J. Phys. Chem. B 2009, 113, 4267–4274
4267
Molecular Simulation of Water/Alcohol Mixtures’ Adsorption and Diffusion in Zeolite 4A Membranes Jian Yang Wu, Qing Lin Liu,* Ying Xiong, Ai Mei Zhu, and Yu Chen National Engineering Laboratory for Green Chemical Productions of Alcohols, Ethers and Esters, Department of Chemical and Biochemical Engineering, College of Chemistry and Chemical Engineering, Xiamen UniVersity, Xiamen, 361005, China ReceiVed: July 5, 2008; ReVised Manuscript ReceiVed: January 8, 2009
The COMPASS (condensed-phase optimized molecular potentials for atomistic simulation studies) force field with two sets of partial atomic charges of water was used to simulate adsorption and diffusion behavior of water/methanol and water/ethanol mixtures in zeolite 4A at 298 K. The adsorption of alcohol first increased and then decreased with increasing pressure, whereas the adsorption of water increased progressively until an adsorption equilibrium was reached. Both the adsorbed molecules and the zeolite framework were treated as a fully flexible model in MD simulations. The simulation results show that the effects of the size and steric hindrance of the diffusing molecules on diffusivity are significant. The diffusivity of water, methanol, and ethanol molecules decreases by 1 order of magnitude in the order of water > methanol > ethanol. The diffusivity of water molecules depends on the mass fraction and the partial charges of water in zeolite 4A. The ethanol and methanol molecules have restricted motion through the R-cages, whereas the water molecules can easily pass through the R-cages window at low feed alcohol concentrations. And the extent of hydrogen bonding increased with increasing water concentration. Introduction Zeolites have been intensively investigated in recent decades because of their unique properties to selectively adsorb and separate molecules from mixtures based on their shape, size, polarity, high selectivity, and high transportability.1-3 For application in separation processes, zeolites have exhibited selectivity as high as 10 000 in the separation of azeotropic water/alcohol mixtures.4 This is consistent with the highly polar nature of water and the strong interaction it shows with zeolite crystal. Zeolite membranes exploit differences in diffusion rates and adsorbed-phase concentrations to differentiate between the different species and separate them. One of the most important industrial zeolites is zeolite 4A. There are two interconnecting, three-dimensional channels in zeolite 4A: (1) connected R-cages or supercages, 11.4 Å in diameter, separated by 4.2 Å apertures, and (2) β-cages or sodalite cages, 6.6 Å in diameter, alternating with the R-cages separated by 2.2 Å apertures.5 There are several reports on the use of zeolite 4A membranes for separating binary gases and liquid mixtures in recent years. It is possible for a given zeolite membrane to be successfully used in fluid separation on the basis of understanding the adsorption and diffusion behavior of feed mixtures. Molecular simulation can provide a useful insight into molecular behavior within zeolite systems that are unavailable from experiments alone. There is an increasing demand for the use of molecular simulation to predict the microproperties of zeolite, especially with the lowering cost of computer simulations. Alten et al.6 developed a molecular model for the adsorption of CO2, N2, and H2, and their mixtures in zeolite 4A. It has been observed that zeolite 4A was strongly selective for CO2 over both N2 and H2 at room temperature. Granato et al.7 used * Corresponding author. E-mail:
[email protected]. Tel: 86-5922183751. Fax: 86-592-2184822.
the configurational-bias Monte Carlo (CBMC) technique to describe the properties of adsorption of propane and propylene onto zeolite 4A. Comparisons between the LTA sodium-free framework and taking the sodium cations and framework partial charges into account have been done. To fit experimental isotherm at 298 K, Jaramillo and Chandross8 studied the adsorption of NH3, CO2, and H2O on zeolite 4A by performing Gibbs ensemble Monte Carlo simulations. They reported that the calculated pressure was about 0.1 kPa in the full hydration states. Faux et al.9,10 performed MD simulations of hydrated zeolite 4A for a range of hydration in which water and Na+ cations dynamics in zeolite 4A were investigated. Furukawa et al.11 investigated the adsorption and diffusion of water/ethanol in zeolite 4A. Jia et al.12 studied the pervaporation separation of liquid mixtures of water/ethanol and water/methanol in three zeolite membranes. It was shown that both selective adsorption and diffusion played an important role in pervaporation-based separations by zeolites. Rutkai and Kristo´f et al.13,14 developed a new effective pair potential parameter for the prediction of the adsorption of mixtures of water/alcohol (methanol and ethanol) in zeolite 4A. However, most developed force fields were only to describe the properties of the adsorption of water/ alcohol mixtures in zeolite 4A but do not include the diffusion behaviors, especially for water/methanol mixtures. It is the purpose of this study to investigate the adsorption/diffusional properties of water/alcohol mixtures in the zeolite 4A framework. MD simulations reported in the literature commonly use the zeolite 4A framework which is treated as rigid or partially rigid in the models. A rigid framework model implies that the framework has negligible effect on the diffusing molecules contained within it. Another aim of this work is to explore the effect of a vibrating framework on the diffusion of molecules within a simulated zeolite matrix. In the present work, the COMPASS force field was used to simulate the adsorption and diffusion behavior of water/methanol
10.1021/jp805923k CCC: $40.75 2009 American Chemical Society Published on Web 03/03/2009
4268 J. Phys. Chem. B, Vol. 113, No. 13, 2009 TABLE 1: Partial Atom Charges on the Water, Methanol, and Ethanol Molecules
and water/ethanol mixtures in zeolite 4A. The preliminary adsorption simulations (results not shown) indicated that the PCFF and CVFF force fields underestimated the equilibrium loadings of water. Furthermore, the PCFF and CVFF force fields do not include the Na+ parameter but only metal sodium. The effect of partial atomic charges of the highly polar water molecules has been taken into account in this work. Particular attention has been paid to deal with electrostatic interactions. We adopted a fully flexible framework model on the basis that few previous studies exist and those which do are for a rigid/ semirigid framework. The target matrix was for zeolite and guest molecules in the MD simulations water/alcohol. The mass fraction dependence of adsorption and diffusion properties was analyzed by GCMC and MD simulations, respectively. Simulated results were then compared with the available experimental data. Simulation Procedures General Simulation Setup. All the simulations were carried out using the Accelrys Material Studio (MS) software simulation package on an SGI Altix work station. The models were built with Visualizer module. The initial structure and topology of water, methanol, and ethanol were generated by the Minimizer submodule of the package, and their electrostatic charges were taken from the previous work,15-18 as summarized in Table 1. The two sets of combination of charges are taken from the
Wu et al. SPE/C and TIP3P water models. The rest of parameters for the simulation are the same as those embedded in the package. Therefore, only atomic charges are different in the (S) and (T) water models. The codes of “water(S)” and “water(T)” denote two sets of partial atomic charges of water used in the present work to explore the effects of the partial atomic charges on the adsorption/diffusion of molecules within a simulated zeolite matrix. Both sets of partial charges of water used in the COMPASS force field are acceptable to calculate the electrostatic energy using Coulombic law; the other combinations (TIP4P, PPC, GCPM, and so on) with the COMPASS force field to describe the interatomic electrostatic interactions underestimated or overestimated the adsorbed amount at equilibrium compared to the “S” and “T” models. A brief discussion of the COMPASS force field and their parameters used in this work are detailed in the Supporting Information. The crystal structure of dehydrated zeolite 4A, stoichiometry Na12Al12Si12O48 per unit cell, was predetermined in the simulation from three-dimensional X-ray diffraction data collected by counter methods reported by Subramanian et al.19 The space group is Pm3m with a lattice parameter a )12.292(2) Å. The schematic diagram of zeolite 4A structure is shown in Figure 1. The 12 Na+ cations are distributed over three crystallographically different sites: eight are on 3-fold axes near the centers of six-rings, three are on eight-ring planes, and the 12th is loosely held in the large cavity on a 2-fold axis opposite a four-ring. The Na+ sites in the zeolite 4A play a very important role in water sorption and transport. The partial atomic charges associated with each atomic species were set as -0.74 for O, +0.80 for Si, +1.42 for Al, and +0.74 for the extra framework cation Na+, which were taken from the work of Dakrim et al.20 Adsorption Simulations. In the present work, the GCMC method21 was employed to simulate the adsorption of water, methanol, and ethanol, and water/methanol and water/ethanol binary mixtures. The simulations were carried out by the Sorption module of the package. The GCMC simulations with periodic boundary conditions were performed for a system consisting of 2 × 2 × 2 unit cells of zeolite NA12-A. The cutoff distance for the calculation of the van der Waals potential energy by atom based technique was taken as 12.0 Å, which was about half the length of the simulation box of 24.584 Å. The electrostatic potential energy was calculated by the Ewald summation technique. The partial pressure required for description of the chemical potential of the adsorbed molecules in the system was calculated using the ideal gas law. The Metropolis algorithm was used in the adsorption simulation. GCMC simulations typically involved performing 1-5 million MC events to equilibrate the system and subsequently a similar length of simulation was used to collect data. Here, the equilibration steps were set to 1 000 000 and the production steps were set to 2 000 000 Monte Carlo moves. Diffusion Simulations. Standard MD simulations with the COMPASS force field and periodic boundary conditions used in the case of the adsorption simulations were performed to study the diffusion of water, methanol, and ethanol, and water/ methanol and water/ethanol binary mixtures in the zeolite using the Discover module in the Materials Studio. The adsorbent model and the calculation of nonbonded interaction are the same as those used in the previous adsorption simulations. Before the production run itself was performed, a minimization and an equilibration run were carried out to bring the system to the most probable configuration. The smart minimizer method was used to perform the energy minimization with fine convergence level. Simulations were carried out in the NVT
Adsorption of Water/Alcohol Mixtures in Zeolite 4A
J. Phys. Chem. B, Vol. 113, No. 13, 2009 4269
Figure 1. (Left) Schematic diagram of 2 × 2 × 2 unit cells of zeolite 4A structure and the free volume (about 4910.46 Å3 with 1.0 Å Connolly radius). (Center) The unit cell of zeolite 4A structure, the occupancy factor of Na1, Na2, and Na3 is 1, 1/4, and 1/12, respectively.19 (Right) One water molecule is near the β-window of β-cage and is normally impeded from diffusing into the β-cage because the Na1 cation is in the middle of the window.
ensemble (constant loading, constant volume, and constant temperature) dynamics using the velocity Verlet algorithm.22 The temperature was controlled using Nose´ method.23,24 A single simulation was undertaken with a 1 fs time step and a 0.5 fs time step to show whether the femtosecond time step impacts the result. In order to examine the effect of lattice vibration as a driving force for sorbate diffusion in zeolite 4A, three types of 4A zeolite models, which are (1) a fully rigid model where all atoms and Na+ cations of zeolite 4A are fixed, (2) a partially rigid model where all atoms except Na+ cations are fixed, and (3) a fully flexible model where all atoms and Na+ cations are treated to move freely, have been employed to simulate the diffusion of five water molecules per unit cell in zeolite 4A. Both the zeolite framework and guest molecules were also treated as a fully flexible model in which they were free to alter their internal configurations. The total simulation time was about 1200 ps, and the trajectories were saved every 1000 steps for further analysis. Several quantities and functions that could aid one in understanding the behavior of guest molecules were evaluated in the simulations. Of structural properties, the guest-guest molecules, guest-zeolite pair correlation functions, dynamical trajectories, and three-dimensional concentration profiles of guest molecules in the zeolite system can be obtained. For the dynamic properties, the mean square displacement (MSD) as a function of time is calculated by
〈X2(t)〉 )
1 NmNt0
∑ ∑ [Xi(t + t0) - Xi(t0)]2 m
(1)
t0
where Nm is the number of diffusing molecules, Nt0 is the number of time origins used in calculating the average, and Xi is the coordinate of the center of mass of molecule i. The self-diffusion coefficients of sorbate can be calculated from the MSD using the well-known Einstein relation:
〈X2(t)〉 ) 6Dt + B
(2)
where D is the self-diffusion coefficient and B is the thermal factor arising from atomic vibration. Results and Discussion Adsorption Properties. Single-Component Adsorption. Zeolite 4A is a hydrophilic crystal and can be used to separate relatively small amounts of water from large amounts of alcohols without unreasonable energy consumption. Figure 2 shows the adsorption isotherms of water and alcohols (methanol and ethanol) in zeolite 4A, respectively, along with the experimental
Figure 2. Experimental and simulated adsorption isotherms of water, ethanol, and methanol (from top to bottom) on zeolite 4A at 298 K.
data,25 the calculated values of Furukawa et al.11 and Kristo´f et al.14 at 298 K. Both sets of partial atomic charges of water provided saturation of water for relative activities (p/ps) higher than 0.1 at 298 K. The adsorption of water(T) from p/ps ) 0.03 to 0.3 was slightly lower than the experimental data, and the saturation adsorption of water(S) was appreciably greater than the experimental data. But at p/ps < 0.1, Furukawa’s calculated data for water were higher than the experimental data as well as our simulations. Our simulation results are only slightly higher than the experimental data when the adsorption equilibrium is achieved. It is probably that water, which has a larger kinetic diameter, cannot enter into the β-cages. However, in MC simulations insertion of a water molecule is made at random in volume V of the simulation cell, and deletion of a water molecule is made by selecting, at random, one molecule among the molecules present in the cell. These trial MC moves are accepted or rejected according to the Metropolis rule. At relative saturation pressure (p/ps ) 1), there are an average of 3.5
4270 J. Phys. Chem. B, Vol. 113, No. 13, 2009
Figure 3. Experimental and simulated isosteric heat of adsorption for methanol as a function of coverage on zeolite 4A at 298 K.
Figure 4. Adsorption isotherm of water/alcohol mixture at different alcohol concentrations in the range of 10-95 wt % with pressure ranging from 0.01 to 100 kPa in the zeolite 4A at 298 K.
water(T) molecules/β-cage and 3.875 water(S) molecules/βcage, similarly noted by Jaramillo and Chandross8 (4.1 molecules/ β-cage) and Faux et al.26 (3.75 molecules/β-cage). These results could slightly affect the quantitative comparison between experimental and simulated data. This is probably because the potential barrier between the R-cages and the β-cages during the experiments prevented the water molecules from entering the internal volume of the β-cages. Figure 2 shows that different partial atomic charges within the COMPASS force field can affect the interaction between water molecules and zeolite framework due to electrostatic interactions. Saturation of ethanol was reached at a very low relative saturation pressure, while that of methanol was reached at a pressure higher than 0.05, which in both cases were all consistent with the experiments. The simulated value of methanol was slightly lower than the experimental observation. This may be explained by the electrostatic interaction between methanol and the zeolite framework being weaker. Kristo´f’s calculated data was higher than the experimental data. But the errors of the present work were smaller than those of Kristo´f. The experimental data of ethanol was lower than our simulation results and Furukawa’s calculated data. It may be that the interaction between ethanol
Wu et al.
Figure 5. MSDs (x, y, and z directions, and total) of (a) water(S) and (b) water(T) molecules as a function of time with different femtosecond time steps (a 1 fs time step and a 0.5 fs time step) for five water molecules per unit cell at 298 K.
Figure 6. MSDs of (a) water(S) and (b) water(T) molecules as a function of time at a loading of five water molecules per unit cell for a fully rigid model (all atoms of zeolite 4A being fixed), a partially rigid model (all atoms except Na+ cations being fixed), and a fully flexible model (all atoms being treated to move freely) at 298 K.
Figure 7. Arrhenius plots of the self-diffusivities of water for 5 and 15 water molecules per unit cell (a comparison of our simulation results to the experimental data reported in the literature32).
and zeolite 4A is stronger than that in reality, and it is apparent that ethanol molecules cannot easily enter into in the R-cages while in the MC simulations insertion of an ethanol molecule according to the Metropolis rule is not a physical process. Table
Adsorption of Water/Alcohol Mixtures in Zeolite 4A
Figure 8. Self-diffusion coefficients of water(S) and water(T) as a function of loading in zeolite 4A at 298 K.
Figure 9. MSD of methanol molecules as a function of time at different loadings in log-log coordinates at 298 K.
J. Phys. Chem. B, Vol. 113, No. 13, 2009 4271
Figure 10. MSD of ethanol molecules as a function of time at different loadings in log-log coordinates at 298 K.
Figure 11. Self-diffusion coefficients of water as a function of time at different water concentrations in the range of 5-90 wt % at 298 K.
TABLE 2: Adsorption of Water at Different Feed Alcohol Concentrations in Zeolite 4A at 298 K sorption (g water/g zeolite) solution ethanol (0.1 wt %) ethanol (10 wt %) ethanol (25 wt %) ethanol (50 wt %) ethanol (75 wt %) ethanol (80 wt %) ethanol (90 wt %) ethanol (92 wt %) ethanol (95 wt %) methanol (10 wt %) methanol (25 wt %) methanol (50 wt %) methanol (75 wt %) methanol (90 wt %) methanol (95 wt %)
water(S)
water(T)
0.26 0.22 0.18 0.16
0.23 0.21 0.17 0.16
0.14
0.15
0.12 0.31 0.29 0.28 0.24 0.19 0.15
0.12 0.29 0.27 0.27 0.24 0.17 0.13
expt 0.6 ( 0.0630 0.36 ( 0.0630 0.18 ( 0.0530 0.16 ( 0.0430 0.1131
1 shows that the absolute values of partial atomic charges on ethanol molecules are greater than those on methanol molecules. But all the errors were nearly within 10%. This result may reflect the limitation of the COMPASS force field for description of the system being studied, since no force field parametrization was performed specifically for the given system. Figure 3 illustrates the isosteric heat of adsorption for methanol as a function of coverage against the experimental data27 at 298 K. The simulated values were lower than the experimental data, whereas the two curves behaved similarly.
It is noteworthy that the isosteric heat of adsorption for methanol in the limit of infinite dilution is larger, which implies that the methanol-zeolite interaction is stronger in the preferred adsorbing sites. The calculated isosteric heat of adsorption for water was 90 ( 15 kJ · mol-1 against the experimental data 100 ( 25 kJ · mol-1 at 298 K.5 The isosteric heat of adsorption for ethanol in our simulation was 90 ( 10 kJ · mol-1 at 298 K, and the simulated value of Rutkai et al.14 was 100 ( 15 kJ · mol-1 at 378 K. Both calculated values are larger than those of the experimental data,28 probably due to the strong interaction of ethanol-thanol and ethanol-zeolite framework. Binary Mixture Adsorption. Binary mixture adsorption is important for practical application but is inconvenient to measure experimentally. It is interesting to simulate the adsorption isotherms of water/alcohol mixtures at various alcohol concentrations. The adsorption isotherms of water/ethanol and water/ methanol mixtures from 10 to 95 wt % alcohols at 298 K are shown in Figure 4. All these results show that the adsorption of alcohols increased up to a maximum and then slowly decreased with increasing pressure. It can be observed that the adsorption of alcohols had a peak around a total pressure of 1.0 kPa. The reason may be that the entropy plays an important role.29 At the same pressure, the adsorption of alcohols increased with increasing the alcohol concentration; similarly, the adsorption of water increased with increasing the water concentration. As to the water, the adsorption increased with increasing the pressure. The adsorption of water(S) was slightly larger than
4272 J. Phys. Chem. B, Vol. 113, No. 13, 2009
Figure 12. MSD of water in the zeolite 4A as a function of time at different feed water concentrations in the range of 5-90 wt % at 298 K.
Figure 13. Distribution profiles of water molecules at different feed ethanol concentrations in the range of 5-90 wt % in the zeolite 4A by MD simulations at 298 K (C(x), C(y), and C(z) denote concentration distribution in three different directions).
that of water(T). The adsorption of water on zeolite 4A in different alcohol mixtures is listed in Table 2 against the available experimental data.30,31 It is shown that the calculated values of water were in good agreement with the experimental data at high feed ethanol concentrations. The experimental adsorption of water in 0.1 wt % ethanol was much higher than our simulation result, as noted by Shah et al.30 Diffusion Properties. Single-Component Diffusion. In order to enhance the understanding of diffusion behavior of molecules within zeolite 4A’s channels and pores, the diffusion of water and alcohol (ethanol and methanol), and their binary mixtures in zeolite 4A was studied. Figure 5 shows the MSDs of water(S) and water(T) molecules as a function of time for five water molecules per unit cell at 298 K, respectively. With both water and zeolite being treated as a fully flexible model, the MSD vs time obtained has a slight difference between a 1 fs time step and a 0.5 fs time step simulations, suggesting that the 1 fs time step is sufficient to yield consistency for MSD. The calculated results (Figure 5) suggest that the fs time steps (1 fs time step and 0.5 fs time step) almost do not impact the result. Therefore, a 1 fs time step was used in further simulations in order to save time. The
Wu et al.
Figure 14. Distribution profiles of water molecules at different feed methanol concentrations in the range of 5-90 wt % in the zeolite 4A by MD simulations at 298 K (C(x), C(y), and C(z) denote concentration distribution in three different directions).
MSDs of water(S) and water(T) as a function of time at the loadings of five water molecules per unit cell for fully rigid, partially rigid, and fully flexible frameworks at 298 K are displayed in Figure 6. Compared to the MSD for the fully rigid framework, the partially rigid framework and the fully flexible framework, the fully flexible framework exhibits strongly enhanced water mobility. It is apparent that a flexible lattice and movable Na+ cations can enhance diffusion by reducing the energy barrier for passing events. From Figure 6 the effect of lattice vibration cannot be negligible. Arrhenius plots of the self-diffusivities of water for 5 and 15 water molecules per unit cell against the experimental data are presented in Figure 7. The simulation results were generally in the same order of magnitude with the experimental data.32 The self-diffusion coefficients increased with increasing temperature. There is no great difference between the water(S) and water(T) with five water molecules per unit cell. The self-diffusion of water(T) was slightly larger than that of water(S) with 15 water molecules per unit cell. Meanwhile, the self-diffusion coefficients of water as a function of loading in the zeolite at 298 K are shown in Figure 8. The self-diffusion of water(T) was larger than that of water(S). All these findings indicate that the diffusivity of the adsorbed water molecules in zeolite 4A is influenced by the partial atomic charges of water. Both exhibited a maximum around 8-10 mmol · g-1 loadings. All the results show that self-diffusion coefficients of water reached a maximum and then slowly decreased with increasing loading, which behaved similarly to the diffusion of methane in LTA zeolite, as described by Beerdsen et al.33 Figures 9 and 10 show the MSDs of methanol and ethanol molecules as a function of time in log-log coordinates under different loadings at 298 K, respectively. Diffusion coefficients of methanol and ethanol molecules in zeolite were estimated from the linear region of the log(MSD) vs log(t) plots. The calculated self-diffusion coefficients of methanol with loadings of 1.25, 2.5, 5, 6.25, and 7.5 molecules per unit cell were 2.02 × 10-11, 2.24 × 10-11, 2.11 × 10-11, 1.26 × 10-11, and 2.28 × 10-11 m2 · s-1, respectively. Unfortunately, there are no reports on experimental or simulated data for the self-diffusivity of methanol in zeolite 4A. The calculated diffusivity of methanol was about 1 order of magnitude lower than that of water. The
Adsorption of Water/Alcohol Mixtures in Zeolite 4A
J. Phys. Chem. B, Vol. 113, No. 13, 2009 4273
Figure 16. H(water)-O(alcohol) and O(water)-H(alcohol hydroxyl) intermolecular pair distribution functions at 10, 50, and 90 wt % alcohol concentrations in zeolite 4A by MD simulations at 298 K. Intermolecular pair correlation functions are presented as they indicate all the hydrogen bondings between molecules (water-water, water-alcohol, and alcohol-alcohol). Figure 15. Typical trajectories of water molecules at 10, 50, and 90 wt % alcohol concentrations in zeolite 4A at 298 K.
self-diffusion coefficients for ethanol with 0.625, 1.875, 2.5, 3.125, and 3.75 molecules per unit cell showed values of 2.13 × 10-12, 1.27 × 10-12, 3.17 × 10-12, 2.53 × 10-12, and 2.78 × 10-12 m2 · s-1 respectively, which were in quite good agreement with the value reported by Pera-Titus et al.,28 (1.15 ( 0.29) × 10-12 m2 · s-1 at zero loading. The diffusivity of ethanol was about 1 order of magnitude lower than that of methanol. This suggests that the kinetic diameter of molecules should be partly responsible for the diffusivity of the adsorbed species. Binary Mixture Diffusion. The MSD of water in the zeolite 4A as a function of time at various feed alcohol concentrations in the range of 10-95 wt % at 298 K is given in Figure 10. The corresponding self-diffusion coefficients of water in the zeolite 4A as a function of time at various feed alcohol concentrations in the range of 10-95 wt % at 298 K are displayed in Figure 11. From Figure 11a and Figure 12a,b, it is observed that the diffusivity of water decreased considerably with decreasing water content in the water/ethanol mixture. The trends can be interpreted as increasing ethanol concentration leads to a reduction of volume for individual water molecule in the zeolite system and consequently a reduction of translational mobility. In addition, steric hindrance between adsorbed molecules passing each other should be taken into account. From Figure 11b and Figure 12c,d, one can note that the diffusivity of water increased considerably with increasing water content in the range of 5-75%. The diffusivity of water did not increase or otherwise decrease afterward with increasing feed water concentration. This result may be explained by that entropy effects should be taken into account and some water clusters may be formed. Moreover, the distribution profiles of water molecules at various feed alcohol contents in the zeolite 4A are shown in Figures 13 and 14, respectively. The water molecules have lower mobility at high alcohol contents than at low alcohol contents. It is interesting to point out that the diffusivity of water(S) is higher than that of water(T) at low water contents, but is reverse at high water contents. The reason may be that the molecular interactions between water(T) and alcohol are stronger than those between water(S) and alcohol.
In order to explain further the diffusion behavior of each component in a mixture at various feed alcohol concentrations, a typical trajectory of the component molecules is shown in Figure 15. It can be seen that the ethanol and methanol molecules are impeded from diffusing through the R-cages window due to the large kinetic diameter of the both molecules, whereas a much wider area is covered by them in their diffusion path at low feed alcohol concentrations. This may be explained by that there are weaker steric influences involved with decreasing alcohol concentration in zeolite 4A and the faster water molecules can speed up the slower alcohol molecules. It is also seen that the water molecules can easily pass through the window at 10 and 50 wt % feed alcohol concentrations. At 90 wt % feed alcohol concentration, the water(T) molecules cover more narrow area than the water(S) molecules. This implies that the water(T) molecules diffuse more slowly than the water(S) molecules. All these findings are in agreement with the calculated diffusion coefficients in Figure 11. Analysis of radial distribution functions (RDFs) can help one understand microscopic behavior of molecules adsorbed in zeolites, such as the average intermolecular distance and the presence of hydrogen bonding. The plots of gOH(r) for the zeolite 4A are shown in Figure 16. Oxygen and hydrogen atoms are on the hydroxyl group of water and alcohol. The typical RDF data were obtained by MD simulations performed at 10, 50, and 90 wt % water concentrations at 298 K. Figure 16 shows that the first peaks around 1.8 Å in gOH(r) are a clear indication of the existence of stable hydrogen bonding. The strength of hydrogen bonding increased with increasing water concentration. The second peaks of gOH(r) around 3.0-4.0 Å, which are lower than the first ones, represent mostly the population of nextnearest neighbors within the cage. The peaks of gOH(r) around 9.0 Å represent the population in adjacent cages. Figure 16d shows that the second peak of gOH(r) at 10 wt % water(T) concentration being much broadened is significantly different from the others. This suggests that the adsorbed molecules within the cage interact through weak hydrogen bonding. Conclusions In the present work, water/alcohol mixtures adsorption and diffusion properties in zeolite 4A at 298 K were studied by
4274 J. Phys. Chem. B, Vol. 113, No. 13, 2009 GCMC and MD simulations using the COMPASS force field, respectively. The simulated mass fraction dependent adsorption isotherms show that the adsorption of alcohol first increased and then decreased with increasing the pressure, and the adsorption of water increased progressively until the adsorption equilibrium was reached. Large equilibrium selectivity for water in accord with experimental observation was observed. The effect of partial charges on the adsorbed water molecules can be noted. The adsorption of water(S) is slightly larger than that of water(T) due to stronger electrostatic interactions. The MD simulation results show that the effects of shape and size of the diffusing molecules on diffusivity were significant. The diffusivities of water, methanol, and ethanol molecules decreased 1 order of magnitude in the order of water > methanol > ethanol. The diffusivity of water molecules in zeolite 4A at 298 K depended on the mass fraction and the partial charges of water. The diffusivity of the adsorbed water molecules in zeolite 4A decreased with increasing feed ethanol concentration, whereas the diffusivity of the adsorbed water molecules in zeolite 4A increased and then decreased with decreasing feed methanol concentration. At high feed alcohol concentrations, the diffusivity of the adsorbed water(S) molecules is much lower than that of the adsorbed water(T) molecules in zeolite 4A. The ethanol and methanol molecules have restricted motion through the R-cages, as observed in their typical trajectories. The water molecules can easily pass through the window at low feed alcohol concentrations. The radial distribution functions show that the first peaks in gOH(r) around 1.8 Å are clearly indicative of the presence of stable hydrogen bonding. The strength of hydrogen bonding increased with increasing water concentration. Acknowledgment. The support of National Nature Science Foundation of China Grant no. 50573063, the Program for New Century Excellent Talents in University, and the research fund for the Doctoral Program of Higher Education (no. 2005038401) in the preparation of this article is gratefully acknowledged. The authors also thank Dr. Ian Broadwell (Xiamen University) for his technical assistance in editing this manuscript for publication. Supporting Information Available: A brief discussion of the COMPASS force field and their parameters used in this work. This material is available free of charge via the Internet at http://pubs.acs.org.
Wu et al. References and Notes (1) Pera-Titus, M.; Llorens, J.; Cunill, F.; Mallada, R.; Santamarı´a, J. Catal. Today 2005, 104, 281. (2) Li, S.; Tuan, V. A.; Falconer, J. L.; Noble, R. D. Ind. Eng. Chem. Res. 2001, 40, 1952. (3) Caro, J.; Noack, M.; Kolsch, P.; Schafer, R. Microporous Mesoporous Mater. 2000, 38, 3. (4) Morigami, Y.; Kondo, M.; Abe, J.; Kita, H.; Okamoto, K. Sep. Purif. Technol. 2001, 25, 251. (5) Breck, D. W. Zeolite molecular sieVes: structure, chemistry, and use; R. E. Krieger: Malabar, FL, 1984; pp 83-90, 593, 634-636, 654. (6) Akten, E. D.; Siriwardane, R.; Sholl, D. S. Energy Fuels 2003, 17, 977. (7) Granato, M. A.; Vlugt, T. J. H.; Rodrigues, A. E. Ind. Eng. Chem. Res. 2007, 46, 321. (8) Jaramillo, E.; Chandross, M. J. Phys. Chem. B 2004, 108, 20155. (9) Faux, D. A. J. Phys. Chem. B 1998, 102, 10658. (10) Faux, D. A. J. Phys. Chem. B 1999, 103, 7803. (11) Furukawa, S.-I.; Goda, K.; Zhang, Y.; Nitta, T. J. Chem. Eng. Jpn. 2004, 37, 67. (12) Jia, W.; Murad, S. Mol. Phys. 2006, 104, 3033. (13) Rutkai, G.; Csa´nyi, E´.; Kristo´f, T. Microporous Mesoporous Mater. 2008, 114, 455. (14) Kristof, T.; Csanyi, E.; Rutkai, G.; Merenyi, L. Mol. Simul. 2006, 32, 869. (15) Noskov, S. Y.; Lamoureux, G.; Roux, B. J. Phys. Chem. B 2005, 109, 6705. (16) Mahoney, M. W.; Jorgensen, W. L. J. Chem. Phys. 2000, 112, 8910. (17) Plant, D. F.; Maurin, G.; Bell, R. G. J. Phys. Chem. B 2007, 111, 2836. (18) Kusalik, P. G.; Svishchev, I. M. Science 1994, 265, 1219. (19) Subramanian, V.; Seff, K. J. Phys. Chem. 1977, 81, 2249. (20) Darkrim, F.; Aoufi, A.; Malbrunot, P.; Levesque, D. J. Chem. Phys. 2000, 112, 5991. (21) Daan, F.; Berend, S. Understanding molecular simulation: From algorithms to applications; Academic Press, Inc.: New York, 1996; p 443. (22) Verlet, L. Phys. ReV. 1967, 159, 98. (23) Nose´, S. Mol. Phys. 1984, 52, 255. (24) Nose´, S. Mol. Phys. 2002, 100, 191. (25) Okamoto, K.-I.; Kita, H.; Horii, K.; Kondo, K. T. Ind. Eng. Chem. Res. 2001, 40, 163. (26) Faux, D. A.; Smith, W.; Forester, T. R. J. Phys. Chem. B 1997, 101, 1762. (27) Izmailova, S. G.; Khvoshchev, S. S.; Shubaeva, M. A. Russ. Chem. Bull. 1992, 41, 1960. (28) Marc, P.-T.; Joan, L.; Javier, T.; Fidel, C. Catal. Today 2006, 118, 73. (29) Lu, L.; Wang, Q.; Liu, Y. Langmuir 2003, 19, 10617. (30) Shah, D.; Kissick, K.; Ghorpade, A.; Hannah, R.; Bhattacharyya, D. J. Membr. Sci. 2000, 179, 185. (31) Bindal, R. C.; Misra, B. M. Sep. Sci. Technol. 1986, 21, 1047. (32) Paoli, H.; Me´thivier, A.; Jobic, H.; Krause, C.; Pfeifer, H.; Stallmach, F.; Ka¨rger, J. Microporous Mesoporous Mater. 2002, 55, 147. (33) Beerdsen, E.; Dubbeldam, D.; Smit, B. J. Phys. Chem. B 2006, 110, 22754.
JP805923K
