ARTICLE pubs.acs.org/Langmuir
Monte Carlo Simulation of Water Adsorption in Hydrophobic MFI Zeolites with Hydrophilic Sites M. G€oktug Ahunbay* Department of Chemical Engineering, Istanbul Technical University, 34469 Istanbul, Turkey
bS Supporting Information ABSTRACT: The effect of strong and weak hydrophilic sites, Al atoms with associated extraframework Na cations and silanol nests, respectively, in high-silica MFI zeolites on water adsorption was investigated using Monte Carlo simulations. For this purpose, a new empirical model to represent potential energy interactions between water molecules and the MFI framework was developed, which reproduced the hydrophobic characteristics of a siliceous MFI-type zeolite, silicalite-1, with both the vapor-phase adsorption isotherm and heats of adsorption at 298 K being in good agreement with experimental data. The proposed model is also compatible with previous hydrocarbon potential models and can be used in the adsorption simulations of VOCwater mixtures. Adsorption simulations revealed that strongly hydrophilic Al sites in Na-ZSM-5 zeolites coordinate two water molecules per site at low coverage, which promotes water clustering in the vicinity of these sites. However, weakly hydrophilic silanol nests in silicalite-1 are in coordination with a single water molecule per site, which does not affect the adsorption capacity significantly as expected. However, even in the presence of 0.125 silanol nest per unit cell, the increase in the heat of adsorption at low coverage is drastic.
1. INTRODUCTION Zeolites are crystalline aluminosilicates with charge-compensating cations in the crystal lattice. Aluminum-free zeolites such as silicalite-1, dealuminated Y, and zeolite-beta are important molecular sieves because they can be used for the separation of organics from water. They are particularly advantageous for their hydrophobicity, stability, incombustibility, and low temperature of regeneration.1 Silicalite-1, which is an all-silica version of MFItype zeolite ZSM-5,2 is especially suitable for some possible environmental applications of the removal and/or remediation of organics, such as chlorinated VOCs3,4 and MTBE5,6 in water. Consequently, the behavior of water confined within silicalite-1 has become a subject of great scientific and technological interest.3,711 However, there is not much agreement among these studies, which may be attributed to the differences in samples used in these experiments, where the effects of various defects, their distribution, and crystal size are significant. A typical example is the work of Oumi et al.,10 where the authors obtained two different isotherms depending on whether the crystals were prepared in the presence of HF. Therefore, it is also important to understand the effect of aluminum atoms and the associated cations as well as the presence of silanol defects on the behavior of water in strongly hydrophobic ZSM-5 structures. Molecular simulation is a powerful tool allowing a better understanding of water behavior in MFI zeolites. Demontis et al.12 used a molecular dynamics (MD) method to study water diffusion in silicalite-1 at different temperatures. They developed an empirical potential model to represent waterwater and waterzeolite interactions, which yielded diffusion coefficients r 2011 American Chemical Society
and heats of adsorption at infinite dilution in good agreement with prior experimental work. Similarly, Bussai et al.13 studied water diffusion in silicalite-1 via MD simulations using a force field that was based on ab initio calculations. Desbiens et al.14 were focused on water condensation in silicalite-1 using grand canonical Monte Carlo simulations (GCMC). In another GCMC study, Puibasset and Pellenq focused on the adsorption of water in silicalite-1.15 Their simulations did not reproduce the expected hydrophobicity of silicalite-1, showing only qualitative agreement. Castillo et al.16 studied various potential models for water adsorption in silicalite and fuajasite in detail. They concluded that the water model and the partial charge on the zeolite atoms were critical parameters to be selected carefully because the dipole moment of water results in completely different behavior than for that of other molecules of similar size but without a dipole moment. They reported also that the framework model had a significant impact on the prediction of water adsorption isotherms. Thomson and his group studied water diffusion and adsorption using MD and GCMC simulations.17,18 However, studies considering high-silica MFI zeolites containing hydrophilic sites are rare. The effect of silanol nests on the water adsorption capacity of silicalite-1 was studied by Ramachandran and co-workers19 via Monte Carlo simulations, and they showed that such hydrophilic defects lead to the adsorption of small amounts of water at low pressures and also to a decrease in the pressure where pore filling occurs. Yazaydin and Thompson20 Received: November 25, 2010 Published: March 23, 2011 4986
dx.doi.org/10.1021/la200685c | Langmuir 2011, 27, 4986–4993
Langmuir
ARTICLE
Table 1. Electrostatic Charges of the Framework Si Atoms Used in the Molecular Simulation of WaterSilicalite-1 Systems authors
Table 2. Intermolecular Potential Parameters for the SPC, MSPC/E, and SPC/exp-6 Water Models
qSi (e)
method
SPC31
model parameter
MSPC/E32
SPC/exp-633
van der Waals term
LJ
LJ
exp-6
þ1.57 to þ1.67
ε/kb (K) σ (Å)
78.21 3.1467
74.68 3.116
159.78 3.195
þ1.2 to þ1.7
R
þ2
qH(e)
0.4100
0.4108
0.3687
GCMC MD
þ0.5 to þ2.05 þ0.89
OH bond (Å)
1.0
0.9839
1.0668
HOH angle (deg)
109.5
109.5
109.5
Ramachandran et al.19
GCMC
þ1.4
Yazaydin and Thompson20
GCMC and MD
þ1.167 to þ1.314
Demontis et al.
MD
þ2
Bussai et al.13
MD
Desbiens et al.14
GCMC
Puibasset and Pellenq15
GCMC
Castillo et al.16 Fleys et al.17
12
used GCMC and MD methods to investigate the effect of silanol groups and extra-framework cations on water adsorption, diffusion, and the structure of water in silicalite-1. In another recent work, MD simulation was used by Arı et al.21 to understand the dynamics of water diffusion and structuring in silicalite-1 and Na-ZSM-5 zeolites. They also investigated the variation in the activation energy of diffusion as a function of the Si/Al ratio. Obviously, the accuracy of simulations on MFIwater systems depends strongly on the selected potential models. Whereas the above-mentioned studies provided an understanding of water behavior in ZSM-5 zeolites, potential models used to describe waterzeolite interactions showed significant variations, especially in the assignment of electrostatic charges of the framework atoms, as summarized in Table 1. Because of the highly polar nature of water molecules, accurate modeling of electrostatic interactions between water molecules and the framework is especially crucial in the adsorption simulations,14,15 which showed that the electrostatic interaction of the framework with confined water is smaller than that with bulk water.15 To account for this difference, either framework14,20 or water15 partial charges were reduced. However, each approach has its own disadvantage: Using a molecular model for confined water that is different from that for bulk water can create a discrepancy in modeling interfaces, such as zeolitewater interfaces, or in modeling water transport across interfaces, such as the interfaces of polymerzeolite hybrid membranes. However, reducing framework charges may lead to inaccurate estimates of heats of adsorption, thus contradicting the hydrophobic character of the zeolite.21 Furthermore, modifying framework charges can be inappropriate in simulating aqueous mixtures in zeolites. When considering the potential use of hydrophobic silicalite-1 for possible environmental applications for the removal and/or remediation of volatile organics in water, accurate molecular simulations for the adsorption of waterorganic mixtures are important. Although there are several potential models representing organicsilicalite-1 interactions that successfully reproduce experimental adsorption isotherms, it is not possible to combine them with the watersilicalite-1 potential models. The problem arises from the fact that most organicsilicalite-1 interaction models are based on the electrostatic charge of Si (qSi) set to þ2e2226 or to þ2.05e.16,20,27 In a recent study, Castillo and co-workers16 were able to reproduce the hydrophobic behavior of water in silicalite-1 even at high pressures with qSi set to þ2.05e using a five-site water model (TIP5P-EW).28 However, this water model was reported to overpredict the saturated vapor pressure considerably.29
12
The objective of this study is to understand the effect of strong and weak hydrophilic sites (i.e., AlNaþ sites and silanol nests, respectively) on water adsorption in high-silica MFI zeolites using Monte Carlo simulations. For this purpose, a new potential model representing the interactions between water molecules and the zeolite framework was developed, which was then used to calculate adsorption isotherms and isosteric heats of adsorption of water in MFI zeolites with different Si/Al ratios and silanol nest contents.
2. METHOD All of the adsorption simulations in the present study were carried out using the Materials Studio 4.1 simulation package (Accelrys Software, San Diego, CA). For the silicalite-1 (silicalite) framework model, the structure with orthorhombic symmetry reported by van Koningsveld30 was adopted in the simulations. To reproduce the hydrophobic behavior of the zeolite, it was necessary to describe watersilicalite interactions with a new potential model that is compatible with semiempirical water models. For this purpose, the watersilicalite interaction model initially proposed by Demontis et al.12 was used as a reference. To represent pairwise interactions between the water atoms (Ow and H) and framework atoms (Si and Oz), 126 Lennard-Jones (LJ) and exp-6 type potential models were tested. The rationale beyond the selection of these potential types was to represent zeolitewater interactions using common functional forms that can be implemented easily into different simulation software. To ensure the use of the new model to simulate the adsorption of waterhydrocarbon mixtures, the framework silicon atom charge (qSi) was kept at 2e. Three different semiempirical rigid water models—SPC,31 MSPC/ E,32 and SPC/exp-6 33—were used in these tests. All three models were proven to reproduce the vapor pressure of water with good accuracy; in particular, SPC/exp-6 matches the experimental value of 3.5 kPa.34 These models consist of a van der Waals term acting between oxygen (Ow) atoms only and a Coulombic term acting between partial charges that are placed on the hydrogen (Hw) and oxygen atoms. The particularity of the SPC/exp-6 model is its partial charges that are smaller in magnitude than those of the other two models. Therefore, the electrostatic contribution to frameworkwater interactions is the lowest for this model, and it has the potential to improve the accuracy of predictions of water adsorption isotherms without sacrificing the accuracy of the bulk water properties. The model parameters are summarized in Table 2. To calculate interaction energies between the framework atoms and the water atoms, the water accessible volume of the silicalite unit cell was divided into grids and Ow and Hw atoms were placed on the grid points in turn. The potential parameters were optimized by minimizing the objective function defined as f ¼ 4987
N
∑ i¼1
UðwÞ UðwÞref Þi 2
ð1Þ
dx.doi.org/10.1021/la200685c |Langmuir 2011, 27, 4986–4993
Langmuir
ARTICLE
Figure 1. Structures of NaþZSM-5 simulation cells: ZSM-5/191 viewed along the (a) a axis and (b) c axis and ZSM-5/95 viewed along the (c) a axis and (d) c axis. Na ions are shown as purple spheres.
Table 3. Properties of Zeolites Used in Adsorption Simulations partial charges (e) hydrophilic ZSM-5 type Si/Al site densitya
Si
¥
þ2
silicalite
191
0.5
ZSM-5/95
95
1
S-0.125D S-0.5D S-1D
a
0
ZSM-5/191
Al
OAl
OAlb
Naþ
H
1
þ2 þ1.7 1 1.175
Figure 2. Structure of S-0.125D containing 0.125 silanol nest per unit cell located on a T12-site viewed along the (a) a axis (sinusoidal channels) and (b) b axis (straight channels).
þ1
0.125 ¥
0.5 1
S-2D
2
S-4D
4
þ2
1
þ0.5
Naþ or silanol nest per unit cell. b Bonded to an Al atom.
where w is either Ow or Hw, U(w)ref is the reference interaction energy value for the water atom considered, which was calculated on the basis of the model proposed by Demontis et al.,12 and N is the total number of grid points. Note that this objective function does not take into account the geometry of the water models. Therefore, the only parameters considered were the electrostatic charges and the van der Waals terms of the water models. Consequently, the OH bond length and HOH angle values had no effect on the potential parameters. Once the parameters were optimized, the heat of adsorption at infinite dilution was calculated at 298 K via 107 MC simulation steps
in the NVT ensemble for a single water molecule in the silicalite framework of 8 (2 2 2) unit cells. A cutoff distance of 10 Å was adopted for the calculation of pairwise van der Waals interactions, and a Ewald summation was used to calculate the electrostatic interactions. Water diffusivity in silicalite was also estimated using the new potential model through MD simulations. To compare the results with previous studies,12,16,19 16 water molecules in a flexible framework of (1 1 2) unit cells were used in these simulations. The intraframework interactions were represented by the PCFF35 force field. A time step of 0.5 fs was used in these simulations and the temperature was controlled through the Nose thermostat. Diffusion coefficients (D) were calculated from the mean square displacement (MSD) obtained from 4 ns MD production runs in the NVT ensemble at 298, 354, and 393 K using the Einstein relationship D¼ 4988
1 d N Æjri ðtÞ ri ð0Þj2 æ lim t f ¥ 6N dt i 1
∑
ð2Þ
dx.doi.org/10.1021/la200685c |Langmuir 2011, 27, 4986–4993
Langmuir
ARTICLE
where N is the number of diffusing molecules, ri(0) and ri(t) are the initial and final positions of the center of mass of molecule i over the time interval t, and Æ|ri(t) ri(0)|2æ is the MSD of the ensemble. Next, the vapor adsorption isotherms were calculated using GCMC simulation by increasing the water pressure gradually from 1 103 up to 50 kPa at 298 K. Water vapor was assumed to be an ideal gas within this pressure range. A total of 1.5 106 MC steps were used for each of the equilibrium and subsequent production runs at each pressure point until the saturation pressure was reached; above this point, 5 106 MC steps were used for each run. To investigate the effect of strong hydrophilic sites on the adsorption of water, simulations were repeated for Na-ZSM-5 zeolites with two different Si/Al ratios: 191 (ZSM-5/191) and 95 (ZSM-5/95). To construct the frameworks, a corresponding number of Si atoms on the T12 sites were substituted by Al atoms,20 which are shown in Figure 1. The neutrality of the aluminosilicate frameworks was maintained by the addition of Naþ ions. The charge on each framework Al atom and its four neighboring O atoms was assigned following the work of Beerdsen et al.27 Locations of the ions were determined through MC runs. To analyze the impact of silanol nests as weak hydrophilic sites in the silicalite framework on water adsorption, T12-silicon atoms20 were removed in the (2 2 2) silicalite framework to construct different structures with defect densities varying between 0.125 and 4 per unit cell. The dangling oxygen atoms were saturated with hydrogen atoms. The electrostatic charges of these H atoms were set to þ0.5 to maintain framework neutrality, and only Coulombic interactions with the water molecules were considered. The properties of the ZSM-5 structures used in the simulation are summarized in Table 3. The geometry of the silanol groups was optimized using the energy-minimization tool available in the software package. The resulting structure of S-0.125D is shown in Figure 2.
3. RESULTS AND DISCUSSIONS Among the various combinations of 126 LJ and exp-6 models considered, the same functional forms yielded the minimum value of eq 1 for all three water models, with slightly different parameter values: A B ð3Þ UOw O ¼ 12 6 r r r UOw Si ¼ C exp ð4Þ ro UH w O
r ¼ R exp Fo
!
r UHw Si ¼ D exp Ro
β r6
ð5Þ
ð6Þ
Similarly, the heats of adsorption at infinite dilution at 298 K for all three water models were close to each other: 33.2, 31.0, and 29.7 kJ/mol for the SPC, MSPC/E, and SPC/exp-6 models, respectively, which are all in good agreement with various experimental work2,3,9,36,37 and with the value of 32.5 kJ/mol calculated by Demontis et al.12 However, when the vapor-phase adsorption isotherms at 298 K were calculated for the three water models, it was seen that waterwatersilicalite three-body interactions had a significant impact on the accuracy of the simulations. Isotherms for the three different models are shown in Figure 3, in comparison with the previous experimental data3,711 that exhibit significant variations due to the differences
Figure 3. Calculated vapor-phase adsorption isotherms of the SPC, MSPC/E, and SPC/exp-6 water models in silicalite at 298 K in comparison with the experimental data. The original figure presenting the experimental data is adapted from ref 20. Copyright 2009 with permission from Elsevier.
Table 4. WaterSilicalite van der Waals Interaction Parameters A (kJ/mol 3 Å12) B (kJ/mol 3 Å6)
1 125 674 4721
C (kJ/mol)
174.8
r0 (Å)
1.0633
R (kJ/mol)
83 330.5
F0 (Å)
3.8838
β (kj/mol 3 Å6)
1709.3
D (kJ/mol)
1.996 107
R0 (Å)
4.9991
in samples used, where the effects of various defects, their distribution, and their crystal size are significant. The steep water uptake in experimental isotherms may be attributed to capillary condensation occurring in the mesoporosity or macroporosity of the zeolite sample.14 From the simulation results, it can be seen that the clustering of SPC and MSPC/E waters begins at pressures much lower than the pure water saturation pressure and thus fails to represent the hydrophobic behavior of the zeolite, in agreement with previous simulation studies.14,15 However, the isotherm of SPC/exp-6 water is in reasonably good agreement with the hydrophobic characteristic of the silicalite, considering the variations in the experimental work. Thus, the significant differences among the calculated isotherms are due to the differences in the magnitude of the electrostatic interactions between the silicalite framework and the water models. The silicalitewater interaction parameters for the best-performing SPC/exp-6 model are presented in Table 4. Note that the NPT simulation of bulk SPC/exp-6 water at the same temperature yielded the heat of vaporization as 44.2 kJ/mol. A combination of the new silicalitewater model with the SPC/exp-6 water model has several advantages. Both the heat of adsorption at infinite dilution and the vapor-phase adsorption isotherm are in agreement with the experimental data, and the water model remains unmodified. Furthermore, it can be used to simulate the adsorption of aqueous mixtures of hydrocarbons in ZSM-5 zeolites using most of the existing hydrocarbon models because the framework electrostatic charges are compatible with them. 4989
dx.doi.org/10.1021/la200685c |Langmuir 2011, 27, 4986–4993
Langmuir
Figure 4. Arrhenius plot of the diffusion coefficients of water in silicalite in comparison with previous simulations and experimental PFGNMR data.
ARTICLE
Figure 6. Density maps for water at 298 K in silicalite, S-0.125D, ZSM5/191, andnZSM-5/95 viewed along the b axis. For the sake of comparison, for each material the plots have the same color scale between dark blue (lowest density) and red (highest density). Na ions are shown as purple spheres.
Figure 5. Adsorption isotherms of water molecules in ZSM-5 zeolites. The inset shows the vapor-phase adsorption region.
The diffusion coefficients of water using the SPC/exp-6 water model were calculated via MD simulations in the NVT ensemble at 298, 354, and 393 K. Production runs of 4 ns were used to ensure the validity of the Einstein relationship. The results presented in Figure 4 are compared to previously reported experimental PFG-NMR38 and simulation12,13,18,20,21 results available in the literature. Considering the lack of general agreement among the various simulation results with the experimental PFG-NMR data, comparison shows the reasonable accuracy of the proposed potential model. It is worth noting that significant disagreement can be seen even in studies where the same force field was used,17,21 which may be attributed to the variations in simulation parameters such as simulation time or cutoff distance. Next, calculations were repeated for ZSM-5/191 and ZSM-5/ 95 with strong hydrophilic sites. The resulting isotherms in Figure 5 show that the water adsorption capacity increases below the saturation pressure with increasing alumina content as the framework becomes more hydrophilic. Water density maps presented in Figure 6 at 3.5 kPa show that sinusoidal channels are preferentially occupied by water molecules in all three
Figure 7. Isosteric heats of adsorption for water in ZSM-5 zeolites. The inset shows the low-coverage region.
structures. Furthermore, the clustering of water molecules can be seen around ionic sites in ZSM-5/191 and ZSM-5/95. As the pressure is further increased, transition into a liquidlike phase begins in all zeolite structures and saturation is reached before reaching 50 kPa. Comparison with the experimental results of Eroshenko et al.,39 where the phase transition begins at 50 MPa, indicates that the potential model used in this study cannot quantitatively reproduce the liquid-phase adsorption behavior of the water. For liquid-phase adsorption studies, the TIP5P-EW model may be a reasonable choice as shown by Castillo et al.,16 even though this model significantly overpredicts the saturated vapor pressure of water. Adsorption simulations showed that water molecules prefer sinusoidal channels at infinite dilution, when the framework is Alfree. As the vapor pressure increases gradually, these channels are filled first, followed by straight channels. However, when a T-site contains an Al atom, it is preferred by water molecules, as expected; next sinusoidal and then straight channels are occupied by water molecules. 4990
dx.doi.org/10.1021/la200685c |Langmuir 2011, 27, 4986–4993
Langmuir
Figure 8. Waterwater (WW) and zeolitewater (ZW) interaction energies in Na-ZSM-5 zeolites.
Isosteric heats of adsorption for the three MFI structures were also calculated and presented in Figure 7. It was found that the heat of adsorption for silicalite increases with coverage, which is a characteristic feature of hydrophobic adsorbents. Noting that adsorption is an exothermic process, this result further validates the accuracy of the proposed interaction model when compared with the results of a recent study,20 where the framework Si atom electrostatic charges were reduced to 1.25e to reproduce the hydrophobicity of the silicalite, resulting in a decrease in the heat of adsorption with coverage contradicting the hydrophobic characteristic of the zeolite. The current result agrees qualitatively with the results reported by Giaya and coworkers.3 When strongly hydrophilic ionic sites are present in the framework, the comparison with silicalite shows that the presence of 0.5 ionic site per unit cell increases the heat of adsorption significantly. However, as the AlNaþ content is increased by two, the heat of adsorption at infinite dilution remains almost constant because only a single Al site can be occupied by the adsorbed water molecule at infinite dilution. For all Al-containing ZSM-5 zeolites, the isosteric heats of adsorption decrease with loading, which is expected as a result of the hydrophobic characteristic of the framework. This finding is in good agreement with previous experimental results.9 Thus, because the framework is rendered hydrophilic by the introduction of an Al site, the sorbateframework interaction energy at this site becomes much higher than that at other sites, and it is preferentially occupied. As the loading increases, adsorption takes place on less-energetic sites, which results in a gradual decrease in the heat of adsorption, which is in contrast with the case of water adsorption in Al-free silicalite. In this case, the very first adsorbed water molecule acts as a seed for the formation of water clusters due to strong waterwater interactions, leading to an increase in the heat of adsorption with loading until the framework is filled homogeneously. However, the heats of adsorption for all structures converge as the loading increases. Furthermore, when the waterwater and zeolitewater interaction energies are compared for ZSM-5/191 and ZSM-5/95 with silicalite, it can be seen in Figure 8 that the waterwater interaction energy for ZSM-5/ 191 and ZSM-5/95 reaches a maximum value at loadings of 0.5 and 1 water molecule per unit cell, respectively. This indicates
ARTICLE
Figure 9. Vapor-phase water adsorption isotherms for defective silicalites containing silanol nests in comparison with defect-free silicalite and Na-ZSM-5 zeolites.
that each Al site is in coordination with two water molecules. The fact that the maximum value of the waterwater interaction energy is positive in magnitude shows the strength of the hydrophilic site. In the case of adsorption in crystals with silanol nests as weak hydrophilic sites, these defects have a less-significant effect on the adsorption capacity than do Al sites in the Naþ-ZSM-5 zeolites, as shown in Figure 9. For the same hydrophilic site densities (0.5 and 1 site per unit cell), the adsorption capacity of the silanol nest containing zeolite is lower at pressures below 3.5 kPa. This behavior can be supported by the lower value of the heat of adsorption at infinite dilution as shown in Figure 10. It is almost constant for all defect densities as a result of the exclusive interactions of water molecules with a single silanol nest until all nests are occupied. The slight increase observed for the density of 4 nests per unit cell is due to the interaction of the water molecule adsorbed at a defect site with the closest second defect site. Consequently, the heat of adsorption curve exhibits a plateau at very low coverage until each silanol nest binds a water molecule and then decreases because of the weakened hydrophilic interactions. As the coverage increases further, the trend becomes similar to that of defect-free silicalite. This behavior can also be seen in the plots of the waterwater and zeolitewater interaction energies in Figure 11, where waterwater interaction energies remain zero until the water density reaches the defect density in the unit cell. The observed difference in the strength of silanol nests and ionic sites agrees with the findings of Yazaydin and Thompson,21 who showed that water adsorption increased significantly only in the presence of two silanol nests per unit cell and above. However, a significant disagreement can be seen in the comparison of heats of adsorption and interaction energies. They reported that the heats of adsorption increased with increasing silanol nest density, which did not show any converging trend with loading. Furthermore, the zeolitewater interaction energies increased and waterwater interaction energies decreased with loading, even below one water molecule per unit cell, which indicated the formation of water clusters instead of the occupation of silanol nests by water molecules one by one. These differences may be due to the potential model adopted by the 4991
dx.doi.org/10.1021/la200685c |Langmuir 2011, 27, 4986–4993
Langmuir
ARTICLE
in the heat of adsorption at low coverage. This information may be used to interpret the experimental results for water adsorption in silicalite better, where trace defects are unavoidably present in experimental samples.
Figure 10. Variation of isosteric heats of adsorption with water loading for defective silicalite structures: silicalite (b), S-0.125D (2), S-0.5D ((), S-1D (9), S-2D (1), and S-4D (þ). The inset shows the lowcoverage region.
Figure 11. Waterwater (filled symbols) and zeolitewater (empty symbols) interaction energies in defective silicalite structures: silicalite (b), S-0.125D (2), S-0.5D ((), S-1D (9), S-2D (1), and S-4D (þ).
4. CONCLUSIONS Samples used in the experimental studies of the vapor-phase adsorption of water in hydrophobic MFI zeolites contain unavoidably strong or weak hydrophilic defects. Consequently, measured isotherms exhibit significant variations. In this study, the effect of these strong and weak hydrophilic sites on water adsorption was investigated using Monte Carlo simulations. As strong hydrophilic sites, Al atoms with associated extraframework Na cations were considered, corresponding to Na-ZSM-5 zeolites with different Si/Al ratios, whereas the weak hydrophilic sites consisted of silanol nest defects in the silicalite-1 framework. To model zeolitewater systems, a new potential model describing interactions between water molecules and the MFI framework was developed and used to calculate the adsorption isotherms and isosteric heats of adsorption of water. Waterwater interactions were represented using the SPC/exp-6 water model. The new model reproduced the hydrophobic characteristics of silicalite-1, both the heat of adsorption and the vaporphase adsorption isotherm, in good agreement with various experimental vapor-phase water adsorption studies. It is also suitable for the simulation of the vapor-phase adsorption of waterhydrocarbon mixtures, such as aqueous mixtures of MTBE and chlorinated VOCs. GCMC simulations in Na-ZSM-5 (Si/Al = 95 and 191) zeolites revealed that strongly hydrophilic Al sites are in coordination with two water molecules per site at low coverage, which promotes water clustering in their vicinity. However, weakly hydrophilic silanol nests are in coordination with a single water molecule per site, which does not affect the adsorption capacity significantly. However, the heat of adsorption at low coverage was increased drastically in the presence of a defect density of as low as 0.125 silanol nest per unit cell. This information would be useful in the interpretation of the experimental results on water adsorption in silicalite-1 because the presence of defects in the crystals is almost unavoidable. ’ ASSOCIATED CONTENT
Table 5. Diffusion Coefficients of Water in Silicalite in Comparison with Previous Simulations and Experimental PFG-NMR Data D (109 m2/s) T (K)
ref 12
ref 13
297
8.6
3.3
354 393
10
6.7
ref 18
ref 20
ref 21
ref 38 (exp) 1.7
8.83
0.6
1.94
10
2.6
3.64
10.55
4.7
6
1.5
this work
bS
Supporting Information. Water and zeolite potential models. Silicalite structures containing silanol nests. This material is available free of charge via the Internet at http://pubs.acs. org.
’ AUTHOR INFORMATION
0.98
Corresponding Author
2.33
*E-mail:
[email protected].
4.28
authors, which also led to a decrease in the heat of adsorption with loading for defect-free silicalite. It can be also noted that whereas the presence of a 0.125 silanol nest per unit cell increases the adsorption capacity of silicalite almost insignificantly, there is a very significant increase
’ ACKNOWLEDGMENT This work is supported in part by the Scientific and Technological Research Council of Turkey (TUBITAK) through grant no. 106M339 and by Istanbul Technical University through the Nano Project. 4992
dx.doi.org/10.1021/la200685c |Langmuir 2011, 27, 4986–4993
Langmuir
’ REFERENCES (1) Garrot, B.; Simonot-Grange, M.-H.; Clausse, B. Use of Dealuminated Faujasite Y and ZSM-5 Zeolites for VOC’s Adsorption. Application of Au Cleaning. In Porous Materials in Environmentally Friendly Processes; Kiricsi, I., Nagy, J. B., Karge, H. G., Palyi, G., Eds.; Studies in Surface Science and Catalysis; Elsevier: Amsterdam, 1999; Vol. 125, pp 683690. (2) Flanigen, E. M.; Bennett, J. M.; Grose, R. W.; Cohen, J. P.; Patton, R. L.; Kirchner, R. L.; Smith, J. V. Nature 1978, 271, 512. (3) Giaya, A.; Thompson, S. W. Microporous Mesoporous Mater. 2002, 55, 265. (4) Uguina, M. A.; Sotelo, J. L.; Delgado, J. A.; Gomez, J. M.; Celemín, L. I. Sep. Purif. Technol. 2005, 42, 91. (5) Erdem-S-enatalar, A.; Bergendahl, J.; Giaya, A.; Thompson, R. W. Environ. Eng. Sci. 2004, 21, 722. (6) Knappe, D. R. U.; Campos, A. A. R. Water Sci. Technol.: Water Supply 2005, 5, 83. (7) Bolis, V.; Busco, C.; Ugliengo, P. J. J. Phys. Chem. B 2006, 110, 14849. (8) Kuznetsov, B. V.; Rakhmanova, T. A. Russ. J. Phys. Chem. 2001, 75, 933. (9) Olson, D. H.; Haag, W. O.; Borghard, W. S. Microporous Mesoporous Mater. 2000, 435, 35. (10) Oumi, Y.; Miyajima, A.; Miyamoto, J.; Sano, T. Stud. Surf. Sci. Catal. 2002, 142, 1595. (11) Soulard, M.; Patarin, J.; Eroshenko, V.; Regis, R. Stud. Surf. Sci. Catal. 2004, 154, 1830. (12) Demontis, P.; Stara, G.; Suffritti, G. B. J. Phys. Chem. B 2003, 107, 4426. (13) Bussai, C.; Hannongbua, S.; Fritzsche, S.; Haberlandt R. Chem. Phys. Lett. 2002, 354, 310. (14) Desbiens, N.; Boutin, A.; Demachy, I. J. Phys. Chem. B 2005, 109, 24071–24076. (15) Puibasset, J.; Pellenq, R. J.-M. J. Phys. Chem. B 2008, 112, 6390. (16) Castillo, J. M.; Dubbeldam, D.; Vlugt, T. J. H.; Smit, B.; Calero, S. Mol. Simul. 2009, 35, 1067–1076. (17) Fleys, M.; Thompson, R. W.; MacDonald, J. C. J. Phys. Chem. B 2004, 108, 12197. (18) Fleys, M.; Thompson, R. W. J. Chem. Theory Comput. 2005, 1, 453. (19) Ramachandran, C. E.; Chempath, S.; Broadbelt, L. J.; Snurr, R. Q. Microporous Mesoporous Mater. 2006, 90, 293. € .; Thompson, R. W. Microporous Mesoporous (20) Yazaydin, A.O Mater. 2009, 123, 169. (21) Arı, M.; Ahunbay, M. G.; Yurtsever, M.; Erdem-Senatalar, A. J. Phys. Chem. B 2009, 113, 8073. (22) Snurr, R. Q.; Bell, A. T.; Theodorou, D. N. J. Phys. Chem. 1993, 97, 13742. (23) Makrodimitris, K. G.; Papadopoulos, K.; Theodorou, D. N. J. Phys. Chem. B 2001, 105, 777. (24) Yue, X.; Yang, X. Langmuir 2006, 22, 3138. (25) Ahunbay, M. G. J. Chem. Phys. Vol. 2007, 127, 44707/1. (26) Selassie, D.; Davis, D.; Dahlin, J; Feise, E.; Haman, G.; Sholl, D. S.; Kohen, D. J. Phys. Chem. C 2008, 112, 16521. (27) Beerdsen, E.; Dubbeldam, D.; Smit, B.; Vlugt, T. J. H.; Calero, S. J. Phys. Chem. B 2003, 107, 12088. (28) Rick, S. W. J. Chem. Phys. 2004, 120, 6085. (29) Lisal, M.; Nezbeda, I.; Smith, W. R. J. Phys. Chem. B 2004, 108, 7412. (30) Van Koningsveld, H.; van Bekkum, H.; Jansen, J. C. Acta Crystallogr. Sect. B: Struct. Sci. 1987, 43, 127. (31) Berendsen, H. J. C.; Postma, J. P. M.; van Gunsteren, W. F.; Hermans, J. In Intermolecular Forces; Pullman, B., Ed.; Reidel: Dordrecht, The Netherlands, 1981. (32) Boulougouris, G. C.; Economou, I. G.; Theodorou, D. N. J. Phys. Chem. B 1998, 102, 1029.
ARTICLE
(33) Errington, J. R.; Panagiotopoulos, A. Z. J. Phys. Chem. B. 1998, 102, 7470. (34) NIST Chemistry WebBook (http://webbook.nist.gov/ chemistry). (35) Hill, J.-R.; Sauer, J. J. Phys. Chem. 1994, 98, 1238. (36) Ison, A.; Gorte, R. J. J. Catal. 1984, 89, 159. (37) Vigne-Maeder, F.; Auroux, A. J. Phys. Chem. 1990, 94, 316. (38) Bussai, C.; Vasenkov, S.; Liu, H.; B€ ohlmann, W.; Fritzsche, S.; Hannongbua, S.; Haberlandt, R.; Karger J. Appl. Catal., A 2002, 232, 59. (39) Eroshenko, V.; Regis, R.-C.; Soulard, M.; Patarin, J. C. R. Phys. 2002, 3, 111.
4993
dx.doi.org/10.1021/la200685c |Langmuir 2011, 27, 4986–4993
