Differential Penetration of Ethanol and Water in Si-Chabazite: High Pressure Dehydration of Azeotrope Solution

This study is aimed to shed light on the mechanisms at the basis of the differential penetration of alcohol and water in hydrophobic zeolites at ambient (P amb ) and non-ambient pressure. Here we report the effects of the penetration of water and alcohol in an all-silica chabazite (Si-CHA) compressed with an ethanol/water azeotrope solution (ethanol : water = 95.63 : 4.37 by mass %). We collected in situ synchrotron X-ray Powder Diffraction (XRPD) data in order to monitor the structural modifications induced by the fluid penetration and to investigate the guest-guest and host-guest interactions. First principles molecular dynamics simulations allowed to complete the structural description at high pressure, providing an atomistic level description of the guest-guest hydrogen bond network. For a comprehensive understanding of the processes involving the Si-CHA + azeotrope interactions, both the zeolite and the alcohol/water solution were firstly investigated separately under pressure. The results obtained prove that both H 2 O and ethanol penetrate Si-CHA porosities even at P amb . However, while in these conditions the H 2 O /ethanol ratio adsorbed inside Si-CHA is similar to that of the external azeotrope solution, under pressure the zeolite extra-framework content corresponds to a Abstract This study is aimed to shed light on the mechanisms at the basis of the differential penetration of alcohol and water in hydrophobic zeolites at ambient (P amb ) and non-ambient pressure. Here we report the effects of the penetration of water and alcohol in an all-silica chabazite (Si-CHA) compressed with an ethanol/water azeotrope solution (ethanol : water = 95.63 : 4.37 by mass %). We collected in situ synchrotron X-ray Powder Diffraction (XRPD) data in order to monitor the structural modifications induced by the fluid penetration and to investigate the guest-guest and host-guest interactions.

This study is aimed to shed light on the mechanisms at the basis of the differential penetration of alcohol and water in hydrophobic zeolites at ambient (P amb ) and non-ambient pressure. Here we report the effects of the penetration of water and alcohol in an all-silica chabazite (Si-CHA) compressed with an ethanol/water azeotrope solution (ethanol : water = 95.63 : 4.37 by mass %). We collected in situ synchrotron X-ray Powder Diffraction (XRPD) data in order to monitor the structural modifications induced by the fluid penetration and to investigate the guest-guest and host-guest interactions. First principles molecular dynamics simulations allowed to complete the structural description at high pressure, providing an atomistic level description of the guest-guest hydrogen bond network. For a comprehensive understanding of the processes involving the Si-CHA + azeotrope interactions, both the zeolite and the alcohol/water solution were firstly investigated separately under pressure. The results obtained prove that both H 2 O and ethanol penetrate Si-CHA porosities even at P amb . However, while in these conditions the H 2 O /ethanol ratio adsorbed inside Si-CHA is similar to that of the external azeotrope solution, under pressure the zeolite extra-framework content corresponds to a composition much richer in H 2 O than the azeotrope one. Hence, our results suggest that a dehydration effect occurred on the azeotrope solution, promoted by pressure. In addition, the experiment performed to test the elastic behavior of Si-CHA with a non-penetrating pressure transmitting medium interestingly indicates that Si-pure chabazite is the most compressible zeolite among those up to now studied in silicone oil.

Introduction
The shape selective properties of zeolites are at the basis of their success in adsorption processes and catalytic activity. All these applications depend on the size and shape of the porous network of the zeolite and on its chemical nature. One of the challenges in renewable energy fuel production is the purification of ethanol from water. Since the biofuel products are typically dilute alcohol-inwater solutions, an energy efficient alcohol-water separation technology is required to generate fuel-grade alcohols. The use of zeolites for ethanol/water separation has been widely explored in the last years [1] [2]. In particular, hydrophilic zeolites are suited for the separation of water from alcohol. For instance, zeolite A membranes are used for industrial-scale dehydration of ethanol to produce fuel-grade ethanol [3]. On the contrary, hydrophobic zeolites are in general exploited for removing ethanol from water, when ethanol is the minority component (e.g. MFI silicalite and ZSM-5 [4]). While the water/ethanol separation in ethanol rich solutions by hydrophilic zeolites is extremely effective (e.g. LTA membranes [5]), the ethanol/water separation performances in H 2 O rich solutions operated by hydrophobic zeolites are definitely worse. This is due to the presence of silanol defects, or Al hydrophilic sites -accidentally present in "nominally silicatic" zeoliteswhich favor the adsorption of water molecules during the purification process.
In the last years, high attention has been devoted to the interaction of water with hydrophobic porous matrices, both in term of applied [6] [7] [8] and fundamental research [9] [10]. It was shown that water can be intruded under pressure into hydrophobic zeolites by using water/alcohols mixtures as pressure transmitting media [11]. Interestingly, it has been demonstrated that the hydrophobic all-silica zeolite ferrierite (Si 36 O 72 ) (Si-FER), once compressed in an alcohols/water mixture -16:3:1 methanol/ethanol/water -shows higher affinity for water than for alcohols. Under such conditions, only water enters the two-dimensional (2D) channel system of ferrierite forming stable aggregates [10]. Another very interesting result was obtained inducing the intrusion of an ethanol/water solution richer in water (ethanol: water = 1:3) into Si-FER under pressure: both components penetrate zeolite cavities, but they are segregated in different channels. While the water molecules occupy only ferrierite 6MR channels, the ethanol molecules are located in the 10MR channels, with the C-C bonds nearly perpendicular to the channel axis, thus forming wires of hydrogen-bonded dimers [9].
These results indicate that the combined effects of pressure and shape constraints can induce the formation of organized arrangements of small molecules in the zeolite porosities and build structural complexity in two dimensions. Such a supramolecular shaping effect, combined with the irreversibility of the encapsulation process, could be a more general feature of the high-pressure behavior of open-framework silicates [12], with possible implications of broad technological relevance for other classes of porous materials, including hybrid-zeolites, ordered mesoporous (organo) silicas, and MOF's [13].
To better understand the zeolite shape-directing action in separating strongly hydrogen-bonded liquid mixtures into their constituents, the influence of different framework geometries should be considered. For this reason, we decided to investigate the behavior of all-silica zeolites with a tridimensional channel system characterized by the presence of large cages. In this paper we report the intrusion of an azeotrope solution (ethanol : water = 95.63 : 4.37 by mass %) in an all silica chabazite (Si-CHA) under pressure. We chose the azeotrope solution due to its peculiar physicalchemical properties and to its applicative interest. For a comprehensive understanding of the interactions involving the Si-CHA/azeotrope system, both the pure zeolite and the azeotrope solution were firstly studied separately under pressure. In situ synchrotron X-ray Powder Diffraction (XRPD) data were collected in order to monitor the structural modifications occurring during the injection of the fluid. In addition, since with this study we intend to shed light on the different affinity of hydrophobic zeolites toward alcohol and water under non-ambient pressure, we have adopted a complementary computational approach. Indeed, computational chemistry can offer a microscopic and local viewpoint that may complete the average structural information obtained by XRPD experiments. Additionally, the computational approach allows one to build suitable model systems, and extract from them atomistic-level information with predictive value [14]. Important progress have been made using both simplified approaches -designated as "geometric methods" (see e.g. [15] [16]), and atomistic-level methodologies (which could be based on empirical forcefields [17] [18] or "first-principles" electronic structure approaches (see [12] for a recent account on first-principles methods applied to zeolites). All these theoretical techniques are nowadays well known and widely adopted tools in the microporous materials communities. The main goals of our simulations have been to elucidate, at molecular level, the guest-guest and host-guest interactions and to compare the structural features of the confined aggregates with the emblematic case of Si-FER. Our integrated experiments and simulations show that a full separation of water from ethanol does not occur in the CHA cavities. Yet, we observe an intriguing phenomenon that, to the best of our knowledge, has been never disclosed to date: the pressure promotes dehydration of the azeotrope solution remaining after compression of the zeolite, and we provide a plausible motivation of its molecular-level cause.

Chabazite Structure
Chabazite framework can be described as an ABC sequence of double 6-rings (D6R) of tetrahedra linked through single 4-rings [19]. The resulting three-dimensional pore system presents cages (3 per unit cell) with 8MR pore openings of 0.38 nm (CHA-cage). The topological symmetry, corresponding to real symmetry, in Si-CHA is rhombohedral 3 .

Sample Preparation
Pure Si-CHA was synthesized using the method reported by Diaz-Cabanas et al. [20], and then characterized by different techniques (thermogravimetric analysis, nitrogen adsorption−desorption and 29 Si solid-state NMR spectroscopy), as reported in detail in the paper of Confalonieri and coworkers [21]. Si-CHA used in this work has chemical formula Si 36

XRPD data Collection
In situ HP-XRPD experiments were performed compressing Si-CHA in a modified Merrill−Basset Diamond Anvil Cell (DAC) [22], using as Pressure Transmitting Medium (PTM) the azeotrope

Data Analysis
Collected images were integrated using Dioptas software [25]. Crystal structure of Si-CHA compressed in EtOH96 and in silicone oil at ambient pressure was refined by Rietveld method, using GSAS [26] package with the EXPGUI interface [27]. The rise of one new peak at 1.42 GPa for Si-CHA compressed in s.o. indicated the occurrence of a phase transition. This pattern was indexed using Expo2014 software [28] and the results indicated a transition from 3 to 1 s.g..
The obtained cell parameters were refined by GSAS-II program [29] using Le Bail method only up to 3.47 GPa due to the partial amorphization of the sample at higher pressure, which was maintained after pressure release too. The unit cell parameters of Si-CHA compressed in EtOH96 were determined for the whole pressure range investigated and upon pressure release to 0.37 GPa.
The structural refinements were performed only up to 1.84 GPa and upon decompression to 0.37 GPa, due to the low quality of the patterns collected at higher pressure. Since no phase transition occurred, the 3 s.g. was adopted for all the structure refinements, using as starting model that proposed by Confalonieri and coauthors [21]. The profile fittings were performed between 2.5 and 26.5 2θ refining scale factor, a Chebyshev polynomial with 30 coefficients for the background, 2θshift and unit cell parameters. Peak profile was refined setting the peak cut-off as 0.1% of the peak maximum and choosing the Thompson pseudo-Voigt function [30]. Framework atoms positions were constrained imposing the Si-O distances to 1.60 Å (e.s.d. 0.02 Å) and the relative weight was gradually decreased during the refinements. Fourier difference map was inspected in order to locate the extra-framework species. The evaluation of the angles and the distances occurring between the maxima of the electronic density allowed distinguishing H 2 O molecules from ethanol ones. In addition, the identification of alcohol molecules was also validated by the good agreement between the refined fractional occupancies of carbon and oxygen atoms belonging to the same ethanol molecule. Restrains, softly weighed (f=10-100), were applied to distances and angles between atoms belonging to a single ethanol molecule. Framework oxygen thermal parameters were constrained to the same value and then refined. The same strategy was used for carbon and oxygen atoms belonging to ethanol molecules and for oxygen atoms of H 2 O molecules. The details of structural refinement parameters are reported in Table S1 of the Supporting Information. Unit cell parameters are reported in Table S2 and in Table 2 and Figure Tables S3, S4. Observed and calculated profiles of the refined patterns are shown in Figure S2, S3, S4 and S5.

DFT calculations and First principles molecular dynamics simulations
Within the Density-Functional-Theory (DFT) formalism, we have modelled the Si-CHA zeolite using a widely adopted exchange correlation functional [31], jointly with D2-dispersion corrections [32]. Such a "PBE-D2" combination of density functional approximation /dispersion correction has been widely used in silicate modeling, ensuring a good accuracy/cost compromise. Indeed, whereas recent benchmark studies [33] indicate slightly better results for the (dispersion-corrected) PBE-sol approach in the computation of zero-K structural parameters of neutral zeotypes, (dispersioncorrected) PBE works slightly better in the case of aluminophosphates [34]. Our choice of selecting PBE-D2 for the present investigation is justified by the fact that, in the case of the pressure-induced water-ethanol incorporation in Si-FER [9] [35], this theoretical approach provided an average roomtemperature framework structure in very good agreement with the X-ray refinements (see [9]).
Moreover, this theoretical protocol has been used in the modeling of several important processes in porous materials, from high-pressure phase transitions [ Calculations with the CPMD code were performed starting from the experimental cell parameters obtained from X-ray refinement at 1.84 GPa (a=b=13.547 Å; c=14.6742 Å, gamma=120). The size of the cell allowed for considering only the Gamma Point in the Brillouin zone sampling [50]. The initial configuration for the First principles molecular dynamics (FPMD) simulation was built using as a guess the atomic positions provided by the X-ray refinement at 1.  After the transition to the triclinic symmetry, a and b parameters and α angle decrease up to the highest investigated pressure, c and β remain almost constant while γ increases in the whole investigated pressure range. As a whole, a total volume contraction of 17% is observed in the pressure range P amb -3.47 GPa (Table 2).

High Pressure behavior of Si-CHA in silicone oil
On the basis of this ∆V variation Si-chabazite results to be the most compressible zeolite among those up to now studied in silicone oil, being even softer than silicalite [64].  Regardless its nature, this phase disappears upon pressure increase.

High pressure behavior of Si-chabazite compressed in EtOH96
Ethanol crystallization occurs between 1.84 and 2.44 GPa. In the corresponding 2D image, the diffraction rings appear very textured, but the peak positions well match those of the pure ethanol at 3 GPa proposed by Allan and Clarke [66]. No evidences of ice crystallization were observed, but this could be due to an ice amount lower than the XRPD detection limits.
The evolution of the cell parameters of Si-CHA compressed in EtOH96 is reported in Figure 4 and Table S1. Once Si-CHA is contacted with the PTM, even at P amb , a contraction along a axis and an expansion along the c direction with respect to the original value obtained in capillary, is observed [21]. During compression, a parameter only slightly increases up to 1.84 GPa and then slightly decreases remaining, however, close to the original value, while c parameter decreases in the whole P range. This behavior, as suggested by Gatta [67] , can be ascribed, to inter-tetrahedral tilting around the oxygens that act as ''hinges". In fact, as observed by [63] in a natural chabazite form The inspection of the Fourier difference map confirms the intrusion of some molecules of the PTM in the zeolite porosities even at P amb .  The comparison of Si-CHA reference structure at P amb [21]  in the upper part of the cage and one in the lower one or vice versa. One of the possible distributions is shown in Figure 5, where the three EtOH molecules interact each other through the hydroxyl groups (dashed line in Figure 5). In addition, also a new bond is established between W1 and one EtOH molecule.

Modeling results and discussion
In this section the computational results obtained on the system compressed at 1.84 GPa will be discussed in comparison to the experimental ones. In addition, the efficiency of CHA in adsorbing and separating ethanol and water will be compared to that of ferrierite, previously studied by our group [9]. We recall that the FER framework acts as a mold, by permanently converting with subnanometric precision a hydrogen-bonded fluid into regular supramolecular nanostructures [9].
The experimental results of this work show that CHA framework absorb both ethanol and H 2 O, but cannot induce a similar separation inside its channels.
This behavior can be interpreted on the basis of the simulation results. Figure 6 shows that the average atomic coordinates derived from FPMD, once symmetrized according to the 3 space group operations, nicely match the experimentally refined structure.
Hence, we may safely trust the atomistic-level information extracted from theory. This will allow obtaining a picture of the local structure, to be compared with the average structure obtained experimentally. The minimum energy structure, illustrated in Figure (Figure 7). These strong hydrogen bonds might well be among the molecular-level factors responsible of the intriguing experimental finding outlined in the previous sections, namely, that at 1.84 GPa the zeolite extraframework content corresponds to a composition much richer in H 2 O than that of the azeotrope, even in the absence of (H 2 O) n clusters.

Summary and concluding considerations
The results of this work show that, when Si-CHA is immersed in the ethanol/water azeotrope solution, it immediately absorbs both H 2 O and ethanol molecules, even at ambient conditions. Both our experimental and computational results demonstrate that, despite this zeolite is itself hydrophobic, the adsorption of alcohol molecules promotes H 2 O co-adsorption through hydrogen bond formation, confirming other literature results [76]. In order to allow the molecules penetration and accommodation, chabazite c axis undergoes a significant increase, while a axis is reduced with respect to the original value ( Figure 4). This behavior is modified when pressure is applied (0.20-1.84 GPa): as H 2 O and EtOH contents rise, the framework contracts along c and expands along a axis. This kind of evolution is correlated to the position of the EtOH molecules, whose C-C bond lies nearly perpendicular to the three-fold axis. According to Daems et al. [71], this specific configuration is adopted by CHA to accommodate high amounts of guest molecules.
This behavior is different from that observed for Si-ferrierite compressed in a mixture of ethanol:water =1:3. In this case, an almost complete separation between H 2 O and EtOH molecules occurs [9] with the penetration of the two species in two distinct parallel channels. In Si-CHA, At 2.44 GPa we observe the crystallization of ethanol from the PTM in the DAC. As discussed before, a higher pressure (more than 4.5 GPa, Figure 2) is necessary to crystallize EtOH and ice from the EtOH96 solution, while 1.9 GPa are needed to crystallize pure ethanol [66]. In a broader context, it would be of interest to gather further microscopic insight on the process investigated in this work, especially considering the innumerable applications of hydrophobic/slightly hydrophilic zeolites not only in water/alcohol separations [77], but also in industrial catalysis [78]. In this perspective, future joint experimental-computational studies of the role of the interface on the pressure-induced penetration of fluids in zeolite cavities would be instrumental in enhancing our knowledge on the mechanisms of forced intrusion processes.
Importantly, literature studies [79] [40] [41] have demonstrated that non covalent interactions between zeolite and adsorbate at the pore interface play a key role in facilitating the physisorption and then the entrance of potential guests inside zeolitic channels. Moreover, the flexibility of the framework -which allows for synchronous host-guest vibrational motions -is of utmost relevance for understanding the guest's behaviour at the pore interface.

Acknowledgements
We dedicate this paper to our friend and colleague Dr. Joel Patarin, internationally known for his pioneering researches in non-wetting fluid intrusion in zeolites. Authors thank the staff of BL04-               Figure S1. Evolution upon compression of cell volume of Si-CHA compressed in silicone oil. The value at ambient pressure is calculated in the rhombohedral setting. Figure S2. Observed (red dash marks) and calculated (green line) diffraction patterns and final difference curve (purple line) from