You are on page 1of 6

Microporous and Mesoporous Materials 90 (2006) 293–298

www.elsevier.com/locate/micromeso

Water adsorption in hydrophobic nanopores:


Monte Carlo simulations of water in silicalite
Chen E. Ramachandran, Shaji Chempath, Linda J. Broadbelt, Randall Q. Snurr *

Department of Chemical and Biological Engineering, Institute for Environmental Catalysis, Northwestern University, Evanston, IL 60208, USA

Received 15 August 2005; received in revised form 8 October 2005; accepted 14 October 2005
Available online 15 December 2005

Dedicated to the late Denise Bartomeuf, George Kokotailo and Sergey P. Zhdanov in appreciation of their outstanding contributions to zeolite science

Abstract

Grand canonical Monte Carlo simulations have been carried out to investigate the adsorption of water from the vapor phase into the
zeolite silicalite. For truly hydrophobic micropores, the simulations predict essentially no adsorption of water at low pressures, followed
by rapid pore filling as pressure is increased. The effect of silanol defects in real silicalite samples was explored through simulations using
‘‘seeded’’ water molecules to represent hydrophilic defects. These defects promote adsorption of some water at low pressures, as mole-
cules form hydrogen-bonded clusters around the defects. The defects also shift the pore filling to a lower pressure than in the completely
hydrophobic material.
Ó 2005 Elsevier Inc. All rights reserved.

Keywords: Water; Silicalite; Defect; Hydrophobicity; Adsorption

1. Introduction which the system oscillates between pores devoid of water


and pores completely filled with water. Hummer et al. [6]
Understanding the behavior of water in hydrophobic observed similar behavior for MD simulations of water
micropores is important in a wide variety of fields, ranging in hydrophobic carbon nanotubes, in which small changes
from biological membrane transport to applications of car- in the interaction between pore walls and water molecules
bon nanotubes, as well as water purification using zeolites profoundly changed water hydration of pores, leading to
and activated carbons. It is also an interesting scientific sharp transitions between filled and empty states on a
question whether water will adsorb in a truly hydrophobic nanosecond timescale. McCallum et al. [7] studied the
pore. adsorption of water on activated carbon experimentally
In the literature, molecular simulations and thermody- as well as through grand canonical Monte Carlo (GCMC)
namic models predict very dramatic water isotherms in simulations and found that the density of active surface
hydrophobic micropores [1–4]. At low pressures, hardly sites contributed significantly to the loading before pore
any water molecules exist in the pores. This persists until filling occurred.
a specific pressure is reached, after which rapid pore filling Silicalite is a zeolite comprised of SiO2 tetrahedra with
occurs [3,4]. This has been observed experimentally for the MFI crystal structure [8,9]. It is widely regarded as
water in activated carbons [3]. A similar steep transition hydrophobic, and the earliest studies on silicalite showed
between empty and filled pores has been observed in a that it readily adsorbs organic molecules over water [8].
molecular dynamics study of hydrophobic pores [5], in These studies, as well as later ones [8,10–12], however,
show that silicalite does adsorb a small amount of water.
*
Corresponding author. Tel.: +1 847 467 2977; fax: +1 847 491 3728. MFI zeolites can also be prepared with various Si/Al
E-mail address: snurr@northwestern.edu (R.Q. Snurr). ratios, and a direct relationship has been established

1387-1811/$ - see front matter Ó 2005 Elsevier Inc. All rights reserved.
doi:10.1016/j.micromeso.2005.10.021
294 C.E. Ramachandran et al. / Microporous and Mesoporous Materials 90 (2006) 293–298

experimentally between increased Al content and increased 2.2. Silicalite model


water loading [12].
In this work, we present results of a GCMC study of The locations of the silicon and oxygen atoms of the
water adsorption in silicalite. The curious behavior zeolite framework were taken from XRD data [17], and
observed in simulations of water confined in hydrophobic the framework was considered to be rigid. The Lennard-
nanopores such as carbon nanotubes was also observed Jones parameters for the zeolite oxygen atoms were
in our simulations of silicalite. This behavior contrasts with obtained from previous work [15] and are listed in Table
the experimentally observed water adsorption isotherms in 1. Since silicon atoms are recessed behind oxygen atoms
silicalite, which do not show any pore filling1 but which do in the pore walls, they are not readily accessible to sorbates.
show low levels of adsorbed water even at low pressures. Therefore, van der Waals interactions between silicon
We will show that the water adsorbed at low pressures is atoms and the sorbate molecules were not considered.
likely associated with adsorption on hydrophilic defects However, Coulombic interactions between sorbates and
[14]. both oxygen and silicon atoms were modeled explicitly.
The silicon atoms carried a + 1.4 partial charge, whereas
2. Simulation methods oxygen atoms carried a 0.7 partial charge [15]. Interpola-
tion from pre-tabulated Lennard-Jones and Coulombic
2.1. GCMC simulations interaction grids was used to speed up the calculation of
zeolite/sorbate interactions.
Grand canonical Monte Carlo simulations are well Three different types of sites can be identified in silica-
suited for investigating adsorption in zeolites and related lite: zig-zag channels, straight channels, and intersections.
materials on the molecular level. In the grand canonical The process of defining sites using a methane probe mole-
ensemble, the volume, temperature, and chemical potential cule is described elsewhere [18]. In our simulations, the
are held constant while the number of molecules fluctuates. regions that were inaccessible to the methane probe but
Simulations were run using the simulation code Music accessible by water molecules were categorized into the
[15,16], which was written in-house. At the low pressures nearest, by distance, defined site.
used in our simulations, it was assumed that water vapor In some simulations, we introduced defect sites within
is an ideal gas and its fugacity is equal to the gas-phase silicalite in a simple way by randomly placing ‘‘seed’’ water
pressure. The simulation box was composed of eight silica- molecules within either zig-zag channels, straight channels,
lite unit cells with periodic boundary conditions. All simu- or intersections. Each defect in a unit cell was initially
lations reported here were at a temperature of 291 K. replicated in the other seven unit cells of the simulation
Properties of interest were calculated based on averaged box. These seeded water molecules are meant to mimic
values over the equilibrated regime for each simulation. intracrystalline defect silanol sites. They were allowed to
Even with biasing of insertion moves, the required number rotate during the GCMC simulation but were not allowed
of Monte Carlo steps for each isotherm point ranged from to be removed or translated. The resulting silicalite struc-
24 million to 360 million. Other researchers have also ture with seeds was then used to generate a water adsorp-
found that unusually long simulations are needed to equil- tion isotherm. Seeded water molecules were not counted
ibrate water in hydrophobic pores [7]. in the reported water loadings.
The van der Waals interactions between atoms were
represented by a Lennard-Jones 12-6 function, and elec- 2.3. Water model
trostatic interactions between point charges placed on the
atoms were calculated using CoulombÕs law. For the silica- Numerous models exist in the literature for bulk water,
lite/water interactions, Ewald summation was used for the each capable of capturing some experimentally observed
Coulombic interactions and the Lorentz–Berthelot mixing characteristics such as the vapor pressure or radial
rules were used to obtain the Lennard-Jones parameters.
For water/water interactions, a spherical cut-off based on
the molecular center of mass was enforced at 13 Å for both Table 1
Lennard-Jones and Coulombic interactions. Potential parameters
ri (Å) ei/kb (K) qi References
Water (SPC/E) O 3.16 78.212 0.848 [23]
H – – 0.424
Water (TIP4P) O 3.15 77.941 – [24]
1 H – – +0.52
As this manuscript was being completed, an experimental report
e – – 1.04
appeared showing pore filling in silicalite at pressures considerably above
Silicalite O 2.80 107.5 0.7 [15]
the bulk saturation pressure [13]. It is interesting to note that Desbiens
Si – – +1.4
et al. used GCMC simulations to predict pore filling above the bulk
saturation pressure, using the TIP4P water model, the same silicalite For the SPC/E model: rO–H = 1 Å, h = 104.5°, and Psat = 2.2 kPa at
partial charges as in Table 1, and Lennard-Jones parameters for the 300 K [33]. For the TIP4P model: rO–H = 0.9572 Å, h = 104.5° rO–e =
silicalite oxygens of r = 3.0 Å and e/k = 93.53 K. 0.15 Å, and Psat = 4.5 kPa at 300 K [27].
C.E. Ramachandran et al. / Microporous and Mesoporous Materials 90 (2006) 293–298 295

Fig. 1. Comparison of water models SPC/E (d) and TIP4P (j) for adsorption in silicalite at 291 K versus experimental data (solid line) from Olson et al.
[12] at 298 K. For the experimental curve, the experimental value of Psat was used, and for the simulation results, the Psat values of 2.2 kPa and 4.5 kPa
were used for the SPC/E and TIP4P models, respectively.

distribution function. As of yet, no universal model has 3. Results and discussion


been developed for this important molecule. Several poten-
tial models have been used in the past to simulate water in The simulated adsorption isotherms of water in silicalite
silicalite [19–22], mostly to elucidate the diffusion behavior. using the SPC/E and TIP4P models are shown in Fig. 1.
For this study, the SPC/E model [23] was chosen to repre- Both isotherms show qualitatively the same behavior: prac-
sent water, and the Lennard-Jones parameters and partial tically no water molecules are adsorbed at low pressure and
charges are listed in Table 1. The SPC/E model has been then the pores fill with water. The onset pressure of the
used extensively in the literature for a wide variety of differ- pore filling is different for the two different models, and it
ent systems since it was first introduced. Several other water appears that this value is very sensitive to the potential
models were also investigated for comparison purposes (i.e., parameters used [13]. Fleys and Thompson [29] also found
TIP3P, TIP4P, and TIP5P) [24–28]. Results are presented extremely low loadings of water in silicalite from GCMC
for the SPC/E model unless otherwise indicated. simulations with somewhat different potential parameters.

Fig. 2. Water adsorption isotherms with the SPC/E model at 291 K with and without seeds to mimic defects: 0.5 seeds per unit cell (n), 4 seeds per unit
cell (m), base case with no seed molecules (d and dashed line). The seeds were placed in the channel intersections. Experimental data from Olson et al. [12]
are also shown (solid line).
296 C.E. Ramachandran et al. / Microporous and Mesoporous Materials 90 (2006) 293–298

They did not report pore filling, possibly because they did the synthesis process. Estimates of as many as four defects
not simulate high enough pressures to observe this behav- per unit cell have been suggested [12].
ior with their model. The maximum loading from our sim- We reasoned that silanol defects could serve as hydro-
ulations is 53 molecules per unit cell, which is similar to philic sites that might affect adsorption of water at low
the expected pore filling value of 57 molecules per unit cell loadings. To test this hypothesis, we determined the water
[12]. isotherms with various numbers of seeded water molecules
Experimental isotherms for water in silicalite have been used to mimic these defects, as described in the previous
reported by several groups [8,10–12]. There are some differ- section. Fig. 2 shows the results with 0.5 seeds per unit cell
ences, but all show low levels of water adsorption (less and 4 seeds per unit cell. The isotherm with only 0.5 seeds
than 10 molecules per unit cell), starting gradually at low per unit cell is essentially identical to the base case. It
pressures. The experimental results of Olson et al. [12] should be noted that the seeded water molecules, which
are shown in Fig. 1 for comparison with the simulation occupy some pore volume, are not included in the loading
results. None of the vapor phase experimental isotherms quantified on the y-axis. Pore blockage by the seed mole-
indicate pore filling [13], although pore filling has been cules might explain some of the differences with the base
observed for adsorption of water in aluminophosphate case observed at the higher loadings, but the differences
molecular sieves [30]. are also indicative of the difficulties in equilibrating the sys-
Clearly, the simulated and experimental isotherms in tem when the pores are completely full.
Fig. 1 show qualitatively different behavior. One possible Using 4 seeds per unit cell produces qualitative changes
reason for this is the presence of silanol and other defects in the isotherm. The loading at low pressures is non-zero
in the experimental silicalite samples. It is believed that
silanol defects arise from the use of templating agents in

Fig. 3. Snapshot of water clustering around defect seed molecules in Fig. 4. Snapshot of water clustering around defect seed molecules in
silicalite. Conditions: 1 seed molecule (located in a straight channel) in 8 silicalite. Conditions: 4 seed molecules per unit cell (located in channel
unit cells, 0.34P/Psat, and 291 K. Oxygen atoms of seed molecules are intersections), 0.136P/Psat, and 291 K. Oxygen atoms of seed molecules
represented by the large blue spheres. Adsorbed water moleculesÕ oxygen are represented by the large blue spheres. Adsorbed water moleculesÕ
atoms are represented by large red spheres, and the hydrogen atoms are oxygen atoms are represented by large red spheres, and the hydrogen
represented by the smaller white spheres. (For interpretation of the atoms are represented by the smaller white spheres. (For interpretation of
references in colour in this figure legend, the reader is referred to the web the references in colour in this figure legend, the reader is referred to the
version of this article.) web version of this article.)
C.E. Ramachandran et al. / Microporous and Mesoporous Materials 90 (2006) 293–298 297

and is more consistent with the experimental data. This


supports the suggestion that defect sites within the pores
influence the adsorption of water at low loading. A truly
hydrophobic pore shows essentially zero loading of water,
but real silicalite samples show finite loadings due to
hydrophilic defect sites. Introduction of seeds also shifts
the pressure at which pore filling begins to a lower value.
A snapshot from a simulation with 1 seed molecule in
the simulation box (eight unit cells) is shown in Fig. 3.
The adsorbed water molecules clearly form clusters that
seem to grow around the seed molecule (note that the clus-
ters at the top/bottom and left/right are connected due to
the periodic boundary conditions). Fig. 4 shows a snapshot
from a simulation with 4 seed molecules per unit cell at low
pressure. Again, the adsorbed water molecules are located
preferentially near the seed molecules. Fig. 5 shows a snap-
shot from a simulation with 4 seed molecules per unit cell
at a higher pressure. Here, it can be seen that the pores
are almost full of water molecules.
Pair distribution functions were also calculated. The
pair distribution function based on the centers of mass
between the seeded molecules and non-seeded water mole-
cules, g[rCOM], showed a very structured configuration of Fig. 5. Snapshot of water adsorption in silicalite near saturation. (Note:
water around the defects at low loading, evident from a For better visualization of seed molecules, this figure is shown as viewed
sharp peak at 2.758 Å [31]. The distance and intensity from a different angle compared to Figs. 3 and 4.) Conditions: 4 seed
molecules per unit cell (located in channel intersections), 0.45P/Psat, and
indicate hydrogen bonding occurs between the seeded 291 K. Oxygen atoms of seed molecules are represented by the large blue
and non-seeded water molecules. Using g[rCOM], the num- spheres. Adsorbed water moleculesÕ oxygen atoms are represented by large
ber of nearest neighbors around the seeded water molecules red spheres, and the hydrogen atoms are represented by the smaller white
could be calculated [32]. For the case with 4 seed molecules spheres. (For interpretation of the references in colour in this figure
per unit cell, results showed on average, approximately one legend, the reader is referred to the web version of this article.)
water molecule surrounding each seed within hydrogen
bonding distance at low pressures (corresponding to The radial distribution function between the oxygen of
Fig. 4) and up to four water molecules surrounding each a seeded molecule and the hydrogen of non-seeded water
seed at high pressures (Fig. 5). molecules, g[rO–H], is shown in Fig. 6 for the case

Fig. 6. Radial distribution function between the oxygen atom of seeded water molecules and the hydrogen atom of adsorbed water molecules. Conditions:
0.5 seed molecules per unit cell (located in channel intersections), 0.14P/Psat, and 291 K.
298 C.E. Ramachandran et al. / Microporous and Mesoporous Materials 90 (2006) 293–298

corresponding to 0.5 seed molecules per unit cell and low [9] International Zeolite Association. Available from: <http://www.iza-
pressure. The first g[rO–H] peak occurs at 1.765 Å and the online.org>.
[10] B.V. Kuznetsov, T.A. Rakhmanova, Russ. J. Phys. Chem. 75 (6)
second at 3.2 Å, again consistent with hydrogen bonding. (2001) 933–938.
[11] A. Giaya, R.W. Thompson, Micropor. Mesopor. Mater. 55 (2002)
4. Conclusions 265–274.
[12] D.H. Olson, W.O. Haag, W.S. Borghard, Micropor. Mesopor. Mater.
The molecular simulations presented in this work predict 35–36 (2000) 435–446.
[13] N. Desbiens, I. Demachy, A.H. Fuchs, H. Kirsch-Rodeschini, M.
that in a perfect silicalite crystal, the pore system is hydro- Soulard, J. Patarin, Angew. Chem. Int. Ed. 44 (2005) 5310–5313.
phobic and essentially no water adsorbs at low pressure. [14] R.R. Cole, M.S. Gruszkiewicz, J.M. Simonson, A.A. Chialvo, Y.B.
At some intermediate pressure, a sharp transition occurs Melnichenko, G.D. Wignall, G.W. Lynn, J.S. Lin, A. Habenschuss,
where the pores become saturated with water molecules. B. Gu, K.L. More, T.D. Burchell, A. Striolo, Y. Leng, P.T.
The presence of hydrophilic defects, such as silanol sites, Cummings, W.T. Cooper, M. Schilling, K.E. Gubbins, H. Trieling-
haus, in: R. Wanty, R. Seal (Eds.), Influence of Nanoscale Porosity
in experimental samples leads to adsorption of small on Fluid Behavior, Water/Rock Interaction, vol. 1, 2004, pp. 735–
amounts of water at low pressures. Water molecules prefer- 739.
entially form hydrogen-bonded clusters around the defects. [15] A. Gupta, L.A. Clark, R.Q. Snurr, Langmuir 16 (2000) 3910–
The presence of hydrophilic defects also leads to a decrease 3919.
in the pressure where pore filling occurs. [16] A. Gupta, S. Chempath, M.J. Sanborn, L.A. Clark, R.Q. Snurr, Mol.
Simul. 29 (2003) 29–46.
[17] D.H. Olson, G.T. Kokotailo, S.L. Lawton, W.M. Meier, J. Phys.
Acknowledgements Chem. 85 (15) (1981) 2238.
[18] R.L. June, A.T. Bell, D.N. Theodorou, J. Phys. Chem. 95 (22) (1991)
This research is supported by the Chemical Sciences, 8866.
Geosciences and Biosciences Division, Office of Basic [19] C. Bussai, S. Hannongbua, R. Haberlandt, J. Phys. Chem. B 105
(2001) 3409–3414.
Energy Sciences, Office of Science, US Department of [20] C. Bussai, S. Fritzsche, R. Haberlandt, S. Hannongbua, J. Phys.
Energy, Grant No. DE-FG02-03ER15457. Additional sup- Chem. B 107 (2003) 12444–12450.
port from the National Science Foundation (CTS-0302428) [21] P. Demontis, G. Stara, G.B. Suffritti, J. Phys. Chem. B 107 (2003)
is also gratefully acknowledged. The authors thank Prof. 4426–4436.
Alain Fuchs for helpful discussions. [22] K.S. Smirnov, D. Bougeard, Chem. Phys. 292 (2003) 53–70.
[23] H.J.C. Berendsen, J.R. Grigera, T.P. Straatsma, J. Phys. Chem. 91
(1987) 6269–6271.
References [24] W.L. Jorgensen, J. Chandrasekhar, J.D. Madura, R.W. Impey, M.L.
Klein, J. Phys. Chem. 79 (1983) 926.
[1] S. Vaitheeswaran, H. Yin, J.C. Rasaiah, G. Hummer, Proc. Nat. [25] W.L. Jorgensen, J.D. Madura, Mol. Phys. 56 (1985) 1381.
Acad. Sci. 101 (49) (2004) 17002. [26] M.W. Mahoney, W.L. Jorgensen, J. Chem. Phys. 112 (20) (2000)
[2] N. Floquet, J.P. Coulomb, N. Dufau, R. Kahn, Stud. Surf. Sci. Catal. 8910–8922.
154 (2004) 1804–1811. [27] R.F. Cracknell, D. Nicholson, N.G. Parsonage, H. Evans, Mol. Phys.
[3] T. Ohba, H. Kanoh, K. Kaneko, J. Phys. Chem. B 108 (39) (2004) 71 (5) (1990) 931.
14964. [28] M.W. Mahoney, W.L. Jorgensen, J. Chem. Phys. 115 (23) (2001)
[4] A. Striolo, K.E. Gubbins, A. Chialvo, P.T. Cummings, Mol. Phys. 10758–10768.
102 (2004) 243–251. [29] M. Fleys, R.W. Thompson, J. Chem. Theory Comput. 1 (2005) 453–
[5] O. Beckstein, M.S.P. Sansom, Proc. Nat. Acad. Sci. 100 (12) (2003) 458.
7063–7068. [30] B.L. Newalkar, R.V. Jasra, V. Kamath, S.G.T. Bhat, Micropor.
[6] G. Hummer, J.C. Rasaiah, J.P. Noworyta, Nature 414 (2001) 188– Mesopor. Mater. 20 (1998) 129–137.
190. [31] C.E. Ramachandran, Ph.D. Thesis, Northwestern University, Evans-
[7] C.L. McCallum, T.J. Bandosz, S.C. McCrother, E.A. Muller, K.E. ton, IL, 2005.
Gubbins, Langmuir 15 (1999) 533–544. [32] R.Q. Snurr, J. Karger, J. Phys. Chem. B 101 (1997) 6469–6473.
[8] E.M. Flanigen, J.M. Bennett, R.W. Grose, J.P. Cohen, R.L. Patton, [33] G.C. Boulougouris, I.G. Economou, D.N. Theodorou, J. Phys.
R.M. Kirchner, J.V. Smith, Nature 271 (1978) 512–516. Chem. B 102 (6) (1998) 1029.

You might also like