Conformational Changes of trans-1, 2-Dichlorocyclohexane Adsorbed

Mar 20, 2012 - Department of Chemistry, The University of Western Ontario, London, ... of Chemical Technology, Taiyuan, Shanxi, 030024, People's Repub...
7 downloads 0 Views 4MB Size
Article pubs.acs.org/JPCC

Conformational Changes of trans-1,2-Dichlorocyclohexane Adsorbed in Zeolites Studied by FT-Raman Spectroscopy and Molecular QM/MM Simulations Andrei Buin,† Jianrong Ma,‡ Yining Huang,*,† Styliani Consta,*,† and Zhang Hui† †

Department of Chemistry, The University of Western Ontario, London, Ontario, Canada N6A 5B7 Shanxi Institute of Chemical Technology, Taiyuan, Shanxi, 030024, People’s Republic of China



S Supporting Information *

ABSTRACT: The conformational behavior of trans-1,2-dichlorocyclohexane (T12D) adsorbed inside zeolites with Faujasite structure (FAU) including sodium Y (Na−Y) and siliceous Y (Si−Y) was investigated by FT-Raman spectroscopy and molecular simulations. The results have clearly shown that the conformational and dynamic properties of T12D strongly depend on the presence of charge-balancing cations as well as Si/Al ratio. Upon loading into Na−Y, the population of the diequatorial (ee) conformer increases compared with pure T12D liquid due to the strong interaction with the extra-framework Na+ ions. Molecular simulations of T12D in Na−Y and Si−Y have also been carried out. The T12D molecule was modeled by quantum and semiempirical quantum chemistry methods and embedded in a zeolite framework that was described by empirical force field. Conformational changes were sampled using quantum mechanics/molecular mechanics molecular dynamics and replica exchange molecular dynamics. Molecular simulations revealed that the ee conformer is preferable versus the diaxial (aa) conformer in both Na−Y and Si−Y frameworks. However, in the Na−Y supercage environment, the Na+ ions polarize the ee conformer stronger than in Si−Y. This leads to a larger shift of the conformational equilibrium in favor of the ee component in Na−Y relative to Si−Y. Computations of the equilibrium population of aa and ee conformers using the dihedral Cl−C−C−Cl angle distributions of aa and ee showed good quantitative agreement with the experimental findings with respect to the dominant ee conformer in both Na−Y and Si−Y.

I. INTRODUCTION Zeolite molecular sieves are microporous aluminosilicate framework materials containing channels and cavities with molecular dimensions.1,2 They are widely used in industry as ion exchangers, sorbents, and catalysts. In the past decade, the increased awareness of the environmental hazards that are caused by chlorinated halocarbons has led to the development of new separation and catalytic conversion processes that utilize zeolites.3−8 The effectiveness of zeolites is based on its porous nature and on the distinct interactions between the zeolitic hosts and adsorbed guest species. Among the chemical and physical transformations that a zeolite may induce to a guest molecule, in this article we consider conformational changes of the adsorbed guest molecule due to electrostatic interactions with the zeolite framework. The existence of different conformers of the guest molecule is one of the factors that affects the separation efficiencies of zeolites. Different conformers can have different electric dipole and quadrupole moments, and these molecular properties can affect the heats of adsorption and diffusion of the guest molecules inside the zeolite hosts. In this study, we examine the conformational behavior of adsorbed trans-1,2-dichlorocyclohexane (T12D) inside zeolites © 2012 American Chemical Society

with Faujasite (FAU) structure including sodium Y (Na−Y) and siliceous Y (Si−Y) using Fourier transform (FT)-Raman spectroscopy and molecular simulations. Previous study of trans-1,4-dichlorocyclohexane (DCC)9 and present study on T12D in FAU discuss two distinct cases of the effect of charge distribution on the same molecular frame in the conformational equilibrium inside FAU zeolites. In the former study, analysis of the thermodynamics from the FT-Raman spectra revealed that the conformational equilibrium is determined almost exclusively by entropy. This is attributed to the fact that both conformers of DCC have zero dipole moment; therefore, the interactions with the zeolitic framework are van der Waals (vdW) and ion-quadrupole moment nature. In the present study, T12D interacts with the zeolitic framework by ion− dipole interactions. The strong electrostatic interactions lead to equilibrium constant between the aa and ee conformers that it is mainly determined by enthalpy change between the conformers. Regarding practical applications, T12D has been Received: December 30, 2011 Revised: March 7, 2012 Published: March 20, 2012 8608

dx.doi.org/10.1021/jp212616h | J. Phys. Chem. C 2012, 116, 8608−8618

The Journal of Physical Chemistry C

Article

moments closer to experimental and high-quality ab initio values.30 RM1 uses a larger training set in the parametrization than AM1 but for chlorinated compounds shows larger deviations in the dipole moment,30 and specifically for chlorinated hydrocarbons it also shows large deviations in solvation free energies.31 Even though we are aware of the deficiencies of RM1 relative to AM1, we also performed simulations with RM1 to show the dependence of the findings on parameters. The zeolite framework was described by empirical force field. We note that the QM/MM approach has been widely used for biological systems; however, it is only recently that this method is explored as an appealing alternative to describe hydrocarbon− zeolite interactions.32 In this study, simulations were performed at low loading for several tens of nanoseconds, which is the longest time that such systems have been simulated so far. Simulations showed that the ee conformer is predominant in both Na−Y and Si−Y. Population analysis showed that the ee conformer is more populous in Na−Y than in Si−Y, in agreement with the experimental findings. The higher population of ee than aa in Na−Y arises from its larger dipole moment and consequently its stronger interactions with the extra-framework Na+ ions. The success of QM/MM using AM1 for the modeling of the quantum subsystem in predicting good quantitative agreement with experimental data provides a practical alternative to empirical force field simulations for low loadings of the adsorbed molecule.

identified as intermediate in the degradation of pentachlorophenol,10 which has been used widely as bactericide and wood preservative. The conformational equilibrium of T12D in various media (vapor, liquid, solid, solutions) has been investigated.11−15 T12D exists in both aa and ee forms, and their relative population depends on the medium that it is found. The ee isomer is predominant in solid phases, whereas in vapor phase, the aa conformer is the major component. In pure liquid at room temperature, the two conformers are in approximately equal amounts. Polar solvents stabilize the ee conformer, which has a larger dipole moment. In nonpolar solvents, the aa isomer is more populous. The conformational behavior of T12D inside the zeolites was monitored by FT-Raman spectroscopy. Raman spectroscopy is a preferred technique for host−guest interaction in zeolites. This is because zeolites are very weak Raman scatters.16−18 Therefore, the Raman spectra obtained are mainly those of the adsorbed organic molecules. Raman has a fast time scale, which allows simultaneous observation of different conformational isomers under ambient conditions. Compared with conventional visible Raman, FT-Raman technique can significantly reduce the fluorescence background (a major problem associated with zeolites) due to utilization of the 1064 nm excitation line from a near-infrared laser. With respect to computations, adsorption of chloromethanes3,7,19 and fluoroethanes20,21 have been studied in Na−X and Na−Y zeolites. In the present study, molecular details of the conformational equilibrium between the aa and ee conformers of T12D were monitored by direct molecular dynamics (MD) and replica exchange MD (REMD)22−25 simulations. REMD is a Monte Carlo (MC) method that allows for dramatic configurational changes in the system by exploiting configurations at elevated temperatures. Even though many simulations on the adsorption of acyclic hydrocarbons and substituted hydrocarbons in zeolites have been reported in the literature,3,7,19−21 to the best of our knowledge, there are no molecular simulations of conformational changes of cyclic aliphatic chlorohydrocarbons. A stumbling-block in these simulations is the modeling of the interactions between the zeolite and the conformers of T12D. In the gaseous state, the aa and ee have dipole moments 1.39 and 3.74 D, respectively. T12D has high polarizability (14.36 Å3 for aa vs 14.70 Å3 for ee in the gaseous state);26 therefore, it is expected that in the polar environment of the zeolite, T12D may acquire considerable induced dipole moment. The strength of the interactions with the extra-framework Na+ ions, which in turn is determined by the charge distribution of T12D, plays a major role in the conformational equilibrium between aa and ee. Therefore, to be able to compare the simulation data with the experimental findings, the molecular model has to capture the induced polarization effect and the changes in the dipole moment of aa and ee in the local minima of the zeolite cages. The importance of polarization of the molecule depending on the zeolite framework has been discussed in the literature.27,28 To allow for variable charge distribution of T12D, we used the quantum mechanics/molecular mechanics (QM/MM) approach for the modeling of the system. T12D was treated as the quantum subsystem and described by quantum chemistry methods such as semiempirical AM129 and RM1,30 density functional theory (DFT), and Moller−Plesset (MP2) perturbation theory. AM1 was chosen because it has been found that for chlorinated hydrocarbons AM1 estimates the dipole

IIA. EXPERIMENTAL SECTION Na−Y (Si/Al = 2.35) and Si−Y (Si/Al = 100) were obtained from Strem Chemicals and Degussa Chemical, respectively. T12D was obtained from Aldrich Chemical and used without purification. The identity, purity, and crystallinity of the zeolite samples were checked by powder X-ray diffraction. Carefully measured aliquots of T12D were added to accurately weighed dehydrated zeolite powders in a glass tube. The tubes were carefully sealed in a flame and then placed in an oven for 3 h at 67 °C for T12D/Na−Y and T12D/Si−Y to disperse the sorbate molecules uniformly throughout the samples. The loadings were 1 molecule/supercage (s.c.) for both Na−Y and Si−Y systems. All Raman measurements were made on a Bruker RFS 100/S spectrometer equipped with a Nd3+/YAG laser operating at 1064.1 nm and liquid-nitrogen-cooled Ge detector. The laser power was typically 80 mW at the sample, and the resolution was 2 cm−1. Low-temperature measurements were carried out by using a Bruker Eurotherm 800 series temperature control unit, which regulated the sample temperature within ±1 °C. Powder XRD patterns were recorded on a Rigaku diffractometer equipped with a graphite monochromator using a Co Kα radiation (a wavelength of 1.7902 Å). IIB. MODELING OF T12D IN NA−Y AND SI−Y A. Modeling of the FAU Framework. T12D molecule in FAU-type zeolites was simulated. Several articles have provided detailed descriptions of the FAU-type zeolite framework. (See, for example, refs 27, 29, 30, and 33.) In general, the FAU structure is composed of a network of cavities with a diameter of ∼12.5 Å that are called supercages and smaller sodalite cages (β cages) with a diameter of ∼6.5 Å (Figure 1). The β cages are arranged in tetrahedral structures where any β cage is found at the center of a tetrahedral defined by four other neighboring β cages. The β cages are connected to each other via hexagonal prisms (also known as double six-rings). The supercages are 8609

dx.doi.org/10.1021/jp212616h | J. Phys. Chem. C 2012, 116, 8608−8618

The Journal of Physical Chemistry C

Article

advantage that in simulations within a unit cell represents in an average manner different spatial distributions of aluminum sites over a number of unit cells. In the modeling we employed, the interactions between the Na+ ions and the oxygen sites of the zeolite framework were described by the Jaramillo−Auerbach (JA) potential function20,45 expressed as U (rij) = ANaO exp( −rij/ρ NaO) −

CNaO rij6

+

ZOZNa rij (1)

+

for a pair i,j of Na and oxygen sites, respectively, that are found at a distance rij. ZO and ZNa represent the partial charges of the oxygen and Na+ sites, respectively. In eq 1, the first two terms is the Buckingham potential function and the third term is the Coulomb interaction between the Na+ ions and the oxygen sites in the zeolite. The parameters are presented in Table 1. Table 1. Parameters of the JA Potential Function in Equation 1 for 57Na−Y and Partial Charges for 57Na−Y and Si−Y

Figure 1. Schematic picture of the structure of Na−Y framework. Typical extra-framework Na+ ion positions are indicated by the black spheres.

57Na−Y

formed by the β-cage structures, and they are connected to each other via 12-ring windows of a diameter of ∼7.4 Å. Every supercage has four 12-ring windows. Depending on the Si/Al ratio, the framework composed of Si, Al, and O sites carries an amount of negative charge that is counter-balanced by extraframework cations such as Na+, K+, or Ba2+ ions. One unit cell is composed of eight supercages and eight β cages. The unit cell that corresponds to Si/Al ratio equal to 2.35 with Na+ as counterions has a composition NaxAlxSi192−xO384, where x = 57.34 This type of zeolite is denoted as 57Na−Y. The Na−Y framework has mainly three types of sites that can be occupied by the cations (Figure 1). Sites I are within the hexagonal prism. Sites I′ are located inside the sodalite cages adjacent to Sites I. Sites II are within the supercages. The Na+ ion sitting at this site coordinates with three framework oxygen atoms from the six-ring window of the β cage. Within each unit cell, there are 32 sodium ions at Sites II, which corresponds to four sodium ions per supercage. As shown below, it is the cations occupying this site that significantly affect both the conformations and the molecular motions of molecules absorbed in Na−Y. For completeness, Sites III are also shown in Figure 1. Sites III are found near the four-rings of the sodalite cage. It has been reported that at Si/Al ≥ 2 Na+ ions are known to occupy sites I, I′, and II.33 The structure of FAU zeolite was taken from the database of zeolites to initiate the simulations.35 The crystalline structure is described in the Fd3m space group, and the dimension of the cubic unit cell is 24.345 Å3. There are two major framework models for FAU zeolite, the “average T-atom” model7,36−43 and the distinguishable Al−Si model by Jaramillo and Auerbach.20,44 The average T-atom model treats the aluminum and silicon atoms as equivalent (T atoms), and the charges of T atoms are determined according to the ratio of Si/Al in the host framework, whereas the distinguishable Si/Al model treats the framework Si/Al atoms using different charges. In our calculations, the zeolite was modeled by the average T model with rigid framework in which the framework atoms were fixed and the sodium cations and guest molecules were allowed to move. It has been reported45 that the average T model has the

Si−Y

ANaO ρNaO CNaO ZNa ZO ZT ZO ZT

1.43567 × 105 kcal/mol (6230.0 eV) 0.2468 Å 2.30446 × 102 kcal/mol Å6 (10.0 eV Å6) +1.00042e −0.829e +1.361e −1.2e +2.4e

The Na+−Na+ and the Na+−T interactions are Coulombic. Even though this was the original force field, for its usage within AMBER version 11,46 the Buckingham function was fitted to ULJ(rij) = (C/rij)12 − (C′/rij)6, where C = 444428.2218 kcal/mol Å12 and C′ = 285.2969 kcal/mol Å6. One unit cell of FAU was used as the simulation box. Periodic boundary conditions were applied in all three dimensions. The loading was one T12D per unit cell (the infinitely dilute limit). Because the molecule is mainly located within the supercage, even though there are infrequent transitions between supercages, the sampling of the conformational changes in one supercage is equivalent to loading of eight noninteracting molecules per unit cell. Correlations and migrations in the motions of sodium due to different loadings may also exist between neighboring supercages, and they are not considered here because there is one molecule in a unit cell. The first stage of simulations involved the equilibration of sodium cations within the zeolite framework.43,45 Equilibration was performed with MD simulations using AMBER version 11 at 300 K. The equations of motion were integrated using a time step of 1 fs with the leapfrog algorithm. Equilibration took place for 40 ns in the NVT ensemble with the temperature maintained by the Andersen thermostat. The reported distribution 7, 18, and 3245 of Na+ ions in sites I, I′, and II, respectively, was reproduced. This distribution was maintained for tens of nanoseconds. Even though metastable states of locations of Na+ ions may also exist, the distribution of Na+ ions in sites II was stable in the duration of the simulations. After equilibration, the obtained structures were used in QM/MM simulations. The same framework as for Na−Y but without Na+ was used for modeling Si−Y. This computational setup was slightly different from the experimental system because in the experiments 8610

dx.doi.org/10.1021/jp212616h | J. Phys. Chem. C 2012, 116, 8608−8618

The Journal of Physical Chemistry C

Article

the number of Na+ ions in Si−Y (Si/Al = 100) is 3% of that in Na−Y (Si/Al = 2.35). B. QM/MM with DFT and MP2 Modeling of T12D. The T12D in Na−Y and Si−Y was modeled with a variety of quantum chemistry methods ranging from semiempirical AM1 and RM1 to DFT and MP2. In DFT, the B3LYP functional47,48 was used with Ahlrichs double-ζ basis functions.49 In MP2, the 6-31G50,51 basis set was used as a good compromise between quality of basis set and speed of calculations. The QM/MM computations were performed with the software GROMACS/ ORCA52 by using electrostatic embedding, where all force fields were implemented. The vdW interaction between guest molecule and T sites was explicitly modeled by force field to mimic the Pauli exclusion principle. In QM/MM calculations, the integration time step was set to 1 fs using the velocity−Verlet integration algorithm. Velocity-rescaling thermostatting, as was developed by Bussi et al.,53,54 with two thermostats, one for QM part and one for the MM system, was also used at 300 K with coupling constant of 0.1 ps, as one thermostat was not sufficient to keep temperature uniform throughout the MM part and QM part. All long-range electrostatic forces were calculated with the Particle Mesh Ewald method. Short-range force cutoff was used at 1 nm. Equilibration time for DFT and MP2 methods was 10 ps starting from various initial configurations, which were representative of the local minima sampled from the semiempirical methods for aa and ee conformations. The production runs of DFT and MP2 were for 50 ps. C. QM/MM with AM1 and RM1 Modeling of T12D. AM1 and RM1 semiempirical methods as implemented in AMBER version 11 were used in the modeling of T12D. The zeolite MM force field, as described for GROMACS/ORCA, was implemented into AMBER. An Andersen thermostat was used to provide uniform temperature throughout the hostT12D system. The time step of the simulations was 1 fs. The analysis of the data provided statistical quantities in the canonical ensemble but not dynamics due to the stochastic nature of Andersen thermostat. The production runs for AM1 and RM1 were 30 ns. To sample dihedral angle distributions for aa and ee and extract the population of the conformers, we performed enhanced sampling by REMD QM/MM. In the calculations, 8−12 replicas were used in the temperature range 300−600 K. The overlap of the potential energy distributions was maintained throughout the calculation so that swapping of configurations between realizations at different temperatures was possible. The time step was 0.2 fs, and exchange attempts were made every 1 ps. Every replica run was for 10 ns (total sampling time 80−120 ns).

Figure 2. FT-Raman spectra of T12D/Na−Y in the regions of (a) 860−780 and (b) 780−660 cm−1 at different temperatures. The top spectrum marked T12D corresponds to pure T12D liquid at room temperature.

symmetric and asymmetric stretching vibration.11,12 For the aa conformer, only the symmetric vibration at 700 cm−1 is observed, and the intensity of the asymmetric mode is predicted to be too weak to observe.55 In the literature, the intensities of the two strong peaks due to the symmetric C−Cl stretching motions of the ee and aa conformer have been used for quantitative conformational analysis of T12D.11,12 Unlike the bands involving C−H motions, these C−Cl stretching modes due to different conformers are well-resolved and not strongly coupled to other vibrations. In the present work, we also used the 735 and 700 cm−1 bands for the assessments of the relative population of the aa and ee conformer. Because the 735 peak overlaps with the band at 743 cm−1, we obtained the integrated intensity of the 735 cm−1 band by spectral deconvolution. In general, the intensity of a Raman band, Ii, for a conformer is given by

III. RESULTS AND DISCUSSION The FT-Raman spectrum of pure T12D liquid was acquired at room temperature. It is consistent with those previously reported.13,15,55 Two spectral regions that are particularly sensitive to the change in conformation are shown in Figures 2 and 3. The first region is between 780 and 860 cm−1, where the ring C−C stretching vibrations appear. The relative intensities of the two bands at 845 and 827 cm−1 assigned to the ee and aa conformers, respectively, can be used qualitatively to monitor the change in conformation. The second one is the C−Cl stretching region (660−780 cm−1). Two C−Cl stretching modes are expected from each conformer because of their low symmetry. For the ee conformer, these two bands are seen at 735 cm−1 and a weak shoulder at 743 cm−1 due to

Ii = σiCi

(2)

where Ci and σi are the concentration and Raman scattering cross section for a given conformer, respectively. For a 8611

dx.doi.org/10.1021/jp212616h | J. Phys. Chem. C 2012, 116, 8608−8618

The Journal of Physical Chemistry C

Article

where ΔS0 is the change in the standard entropy and ΔH0 is the change in the standard enthalpy, it is found that ln(Iaa/Iee) = ( −ΔH 0/RT ) + const

Therefore, the conformational enthalpy change can be obtained from a variable-temperature study without prior knowledge of ΔS0. T12D/Na−Y. FT-Raman spectra of the T12D molecules adsorbed in Na−Y at a loading level of 8 molecules/u.c. (corresponding to 1 molecule per supercage) were measured as a function of temperature and are shown in Figure 2. At room temperature, the intensity of the C−Cl symmetric stretching mode of the aa conformer seen at 700 cm−1 in pure liquid decreased remarkably relative to its ee counterpart upon loading, which indicates that the conformational equilibrium shifted toward the ee configuration. Analyses of the peak intensities reveal that the relative populations of the aa and ee conformers are 11 and 89%, respectively. (The numbers for the pure liquid are 50%.) A similar change was also observed in the ring C−C stretching region, where the intensity of the ee mode increased significantly upon adsorption. It appears that the electric field generated by the cations in the framework stabilizes the larger dipole moment of the ee isomer, leading to the observed difference in conformational population. In addition, we observed that the position of the ee C−Cl stretching mode shifted to the lower energy by 7 cm−1 from 735 cm−1 in pure liquid to 728 cm−1 in Na−Y, which further indicates a strong interaction between the Cl atom of T12D and the Na+ ions in Na−Y. Such interaction weakens the C−Cl bond, resulting in the observed low-frequency shift. To examine the possible molecular motions of the T12D in Na−Y, the FT-Raman spectra of T12D/Na−Y were also measured at low temperatures (Figure 2). No dramatic change in relative intensities for the bands associated with different conformers was observed when the temperature was lowered. Careful analyses of the Raman intensities of the C−Cl stretching bands reveal that the population of the ee conformer increased slightly with respect to the corresponding aa isomer. The relative intensities (Iaa/Iee) of the C−Cl bands over a range of temperature were used to determine the conformational enthalpy change ΔH0 using eq 5. The plot of ln(Iaa/Iee) versus 1/T is approximately linear (Figure 4), yielding a very small enthalpy difference of 1.31 kJ/mol. Another parameter reflecting the dynamics of the guest species is the full width at half-height (fwhh) of the C−Cl mode of the T12D adsorbed in Na−Y. In pure T12D liquid,

Figure 3. FT-Raman spectra of T12D/Si−Y in the regions of (a) 860−780 and (b) 780−660 cm−1 at different temperatures. The top spectrum marked T12D corresponds to pure T12D liquid at room temperature.

conformational equilibrium ee ⇌ aa, the equilibrium constant K can be expressed by K = Caa/Cee = (Iaa/Iee)σ

(3)

where σ = σaa/σee. For T12D, the σ value was previously estimated to be 0.95,11 and this value was also used in this study to calculate the conformational population of T12D adsorbed in a zeolite at various temperatures, assuming that the Raman scattering cross section in the zeolite is the same as that in solution and is temperature-independent. Furthermore, using the fundamental expression of thermodynamics ln K = ( −ΔH 0/RT ) + ΔS 0/R

(5)

(4)

Figure 4. Plots of ln(Iaa/Iee) versus 1/T for T12D adsorbed in (a) Na−Y and (b) Si−Y. 8612

dx.doi.org/10.1021/jp212616h | J. Phys. Chem. C 2012, 116, 8608−8618

The Journal of Physical Chemistry C

Article

Figure 6. Typical snapshots of ee conformer at local minima in 57Na− Y at 300 K. The representation of the atomic sites is as in Figure 5. (a) ee conformer interacting with two Na+ ions and (b) ee conformer interacting with one Na+ ion. “inter” denotes the local minimum between two supercages.

Figure 5. Typical snapshots of aa conformer at local minima in 57Na−Y at 300 K (a) within a supercage and (b) in the window between two supercages. The blue spheres represent Na+ ions, the yellow spheres represent the T sites, and the red spheres represent the oxygen sites. The carbon framework of T12D molecule is shown by the light-blue spheres, and the gray spheres show the chlorine sites and the white spheres show the hydrogen sites. The four Na+ ions in the supercage are connected by blue lines to show that they are located at the vertices of a tetrahedral. T12D may be interlocked between any of the edges of the tetrahedral.

ment) of the T12D molecule inside each supercage is slightly different. Consequently, the frequency of a given vibrational mode also varies from cage to cage, resulting in an observed broad band due to the distribution of the frequency. This argument is supported by the fact that cooling the system to 153 K results only in a small reduction in the fwhh of the C−Cl stretching mode. T12D/Na−Y with Molecular Simulations. Figures 5 and 6 show typical snapshots of aa and ee conformers at local minima that were identified by using REMD with QM/MM, where the QM part was modeled by AM1 and RM1 and verified by direct MD QM/MM, where the QM part was modeled by DFT and MP2. AM1 and RM1 with REMD captured the majority of the minima. The minima were verified by using the semiempirical configurations as initial conditions for DFT and MP2 MD runs for ∼50 ps. Even though we sampled the vicinity of the minima for 50 ps, semiempirical modeling indicated that the lifetime in a local minimum is much longer than 50 ps. The four Na+ ions within a supercage form a tetrahedral geometry, as indicated by the blue tetrahedron in Figures 5a and 6. The aa conformer interlocks between any

the fwhh’s are 9 and 10 wavenumbers for the 700 and 735 cm−1 band, respectively. Upon being adsorbed inside Na−Y, they increased from 13 and 11 wavenumbers to 22 and 21 wavenumbers for the respective aa and ee conformers. In general, the line-width of vibrational modes of a small organic molecule in solution is much broader than that of an ordered crystalline solid due to isotropic motion. In the present case, the fact that the fwhh’s of the C−Cl stretching mode for the T12D inside Na−Y are more than double the values of pure liquid strongly suggests that the origin of the line-broadening is not molecular motion because compared with the liquid state the degree of motion for the T12D trapped inside Na−Y must be reduced due to the spatial confinement imposed by the framework. Therefore, we suggest that the orientation of the T12D molecules among supercages in the lattice is statically disordered. This means that the orientation (i.e., the environ8613

dx.doi.org/10.1021/jp212616h | J. Phys. Chem. C 2012, 116, 8608−8618

The Journal of Physical Chemistry C

Article

Figure 7. Radial distribution function (gCl−Na(r)) of Cl−Na+ in 57Na−Y at 300 K calculated by DFT, MP2, AM1, and RM1 methods. (a) aa conformer and (b) ee conformer. gCl−Na(r) for aa using DFT and MP2 has been computed in the minimum that corresponds to Figure 5a and for ee in the minimum shown in Figure 6a.

Figure 8. Normalized C−Cl bond length distribution, ρ(rC−Cl), of T12D found in the Na−Y zeolite and gaseous state calculated by DFT method at 300 K. (a) For aa conformer and (b) for ee conformer.

of the first peak. RM1 and AM1 differ by 0.04 Å and 0.07 Å from MP2 data, respectively. The deviations are larger for the ee conformer but still within acceptable range for the predictions of various quantum chemistry methods.56 The agreement in the first peak of AM1 with DFT and MP2 indicates that AM1 captures the same Cl−Na+ distance fluctuations as reliably as more rigorous quantum chemistry methods. RM1 does not capture the same fluctuations as AM1. The frequent adulations after 0.4 nm arise from the fact that there is more noise in the data of DFT and MP2 due to substantially shorter sampling times. Distribution of the C−Cl bond length (Figure 8) relative to the gaseous state shows that the C−Cl bond weakens, which corresponds to shifting of the Raman C−Cl frequency to lower values in the experiments. In the aa conformer, the weakening of the bonds is not the same for both C−Cl bonds, which means that two Cl are not equivalent. For the ee isomer, the weakening of the bonds appears to be the same for both Cl. Dipole moments of aa and ee in the various local minima were computed using MP2 (Figure 9) and DFT (Supporting Information) in GROMACS/ORCA. For testing the methods, initially, the dipole moment in the gaseous state of aa and ee conformers was reproduced very closely to the values reported in ref 12.14 Inside Na−Y, the dipole moment of ee conformer was estimated to be almost twice (7.05 D) the dipole moment of aa (3.52 D). The aa and ee dipole moments are considerably larger than in the gaseous state because of the zeolite polar

of the two Na+ ions that give rise to several local minima within the supercage. Intercage locations were also found that survive for several tens of picoseconds. For the ee conformer, two types of local minima were identified: In one of the minima, the chlorine sites of the ee conformer are closer to two Na+ ions, whereas in the other minimum, they interact stronger with one Na+ ion only. During the realizations, there were no migrations of Na+ ions among the various sites of Na−Y zeolite. The picture that emerges is that T12D exhibited jump diffusion within the zeolite supercage, which is characterized by long periods spent within low free-energy adsorption sites punctuated by rapid jumps to adjoining adsorption sites. The fact that T12D can be found in several long-living local minima contributes toward the static disordering that is discussed in the experimental part as the reason for the broadening of the fwhh’s of the C−Cl stretching mode relative to the pure liquid. Furthermore, the local minima are temperature-independent, and this is consistent with the fact that the fwhh of the C−Cl stretching mode changes only slightly by reducing the temperature. Radial distribution functions (g(r)) of Cl−Na+ were computed to estimate their average distances and verify the applicability of the semiempirical methods. The g(r) profiles are shown in Figure 7. The g(r) profiles computed by DFT and MP2 agree very well with each other in the location and width 8614

dx.doi.org/10.1021/jp212616h | J. Phys. Chem. C 2012, 116, 8608−8618

The Journal of Physical Chemistry C

Article

interaction for aa accounts for 18% of the total binding energy and for ee 32%. The equilibrium population of aa and ee conformers was quantified by estimating the dihedral Cl−C−C−Cl angle distributions of aa and ee using REMD with AM1. The dihedral angle distribution is shown in Figure 10a. To monitor convergence of the distribution, we started REMD realizations independently from aa and ee conformations. Moreover, the convergence was gauged by having time-independent distribution profiles. To confirm that the AM1 modeling is close to what is predicted by more rigorous quantum chemistry methods, the dihedral angle profiles at local minima explored by MP2 were also used for comparison (Figure 10a). The difference between AM1 and MP2 in the average values is 1.6% for ee and 5.8% for aa. The sensitivity of the results to the parameters is demonstrated by using RM1 dihedral angle distributions (Figure 10b). The calculations show that the equilibrium between aa and ee is completely reversed. The average differences in dihedral angles are 31% for ee and 2.4% for aa conformer. The substantially larger error for the ee conformer contributes to the different prediction relative to AM1. Integration of the dihedral angle distributions gives the population profile shown in Figure 10c. The percentage of aa versus ee is estimated to be 8−92. The finding is in close agreement with the experimental values (11 to 89).

Figure 9. Normalized dipole moment distributions (ρ(μ)) of aa and ee conformers in 57Na−Y calculated by MP2 method at 300 K. “intra” in the legend denotes locations of T12D inside the supercage and “intra 1” the local minimum of ee when it is closer to one Na+. “inter” denotes the local minimum between two supercages.

environment. The consequence of the larger dipole moment of ee is that the ee conformer is favored in the polar environment of the Na−Y zeolite. We estimated that in Na−Y the vdW

Figure 10. Normalized dihedral angle distribution (ρ(Θ)) of T12D in Na−Y zeolite at 300 K computed by: (a) REMD AM1 method (the MP2 distributions at local minima separately for the aa and ee conformers have been included for comparison) and (b) REMD RM1 method. The raw data to build the distributions were collected from the lowest-lying replica at 300 K over 10 ns production run. In the legends, “initial aa” and “initial ee” denote the distributions that were generated starting from different aa and ee initial conditions. (c) Population distribution of aa and ee, as it is delivered from panel a. 8615

dx.doi.org/10.1021/jp212616h | J. Phys. Chem. C 2012, 116, 8608−8618

The Journal of Physical Chemistry C

Article

T12D/Si−Y. For comparison, we also examined the Raman spectra of the T12D molecules adsorbed in highly siliceous zeolite Y (Si−Y). Si−Y has the same framework topology as Na−Y, yet it possesses a much larger Si/Al (= 100). A framework with such a Si/Al ratio has only a very small number of extra-framework sodium cations. The Raman spectrum of the T12D/Si−Y system acquired at room temperature is shown in Figure 3. Interestingly, we also see a significant increase in the ee conformer from both the C−Cl and ring C−C stretching regions. The relative intensity analyses of the C−Cl bands indicate that the population of the aa and ee conformers is approximately 23 and 77%, respectively. (These numbers are 11 and 89% for the T12D in Na−Y.) The results show that the framework of Si−Y also favors the ee conformation. However, the origin of the host−guest interaction appears to be different. In Na−Y, the main interaction is electrostatic in nature, involving Na+ and Cl of T12D. Such interaction is unlikely to be the dominate force because the number of Na+ ions in Si−Y (Si/Al = 100) is only 3% of that in Na−Y (Si/Al = 2.35). This is consistent with the observation that unlike in Na−Y the frequency of the C−Cl stretching band due to the ee conformer in Si−Y did not shift to the low energy upon absorption. One possible sorbate−sorbent interaction in siliceous zeolitic framework is the vdW interaction. Therefore, the ee configuration with a slight larger molecular volume allows maximization of the attractive interaction with the framework. An additional factor that may contribute to the observed increase in the ee population is the possibility of existing some Si−OH groups as defects resulting from the fracture of Si−O−Si bonds. The silanol groups possibly in the interior of supercage provide a slightly polar environment, which also enhances the ee population. Upon cooling, a further decrease in the aa population was observed. From the van’t Hoff plot in Figure 4, the conformational enthalpy difference was estimated to be 5.69 kJ/mol. Another interesting observation is that the C−Cl stretching mode of the ee conformer did not undergo significant line-broadening. This can be explained by the lack of Na+ ions in the framework. As previously mentioned, the strong interaction between Na+ and Cl atom localizes the T12D molecule, leading to a disordered distribution of the guest species among different supercages. In Si−Y, the absence of such interaction allows the T12D molecule to move more freely within each cage. Consequently, no static disordering was observed. T12D/Si−Y with Molecular Simulations. In Si−Y, the T12D molecule resides mainly close to the center of the supercage. The motion of T12D in the Si−Y supercage is less restricted than that in Na−Y because it is not bound in longliving local minima as in the presence of Na+. This behavior is more consistent with the smaller broadening of the fwhh of the C−Cl mode than that of Na−Y. The dipole moment distributions of the aa and ee conformers in Si−Y are shown in Figure 11. In Si−Y, the average dipole moment of aa and ee is found to be 2.0 and 6.0 D, respectively, which is smaller than 2.7 and 7.0 D for these conformations in Na−Y. The smaller dipole moment in Si−Y contributes toward the smaller increase in the aa component relative to ee. We estimated that in Si−Y the vdW interaction for aa accounts for 62% of the total binding energy and for ee 58%. The distribution of the dihedral angle is shown in Figure 12a. The integration of the profiles gives population of aa versus ee of 15 to 85%. The trend in the way that the zeolite framework affects the aa versus ee population is the same way as that detected in the Raman spectra.

Figure 11. Normalized dipole moment distributions (ρ(μ)) of aa and ee conformers in Si−Y calculated by MP2 method at 300 K.

Figure 12. (a) Normalized dihedral angle distribution (ρ(Θ)) of T12D in Si−Y zeolite using AM1 method. (b) As in Figure 10 c but for Si−Y. 8616

dx.doi.org/10.1021/jp212616h | J. Phys. Chem. C 2012, 116, 8608−8618

The Journal of Physical Chemistry C

Article

IV. CONCLUSIONS In this work, we have investigated conformational properties of trans-1,2-dichlorocyclohexane adsorbed inside two zeolites with the same FAU framework topology but different Si/Al ratio (Na−Y and Si−Y) by variable-temperature FT-Raman spectroscopy and molecular simulations. The results demonstrate that the conformational behavior of T12D strongly depends on the Si/Al ratio and the presence of charge compensating cations. For zeolite Na−Y that has charge-balancing cations, the ee conformer is overwhelmingly preferred because of the interaction of the T12D molecules with the cations. The cations stabilize the larger dipole moment of the ee isomer. Molecular simulations of T12D in Na−Y (Si/Al = 2.35) and Si−Y were performed using QM/MM method. QM/MM was used to allow for the charge distribution of T12D to vary as the molecule jumps in the various local minima inside the supercage. The change in the charge distribution was achieved by modeling T12D by semiempirical quantum chemistry methods (AM1 and RM1) as well as DFT and MP2. The current simulations show that AM1 modeling for chlorinated hydrocarbons may produce equilibrium populations of the conformers close to those found in the presented experiments. The QM/MM implementation in AMBER version 11 is efficient for the performance of the simulations. Contribution of the polarization energy has been studied for p-xylene and m-xylene in Na−Y27 that have similar polarizability to T12D. It has been shown that polarization energy accounts for 25% of the potential energy. For Xe in Na−Y, the polarization energy accounts for 17% of the total energy.28 It has also been established that the amount of the induced polarization energy depends on the zeolite framework.28 In our modeling, electrostatic energy accounts for all possible multipole interactions of the charge density of the QM subsystem. Our computations suggest that the polarization induced by the FAU framework on the highly polarizable T12D molecule is the reason for the shift in the ee and aa equilibrium in favor of the ee conformer in both Si−Y and Na−Y. Simulations indicate that in Na−Y, Na+ ions have an additional effect in shifting the equilibrium in favor of the ee conformer. It has been reported27 that the electrostatic field ranges between 5 and 18 V/nm in a Na−Y zeolite (Si/Al = 3.0) at a distance ∼3 Å from the Na+ sites. Because polarizability changes slightly between aa and ee (14.36 Å3 for aa vs 14.70 Å3 for ee)26 in the gaseous state, assuming gaseous state polarizability one finds that the induced dipole moment ranges between 2.4 and 8.7 D. This range is in good agreement with our calculated data of the induced dipole moment from DFT and MP2. Even though the loading of DCC in FAU is larger than that of T12D, and therefore there are additional intermolecular interactions, FT-Raman spectra reveal that in both DCC/FAU and T12D/FAU the chlorine sites interact with the Na+ sites, and as a result the C−Cl bond weakens. Qualitatively, this is not a surprising result. In both DCC and T12D, the fwhh’s of the C−Cl stretching mode is attributed to static disorder relative to the liquid state. The reported simulations of T12D provided direct evidence of the static disorder, and moreover, the various docking conformations (local minima) of T12D between Na+ were identified. Good description of the zeolitic environment−guest interactions is one of the key components for making simulations a predictive method for adsorption of guest molecules in zeolites. It is generally believed that DFT or MP2 modeling of

polyatomic organic molecules in long-time MD or long MC realizations that achieve good sampling may be inefficient. To improve efficiency, one may consider a MC methodology that allows for sampling using high-quality quantum chemistry methods for modeling the interactions utilizing configurations produced by a reasonably good way of modeling.57 This methodology has been applied in grand canonical ensemble sampling for cluster distributions in the vapor phase,57 and one may consider its extension in zeolitic environment. Another approach is one to consider the usage of QM/MM with semiempirical modeling of organic molecules in zeolites as an attractive alternative to fixed charge empirical force fields.



ASSOCIATED CONTENT

S Supporting Information *

Comparison of the dipole moment distributions of T12D aa and ee conformers at local minima using DFT and MP2. This material is available free of charge via the Internet at http:// pubs.acs.org.



AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]; [email protected]. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS J.M. is a visiting scientist of the University of Western Ontario and she acknowledges the financial support of Shanxi Scholarship Council of China. Y.H. acknowledges the financial assistance from Natural Science and Engineering Research Council of Canada (NSERC) for a research grant and the award of an FT-Raman spectrometer. S.C. thanks the Accelerator Grant for Exceptional New Opportunities (AGENO)-NSERC and the NSERC Discovery Grant. We also thank SHARC-Net, Scinet, and Westgrid for providing the computing facilities to perform the simulations.



REFERENCES

(1) Introduction to Zeolite Science and Practice; van Bekkum, H., Flanigen, E. M., Jacobs, P. A., Jansen, J. C., Eds.; Elsevier: Amsterdam, 2001. (2) Szostak, R. Handbook of Molecular Sieves; Van Nostrand Reinhold: New York: 1992. (3) Mellot, C. F.; Cheetham, A. K.; Harms, S.; Savitz, S.; Gorte, R. J.; Myers, A. L. J. Am. Chem. Soc. 1998, 120, 5788−5792. (4) Giaya, A.; Thompson, R. W.; Denkewicz, R. Micropor. Mesopor. Mat. 2008, 40, 205−218. (5) Clausse, B.; Garrot, B.; Cornier, C.; Paulin, C.; Simonot-Grange, M. H.; Boutros, F. Microporous Mesoporous Mater. 1998, 25, 169−177. (6) Gonzalez-Velasco, J. R.; Lopez-Fonseca, R.; Aranzabal, A.; Gutierrez-Ortiz, J. I.; Steltenpohl, P. Appl. Catal., B 2000, 24, 233− 242. (7) Mellot, C. F.; Davidson, A. M.; Eckert, J.; Cheetham, A. K. J. Phys. Chem. B 1998, 102, 2530−2535. (8) Zhou, J. M.; Zhao, L.; Huang, Q. Q.; Zhou, R. X.; Li, X. K. Catal. Lett. 2009, 127, 277−284. (9) Huang, Y. N.; Leech, J. H.; Wang, H. Y. J. Phys. Chem. B 2003, 107, 7632−7639. (10) Wu, T.-N. Water Sci. Technol. 2010, 62, 2313−2320. (11) McClain, B. L.; Ben-Amotz, D. J. Phys. Chem. B 2002, 106, 7882−7888. (12) McClain, B. L.; Ben-Amotz, D. J. Chem. Phys. 2002, 117, 6590− 6598. 8617

dx.doi.org/10.1021/jp212616h | J. Phys. Chem. C 2012, 116, 8608−8618

The Journal of Physical Chemistry C

Article

(13) Ulyanova, O. D.; Ostrovsk, M. K.; Pentin, Y. A. Russ. J. Phys. Chem. 1970, 44, 563−564. (14) Wilberg, K. B. J. Org. Chem. 1999, 64, 6387−6393. (15) Klaeboe, P.; Lothe, J. J.; Lunde, K. Acta Chem. Scand. 1957, 11, 1677−1691. (16) Crawford, M. K.; Dobbs, K. D.; Smalley, R. J.; Corbin, D. R.; Maliszewskyj, N.; Udovic, T. J.; Cavanagh, R. R.; Rush, J. J.; Grey, C. P. J. Phys. Chem. B 1999, 103, 431−434. (17) Huang, Y. N.; Leech, L. J. Phys. Chem. A 2001, 105, 6965−6970. (18) Huang, Y. N.; Leech, L. H.; Havenga, E. A.; Poissant, R. R. Microporous Mesoporous Mater. 2001, 48, 95−102. (19) Mellot-Draznieks, C.; Rodriguez-Carvajal, J.; Cox, D. E.; Cheetham, A. K. Phys. Chem. Chem. Phys. 2003, 5, 1882−1887. (20) Jaramillo, E.; Grey, C. P.; Auerbach, S. M. J. Phys. Chem. B 2001, 105, 12319−12329. (21) Lim, K. H.; Jousse, F.; Auerbach, S. M.; Grey, C. P. J. Phys. Chem. B 2001, 105, 9918−9929. (22) Frantz, D. D.; Freeman, D. L.; Doll, J. D. J. Chem. Phys. 1990, 93, 2769−2784. (23) Geyer, C. J. Computing Science and Statistics Proceedings of the 23rd Symposium on the Interface; American Statistical Association: New York, 1991, p 156. (24) Xu, H.; Berne, B. J. J. Chem. Phys. 1999, 110, 10299−10306. (25) Mitsutake, A.; Sugita, Y.; Okamoto, Y. Biopolymers 2001, 60, 96−123. (26) Boyd, R. H.; Kesner, L. J. Chem. Phys. 1980, 72, 2179−2190. (27) Lachet, V.; Boutin, A.; Tavitian, B.; Fuchs, A. H. J. Phys. Chem. B 1998, 102, 9224−9233. (28) Pellenq, R. J.-M.; Tavitian, B.; Espinat, D.; Fuchs, A. H. Langmuir 1996, 12, 4768−4783. (29) Dewar, M. J. S.; Zoebisch, E. G.; Healy, E. F. J. Am. Chem. Soc. 1985, 107, 3902−3909. (30) Rocha, G. B.; Freire, R. O.; Simas, A. M.; Stewart, J. J. P. J. Comput. Chem. 2005, 27, 1101−1111. (31) Ibrahim, M. A. A. J. Chem. Inf. Model. 2011, 51, 2549−2559. (32) Zimmerman, P. M.; Head-Gordon, M.; Bell, A. T. J. Chem. Theory Comput. 2011, 7, 1695−1703. (33) Buttefey, S.; Boutin, A.; Fuchs, A. H. Mol. Simul. 2002, 28, 1049−1062. (34) Calero, S.; Dubbeldam, D.; Krishna, R.; Smit, B.; Vlugt, J. H. T.; F. M. Denayer, J.; A. Martens, J.; L. M. Maesen, T. J. Am. Chem. Soc. 2004, 126, 11377−11386. (35) Baerlocher, Ch.; McCusker, L. B. Database of Zeolite Structures. http://www.iza-structure.org/databases/. (36) Mellit, C. F.; Cheetham, A. K.; Harm, S.; Savitz, S.; Gorte, R. J.; Myers, A. L. J. Am. Chem. Soc. 1998, 120, 5788−5792. (37) Moon, G. K.; Choi, S. G.; Kim, H. S.; Lee, S. H. Bull. Korean Chem. Soc. 1993, 14, 356−362. (38) Lee, S. H.; Moon, G. K.; Choi, S. G.; Kim, H. S. J. Phys. Chem. 1994, 98, 1561−1569. (39) Smirnov, K.; LeMaire, M.; Bremard, C.; Bougeard, D. Chem. Phys. 1994, 179, 445−454. (40) Krause, K.; Geidel, E.; Kindler, J.; Forster, H.; Smirnov, K. S. Vibr. Spectrosc. 1996, 12, 45−52. (41) Schrimpf, G.; Schlenkrich, M.; Brickmann, J. J. Phys. Chem. 1992, 96, 7404−7410. (42) Mellot, C. F.; Cheetham, A. K. J. Phys. Chem. B 1999, 103, 3864−3868. (43) Beauvais, C.; Guerrault, X.; Coudert, F. X.; Boutin, A.; Fuchs, A. H. J. Phys. Chem. B 2004, 108, 399−404. (44) Jaramillo, E.; Auerbach, S. M. J. Phys. Chem. B 1999, 103, 9589− 9594. (45) Buttefey, S.; Boutin, A.; Mellot-Draznieks, C.; Fuchs, A. H. J. Phys. Chem. B 2001, 105, 9569−9575. (46) Case, D.;et al. AMBER 11; University of California: San Francisco, 2010. (47) Kossmann, S.; Neese, F. J. Chem. Theory Comput. 2010, 6, 2325−2338.

(48) Neese, F.; T. Schwabe, T.; Kossmann, S.; Schirmer, B.; Grimme, S. J. Chem. Theory Comput. 2009, 6, 3060−3073. (49) The Ahlrichs DZ basis set was obtained from the TurboMole basis set library under ftp.chemie.uni-karlsruhe.de/pub/basen. (50) Hehre, W. J.; Ditchfield, R.; Pople, J. A. J. Chem. Phys. 1972, 56, 2257−2261. (51) Francl, M. M.; Pietro, W. J.; Hehre, W. J.; Binkley, J. S.; M. S. Gordon, D. J. D.; Pople, J. A. J. Chem. Phys. 1982, 77, 3654−3665. (52) Neese, F.; T. Schwabe, T.; Kossmann, S.; Schirmer, B.; Grimme, S. J. Chem. Theory Comput. 2009, 6, 3060−3073. (53) Bussi, G.; Donadio, D.; Parrinello, M. J. Chem. Phys. 2007, 126, 014101. (54) Berendsen, H. J. C.; Postma, J. P. M; van Gunsteren, W. F.; DiNola, A.; Haak, J. R. J. Chem. Phys. 2009, 81, 3584−3590. (55) Kozima, K.; Sakashita, K.; Maeda, S. J. Am. Chem. Soc. 1954, 76, 1965−1969. (56) McCaffrey, P. D.; Mawhorter, R. J.; Turner, A. R.; Brain, P. T.; Rankin, D. W. H. J. Phys. Chem. A 2007, 111, 6103−6114. (57) Shi, Y. J.; Consta, S.; Das, A. K.; Mallik, B.; Lacey, D.; Lipson, R. H. J. Chem. Phys. 2002, 116, 6990−6999.

8618

dx.doi.org/10.1021/jp212616h | J. Phys. Chem. C 2012, 116, 8608−8618