will allow us to approach
a more comprehensive description of the key reactions that occur
during the pre-nucleation phase of zeolites and other silicates.
Acknowledgment. We acknowledge Dr. Jörg A. Becker and
Dr. Jürgen Woenckhaus for stimulating discussion regarding
this work. M.J.M.-F. is greatly indebted to CONACYT Mexico
and UCL for financial support. These calculations were
18158 J. Phys. Chem. C, Vol. 111, No. 49, 2007
performed on a computational resource provided by the EPSRC
to D.W.L. through grant GR/S06233/01.
References and Notes
(1) Cundy, S. C.; Cox, A. P. Chem. ReV. 2003, 103, 663.
(2) Cundy, C. S.; Cox, P. A. Microporous Mesoporous Mater. 2005,
82, 1.
(3) Kinrade, S. D.; Knight, C. T. G.; Pole, D. L.; Syvitski, R. T. Inorg.
Chem. 1998, 37, 4272.
(4) Kirschhock, C. E. A.; Kremer, S. P. B.; Grobet, P. J.; Jacobs, P.
A.; Martens, J. A. J. Phys. Chem. B 2002, 106, 4897.
(5) Mora-Fonz, M. J.; Catlow, C. R. A.; Lewis, D. W. Angew. Chem.,
Int. Ed. 2005, 44, 3082.
(6) Iler, R. K. The Chemistry of Silica: Solubility, Polymerization,
Colloid and Surface Properties, and Biochemistry; Wiley: New York, 1979.
(7) Brinker, C. J.; Scherer, G. W. Sol-Gel Science: The Physics and
Chemistry of Sol-Gel Processing; Harcourt Brace Jovanovich: Boston, 1990.
(8) Sefcik, J.; McCormick, A. V. AIChE J. 1997, 43, 2773.
(9) Schuurmann, G. J. Chem. Phys. 1998, 109, 9523.
(10) Pereira, J. C. G.; Catlow, C. R. A.; Price, G. D. Chem. Commun.
1998, 1387.
(11) Tossell, J. A. Geochim. Cosmochim. Acta 2005, 69, 283.
(12) Catlow, C. R. A.; Coombes, D. S.; Lewis, D. W.; Pereira, J. C. G.
Chem. Mater. 1998, 10, 3249.
Mora-Fonz et al.
(13) Sefcik, J.; Goddard, W. A. Geochim. Cosmochim. Acta 2001, 65,
4435.
(14) Trinh, T. T.; Jansen, A. P. J.; van Santen, R. A. J. Phys. Chem. B
2006, 110, 23099.
(15) Delley, B. J. Chem. Phys. 2000, 113, 7756.
(16) Baldridge, K.; Klamt, A. J. Chem. Phys. 1997, 106, 6622.
(17) Aue, D. H.; Webb, H. M.; Bowers, M. T. J. Am. Chem. Soc. 1976,
98, 318.
(18) Tuckerman, M.; Laasonen, K.; Sprik, M.; Parrinello, M. J. Chem.
Phys. 1995, 103, 150.
(19) Tuckerman, M. E.; Marx, D.; Parrinello, M. Nature 2002, 417, 925.
(20) Dewar, M. J. S.; Zoebisch, E. G.; Healy, E. F.; Stewart, J. J. P. J.
Am. Chem. Soc. 1985, 107, 3902.
(21) Pearson, R. G. J. Am. Chem. Soc. 1986, 108, 6109.
(22) Tossell, J. A.; Sahai, N. Geochim. Cosmochim. Acta 2000, 64, 4097.
(23) Mora-Fonz, M. J.; Catlow, C. R. A.; Lewis, D. W. Stud. Surf. Sci.
Catal. 2005, 158, 295.
(24) Caullet, P.; Guth, J. L. Observed and Calculated Silicate and
Aluminosilicate Oligomer Concentrations in Alkaline Aqueous Solutions.
In Zeolite Synthesis; Occelli, M. L., Robson, H. E., Eds.; American Chemical
Society: Washington, DC, 1989.
(25) Sillanpää, A. J.; Päivärinta, J. T.; Hotokka, M. J.; Rosenholm, J.
B.; Laasonen, K. E. J. Phys. Chem. A 2001, 105, 10111.