Insight into the adsorption mechanism of benzene in HY zeolites: the effect of loading

Abstract

An interesting two-stage adsorption mechanism was first proposed for the benzene/HY system by Metropolic Monte Carlo (MMC) simulations at loadings below and above an “inflection point”, and were composed of processes labeled “ideal adsorption” and “insertion adsorption”, respectively. Below the inflection point (from infinite dilution up to 32 molecule/UC for all Si : Al ratios), benzenes were located on the sorption sites inside the supercages with an ideal adsorption geometry configuration, which is in accordance with previous studies. Above the inflection point, the benzene molecule tended to insert into the space between existing adsorbed benzenes, and no obvious rearrangement was observed for previously adsorbed benzenes. It was found that the proposed adsorption mechanism existed independently of the Si : Al ratio, while the inflection point shifted to a higher loading for zeolite with a lower Si : Al ratio. This is due to increased utilization of the 12-T ring caused by the contribution of the H1 site in zeolite with a lower Si : Al ratio, which result in less crowed adsorption at loadings approaching saturation.

---

1. Introduction

Zeolites are one of the most important heterogeneous catalysts for environmental and industrial applications.^1–3 Particularly, Y-type zeolites have many applications in catalysis,^4,5 especially in petroleum cracking processes.^6,7 The adsorption of the reactant is the very first step of all catalysis in zeolites. Therefore, an elaborate understanding of the adsorption mechanisms of adsorbates inside Y zeolites is a prerequisite for clarifying the underlying mechanism of processes such as catalytic cracking or hydroisomerization, as well as for exploring the further applications of zeolites as catalysts and adsorbents.⁸

Extensive researches have been performed including experimental^9–14 and computational,^15–22 on the adsorption process of benzene and other aromatics in Y-type zeolites. As for benzene/HY system, two types of sites were responsible for benzene adsorption,^23–30 the energetically favorable S site within the supercage and the W site centered in the circular window that links together neighboring supercages. However, our previous computational study²⁸ revealed that even at loading as low as one molecule per supercage for HY zeolite, benzene adsorption still occurred on these two sites despite the energy difference, which is consistent with experimental studies of benzene in HY³¹ and H-SAPO-37.³²

For the sequence of adsorption, literatures suggested that with increasing pore filling,²⁹ the S site, which have been claimed to be the site of preferred location of benzenes, will be saturated at first in tetrahedral arrangement (accommodating 4 molecules per supercage). At saturated loading,³³ at least one extra molecule have to be adsorbed each supercage on average (accommodating 5–6 molecules per supercage). However, the adsorption sites of extra molecules are still controversial. Some studies revealed extra benzene could get into the supercage and influence the adsorption distribution of pre-adsorbed molecules.³⁴ Recent researches suggested the extra benzene molecules gradually occupied the available W site according to the prediction based on the relative energy of sites.³⁵

Most studies engaging in the molecular-based adsorbate focused on relatively low coverage^38,39 and they explained the loading dependent diffusion behavior for benzene/Y system by site-hopping mechanism. In the meantime, energetic properties such as heat of adsorption and interaction energies, which can only provide indirect information about the nature of adsorption, were mostly discussed in revolution of adsorption mechanism with loading.^35–37 However, the complexity of the adsorption system may greatly increase as pore fillings approach saturation,^36,37 which is attributed to the guest–guest interactions and the packing effects of the guest molecules in relation to the size and shape of the adsorbates and their fitting capability within the zeolites.³⁸ Therefore, the unclear adsorption mechanism in whole loading range, especially at high loadings, needs to be clarified before the research results are utilized to design new processes in zeolites.

Monte Carlo (MC) simulation can provide density profiles,^2,39 radial distribution functions^40,41 and snapshots of energetically stable adsorbate configurations,^42,43 which is a very suitable method for the study of such subject at the molecular lever. Peralta et al.⁴⁴ performed MC calculation to study the adsorption behavior of xylene on a metal–organic framework (MOF) for different pore fillings. Together with a sharp increase of the guest–guest interaction energy and a shoulder which appears at 0.4 nm in the radial distribution function, a π-stacking adsorption mechanism was proposed for xylene in MOF and confirmed by different snapshots corresponding to minimum energy configurations.

In this study, we performed MC simulations to elaborate the loading dependence of the adsorption mechanism for benzene in HY zeolite. The concentration plots, radial distribution functions (RDFs), and energetic stable configurations of the benzene/HY system present a consistent body of evidence that supports change of adsorption mechanism for benzene with different loading ranges.

2. Methods

2.1 Models of zeolite

In this study, the chemical composition of HY zeolite Si_192−xAl_xH_xO₃₈₄ was considered for x = 0, 14, 28, and 56, namely 0Al, 14Al, 28Al, and 56Al, respectively. In this manner, zeolite models with Si : Al ratios of ∞, 12.71, 5.86, and 2.43 were obtained, such that the 56Al model is in agreement with experimentally observed HY zeolite.⁴⁵

As is known, there are four different positions for O atoms in the FAU framework (O_z),¹⁸ namely O1, O2, O3, and O4, and the protons (H_z), which are attached to various O_z atoms, are denoted as H1, H2, H3, and H4, as illustrated in Fig. 1. In this study, the occupation of H_z for each model is based on that found in ref. 12 and has been idealized for simplicity, as listed in Table SI1 (ESI†). Among the three generally detected H_z atoms (H1, H2, and H3), only O_z–H1 and O_z–H2 bonds lie approximately in the plane of a 12-T (T = Si or Al atom) ring window and in the plane of a 6-T ring of the sodalite cage (SOD), which can interact with adsorbed molecules. More details about construction of the HY zeolite models can be found in our previous paper,²⁸ as well as in the ESI.†

Fig. 1 A cut of the FAU framework showing four different positions for O atoms. Blocked SOD is colored gray.

The pore landscape illustrated in Fig. 1 reveals the presence of a SOD. Benzene molecules can occupy such pockets in a MC simulation, even though it is impossible for benzene molecules to migrate through the six-ring windows of the SOD in practice.^46–50 The procedure used in this study is to segregate the underlying field into separated subfields, based on the SOD and supercage, such that MC calculations can avoid benzene interaction with the SOD.

The simulation box consisted of a single unit cell of HY zeolite. A separate simulation in a larger box with 2 × 2 × 2 unit cells was performed and gave very similar results for both adsorbate distribution and energy (within statistical error) of adsorption at both low and high adsorbate loadings, implying a negligible size effect. All zeolite models were considered to be rigid in our simulations, as is the case in most simulation studies of adsorption in zeolites. Because there is not experimental evidence of large structural changes of FAU zeolite at the temperature under study (298 K), indicating that flexibility should have minor effects on the static properties.⁵¹ Nevertheless, we evaluated the effect of framework flexibility^52,53 by comparing the results of this simulation study (MC) with a rigid zeolite structure and molecular dynamic simulation with a flexible structure.^54,55 One can confidently say that the main conclusions in this simulation are scarcely influenced by flexibility of the zeolite frameworks.

2.2 Simulation details

The commercial software Materials Studio from Accelrys, Inc. was used for energy minimization and MC simulations. Prior to energy related calculations, it was necessary to assign force field types to all particles in the system. In all of our simulations, the COMPASS (Condensed-phase Optimized Molecular Potentials for Atomistic Simulation Studies) force field⁵⁶ was used to define the interactions between the atoms. The COMPASS force field has been successfully applied to explore the adsorption of many adsorbate/zeolite systems,^57–63 including the benzene/HY zeolite.^64,65 A wide ranges of charge parameters have been used by different authors. The partial charges used in this paper was taken from ref. 66, which can be found in Table SI2, ESI.†

The simulation scheme employed in this work is summarized briefly in this section. First, the structures of benzene molecules and the 0Al, 14Al, 28Al, and 56Al models were optimized. Then MC simulations in the canonical ensemble were carried out for benzene in each HY zeolite model at 298 K. 100 000 configurations frames are saved in this study so that accurate radial distribution properties can be obtained. The equilibration MC steps are 5 × 10⁶ followed by another 5 × 10⁶ MC steps for production. In this study, one MC step is defined as the attempt to move each of the adsorbates, once. A typical mix of MC moves was used as previously reported²⁸ for all the loadings considered: from 1 to 44 molecule per UC, since the maximum possible loading has been documented to be between 5 and 6 benzene molecules per supercage.³³ Several loading points of our simulations were recomputed with 50 × 10⁶ equilibration MC steps and 50 × 10⁶ MC production steps to check on the convergence used. The variance of the energy of adsorption was typically found to be less than 2%. The electrostatic potential energy was calculated by the Ewald summation method, with a calculation accuracy of 4.184 J mol⁻¹. The cutoff distance for the calculation of the van der Waals potential energy was taken to be 1.2 nm.⁶⁷ The periodic boundary conditions were applied in the three coordinate directions to form an infinite sorbent structure without open surfaces.²⁷

2.3 Energetics of adsorption

The total adsorption energy of the model system (U_ad) is an important parameter for the adsorption process. It is determined by summing all interactions between pairs of adsorbate molecules (U_ads–ads), as well as all adsorbate interactions with the zeolite framework (U_ads–zeo):

U_ad = U_ads–ads + U_ads–zeo (1)

The energy of adsorption is more generally reported in terms of the isosteric heat of adsorption (Q_st). From the loading dependence of the total adsorption energy U_ad, Q_st is calculated as the difference between the partial molar enthalpy of the adsorbate in the gas-phase minus the partial molar energy of the adsorbed phase:

where U_ad includes contributions from both U_ads–ads and U_ads–zeo, and U_intra is the intramolecular energy of the adsorbate molecules. Thus, there was a positive correlation between Q_st and the absolute value of U_ad.

3. Results and discussion

3.1 Adsorption isotherms and isosteric heats

The properties of adsorption isotherms and isosteric heats are important for understanding the adsorption behavior of adsorbates in micropores. They were often used to verify the reliability of force field parameters and models adopted in calculations. Take 0Al model as an example, the experimental and simulated adsorption isotherms of benzene are reported in Fig. 2. The simulation results show a good agreement with the experimental values.^20,21 Moreover, the adsorption isotherm exhibits an inflection point, in which the curvature changes from positive (increasing slope) to negative (decreasing slope). This change also could be found in experimental results for benzene in an all-silica FAU system and simulation results for other single-aromatics.⁶⁸ It should relate to the fact that the adsorption mechanism changes with the loading of benzene.

Fig. 2 Adsorption isotherm of benzene in 0Al models at 298 K with the experimental and simulation results.

Furthermore, the Q_st were calculated in 0Al, 28Al, and 56Al models at 298 K. Each data was the average value of three times repeat calculations and relative error were less than 1%. In the whole loading range, the Q_st of benzene in 0Al model is in the range of 9.5 to 17.1 kcal mol⁻¹, well agree with the result of ref. 21 (in the range of 10.0 to 15.5 kcal mol⁻¹).

3.2 Contour map of density

Both simulation^23–28 and experiment^29,30 reveals that the distribution of benzene in HY zeolite can be divided into two regions: the dominant site which lies in the supercage (S site) and the second site which is centered in the 12-ring window (W site). The S site is a general name that includes all the sites located inside the supercage (i.e., benzene molecule adsorbed onto 4-T rings, and H_z of HY zeolite). Contour map of density is typically used to study the distribution of adsorbates in zeolites. Fig. 3 shows contour maps of density of benzene, with respect to the zeolite framework of the 28Al model. As can be observed in the figure, the spatial distribution of benzene in HY zeolite is roughly territorial and the adsorptions of benzene on both sites are observed.

Fig. 3 Contour maps of density of benzene molecules in 28Al model at (a) 1 molecule per UC, (b) 4 molecule per UC, (c) 8 molecule per UC, (d) 12 molecule per UC, (e) 16 molecule per UC, (f) 20 molecule per UC, (g) 24 molecule per UC, (h) 28 molecule per UC, (i) 32 molecule per UC, (j) 36 molecule per UC, (k) 40 molecule per UC, (l) 44 molecule per UC. Lines are zeolite framework of 28Al model. The unit of density scale is the number of adsorbate molecules per ų.

One can observe that the benzene distributions at both sites are related with loading, as exhibited in Fig. 3. At relatively low loadings, the S sites are predominately occupied. This reveals that S site adsorption is initially energetically favorable, which can be confirmed by detailed investigation of types and energetics of adsorption sites of benzene/HY system in our previous study.²⁸ The density of benzene on the S site increased with the number of adsorbate molecules, agreeing with the unified understanding of earlier studies.⁶⁹ However, the loading dependence of the W site is more complex. As shown in Fig. 3(a)–(i), the occupancy of the W site increases at loadings from 1–36 molecules per UC. This is attributed to the saturation of energetically favorable S sites, as well as increased cooperative interaction between the adsorbed benzene molecules.²⁸ However, this trend is not followed by loadings approach to saturation, as shown in Fig. 3(j)–(l), showing that the occupation of W site declines with increasing loading. Therefore, overall, the density distribution of benzene molecules on the W sites follows a non-monotonous evolution with loading, first increases and then decreases with increasing loading. This finding is different from what commonly assumed for benzene/Y system,³⁵ in which considered that more benzene molecules are adsorbed on the W site at saturation loading. Although, their prediction was based on the relative energy of sites without verification from molecular adsorbate configurations. In order to further investigate the changing rule of W site occupancy versus loading for HY zeolite, the slice data generate from the density contours of benzene are discussed in the following section.

3.3 Slice data from contour maps of density

Fig. 4(a)–(c) illustrate the position of the slice, which is specially chosen to reveal the loading dependence of W site. As we can see, the slice is perpendicular to the xz plane and crosses the centroids of six 12-T rings. The representative density contour of benzene on the slice are shown in Fig. 3(d), with numbers denoting six centroid of 12-T ring passed by the slice.

Fig. 4 (a) The position of the slice, which is perpendicular to the xz plane of the unit cell and crosses the centroids of six 12-T rings, (b) the centroid of a 12-T ring, (c) the centroid of 12-T rings in a unit cell, six of them passed by the slice are colored purple and others are colored yellow for clarity. (d) The density contour of benzene on the slice with numbers showing positions of six centroid of 12-T ring passed by the slice. Accessible and blocked places of the structure are shown as white and black.

Fig. 5 shows density contours of benzene molecules on the slice reveals in Fig. 4 for four zeolite models at different loadings. As visualized in Fig. 5, the adsorption probability of benzene molecules on the W site tends to grow and decline with increasing pore filling for all the zeolite models. This observation strongly supports the results given in Fig. 3. Meanwhile, because of the decrease of occupancy on the W site, the occupancy of adsorbates inside the supercage must be more intensive at high loadings. All these indicate that there might be a change of adsorption mechanism with loading.

Fig. 5 Density contours of benzene molecules on the slice through the system of four zeolite models, at loadings of (a) 12 molecule per UC, (b) 24 molecule per UC, (c) 32 molecule per UC, (d) 36 molecule per UC, (e) 40 molecule per UC, (f) 44 molecule per UC, respectively. Accessible and blocked places of the structure are shown as white and black. The unit of density scale is the number of adsorbate molecules per ų.

Generally speaking, three possible distributions exist for benzene molecules as illustrated in Scheme 1: (A) adsorbed benzene molecules remain at their original positions and the additional benzene is located ideally at another site, i.e., ideal adsorption; (B) adsorbed benzene molecules remain at their original positions and the later one is inserted into the space between any two of them, named “insertion adsorption”; (C) the later benzene significantly affects the adsorption of the adsorbed adsorbates, named “rearrangement adsorption”. Till now, the detailed adsorption mechanism of benzene/HY system is still ambiguous, and a further investigation of the benzene distribution inside the supercage is needed.

Scheme 1 Possible adsorption distribution patterns for benzene/HY zeolite system, (A): ideal adsorption; (B): insertion adsorption with two situation, (B1) represent insertion on supercage center side which close to the center of the supercage, (B2) represent insertion on zeolite wall side near the wall of zeolite; (C): rearrangement adsorption. The benzene molecule adsorbed later is yellow.

3.4 Relative concentration plots

As a concise but comprehensive illustration of the distribution of adsorbate, the relative concentration plots P(x) of the center of mass of benzene (COM) along the x-direction of the zeolite models are illustrated in Fig. 5. At low and medium loadings (from infinite dilute up to above 32 molecule per UC) for all the models, P(x) exhibits identical distributing curves for each zeolite model, indicating that the adsorption sites revealed at infinite dilution remain unchanged in the vast majority of loadings. Each unit cell of FAU is composed of eight supercages, which affirms that at least four benzene molecules are able to adsorb on the adsorption site of HY zeolite almost unperturbed. At this stage, the adsorption mechanism is clearly belonging to “ideal adsorption”. However, at loadings approach to saturation (corresponding loadings are colored red in Fig. 5), the P(x) changes obviously, representing different adsorption behavior compared with lower loadings. This confirms that the adsorption mechanism of benzene in HY zeolite is comprised of two stages for different loadings, separated by an inflection point (named “I-P” hereafter). This finding has never been reported so far.

3.5 Radial distribution functions

Different adsorbate distributions result in different radial distribution functions (RDFs), which can provide a good measure of the probability that, given the presence of specie at the origin of arbitrary reference specie. The resulting function is commonly given as g(r), and is calculated as defined from:where r is the distance between species i and j, ΔN_ij(r, r + Δr) is the ensemble averaged number of species j around i within a shell from r to r + Δr, V is the system volume, and N_i and N_j are the numbers of species i and j, respectively.

Fig. 7 shows the COM–COM RDFs of benzene molecules for the four models. In Fig. 7, the main peak centered at 0.55 nm is taken to be benzene molecules adsorbed on a conventional site. The position of this peak is uniform for all g(r) regardless of loading, which demonstrates that the adsorbed benzene molecules at loadings above the inflection point were incapable of changing the existing adsorbate distribution. Therefore, the rearrangement adsorption mechanism (pattern C in Scheme 1) characterized by a variation of distance between benzene molecules can be eliminated.

For loadings above the I-P, two small peaks around 0.45 nm and 0.78 nm appear. As anticipated, the changing features of the COM–COM RDFs versus loading share the I-P with the P(x) of Fig. 6, which further validates the two-stage mechanism. These small peaks represent later adsorbed benzene molecules located at positions closer than the former positions, which excludes the ideal adsorption mechanism (pattern A) leading to an identical RDF feature. The new peaks cannot be attributed to adsorption on W sites as well, because of the lower W site adsorption density demonstrated in Fig. 3 and 5. Therefore, adsorption of these later benzene molecules must be represented by the insertion mechanism (pattern B). This configuration was considered to be due to confinement of the supercage and the stronger repulsive adsorbate–adsorbate interaction, preventing more benzene molecules from perfectly locating in the unit cell.

Fig. 6 Relative concentration P(x) of the center of mass of benzene (COM) at different loadings along the x-direction of the unit cell of (a) 0Al, (b) 14Al, (c) 28Al, and (d) 56Al. Black and red numbers given in the parentheses are loadings corresponding with the curves below and above the I-P in molecule per UC.

Fig. 7 RDFs of COM–COM for four model at various loadings.

For the newly reported insertion mechanism, a detailed description of the position of the inserted benzene molecule (inserted on the supercage center side or wall side), as shown by B1 and B2, respectively, in Scheme 1 is also of interest. In order to explore the location of the benzene molecule, the RDFs between COM and the supercage center (Supercage^c) as well as the wall atoms of the HY zeolite were examined. The RDFs of COM–Supercage^c for four models are shown in Fig. 8. The population of benzene near the cage center is generally observed as zero for all loadings. This finding is confirmed by density contour planes passing through Supercage^c at loadings of 8, 20, and 40 molecules per UC, as shown in Fig. 9, which indicate that the cage center is energetically very unfavorable. Changes of g(r) in Fig. 8 can also be separated by the I-P. Under the I-P, the one and only peak location remained at 0.38–3.9 nm and increasingly broadened with increasing loading. This result indicates that benzene adsorbed on the same sites 0.38–0.39 nm away from Supercage^c with a gain in intensity of adsorption closer to the cage center, which is consistent with Fig. 8(a) and (b). Above the I-P, a shoulder at around 0.31–0.32 nm is observed to appear in Fig. 8. Although the location of this shoulder seems similar to the broadening of the one and only peak shown below the I-P, it can be further identified by the RDFs of C_ben–Supercage^c (Fig. SI1, ESI†) with an obvious new peak centered around 0.20 nm at a loading of 40 molecules per UC compared with the single peak observed at a loading of 1 molecule per UC. This shoulder represents insertion molecules inside the supercage, which are adsorbed closer to Supercage^c than the benzene molecules on stable sites. That is, the inserted molecules locate on the center side of the supercage, which is in line with adsorption pattern B1 in Scheme 1. Meanwhile, the main peak in Fig. 8 shift to longer distance from Supercage^c, this is attributed to the inserted benzene pushing adsorbed molecules towards the zeolite framework. This observation can be further confirmed by snapshots of adsorption, as well as RDFs of COM and H1 protons of HY zeolite in ESI (see Fig. SI2†). Remarkably the main peak become more pronounced for loadings following insertion mechanism, indicating a less broad probability distribution within the pore, consistent with more localized density contour at 40 molecule per UC (Fig. 9(c)) compared with 20 molecule per UC (Fig. 9(b)). This is somewhat unusual, manifesting that the benzene molecules capture insertion sites have positive consequences on the distribution order of adsorbates. We argue that this is due to inserted benzene molecule taking other adsorbate capture ideal sites as a whole, which restricts independent change of adsorption in this whole.

Fig. 8 RDFs of COM–Supercage^c for benzene in each model at various loadings.

Fig. 9 Density contours of benzene on a plane passing through the Supercage^c of 28Al at three loadings: (a) 8 molecule per UC, (b) 20 molecule per UC, and (c) 40 molecule per UC.

From the study of RDFs and snapshots of adsorbates, a two-stage adsorption mechanism is proposed as ideal adsorption and insertion adsorption for loadings below and above the inflection point, as illustrated in Fig. 10(a). At loadings below the inflection point, benzene molecules are adsorbed stably on the conventional sites. Representative equilibrium geometry at a loading of 32 molecules per UC is shown in Fig. 10(b). A greater tendency for locating at a W site as well as near the Supercage^c is also observed, which, however, has no influence on benzene adsorption at the S site. Representative equilibrium geometry at a benzene loading of 40 molecules per UC is shown in Fig. 10(c). This snapshot further confirms that the inserted benzene is adsorbed on the Supercage^c side, which leads to no obvious rearrangement but pushes the originally adsorbed benzene molecules closer to the supercage.

Fig. 10 (a) Two-stage adsorption mechanism for benzene in HY zeolite. Snapshot of a representative equilibrium geometries of benzene in a supercage of 28Al model at the loading of (b) 32 molecule per UC denoted ideal adsorption, (c) 40 molecule per UC denoted insertion adsorption.

3.6 Dependence of I-P on Si : Al

Besides the revealing of adsorption mechanism, one can also abserve that the position of the I-P in Fig. 6 and 7 is slightly shifted to higher benzene concentration with decreasing Si : Al ratio: from 32 molecules per UC (4 molecules per supercage) to 40 molecules per UC (5 molecules per supercage) for 0Al, 14Al, 28Al, and 56Al, indicating that the adsorption of the adsorbates is less crowded in the model with lower Si : Al. On the basis of a simple reasoning in terms of zeolite Si : Al ratio, this is rather unexpected, since the available space in the framework decreases with decreasing Si : Al ratio, due to increase number of protruded H1 protons, which means the adsorption of the adsorbates should be more crowded in the model with lower Si : Al, and the inflection point should come at lower loading. The underlying reason of the dependance of I-P will be discussed in the following.

Based on the framework structure of FAU (cages connected by narrow 12-T windows), the utilization of the 12-T rings become the center of interest in the loading dependence of I-P for different models. An examination of the equilibrium snapshots of adsorbates reveals another site in addition to W site (Fig. 11(a)) that is able to make use of the 12-T ring space: benzene adsorption facially on the H1 atoms on the 12-T ring and inserting into the 12-T ring (H1 site, Fig. 10(b)) for 14Al, 28Al and 56Al model. This can be visualized by density contours concerning the 12-T ring in Fig. 5 and 12, which are directly affected by the location of H1 atoms on the 12-T ring. As shown in Fig. 5 and 12, at loadings near saturation, the adsorption probability of benzene molecules around the 12-T window tends to decline with increasing loading. Nevertheless, the magnitude of declination is different for four zeolite models. That is, the utilization of 12-T ring increased with decreasing Si : Al, which resulted in less crowded adsorption in the model with lower Si : Al, and might be the reason of the unexpected phenomena: I-P comes at lower loading for zeolite with higher Si : Al. What's worth mentioning is that Fig. 5 showed roughly that the occupation on W site decrease to almost zero at loadings near saturation for all the four models, therefore the increased utilization of 12-T ring with decreasing Si : Al, is mainly caused by increased contribution of H1 site. The detailed distribution of benzene around the 12-T ring can be further elucidated from the benzene radial distribution functions g(r).

Fig. 11 Two representative snapshots of benzene molecule on 12-T sites from 14Al at 24 molecule per UC: (a) W site, (b) H1 site. The dashed lines denote the 12^c.

Fig. 12 Density contours of benzene on the plane of a representative 12-T ring at (a) 1 molecule per UC, (b) 4 molecule per UC, (c) 8 molecule per UC, (d) 12 molecule per UC, (e) 16 molecule per UC, (f) 20 molecule per UC, (g) 24 molecule per UC, (h) 28 molecule per UC, (i) 32 molecule per UC, (j) 36 molecule per UC, (k) 40 molecule per UC, and (l) 44 molecule per UC in four models.

The g(r) for the COM and the centroid of the 12-T ring (12^c): COM–12^c are shown as Fig. 13. For all models, two separated peaks at r = 0 and 0.33–0.45 nm exists in the g(r) of COM–12^c referring to adsorption of benzene onto W and S site, in accordance with the observation in Fig. 3. For 14Al, 28Al, and 56Al models containing proton atoms, another peak is seen at distance slightly longer than the window site (r = 0.14–0.16 nm), corresponding with H1 site that can take advantage of the 12-T window. As shown in Fig. 13, the first peak decrease to almost zero at loadings near saturation, this confirms decreased adsorption on W site illustrating by Fig. 3 and 5. With the decline of Si : Al, the relative magnitude of the second peaks of Fig. 13 corresponding with H1 site increases compared with the first one at a fixed loading, especially for loadings near saturation. This confirms that at high loadings, the increased utilization of 12-T ring caused by the contribution of H1 site, can result in more available space for zeolite with lower Si : Al, which makes the I-P comes at higher loading for zeolite with lower Si : Al.

Fig. 13 RDFs for the COM around the 12^c at various benzene loadings for (a) 0Al, (b) 14Al, (c) 28Al, and (d) 56Al.

3.7 Fraction on each site

Although we focus on the changing of adsorption mechanisms in this paper, the exact fraction of benzene molecules found at different sites are also of interest. We believe that our analysis confirmed an ideal adsorption for loadings below the I-P, which means almost all of the benzene, would like to adsorbed on the S site rather than W site. That is to say, in an ideal circumstances, the action of benzene molecules found at S, insertion (IN), and W at loading X (X < I-P) would be 1 : 0 : 0. For loadings higher than I-P, the latter adsorbed benzene tends to insert into the supercage (IN site) rather than adsorbed on W site. That is to say, ideally the action of benzene molecules found at S, IN, and W at loading X (X > I-P) would be I: (X-I): 0 (I = the loading of I-P, that is 32 molecule per UC for 0Al model).

Fig. 14 shows represented low energy snapshots of benzene on 0Al model at 298 K. Table 1 provide the fraction of benzene molecules at S, W, and the IN sites for ideal circumstances as well as for the represented low energy snapshots shown in Fig. 14. As we can see, the proportions of sites for low energy snapshots are roughly consistent with ideal circumstances and become more complicated. This further confirms our results: almost all of the benzene molecules follow ideal adsorption mechanism at loadings below I-P; and benzene tends to insert into the supercage for higher loadings.

Fig. 14 Represented low energy snapshots of benzene on 0Al model at 300 K: (a) 1 molecule per UC, (b) 4 molecule per UC, (c) 8 molecule per UC, (d) 12 molecule per UC, (e) 16 molecule per UC, (f) 20 molecule per UC, (g) 24 molecule per UC, (h) 28 molecule per UC, (i) 32 molecule per UC, (j) 36 molecule per UC, (k) 40 molecule per UC, and (l) 44 molecule per UC. Benzenes on S sites are colored by atoms (grey for C atoms and white for H atoms); on W and IN sites are colored by green and purple, respectively.

Table 1 The fraction of benzene molecules at S, IN and W sites for ideal circumstances as well as for represented low energy snapshots of 0Al

Loading (mole. per UC)	Ideal circumstances			Snapshots
	S	IN	W	S	IN	W
1	1	0	0	1	0	0
4	1	0	0	1	0	0
8	1	0	0	1	0	0
12	1	0	0	1	0	0
16	1	0	0	0.87	0.13	0
20	1	0	0	0.90	0.05	0.05
24	1	0	0	0.92	0.08	0
28	1	0	0	0.89	0.11	0
32	1	0	0	0.91	0.09	0
36	0.89	0.11	0	0.89	0.08	0.03
40	0.80	0.20	0	0.82	0.18	0
44	0.73	0.27	0	0.77	0.23	0

---

4. Conclusion

MMC was revealed to be a powerful tool for study of the benzene adsorption distribution with respect to benzene loading in HY zeolites. On the contrary to what commonly assumed, the density distribution of benzene molecules on the W sites first increased then decreased with increasing of loading at loadings approach saturation. We also found that adsorbate tend to be more ordered at loadings following insertion mechanism, which is hard to envision. The adsorption mechanism for benzene in HY zeolites was positively identified. The adsorption of benzene is dominated by an ideal adsorption mechanism for the vast of the loading range. This finding was confirmed by RDFs that demonstrated invariable main peak positions versus adsorbate loading. For loadings approach to saturation, an insertion mechanism was suggested for benzene, so called because the adsorbates keep organized and no obvious rearrangement of the originally adsorbed benzene molecules was observed. The later benzene molecules were found to be located in the supercage near the cage center and inserted into the clusters formed by the adsorbed adsorbates, which resulted in the adsorbed molecules moving toward the wall of the supercage. The above viewpoints were generally observed for all zeolite models with different Si : Al ratios. However, the inflection point (loading) for the adsorption mechanism increases with decreasing Si : Al ratio. This is attributed to benzene adsorbed on the H1 site inserting into the 12-T ring, which result in more available space for zeolite with lower Si : Al.

Although the I-P comes at quite high loading for benzene in HY zeolite, this work suggests that it may be possible to realize change of adsorption mechanisms at lower loadings for other aromatics larger than benzene. This study points out the necessary of adsorption investigations at high loadings to get an explicit adsorption mechanism, especially for adsorbates with suitable size in cage-like zeolite. As a consequence, this may affect the dynamics and reaction phenomena of guest molecules inside the pore matrix, the impact of our findings for future applications deserves further investigation.

Acknowledgements

The authors acknowledge the financial support of the National Natural Science Foundation of China (Grant No. 21236009, 21336011, and 21476260).

References

1. R. A. Van Santen and G. J. Kramer, Chem. Rev., 1995, 95, 637.
2. J. Pisson, N. Morel-Desrosiers, J. P. Morel, A. de Roy, F. Leroux, C. Taviot-Gueho and P. Malfreyt, Chem. Mater., 2011, 23, 1482.
3. R. Millini, F. Frigerio, G. Bellussi, G. Pazzuconi, C. Perego, P. Pollesel and U. Romano, J. Catal., 2003, 217, 298.
4. H. Hattori, Chem. Rev., 1995, 95, 537.
5. R. Gounder, A. J. Jones, R. T. Carr and E. Iglesia, J. Catal., 2012, 286, 214.
6. G. Busca, Chem. Rev., 2007, 107, 5366.
7. A. T. To, R. E. Jentoft, W. E. Alvarez, S. P. Crossley and D. E. Resasco, J. Catal., 2014, 317, 11.
8. F. C. Jentoft, J. Krohnert, I. R. Subbotina and V. B. Kazansky, J. Phys. Chem. C, 2013, 117, 5873.
9. M. F. Ciraolo, J. C. Hanson, B. H. Toby and C. P. Grey, J. Phys. Chem. B, 2001, 105, 12330.
10. J. H. Lunsford, P. N. Tutunjian, P. J. Chu, E. B. Yeh and D. J. Zalewski, J. Phys. Chem., 1989, 93, 2590.
11. C. E. Bronnimann and G. E. Maciel, J. Am. Chem. Soc., 1986, 108, 7154.
12. M. Czjzek, H. Jobic, A. N. Fitch and T. Vogt, J. Phys. Chem., 1992, 96, 1535.
13. C. Pichon, A. Méthivier, M.-H. Simonot-Grange and C. Baerlocher, J. Phys. Chem. B, 1999, 103, 10197.
14. E. N. Coker, C. Jia and H. G. Karge, Langmuir, 2000, 16, 1205.
15. B. Smit and J. I. Siepmann, J. Phys. Chem., 1994, 98, 8442.
16. B. Smit and A. Krishna, Chem. Eng. Sci., 2003, 58, 557.
17. R. Q. Snurr, A. T. Bell and D. N. Theodorou, J. Phys. Chem., 1993, 97, 13742.
18. H. Klein, C. Kirschhock and H. Fuess, J. Phys. Chem., 1994, 98, 12345.
19. V. Lachet, A. Boutin, B. Tavitian and A. H. Fuchs, Langmuir, 1999, 15, 8678.
20. A. Takahashi, F. H. Yang and R. T. Yang, Ind. Eng. Chem. Res., 2002, 41, 2487.
21. Y. P. Zeng, S. G. Ju, W. H. Xing and C. L. Chen, Ind. Eng. Chem. Res., 2007, 46, 242.
22. A. H. Fuchs and A. K. Cheetham, J. Phys. Chem. B, 2001, 105, 7375.
23. P. Santikary, S. Yashonath and G. Ananthakrishna, J. Phys. Chem., 1992, 96, 10469.
24. L. M. Bull, N. J. Henson, A. K. Cheetham, J. M. Newsam and S. J. Heyes, J. Phys. Chem., 1993, 97, 11776.
25. A. De Mallmann and D. Barthomeuf, J. Phys. Chem., 1989, 93, 5636.
26. C. Bremard, G. Ginestet, J. Laureyns and M. Le Maire, J. Am. Chem. Soc., 1995, 117, 9274.
27. C. Brémard, G. Ginestet and M. Le Maire, J. Am. Chem. Soc., 1996, 118, 12724.
28. H. M. Zheng, L. Zhao, Q. Yang, J. S. Gao, B. J. Shen and C. M. Xu, Ind. Eng. Chem. Res., 2014, 53, 13610.
29. A. Fitch, H. Jobic and A. Renouprez, J. Phys. Chem., 1986, 90, 1311.
30. A. De Mallmann and D. Barthomeuf, Stud. Surf. Sci. Catal., 1986, 28, 609.
31. B. L. Su and D. Barthomeuf, J. Catal., 1993, 139, 81.
32. A. K. C. L. M. Bull, B. M. Powell, J. A. Ripmeester and C. I. Ratcliffe, J. Am. Chem. Soc., 1995, 117, 4328.
33. L. J. Ma, S. B. Liu, M. W. Lin, J. F. Wu and T. L. Chen, J. Phys. Chem., 1992, 96, 8120.
34. A. N. F. Hervé Jobic and J. Combet, J. Phys. Chem. B, 2000, 104, 8491.
35. I. Daems, A. Méthivier, P. Leflaive, A. H. Fuchs, G. V. Baron and J. F. Denayer, J. Am. Chem. Soc., 2005, 127, 11600.
36. F. Jousse, S. M. Auerbach and D. P. Vercauteren, J. Phys. Chem. B, 1998, 102, 6507.
37. J. J. M. Beenakker and I. Kuscer, Zeolites, 1996, 17, 346.
38. D. J. Keffer, H. T. Davis and A. V. McCormick, Adsorption, 1996, 2, 9.
39. L. J. Criscenti, R. T. Cygan, A. S. Kooser and H. K. Moffat, Chem. Mater., 2008, 20, 4682.
40. R. Babarao and J. Jiang, J. Am. Chem. Soc., 2009, 131, 11417.
41. J. W. Singer, A. O. Yazaydin, R. J. Kirkpatrick and G. M. Bowers, Chem. Mater., 2012, 24, 1828.
42. A. Nalaparaju, X. Zhao and J. Jiang, J. Phys. Chem. C, 2010, 114, 11542.
43. T. P. Caremans, T. S. van Erp, D. Dubbeldam, J. M. Castillo, J. A. Martens and S. Calero, Chem. Mater., 2010, 22, 4591.
44. D. Peralta, K. Barthelet, J. Pérez-Pellitero, C. Chizallet, G. Chaplais, A. Simon-Masseron and G. D. Pirngruber, J. Phys. Chem. C, 2012, 116, 21844.
45. P. Gallezot, R. Beaumont and D. Barthomeuf, J. Phys. Chem., 1974, 78, 1550.
46. S. P. Bates, W. J. van Well, R. A. van Santen and B. Smit, J. Am. Chem. Soc., 1996, 118, 6753.
47. D. Dubbeldam and B. Smit, J. Phys. Chem. B, 2003, 107, 12138.
48. M. A. Granato, T. J. Vlugt and A. E. Rodrigues, Ind. Eng. Chem. Res., 2007, 46, 321.
49. R. Krishna, J. Phys. Chem. C, 2009, 113, 19756.
50. Y. S. Bae, A. O. z. r. Yazaydın and R. Q. Snurr, Langmuir, 2010, 26, 5475.
51. T. J. Vlugt and M. Schenk, J. Phys. Chem. B, 2002, 106, 12757.
52. W. Zhu, F. Kapteijn, J. Moulijn, M. Den Exter and J. Jansen, Langmuir, 2000, 16, 3322.
53. L. A. Clark and R. Q. Snurr, Chem. Phys. Lett., 1999, 308, 155.
54. T. J. Vlugt and M. Schenk, J. Phys. Chem. B, 2002, 106, 12757.
55. R. J. Pellenq, B. Tavitian, D. Espinat and A. H. Fuchs, Langmuir, 1996, 12, 4768.
56. H. Sun, J. Phys. Chem. B, 1998, 102, 7338.
57. S. M. Auerbach, N. J. Henson, A. K. Cheetham and H. I. Metiu, J. Phys. Chem., 1995, 99, 10600.
58. S. M. Auerbach, L. M. Bull, N. J. Henson, H. I. Metiu and A. K. Cheetham, J. Phys. Chem., 1996, 100, 5923.
59. C. F. Mellot, A. K. Cheetham, S. Harms, S. Savitz, R. J. Gorte and A. L. Myers, J. Am. Chem. Soc., 1998, 120, 5788.
60. C. F. Mellot, A. M. Davidson, J. Eckert and A. K. Cheetham, J. Phys. Chem. B, 1998, 102, 2530.
61. Z. Du, G. Manos, T. J. Vlugt and B. Smit, AIChE J., 1998, 44, 1756.
62. B. Smit and J. I. Siepmann, Science, 1994, 264, 1118.
63. I. Kusaka, M. Talreja and D. L. Tomasko, AIChE J., 2013, 59, 3042.
64. S. S. Jirapongphan, J. Warzywoda, D. E. Budil and A. Sacco, Microporous Mesoporous Mater., 2007, 103, 280.
65. M. Fermeglia and S. Pricl, AIChE J., 2001, 47, 2371.
66. F. Jousse, S. M. Auerbach and D. P. Vercauteren, J. Phys. Chem. B, 2000, 104, 2360.
67. P. Bai, J. I. Siepmann and M. W. Deem, AIChE J., 2013, 59, 3523.
68. H. M. Zheng, L. Zhao, J. J. Ji, J. S. Gao, C. M. Xu and F. Luck, ACS Appl. Mater. Interfaces, 2015, 7, 10190.
69. Y. Zeng, S. Ju, W. Xing and C. Chen, Ind. Eng. Chem. Res., 2007, 46, 242.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: 10.1039/c6ra02338j

---