Adsorption equilibria of alkane-α, ω-diols (propane-1,3-diol, butane-1,4-diol, pentane-1,5-diol, and hexane-1,6-diol) from aqueous solution onto an all-silica zeolite of the type mordenite framework inverted (MFI, also known as silicalite-1) are obtained by simulations and experiments at T = 323 K and also for pentane-1,5-diol (C5) at 348 and 383 K. After an initial slow rise, isotherms at T = 323 K exhibit steep changes in loading, reaching saturation at 10, 9, 8, and 7 molec/uc as the number of carbon atoms of the diols increases from 3 to 6. The abrupt change in loading corresponds to a minimum in the free energy of adsorption (from vapor to zeolite) that is associated with a rapid rise in the number of hydrogen bonds per sorbate molecule due to the formation of large clusters. For C5 at low loading, the centers-of-mass primarily occupy the channel intersections with oxygens oriented along the straight channels where intermolecular hydrogen bonds are formed. At saturation loading, the C5 centers-of-mass instead occupy the straight and zig-zag channels, and nearly all C5 molecules are involved in a percolating hydrogen-bonding network (this also occurs for C6). With increasing temperature, the C5 isotherm decreases in steepness as the minimum in free energy of adsorption decreases in depth and a less-ordered structure of the adsorbed molecules results in a lower number of diol–diol hydrogen bonds. However, the C5 isotherm does not shift significantly in concentration of the adsorption onset, as the free energies of solvation and adsorption increase by similar and compensating amounts. At T = 323 and 348 K, the steep change for the C5 adsorption isotherm is found to be a phase transition (as indicated by a bimodal distribution of unit cell occupancies at intermediate loading) from a less-dense phase with only small hydrogen-bonded clusters to an ordered solid phase with loadings of 8 molec/uc. At T = 383 K, the sorbates are less ordered, the distribution of occupancies becomes unimodal at intermediate loading, and the loading rises more gradually with concentration. Several different enhanced sampling methods are utilized for these simulations.

Zeolites are crystalline materials composed of SiO4 and AlO4 tetrahedra, providing them with well-defined pores that can distinguish between alkanes of various lengths and degrees of branching. This, combined with their stability, has led to their use industrially in fluid catalytic cracking, isomerization, and alkylation processes.1,2 With recent interest in replacing petrochemical processes with renewable ones and improvement in material synthesis, high-silica zeolites (hydrophobic) have been exploited in the processing of biomass-derived mixtures.3–7 

Adsorption equilibria of mixtures derived from biomass are more complicated than those derived from petroleum. Accurate measurement is challenging because the mixtures are typically liquids and only the bulk solution phase is open to direct measurement. Accurate prediction of mixture adsorption is challenging because solvents, intermediates, and products are usually capable of forming hydrogen bonds which can lead to large deviation from ideal solution behavior. Understanding the overall transfer thermodynamics requires both adsorption equilibria and solvation equilibria, which are not always available.

The majority of studies investigating adsorption of mixtures typically encountered in biorefineries onto high-silica zeolites have involved those of primary alcohols from aqueous solutions.8–18 For aqueous solutions of poly-oxygenated compounds, correlations of Henry’s constants of adsorption (i.e., at very low loadings) with octanol–water partition coefficients (estimated from group contribution methods) is one main finding.4 Methods for more reliably calculating solute and multisolute adsorption have been reported.5,17,19,20 Simulation studies have been undertaken for the more chemically complex glucose but were only able to consider low loadings.21 Complete isotherms of several polyols with two to six carbons have been simulated, but the simulations at high loadings neither included solvent adsorption nor solvation effects.8 A better understanding of adsorption could enable exploitation of shape selectivity in biomass conversion, which is utilized less than that in petrochemical conversion.22 

Linear alkane-α, ω-diols (also called terminal diols or alkane-1,n-diols; hereafter, referred to as diols) with 3–6 carbon atoms are high-value chemicals with many applications in the polymer and solvent industries,23 making their renewable production attractive.24–32 Despite the use of zeolites in separating diols in renewable processes,33–35 in catalyzing their amino cyclization,36 or selective cyclodehydration,37,38 or the use of diols as cheaper and less toxic structure directing agents for zeolite synthesis,39–42 the adsorption and confinement of diols in zeolites has received little attention.

The adsorption equilibria of alkane-α, ω-diols (diols) with 2, 4, and 6 carbons were reported to exhibit very steep changes in adsorption with slight changes in solution concentration onto all-silica zeolite of the type mordenite framework inverted (MFI, also referred to as silicalite-1).43 This is different than the isotherms for n-hexane and n-heptane, which show a step or kink at a loading of 4 molec/uc44–48 in MFI and exhibit two distinct regimes of adsorption.49,50 While this adsorption of n-heptane has been termed “commensurate freezing,”48 this phenomena, with two distinct regimes of adsorption, really only corresponds to the final adsorption state with a loading of 8 molec/uc.51 While the adsorption isotherms onto all-silica MFI-type zeolite of propan-1-ol,8–12 butan-1-ol,8,9,12,13 and pentan-1-ol8,9 from aqueous solution and those of alkane-1,2-diols with 4 or 6 carbons8 from gas-phase do not exhibit steep changes in loading, the adsorption of water is known to take place through a first-order phase transition.52–55 Since the steepness of an isotherm is typically explained in part from more favorable sorbate–sorbate interactions relative to sorbate–sorbent interactions,53,56–58 one might expect the diols to behave similarly to water.

In this work, molecular simulations with enhanced sampling techniques and experiments are employed to obtain the binary adsorption equilibria of diols with 3–6 carbons (hereafter referred to as C3, C4, C5, and C6) from aqueous solution at T = 323 and 348 K and p = 1.0 bar. The effect of increasing the temperature up to 383 K (at an elevated pressure of 3.0 bars), which is difficult to be measured with conventional equipment but is necessary for the design of liquid adsorption systems,59 is probed by simulation. The catalytic production of these diols from renewable resources occurs at elevated temperates (usually above 373 K)29–32 and adsorption at ambient conditions would increase the energy required for cooling of the reaction mixture and heating of the zeolite phase for subsequent desorption. Thermodynamics and quantitative structure analysis are used to explain the unique behavior of each diol in this homologous series confined in MFI. As the diols have been used to aid in the understanding of hydrogen-bonding behavior in solution of carbohydrates,60–62 one might anticipate that this analysis would be quite informative for the adsorption mechanism of more complex polyols.

1. Molecular models

Molecular representations of water and diols were obtained from the TIP4P (Transferable Intermolecular Potential, 4 Points) model63 and the united-atom version of the Transferable Potentials for Phase Equilibria (TraPPE–UA) force field,64–66 respectively. Following the TraPPE–UA force field for ethers and glycols,66 intramolecular Coulombic interactions between pseudo-atoms separated by three bonds were scaled by a factor of 1/2. The unary vapor–liquid equilibria (VLE) of C3, C4, C5, and water have been computed in previous studies65–68 using the TraPPE–UA and TIP4P models, and the predictions are quite satisfactory. The predicted normal boiling points of C3 and C5 fall 3 K above and 4 K below, respectively, the experimental values, but agree within the combined statistical uncertainties. The normal boiling points of C4 and water are underpredicted by 8 and 10 K, respectively. Experimental data for binary VLE of water with (C5 or C6) are not available. The simulations including an explicit solution-phase reservoir (see below) were used to estimate the relative volatilities for C3/water and C4/water as a function of liquid-phase concentration at a fixed temperature of 323 K, but experimental data are only available at fixed pressure. For C3, the simulations yield relative volatilities in the range from 580 ± 90 to 1500 ± 600 as the liquid-phase mole fraction of the diol is increased from 5.68 × 10−4 to 0.306 ± 0.001, and the experimental data at p = 0.08 bar yield relative volatilities of 390 at a liquid-phase mole fraction of 0.3707 (and T = 324.55 K) and 340 at a liquid-phase mole fraction of 0.4584 (and T = 327.85 K).69 For C4, the simulations yield relative volatilities in the range from 260 ± 60 to 1000 ± 300 as the liquid-phase mole fraction of the diol is increased from 0.009 ± 0.0010 to 0.1553 ± 0.0007, and the experimental data at p = 0.13 bar yield relative volatilities of 130 at a liquid-phase mole fraction of 0.026 (and T = 318.85 K) and 840 at a liquid-phase mole fraction of 0.43 (and T = 324.15 K).70 The over and underpredictions of the relative volatilities reflect the small deviations in the normal boiling points of the pure compounds. It is noteworthy that the relative volatilities increase with increasing diol concentration.

The ortho form (ORTHO) of all-silica MFI with Pnma space group and orthorhombic symmetry71 was utilized for the majority of simulations because it has been identified as the standard polymorph of MFI by the Structure Commission of the International Zeolite Association.72 The unit cell was replicated 2, 2, and 3 times in the a, b, and c-directions, respectively, creating a simulation box with dimensions of 4.00 × 3.98 × 4.01 nm, in that order. In all simulations, the interactions of sorbate molecules with the rigid zeolite framework were described by the TraPPE–zeo force field73 and were implemented with pretabulation and interpolation during simulation74 using a grid spacing of 0.1 Å. To obtain structural information in the adsorbed phase, the intersection of the straight channels and zig-zag channels was defined as a sphere of radius 4.0 Å centered at (a/2, b/4, 0) and symmetrically equivalent positions. This yielded a supercell with the pore topology defined as in Fig. 1, with ratios of pore volumes in the straight, zig-zag, and intersections being approximately 1.7:2.5:1.0, respectively. To investigate the effects of slight changes in the structure of the MFI framework on the adsorption properties, simulations for the adsorption of C5 and C6 from aqueous solution were also carried out for two additional structures: monoclinic symmetry with P21/n11 space group (MONO,75 replicated 2, 2, and 3 times in the a, b, and c-directions to yield the simulation supercell) and orthorhombic symmetry with P212121 space group (PARA,76 replicated 2, 2, and 3 times in the a, b, and c-directions).

FIG. 1.

Pore channels within the framework for a 2 × 2 × 3 supercell of all-silica MFI-type zeolite. The zig-zag channels, straight channels, and intersections are filled in green, black, and white, respectively.

FIG. 1.

Pore channels within the framework for a 2 × 2 × 3 supercell of all-silica MFI-type zeolite. The zig-zag channels, straight channels, and intersections are filled in green, black, and white, respectively.

Close modal

2. Enhanced sampling algorithms and simulation parameters

Monte Carlo (MC) simulations in the isobaric–isothermal version of the Gibbs ensemble77–79 (NpT GEMC simulations) were employed to obtain adsorption equilibria from aqueous solution. This approach allows for improved rates of equilibration than comparable molecular dynamics simulations,80 because an explicit interface does not need to be represented between the adsorbed phase and solution phase and special MC moves allow for sampling of relevant configurations that otherwise would become kinetically trapped when following Newton’s equations of motion. For all simulations, the set of trial moves included rigid-body translations and rotations around the center-of-mass (COM) for all sorbate types and coupled-decoupled81 configurational bias Monte Carlo82–84 (CD-CBMC) moves for all sorbate types except water. CD-CBMC strategies were also employed for particle transfer moves between two phases.28,85,86 Volume moves were performed on all simulation boxes except those representing a zeolite.

Adsorption of C3, C4, C5, or C6 from aqueous solution was conducted at p = 1.0 bar and 323 K with a total of Ntot = 1100 molecules. Adsorption of C5 was also conducted at 348 K and 1.0 bar and 383 K and 3.0 bars to investigate the effect of temperature. Isotherms were obtained by varying the overall composition such that Ntot included 2–400 diol molecules and the remainder solvent molecules. All of the aforementioned simulations were implemented with three different simulation boxes in thermodynamic but not physical contact to represent the three distinct phases.

For simulations including a liquid phase, direct zeolite–liquid transfers of molecules were replaced with liquid–vapor and vapor–zeolite transfers as the rate of acceptance for transfer moves is much higher between a dense phase and vapor phase than between two dense phases.87 A suitable volume of the vapor phase was achieved through addition of a fixed amount of ideal (i.e., noninteracting) particles so that the number of diol molecules in the vapor phase was >2 or approximately equal in both solution and vapor phases. The fluctuations of diol molecules in the zeolite are constrained by the total number of diol molecules. When adsorption occurs at very low concentrations, with very few diol molecules being in the vapor and solution phase, the fluctuation of diol molecules in the zeolite phase is very small. This advantage is similar to that provided by the gauge cell Monte Carlo simulation method88,89 except an explicit solution concentration is obtained.

The total number of molecules, Ntot, in simulations involving a liquid phase included 2 or 3 types of normal alcohols with smaller numbers of carbons than the diol as “impurity” molecules, with a total of 2 molecules for each type. These “impurities” aid in the sampling of molecule transfers through identity switch moves82,90 that change molecule identities and boxes simultaneously. Uniform biasing potentials on each impurity type were used in the transfer boxes to yield on average about one impurity molecule in both solution and vapor phases. Due to the high volatility of water relative to the diols, a uniform biasing potential was used on water in the vapor boxes so that the number of molecules in this box did not exceed ≈30. Impurity molecules were not allowed to transfer into the zeolite phase.

The isotherms of C5 and C6 at 323 K include diol adsorption at very dilute concentrations. At these concentrations, there was no need to represent the solution phase explicitly, as equilibria between the solution and vapor phases could be described with a Henry’s constant relation. Therefore, for a given solution concentration, the corresponding vapor phase (assuming the Henry’s relation) was represented by two separate water (NW = 100) and diol (ND = 200) gas-phase reservoirs with suitable applied pressures. Henry’s constant relation was obtained from the low-concentration free energies of hydration from the aforementioned three box simulations as well as simulations implemented identically except with only two boxes representing a vapor and liquid phase. It was found that deviations from ideal gas behavior at the conditions of interest are negligible compared to the uncertainties in Henry’s constant and, hence, the vapor phases were treated as ideal.

To assess the influence of water on the gas-phase adsorption of diols, NpT GEMC simulations were performed for unary adsorption of C3, C4, C5, and C6 diols (D) with Ntot = ND = 200 and two simulation boxes representing an ideal gas phase and a zeolite phase.

The nonbonded interactions in solution and zeolite phases, including pairwise additive Lennard-Jones (LJ) 12–6 and Coulombic potentials, were implemented following the same procedure as previous studies,6,10,13 with the LJ interactions being spherically truncated at a distance of rcut = 14.0 Å and analytical tail corrections to estimate interactions beyond this distance. Coulombic interactions were described using the Ewald summation method91 with a screening parameter of κ = 3.2/rcut and an upper bound of the reciprocal space summation at Kmax=κLbox. For simplicity and to prevent the large computational cost associated with the reciprocal Ewald space summation, all vapor boxes were implemented as ideal and any intramolecular Coulombic interactions were calculated explicitly.

To investigate sorbed phase coexistence of C5, Monte Carlo simulations were conducted in the isochoric–isothermal version of the Gibbs ensemble with a zeolite box and a vapor box. The unary system was initialized with Ntot = NC5 = 320 molecules at 323, 348, or 383 K at near saturation loading in a supercell of 2 × 6 × 3 unit cells. A cut-off distance of rcut = 3c/2 was used and the size of the vapor box was tuned to afford ≈2 diol molecules. Instead of using only an adsorbed phase,92 this setup allows for improved sampling of the spatial distribution by means of CD-CBMC particle transfer moves between the vapor phase and zeolite phase. The initial structure was generated by elongating the b-direction with 6 more unit cells, making a supercell with 2, 12, and 3 unit cells in the a, b, and c-directions, respectively, creating an elongated simulation box with dimensions of 4.0 × 23.9 × 4.0 nm, in that order. Additional simulations were carried out for a much larger system with a near-cubic supercell consisting of 12 × 12 × 18 unit cells (24.0 × 23.9 × 24.1 nm) with Ntot = NC5 = 10 000 molecules (average loading of 3.86 molec/uc) and cutoff at 40 Å applied to Lennard-Jones and Coulombic interactions of neutral groups (CH2OH).

Eight independent simulations were carried out at each state point, and the statistical uncertainties are reported as the 95% confidence intervals estimated by multiplying the standard error of the mean by a factor of 2.4. The lengths of the simulations are given in the number of Monte Carlo Cycles (MCCs), each consisting of Ntot randomly selected trial moves. For equilibration of the mixtures containing C3, C4, and C5, 100 000 to 850 000 MCCs were used, with longer periods being required for simulations at low concentrations and/or high loadings. The lengths of equilibration for C6 ranged from 1.0 to 2.5 × 106 MCCs. Production periods consisted of 100 000–450 000 MCCs.

1. Synthesis of all-silica MFI zeolite

All-silica MFI zeolite was synthesized4 in fluoride media with a starting molar composition of 1.0 SiO2: 0.08 TPABr: 0.4 NH4F: 20 H2O. Typically, 1.66 g of tetrapropylammonium bromide (98%, Sigma Adrich) was added to 27.62 g of distilled water, followed by the addition of 1.15 g of ammonium fluoride (98%, J.T. Baker) and vigorously stirring at ambient conditions for 15 min. 4.61 g of silica (SiO2, Cabosil M5, Riedel de Haën) was added to the solution and hand-mixed with a spatula until a homogeneous solution was obtained. The solution was transferred into a Teflon-lined stainless steel autoclave and heated at 448 K for 7 days under static condition. The resulting MFI crystals were washed by repeated centrifugation at a relative centrifugal force of 20 000 for 20 min until the pH of the supernatant became neutral. The solid product was dried in a convection oven at T = 343 K overnight. The dried MFI crystals were calcined in a tubular furnace at T = 823 K at a ramp rate of 1 K/min for 20 h under a dry air flow of 100 ml/min. The synthesized material (Si[F]) was characterized previously:4 it has monoclinic crystal symmetry at ambient conditions and is free of silanol defects, with an average crystal size of 50 μm along the c-direction.

2. Liquid phase adsorption

Adsorption experiments were conducted from aqueous solutions at T = 323 ± 0.5 K or T = 348 ± 0.5 K. Among all experiments, the ratio of the initial adsorbate solution to adsorbent ranged between 4 and 8 ml/g. Approximately 100 mg of zeolite with an appropriate amount of diol solution was added to glass vials (C4011-1, crimp seal, Thermo Scientific™) and then the vials were rotated at 20 rpm in a ProBlot™ 12 hybridization oven until equilibrium was reached. The supernatant solutions were filtered using a Monoject syringe fitted with a 0.2 μm hydrophilic polypropylene (GHP) syringe filter to remove the zeolite particles. The filtrate concentrations were analyzed with an Agilent 1200 high performance liquid chromatography (HPLC) equipped with an ion exclusion column (Aminex HPX-87H, Bio-Rad) and a refractive index detector (RID). The column temperature was set at 333 K with the RID at 323 K. Injection volumes of 10, 20, 30, and 40 μl were eluted using a mobile phase of 0.005M sulfuric acid in distilled water with a flow rate of 0.6 ml/min. The relative signal intensities of the adsorbate and a glycerol (99.5%, Aldrich) internal standard were used to determine the final concentrations. Each sample was injected three times and the results were averaged.

The aqueous solution densities for each diol investigated were obtained as a linear fit to dilute concentrations,93,94 allowing solution nonidealities to be captured. The amount adsorbed in the zeolite was calculated using the pore-filling model, which assumes that the entire pore volume is filled upon adsorption.5,19 In computation of the amount adsorbed with this model, the value assumed for the total pore volume was 0.125 ml/g. This value was obtained by estimating the pore volume from the maximum amount of water that can adsorb in silicalite-1 using an average maximum water loading73,95 of 40.4 molec/uc and assuming that the adsorbed water phase has the same density as water in the (low-pressure) bulk liquid phase (0.9894 g/ml at T = 323 K and p = 1.0 bar)96 and determining the mass of a unit cell from 192 O atoms and 96 Si atoms. It should be noted that water does not pack efficiently in the 5 Å channels of silicalite-1 with most molecules forming only two hydrogen bonds.97 

The uptake of diols from aqueous solution onto silicalite-1 at T = 323 K is presented in Fig. 2. In comparison to the uptake of C4 and C6 reported by Fegan and Lowe at T = 298 K,43 the experimentally measured isotherms of the present work are slightly shifted to higher concentrations. The concentration at which diol adsorption becomes larger than 0.5 molec/uc, referred to as the concentration of the initial onset, differs between diols by approximately an order of magnitude with increasing number of methylene segments. As the concentration of the initial onset is reached, steep changes in loading with slight differences in the solution concentration are observed. At higher solution concentrations, the C3 and C4 isotherms exhibit a “knee,” suggesting an increased entropic penalty. The C5 and C6 isotherms, on the other hand, are step functions which nearly reach saturation loading with the initial steep change in loading.

FIG. 2.

Adsorption isotherms of (a) diols and (b) water for C3, C4, C5, and C6 mixtures from aqueous solution, as determined by simulation (using the ORTHO structure) and experiment at T = 323 K and p = 1.0 bar. Simulation points are connected with lines to guide the eye.

FIG. 2.

Adsorption isotherms of (a) diols and (b) water for C3, C4, C5, and C6 mixtures from aqueous solution, as determined by simulation (using the ORTHO structure) and experiment at T = 323 K and p = 1.0 bar. Simulation points are connected with lines to guide the eye.

Close modal

The maximum diol loadings and associated water loadings obtained from the simulations are listed in Table I. Numerical data for all concentrations are provided in Tables S1–S4 in the supplementary material. At saturation, the number of C5 or C6 molecules is nearly an integer per unit cell, suggesting a configuration commensurate with the pore topology of the unit cell. While the number of molecules per unit cell at saturation decreases with increasing chain length, the loading on a mass basis increases with increasing chain length from C3 to C5, but is nearly the same for C5 and C6.

TABLE I.

Adsorption data for the concentrations yielding maximum diol loadings and maximum water loadings investigated by simulation at T = 323 K.

Maximum diol loadingMaximum water loading
SystemStructureC (mg/ml)QD (molec/uc)QW (molec/uc)C (mg/ml) QD (molec/uc)QW (molec/uc)
C3 ORTHO 659 ± 1.0 9.6 ± 0.10 2.8 ± 0.13 50 ± 10 6.8 ± 0.2 4.9 ± 0.2 
C4 ORTHO 473 ± 1.4 8.80 ± 0.07 1.0 ± 0.10 3 ± 1 5.21 ± 0.02 2.7 ± 0.13 
C5 ORTHO 83 ± 9 7.9997 ± 0.0002 0.8 ± 0.10 0.19 ± 0.08 5.327 ± 0.002 2.2 ± 0.2 
C5 MONO 130 ± 10 8.01 ± 0.02 1.14 ± 0.09 0.9 ± 0.14 7.79 ± 0.04 1.6 ± 0.2 
C6 ORTHO 190 ± 11 7.06 ± 0.03 2.6 ± 0.2 0.197 ± 0.009 6.5 ± 0.3 3.1 ± 0.2 
C6 MONO 194 ± 7 7.21 ± 0.08 1.9 ± 0.3 0.1 ± 0.01 5.5 ± 0.4 3.1 ± 0.2 
Maximum diol loadingMaximum water loading
SystemStructureC (mg/ml)QD (molec/uc)QW (molec/uc)C (mg/ml) QD (molec/uc)QW (molec/uc)
C3 ORTHO 659 ± 1.0 9.6 ± 0.10 2.8 ± 0.13 50 ± 10 6.8 ± 0.2 4.9 ± 0.2 
C4 ORTHO 473 ± 1.4 8.80 ± 0.07 1.0 ± 0.10 3 ± 1 5.21 ± 0.02 2.7 ± 0.13 
C5 ORTHO 83 ± 9 7.9997 ± 0.0002 0.8 ± 0.10 0.19 ± 0.08 5.327 ± 0.002 2.2 ± 0.2 
C5 MONO 130 ± 10 8.01 ± 0.02 1.14 ± 0.09 0.9 ± 0.14 7.79 ± 0.04 1.6 ± 0.2 
C6 ORTHO 190 ± 11 7.06 ± 0.03 2.6 ± 0.2 0.197 ± 0.009 6.5 ± 0.3 3.1 ± 0.2 
C6 MONO 194 ± 7 7.21 ± 0.08 1.9 ± 0.3 0.1 ± 0.01 5.5 ± 0.4 3.1 ± 0.2 

Good agreement between simulation and experiment is observed for diols with three to five carbons. However, the inflection in C6 loading is not observed in the simulation using ORTHO silicalite-1. The simulations are not thought to exhibit a sampling issue because this behavior remained after conducting the C6 simulations for >2 × 106 MCCs. Instead, the discrepancy could result from the larger chain length of C6, resulting in a higher propensity to induce framework flexibility upon adsorption. It should also be noted that the Si[F] silicalite-1 samples used here were found to be in the monoclinic structure.4 While no change in the monoclinic structure was observed upon the adsorption of n-hexane,98 a difference in structure symmetry upon adsorption of n-heptane has been reported.51 

The influence of the framework structure (ORTHO, MONO, and PARA) on the simulated adsorption equilibria was investigated for the C5/water and C6/water mixtures (see Figs. S1 and S2 in the supplementary material, respectively). Compared to the ORTHO structure, used in all other simulations in this work, the adsorption of both solutes into the MONO structure occurs at slightly higher concentrations. The initial onset of the experimental isotherms falls in between the simulated isotherms for the ORTHO and MONO structures, but somewhat closer to those for the MONO structure. While the simulated C6 isotherm onto the ORTHO structure is not stepped, the isotherm for the MONO structure is stepped. The PARA structure, on the other hand, exhibits adsorption shifted to higher concentrations than the MONO structure, at about an order-of-magnitude higher than the ORTHO structure. This structure has only been shown to occur upon adsorption of aromatic hydrocarbons like p-xylene at loadings higher than 4 molec/uc.76,99–101 On the basis of these results, it is clear that the MONO structure is a better representative to match the simulation and experimental results presented here, but it cannot be ruled out that the experimental Si[F] sample (upon adsorption) is some combination of the MONO and ORTHO forms4 or that deficiencies in the force fields require a mix of structures to agree with high-quality experiments (see also Ref. 73 for a similar observation for water adsorption).

The associated uptake of water obtained from simulations with the ORTHO structure is presented in Fig. 2(b) (and Figs. S1 and S2 in the supplementary material for a comparison with MONO and PARA structures). The water loading in the ORTHO structure is ≈0.4 molec/uc (and lower by a factor of about 2 in the MONO and PARA structures) at the lowest solution concentrations, reflecting the hydrophobicity of the all-silica zeolite. This loading is slightly increased compared to a unary adsorption of 0.2 molec/uc in the ORTHO structure at T = 300 K and p = 70 bars,73 but is consistent with an increased vapor pressure.

Increases in water loadings with solution concentration are observed to coincide with increases in diol uptakes observed in Figs. 2(a) and 2(b). Water adsorbed with diols with 3–5 carbons exhibits a distinct maximum in adsorption. The loadings at the maxima of water adsorption are provided in Table I. The heights of the maxima increase with decreasing diol hydrophobicity (decreasing number of carbons). These maxima coincide with the concentration at which the diols reach about two-thirds of their saturation loading. For C5 in the ORTHO structure, the water adsorption appears to be a jump discontinuity, increasing to ≈2.2 molec/uc with the diol step change and decreasing to ≈0.8 molec/uc thereafter. However, the jump is less pronounced for the MONO structure with a maximum of only ≈1.6 molec/uc and a decrease to ≈1.2 molec/uc at high diol concentrations. For C3 and C4, the peak in water loading is broader. For C3, the water loading at the highest concentration is ≈3 molec/uc, whereas the water loading for C4 decreases to ≈1 molec/uc like C5. For the ORTHO structure, the water adsorption associated with C6 does not exhibit a clear maximum and remains above 80% of maximum loading even at the highest C6 concentration; this suggests that C6 cannot efficiently pack in the ORTHO structure to prevent coadsorption of water at saturation loading. Packing is even more disrupted in the PARA structure. However, simulations with the MONO structure yield a pronounced maximum in water loading also for C6, in agreement with the stepped C6 isotherm, indicating that C6 packs well in this polymorph.

The adsorption of diols onto ORTHO silicalite-1 as a function of gas-phase partial pressure is found to be slightly different in the presence and absence of water (see Fig. S3 in the supplementary material). The presence of water causes a shift in the initial onset of the adsorption isotherm to lower pressures because the water molecules in the zeolite can act as favorable adsorption sites for the diols. The magnitude of the shift in the initial onset due to the presence of water is related to the hydrophilicity of the diol, with more hydrophilic diols shifting more toward lower pressures in the presence of water. The shape of the isotherms is slightly different for C3 with the presence of water leading to a more pronounced knee and a downward shift in the C3 loading at high pressures [i.e., the presence of 3–5 water molecules per unit cell increases the entropic penalty associated with increasing C3 loading (displacing water) and reduces the pore space available for C3]. A similar change in the isotherm shape, but to a much smaller extent, can also be observed for C4. By contrast, the maximum observed loading of C5 and C6 is not influenced by the presence of water co-adsorption, i.e., it is governed by the pore structure of the zeolite.

The diol and water loadings and associated concentrations obtained from simulation can be converted to adsorbed diol mole fractions (number of adsorbed diol molecules divided by the total number of adsorbed molecules) and diol selectivities [SD/W=(xDZ/xDS)/(xWZ/xWS), where the superscripts Z and S denote the zeolite and solution phases and subscripts D and W denote diol and water, respectively] with respect to concentration. As shown in Fig. 3(a), the diol mole fractions in the adsorbed phase reach values of 0.8, 0.9, 0.9, and 0.7 for C3, C4, C5, and C6, respectively, at high diol concentrations in the solution phase. In Fig. 3(b), very high selectivities over water are observed at dilute concentrations, with SD/W>10NC1, where Nc is the number of diol carbons. Overall, the adsorption selectivities decrease as the solution-phase concentration increases, which is in marked contrast to the trend of the relative volatilities. The adsorption selectivity order for each diol at higher concentrations follows the peaks and widths observed in the water isotherms. For C5, the maximum in selectivity corresponds to the solution concentration just above the step in the adsorption isotherm, which is also after the decrease in water adsorption. The effect of selectivity on the framework adopted in the simulations is examined in Figs. S4 and S5 in the supplementary material for the adsorption of C5 and C6 from aqueous solution at 323 K.

FIG. 3.

(a) Adsorbed mole fraction of alkane-α, ω-diols and (b) diol selectivity over water obtained by simulation at T = 323 K.

FIG. 3.

(a) Adsorbed mole fraction of alkane-α, ω-diols and (b) diol selectivity over water obtained by simulation at T = 323 K.

Close modal

The influence of increasing temperature on the adsorption of C5 from aqueous solution up to 383 K is examined in Fig. 4. Simulations of C5 adsorption at 323 and 348 K agree with experiment. Experiments were not carried out at above the normal boiling point of water, but this condition is relevant for reaction processes.30 As the temperature is increased, the isotherm becomes less steep. The initial onset of adsorption from solution, however, does not change significantly with increasing temperature like alkanes do from gas phase.102 Instead, this behavior is similar to that observed for pure water.54 

FIG. 4.

(a) C5 adsorption from aqueous mixtures at 323 K and 1.0 bar, 348 K and 1.0 bar, and 383 K and 3.0 bars. (b) Associated water adsorption for the mixtures in (a). Simulation points are connected with lines to guide the eye.

FIG. 4.

(a) C5 adsorption from aqueous mixtures at 323 K and 1.0 bar, 348 K and 1.0 bar, and 383 K and 3.0 bars. (b) Associated water adsorption for the mixtures in (a). Simulation points are connected with lines to guide the eye.

Close modal

When temperature is increased from aqueous solution, the height and width of the peak in the water isotherm increase, with the maximum height observed increasing ≈1 molec/uc between 323 and 383 K. However, as the solution concentration increases and the diol reaches saturation, the amount of water decreases to the same value at saturation, slightly less than 1 molec/uc. This suggests that the diol packing is similar near diol saturation where water can only adsorb in locations that are not occupied by the diol. The trends in adsorbed mole fraction and selectivity as a function of C5 concentration at the various temperatures are shown in Fig. S6 in the supplementary material.

The role that water plays on the C5 adsorption from gas-phase is examined by comparison of the C5 isotherms in the presence and absence of water (i.e., the unary and binary isotherms, respectively). The C5 isotherms are shifted to lower onset pressures in the presence of water at the same temperature (see Fig. S7 in the supplementary material). The magnitude of this shift increases with increasing temperature, and this trend with increasing temperature is similar to that of increasing diol hydrophilicity at constant temperature (see Fig. S3 in the supplementary material). With increasing temperature (and, for T = 383 K and p = 3.0 bars, also increased pressure), the water adsorption increases. This causes a larger shift in the initial onset of adsorption. At 383 K, the unary C5 adsorption becomes higher than the binary C5 adsorption at loadings above 4 molec/uc (i.e., the isotherm is more steep). This is along the same lines as the comparison for C3 (see Fig. S3 in the supplementary material), where the diol loading can increase more drastically in the unary case because there is no entropic penalty for exclusion of water.

For rational application of zeolites in the processing of biomass-derived mixtures, it is necessary to be able to explain the distinct features of adsorption equilibria. However, the isotherms from solution alone do not reveal whether the notable differences in adsorption with diol chain length or temperature are primarily a result of interactions in the adsorbed phase or in the solution phase. The Gibbs ensemble simulations directly output densities in each phase, which allows for calculation of the free energies of transfer between pairs of phases.103,104

The resulting free energies of solvation (vapor to liquid) and of adsorption (vapor to zeolite) for C3, C4, C5, and C6 in water at 323 K are shown in Figs. 5(a)–5(d), respectively. At very low concentrations below the initial onset of diol adsorption, the diol adsorption free energy is independent of concentration. In this region, the magnitude of the adsorption free energy increases by ≈6 kJ/mol between C3 and C4, ≈8 kJ/mol between C4 and C5, and ≈7 kJ/mol between C5 and C6. The deviation from the regular trend observed for C4 is related to its ability to form intramolecular hydrogen bonds in the vapor phase (see below). The changes in hydration free energy are much smaller, and the value of ≈1 kJ/mol for the incremental free energy per methylene group is consistent with experimental and simulation data for primary alcohols.105,106 (Note that C4 is again an outlier.) Overall, the difference in the initial onset concentration of diol adsorption is attributed to be mainly a result of favorable adsorption into the zeolite, as opposed to desolvation from the gas phase.

FIG. 5.

Free energies of hydration (VL, vapor to aqueous solution, open symbols) and of adsorption (VZ, vapor to zeolite, filled symbols) for (a) C3, (b) C4, (c) C5, and (d) C6 in water at T = 323 K. Values obtained for the hydration free energies at infinite-dilution for C5 (−39.3 ± 0.4 kJ/mol) and C6 (−39.2 ± 0.3 kJ/mol) are shown as horizontal green and blue lines, respectively.

FIG. 5.

Free energies of hydration (VL, vapor to aqueous solution, open symbols) and of adsorption (VZ, vapor to zeolite, filled symbols) for (a) C3, (b) C4, (c) C5, and (d) C6 in water at T = 323 K. Values obtained for the hydration free energies at infinite-dilution for C5 (−39.3 ± 0.4 kJ/mol) and C6 (−39.2 ± 0.3 kJ/mol) are shown as horizontal green and blue lines, respectively.

Close modal

Near the steps in the adsorption isotherms, an insignificant change in the diol free energy of solvation is observed (i.e., the hydration is in the Henry’s law region). The adsorption free energy, however, decreases significantly. For C5, C4, and C3, the adsorption free energy becomes more favorable by ≈10 kJ/mol from its initial value. The minimum in adsorption free energy is the point on the isotherm where ≈2/3 of saturation loading is reached and where the water loading is near its maximum. At concentrations higher than the step, the adsorption free energy increases significantly due to the entropic penalty associated with the reduction of available pore volume. The solvation free energies of C4, C5, and C6 are observed to slightly decrease at high concentrations because of the diol being a better solvent in itself.

Since the differences in the initial onset concentration of diol adsorption and the steep region of the isotherms are due to the differences in free energies of transfer from vapor to zeolite, this free energy was decoupled for the unary simulations into enthalpic and entropic terms (see Fig. S8 in the supplementary material). In the limit of low pressures, the magnitude of the enthalpy of adsorption increases by ≈14 kJ/mol between C6 and C5, while the entropic term (−TΔS) decreases by ≈6 kJ/mol. The magnitude of the enthalpy of adsorption increases by ≈11 kJ/mol between C5 and C4, while the entropic term decreases by ≈4 kJ/mol. Between C4 and C3, the enthalpy of adsorption increases by ≈9 kJ/mol, while the entropic term increases by ≈3 kJ/mol. This demonstrates that differences in initial adsorption from the gas-phase are due to both differences in enthalpy and entropy of adsorption, but primarily due to differences in enthalpy.

Decoupling the free energy of adsorption into entropic and enthalpic terms also provides insight into the associations with the minima in the free energies of adsorption. The steep changes in the simulated isotherms for C3, C4, and C5 diols are associated with steep decreases in adsorption enthalpy and also steep increases in −TΔS. For C6, however, whose simulated isotherm does not exhibit a clear step onto ORTHO silicalite-1, a drastic change in enthalpy of adsorption is not observed.

In Fig. S9 in the supplementary material, the free energies of transfer for water associated with the diols at 323 K are reported. At diol concentrations below ≈0.1 g/ml (i.e., the majority of the concentrations investigated in this work), the free energy of water solvation is nearly independent of solute identity. At higher concentrations, the solvation free energy increases (becomes less favorable) as the solution contains more diols with hydrophobic alkane chains. The water free energy of adsorption is analogous to the water adsorption isotherms. The free energy decreases (becomes more favorable) with the onset of diol adsorption and reaches a minimum that corresponds to the maximum in water adsorption. At higher concentrations where diols occupy a large fraction of the pores, the decreasing water adsorption corresponds to an increasing free energy of adsorption.

The effects of temperature on the free energies of transfer of C5 are shown in Fig. 6. Both the free energies of solvation and adsorption become less favorable with increasing temperature. Since both curves are shifted by similar amounts, the overall free energy of transfer from solution to zeolite changes little with temperature. This explains why the onset of C5 adsorption does not shift significantly with increasing temperature. The minimum in adsorption free energy also becomes less shallow and more wide with increasing temperature. This change is associated with the isotherm becoming less steep.

FIG. 6.

Free energies of solvation (VL, vapor to solution, open symbols) and of adsorption (VZ, vapor to zeolite, filled symbols) for C5 at T = 323 K and p = 1.0 bar, T = 348 K and p = 1.0 bar, and T = 383 K and p = 3.0 bars.

FIG. 6.

Free energies of solvation (VL, vapor to solution, open symbols) and of adsorption (VZ, vapor to zeolite, filled symbols) for C5 at T = 323 K and p = 1.0 bar, T = 348 K and p = 1.0 bar, and T = 383 K and p = 3.0 bars.

Close modal

The effects of temperature on the corresponding free energies of transfer of water are similar to those for C5 (see Fig. S10 in the supplementary material). However, the solvation free energy of water increases (desolvation becomes more favorable) more with increasing temperature than does the adsorption free energy. In other words, the increased water adsorption at higher temperatures results from the desolvation of water becoming more favorable with increasing temperature than does the adsorption of water.

The degree of steepness of an isotherm is thought to be related to the strength of sorbate–sorbate interactions relative to sorbate–sorbent interactions.53,56–58 Since diols can only orient with the pores of MFI, the primary interactions involving diols are hydrogen bonds. Therefore, the number of hydrogen bonds between a pair of sorbates i and j per molecule of i, Hij, was determined for the adsorption of diols from aqueous mixtures at 323 K (see Fig. 7). Hydrogen bonds were defined using a set of geometric criteria of rHO < 2.5 Å and ∠OHO > 130°. Two-dimensional histograms of rHO and ∠OHO are shown for the four aqueous systems at 323 K in Fig. S11 in the supplementary material and demonstrate that these criteria result in a slight estimation of hydrogen-bonds. The number of diol–diol hydrogen bonds per molecule of diol adsorbed, HD–D, changes steeply with solution concentration and reaches values of 2.02 ± 0.04, 1.87 ± 0.03, 1.54 ± 0.02, and 1.36 ± 0.03 for C6, C5, C4, and C4, respectively. These steep changes parallel those of the isotherms, and for C6 there is an inflection point with about 0.9 HD–D at 10−4 g/ml. As observed in HW–D, water forms more hydrogen bonds per molecule with the diols with increasing concentration. For C3, C4, and C5, the number of diol–water hydrogen bonds per molecule of diol, HD–W, and the number of water–water hydrogen bonds per water molecule, HW–W, exhibit peaks with solution concentrations like the corresponding water loadings. The HW–W peak occurs at solution concentrations slightly lower than HD–W. The HW–W peak is coincident with the maximum in water adsorption observed in the isotherms. For C6, which does not exhibit a clear maximum in water adsorption, a peak in HW–W is also not observed. Since C6 can occupy only 3/4 straight channels per unit cell at maximum loading, water occupies the remaining parts of straight channels and hydrogen bonds more with itself. The diols with larger chain lengths fill up the pores with fewer molecules per unit cell and become more coordinated with themselves via hydrogen bonding. As a result, HD–D increases at an increasing rate with increasing chain length. At maximum loading, HD–D increases by 0.15–0.35 for each additional methylene segment. It should also be noted that at the lowest diol concentrations, HD–W is about 0.5 and about twice as large as HW–W. The dependence of these results on several different hydrogen-bonding criteria, which will always remain ambiguous for the relatively simple force fields used in this work, is shown in Fig. S12 in the supplementary material. Different criteria change the absolute numbers of hydrogen bonds, but do not change any of the trends.

FIG. 7.

Number of hydrogen bonds (a) between diol and diol per diol molecule, (b) between water and diol per water molecule, (c) between diol and water per diol molecule, and (d) between water and water per water molecule for adsorbed aqueous mixtures of diols at 323 K.

FIG. 7.

Number of hydrogen bonds (a) between diol and diol per diol molecule, (b) between water and diol per water molecule, (c) between diol and water per diol molecule, and (d) between water and water per water molecule for adsorbed aqueous mixtures of diols at 323 K.

Close modal

In the adsorption of the C5 mixture, HD–D decreases from 1.87 ± 0.03 to 1.58 ± 0.015 at saturation loading as the temperature is increased from 323 to 383 K, and the change in HD–D with temperature follows that of the isotherm (see Fig. S13 in the supplementary material). This suggests that since hydrogen-bonds are less dominant at higher temperatures, the isotherm becomes less steep.

In the solution phase, no difference in hydrogen bonding is found for the different diol solutes at the same mole fraction (see Fig. S14 in the supplementary material). For xD < 0.01, HW–W and HD–W remain constant at 3.1 and ≈4.0, respectively (note that the diols possess 4 acceptor and 2 donor sites), and HD–D < 0.1, i.e., diols are well solvated and essentially isolated from each other. At higher concentrations, the decrease in HD–W is nearly exactly balanced by an increase in HD–D; thus, the slight decrease in ΔGVL is due to more favorable dispersive interactions of the diols and a reduction in the cavitation entropy as diols start to aggregate. In the vapor phase, the number of diol–diol hydrogen bonds for C4 is 0.5 ± 0.11 and about factors of 6, 12, and 17 larger than for C3, C5, and C6, respectively (see Table S5 in the supplementary material). C4 can form intramolecular hydrogen bonds with a ring of 6 bonds and little strain whereas bond angle strain and larger entropic penalties (only for C5 and C6) lead to fewer hydrogen bonds. This explains why the solvation thermodynamics do not follow regular methylene increments. The ability of C4 to form significant intramolecular hydrogen bonds in the gas phase was also reported previously.60,107–111 Unary NVT GEMC simulations with a liquid-box and a vapor-box of C3, C4, C5, and C6 with 2520 interaction sites and 3 “impurity” molecules yield HD–D = 3.14 ± 0.02, 3.14 ± 0.04, 3.06 ± 0.03, and 3.03 ± 0.03 for C3, C4, C5, and C6, respectively, in the bulk liquid phase.

In Fig. 8, a quantitive description of the adsorption mechanism is provided for the C6 and C5 diols in the left column [Figs. 8(a), 8(c), and 8(e)] and right column [Figs. 8(b), 8(d), and 8(f)], respectively. Here, the amount located in the zig-zag channels, straight channels, and channel intersections per unit cell for diol centers-of-masses and diol oxygens are shown in Figs. 8(a) and 8(b) and Figs. 8(c) and 8(d), respectively. The fractions of diols in the all-trans conformation for dihedrals involving rotation about two carbons (fall–trans) in the various channels (determined by location of COM) are shown in Figs. 8(e) and 8(f).

FIG. 8.

Channel-specific adsorption of C6 diol from aqueous solution at 323 K (left column) and C5 diol from aqueous solution at 323 K (right column). Rows from top to bottom correspond to the location of diol center-of-mass [(a) and (b)], location of diol oxygen [(c) and (d)], and fraction of diols in the all-trans form at the center-of-mass location [(e) and (f)]. The total adsorption of diol corresponds to the simulation data shown in Fig. 2.

FIG. 8.

Channel-specific adsorption of C6 diol from aqueous solution at 323 K (left column) and C5 diol from aqueous solution at 323 K (right column). Rows from top to bottom correspond to the location of diol center-of-mass [(a) and (b)], location of diol oxygen [(c) and (d)], and fraction of diols in the all-trans form at the center-of-mass location [(e) and (f)]. The total adsorption of diol corresponds to the simulation data shown in Fig. 2.

Close modal

At low loadings, C5 and C6 diols prefer to adsorb in the channel intersections with their oxygens along the straight channels. For comparison, normal alkanes with 7 or fewer carbon atoms preferentially adsorb first in the zig-zag or straight channels, whereas longer normal alkanes prefer the straight channel.112–115 At loadings of above 2 molec/uc, where HD–D ≈ 0.75, the diol COMs reach a maximum loading in the intersections while, simultaneously, the diol oxygens reach a maximum loading in the straight channels. Intermolecular hydrogen-bonding with water along the straight channel occurs as ≈ 1 water COM is located in the straight channels (see Fig. S15 in the supplementary material) and HD–W ≈ 0.5. Compared to C6, whose maximum of diol COM in the intersection occurs at a total loading of ≈4 molec/uc and a height of ≈1.7 molec/uc, the maximum of C5 COM in the intersection occurs at ≈2.7 molec/uc and a height of ≈2.0 molec/uc. As a result, the adsorbed configuration of C5 at this maximum is much more ordered, with the total amount of C5 not adsorbed with COM in the intersection and oxygens along the straight channel being only ≈0.7 molec/uc. This configuration also affords C5 diols with COM in the intersection to have fall–trans ≈ 35% (another maximum for the intersections), while the less-ordered C6 diols only have fall–trans ≈ 20%. The channel-specific adsorption of C5 from aqueous solution at 383 K and 3.0 bars, which does not possess a step change, is similar to the channel-specific adsorption of C6 from aqueous solution at 323 K onto the ORTHO structure (see Fig. S15 in the supplementary material). The preference for adsorption of C5 along the straight channels with COM in the intersections is decreased considerably. As a result, the mechanism behind the steep isotherms for diols with 5 or more carbons is attributed to result from the initial adsorption of COM in the intersections with oxygens along the straight channels, allowing for the formation of intermolecular hydrogen-bonds. Since the adsorption of C6 in the MONO structure results in a stepped isotherm, the MONO structure is expected to allow hydrogen-bonded chains of C6 along the straight channels at low loading to be more favorable.

As the total loading increases, the preference of C5 and C6 COM adsorption exhibits a complete reversal, with the primary adsorption being in the zig-zag and straight channels. C5 adsorption also leads to a reversal in water loading, with the preference changing from the straight channels to the intersections at high loading (see Fig. S15 in the supplementary material). For C6, a reversal in water location is not observed and water adsorbs preferentially in the straight channels even as C6 saturation loading is approached (see Fig. S15 in the supplementary material). The C5 COM occupation in straight channels at saturation is 3.85 ± 0.08, while the C6 COM occupation reaches only 2.96 ± 0.04. This absence of filling all straight channels leads to the high amount of water adsorption in the straight channels at high loading (see Fig. S15 in the supplementary material) and also explains the aforementioned lack of a decrease in water adsorption after saturation of C6 is approached. For the PARA structure, the water adsorption decreases by ≈1 molec/uc as the C6 loading increases from 7.08 ± 0.04 to 7.4 ± 0.2 molec/uc. These results suggest that the drastic decrease in water loading is due to a higher occupation of C6 in the straight channel, decreasing the water adsorption in the straight channel. As the zig-zag and straight channels increase in loading, the fraction of linear diol conformations in the straight channels increases for C5 and C6. The fraction of linear diol conformations in the zig-zag channels for C5 increases up to ≈45% with increasing loading as the number of oxygens in the zig-zag channels approaches 8, while C6 all-trans conformations remain at ≈5% and oxygen count approaches 7.24 ± 0.03. However, despite the longer chain lengths of C6 diols, they have a similar or higher fraction of all-trans configurations adsorbed in the straight channels or intersections at high loading.

While the hydrogen-bonding profiles for C3 and C4 with concentration (and loading) follow the same trends as those for the larger diols, the channel-specific adsorption profile, shown in Fig. 9, is quite contrasting. When the total loading is below the point at which the initial steep change in loading occurs, the amounts of diol oxygens and COMs in each region vary linearly with total loading, indicating a similar preference in location and orientation. This indicates that the diols with smaller chain lengths can still have steep isotherms without having highly ordered configurations that are conducive to hydrogen-bonding. At the point at which the steep rise of adsorption stops, where the water loading is at a maximum, the amount of diol COMs in the intersection is saturated. This suggests that more hydrogen bonds can be formed between water and hydroxyl groups when the diol COMs are in the intersection. At higher loadings, the increased diol COM presence in the zig-zag and straight channels forces water out of these channels (see also Fig. S16 in the supplementary material). The extent to which C3 and C4 diols can pack in the zig-zag and straight channels at high loadings therefore determines the width of the peak in the water loading. The high fraction (approaching 70% at high loading) of linear conformations observed for C3 in the zig-zag channels suggests that the length of C3 is short enough that it can fit in between the bends of the channels. However, by maintaining a linear form, more room is available for water to adsorb (see also Fig. S16 in the supplementary material) in the zig-zag channels.

FIG. 9.

Channel-specific adsorption of C4 diol from aqueous solution at 323 K (left column) and C3 diol from aqueous solution at 323 K (right column). Rows from top to bottom correspond to the location of diol center-of-mass [(a) and (b)], location of diol oxygen [(c) and (d)], and fraction of diols in the all-trans form at the center-of-mass location [(e) and (f)]. The total adsorption of diol corresponds to the simulation data shown in Fig. 2.

FIG. 9.

Channel-specific adsorption of C4 diol from aqueous solution at 323 K (left column) and C3 diol from aqueous solution at 323 K (right column). Rows from top to bottom correspond to the location of diol center-of-mass [(a) and (b)], location of diol oxygen [(c) and (d)], and fraction of diols in the all-trans form at the center-of-mass location [(e) and (f)]. The total adsorption of diol corresponds to the simulation data shown in Fig. 2.

Close modal

The packing along the zig-zag channels at the highest concentrations investigated is well captured by Figs. 10(a), 10(c), 10(e), and 10(g), two-dimensional probability histograms of diol oxygens on the ac-plane, and Table II, which allows for assignment of the total number of oxygens that correspond to well-defined peaks in Fig. 10. Analogous probability histograms for the diol COMs are presented in Fig. S17 in the supplementary material. The order for each mixture along the zig-zag channels appears qualitatively to be C5 > C6 > C4 > C3. While mixtures with diols with more than 3 carbons all have very close to 4 COMs per unit cell in the zig-zag channels, only C5 and C6 mixtures have close to 8 oxygens per unit cell in the zig-zag channels (see Table II). C4 [see Fig. 10(c)] adsorbs to a significant extent with its COM in the intersections, likely leading to the extra oxygen atoms in the zig-zag channels. C5 [see Fig. 10(e)] can extend along the zig-zag channels without its oxygens adsorbing significantly in the intersections. This explains the high fall–trans observed in Fig. 8(f). The C6 oxygens [see Fig. 10(g)] have a significant probability to be located at the edge of the intersections, explaining why NO in the zig-zag channels is <8. While the smaller fall–trans for C6 compared to C5 in the zig-zag channels at saturation results from a longer chain length and one additional dihedral angle (i.e., even in a bulk liquid phase, fall–trans is lower for C6 than C5), the channel geometry imposes a stronger confinement on C6 in the zig-zag channels (the oxygen atoms of C6 in the zig-zag channels occupy similar locations as for C5 in the zig-zag channels) which also leads to more conformational defects to reduce the length extended along the zig-zag channels.

FIG. 10.

Oxygen location probability for (a) and (b) C3, (c) and (d) C4, (e) and (f) C5, and (g) and (h) C6 molecules confined in the pores of MFI from aqueous solution at the highest concentration investigated. [(a), (c), (e), and (g)] Distributions averaged over the ac-plane with zig-zag channels running horizontally (along x) and straight channels running perpendicular to the plane of the page. [(b), (d), (f), and (h)] Distributions averaged over the bc-plane with straight channels running horizontally (along y) and zig-zag channels running perpendicular to the plane of the page. The boundaries of the channel intersections are depicted in red.

FIG. 10.

Oxygen location probability for (a) and (b) C3, (c) and (d) C4, (e) and (f) C5, and (g) and (h) C6 molecules confined in the pores of MFI from aqueous solution at the highest concentration investigated. [(a), (c), (e), and (g)] Distributions averaged over the ac-plane with zig-zag channels running horizontally (along x) and straight channels running perpendicular to the plane of the page. [(b), (d), (f), and (h)] Distributions averaged over the bc-plane with straight channels running horizontally (along y) and zig-zag channels running perpendicular to the plane of the page. The boundaries of the channel intersections are depicted in red.

Close modal
TABLE II.

Channel-specific loadings for alkane-α, ω-diols at maximum loading.

NCOM (per uc)
SorbateT (K)StraightZig-zagIntersection
C3 323 3.41 ± 0.02 4.7 ± 0.1 1.5 ± 0.1 
C4 323 2.71 ± 0.06 3.988 ± 0.004 2.11 ± 0.03 
C5 323 3.85 ± 0.08 3.9995 ± 0.0002 0.15 ± 0.08 
C6 323 2.96 ± 0.04 3.9998 ± 0.0002 0.10 ± 0.04 
C5 383 3.77 ± 0.02 3.988 ± 0.003 0.21 ± 0.03 
NCOM (per uc)
SorbateT (K)StraightZig-zagIntersection
C3 323 3.41 ± 0.02 4.7 ± 0.1 1.5 ± 0.1 
C4 323 2.71 ± 0.06 3.988 ± 0.004 2.11 ± 0.03 
C5 323 3.85 ± 0.08 3.9995 ± 0.0002 0.15 ± 0.08 
C6 323 2.96 ± 0.04 3.9998 ± 0.0002 0.10 ± 0.04 
C5 383 3.77 ± 0.02 3.988 ± 0.003 0.21 ± 0.03 
NO (per uc)
SorbateT (K)StraightZig-zagIntersection
C3 323 6.67 ± 0.07 9.4 ± 0.1 3.20 ± 0.04 
C4 323 5.95 ± 0.06 10.04 ± 0.06 1.61 ± 0.09 
C5 323 3.88 ± 0.08 8.10 ± 0.04 4.0 ± 0.1 
C6 323 1.9 ± 0.2 7.88 ± 0.03 4.3 ± 0.2 
C5 383 3.96 ± 0.05 8.139 ± 0.006 3.83 ± 0.05 
NO (per uc)
SorbateT (K)StraightZig-zagIntersection
C3 323 6.67 ± 0.07 9.4 ± 0.1 3.20 ± 0.04 
C4 323 5.95 ± 0.06 10.04 ± 0.06 1.61 ± 0.09 
C5 323 3.88 ± 0.08 8.10 ± 0.04 4.0 ± 0.1 
C6 323 1.9 ± 0.2 7.88 ± 0.03 4.3 ± 0.2 
C5 383 3.96 ± 0.05 8.139 ± 0.006 3.83 ± 0.05 

The order along the straight channels is observed from the two-dimensional probability histograms of diol oxygens on the bc-plane [see Figs. 10(b), 10(d), 10(f), and 10(h)]. Analogous probability histograms for the diol COMs are presented in Fig. S17 in the supplementary material. While C3 [see Fig. 10(b)] and C4 [see Fig. 10(d)] show more disordered configurations, almost all of the adsorbed locations of C5 [see Fig. 10(f)] and C6 [see Fig. 10(h)] oxygens in the intersections are realized as distinct peaks with PO > 0.01. Since the sum of C6 COMs at this loading in the intersections or straight channels is nearly 3, it is postulated that two consecutive straight channels are likely populated by either 2 or 1 C6 diols. The 4.0 ± 0.1 oxygens of C5 in the intersections are noted by 4 peaks with PO > 0.01 inside the intersections in Fig. 10(f). This suggests that each peak of PO > 0.01 inside the intersections corresponds to ≈1 diol oxygen per unit cell. At the same time, the 3.85 ± 0.08 oxygens in the straight channels are noted by 8 different peaks with PO > 0.01 in the straight channels of in Fig. 10(f), suggesting that these peaks correspond to ≈0.5 diol oxygens. These observations are attributed to C5 molecules adsorbing along the straight channels with one of its oxygen atoms in the intersection and its adjacent oxygen residing in the straight channel either in the positive or negative y-direction, with similar probability. Indeed, the step change isotherms of C5 and C6 which lead to saturation loadings of near integer numbers of diols per unit cell lead to very well ordered configurations.

HD–D approaches 2 for C5 and C6 at maximum loading, suggesting that a hydrogen-bonding network is formed. To quantify the presence of diol–diol hydrogen-bonding networks, a graph was constructed using NetworkX116 between diols involved in hydrogen bonds. The vertexes chosen for the graph were diol oxygens. Two diol oxygens were connected with an edge if they were on the same molecule or if their hydroxyl groups participated in a hydrogen bond, enabling a diol cluster size distribution to be calculated. The cluster size distribution at maximum loading for the aqueous systems at 323 K is presented in Fig. 11. With increasing number of methylene segments, the peak at a cluster size of one decreases. The probability for an adsorbed C3 or C4 molecule to belong to a cluster of size >25 is quite small, being less than 0.01. For C5, however, which has a similar adsorbed composition as C4, a significant probability of being in a large cluster size is observed. A peak with a size of ≈80 is observed which extends to almost the total amount of molecules present in the simulation box. For C6, despite having much more coadsorbed water at high loading, which could hinder diols from hydrogen bonding with other diols, another peak is observed at high loading corresponding to ≈75 molecules. The effect of increasing the temperature on the C5 cluster distribution parallels that of decreasing the solute chain length (see Fig. S18 in the supplementary material). With increasing temperature, the peak at the high cluster size decreases and the peak at the low cluster size increases.

FIG. 11.

Probability for an adsorbed alkane-α, ω-diol (diol) to be in a given diol cluster size in the zeolite box for the simulated aqueous systems at 323 K. The total number of diol molecules in the box is 115 ± 1, 105.6 ± 0.8, 95.996 ± 0.002, and 84.7 ± 0.4 for C3, C4, C5, and C6, respectively. The associated number of water molecules in the box is 35 ± 1, 12 ± 1, 10 ± 1, and 35 ± 1 for C3, C4, C5, and C6, respectively.

FIG. 11.

Probability for an adsorbed alkane-α, ω-diol (diol) to be in a given diol cluster size in the zeolite box for the simulated aqueous systems at 323 K. The total number of diol molecules in the box is 115 ± 1, 105.6 ± 0.8, 95.996 ± 0.002, and 84.7 ± 0.4 for C3, C4, C5, and C6, respectively. The associated number of water molecules in the box is 35 ± 1, 12 ± 1, 10 ± 1, and 35 ± 1 for C3, C4, C5, and C6, respectively.

Close modal

Simulations of C5 (unary, with no water) in an elongated zeolite box in the b-direction or in a very large near-cubic box were undertaken to ascertain whether or not the step-change isotherm corresponds to an adsorbed phase coexistence between low-density and high-density phases. Snapshots of the adsorbed phase in the elongated box at T = 323 and 383 K are presented in Figs. 12(a) and 12(b), respectively. At 323 K, clear regions of high and low adsorbed density are observed (and also visually less clear regions of high and low adsorbed density are observed for the near-cubic box, see Fig. S19 in the supplementary material).

FIG. 12.

Snapshots of adsorbed C5 in an elongated supercell at (a) 323 K and (b) 383 K. The zig-zag channels run vertically and the straight channels run horizontally.

FIG. 12.

Snapshots of adsorbed C5 in an elongated supercell at (a) 323 K and (b) 383 K. The zig-zag channels run vertically and the straight channels run horizontally.

Close modal

To quantitatively determine whether phase coexistence was observed, the probability distributions of carbon loadings in each of the 72 (or 2592) unit cells for the elongated (or near-cubic) were calculated (see Fig. 13). The distribution of carbon loading is more informative than that of COM loading because small fluctuations in the COM position in the straight channel can cause large changes in the number of COM found in a specific unit cell. The distribution at 323 K exhibits clear peaks at loadings of 5 carbon/uc and at 40 carbon/uc, demonstrating that adsorbed phase coexistence is indeed present. Due to the much larger interfacial area in the cubic box, intermediate loadings are observed with a higher probability. With increasing temperature from 323 K to 348 K, the height of the peak at 40 carbon/uc decreases and the low-density peak shifts to 5-10 carbons/uc. At 383 K, a unimodal distribution is observed, suggesting that the loading is in the one-phase region.

FIG. 13.

Probability of observing a unit cell with a given loading of carbon atoms per unit cell for (a) the elongated supercell with 2 × 12 × 3 unit cells and (b) the near-cubic supercell with 12 × 12 × 18 unit cells. The overall average C5 loadings at 323 K, 348 K, and 383 K are 4.385 ± 0.002, 4.369 ± 0.002, and 4.280 ± 0.005 molec/uc, respectively, in the elongated box and 3.855 88 ± 0.000 09, 3.8556 ± 0.000 12, and 3.8542 ± 0.000 11, molec/uc, respectively, in the cubic box.

FIG. 13.

Probability of observing a unit cell with a given loading of carbon atoms per unit cell for (a) the elongated supercell with 2 × 12 × 3 unit cells and (b) the near-cubic supercell with 12 × 12 × 18 unit cells. The overall average C5 loadings at 323 K, 348 K, and 383 K are 4.385 ± 0.002, 4.369 ± 0.002, and 4.280 ± 0.005 molec/uc, respectively, in the elongated box and 3.855 88 ± 0.000 09, 3.8556 ± 0.000 12, and 3.8542 ± 0.000 11, molec/uc, respectively, in the cubic box.

Close modal

To determine whether the high-density phase is a liquid-crystalline phase with positional disorder in the straight channels or a fully crystalline phase, structure factors were calculated117 using the following equation:

(1)

where q is the wave vector, rj is the position of j, and Nsite is the number of interaction sites considered. Both locations of diol COMs and oxygen atoms were considered for q = {011}, {101}, {200}, and {020} (see Fig. 14). At {101}, which only encompasses the zig-zag channels, S for C6 is higher than S for C5, and C5 does not exhibit a significant temperature dependence. At {200}, which runs along the straight channels (and therefore also includes some locations in the intersections and zig-zag channels), S for C5 is larger than that for C6 and exhibits a temperature dependence. The {011} (which includes locations in the zig-zag channels and centers of straight channels) and {020} (which only consists of the centers of straight channels) planes exhibit S ordering of C5 > C6 and a decrease in S for C5 with increasing temperature. This demonstrates that the order of the high-density phase for C5 does not exhibit significant temperature dependence in the zig-zag channels. The order along the straight channels and in the center of the straight channels is, however, decreased with temperature and with an increase in methylene segments.

FIG. 14.

Structure factor of adsorbed locations of diol center-of-mass and oxygen atoms for adsorbed aqueous mixtures of C5 at 323 K and 383 K and of C6 at 323 K at the highest concentrations investigated.

FIG. 14.

Structure factor of adsorbed locations of diol center-of-mass and oxygen atoms for adsorbed aqueous mixtures of C5 at 323 K and 383 K and of C6 at 323 K at the highest concentrations investigated.

Close modal

Simulations and experiments were used to obtain the adsorption isotherms for alkane-α, ω-diols (diols) with 3–6 carbons from aqueous solutions onto all-silica MFI. Good agreement was found between simulation onto the MONO structure and experiment, whereas adsorption onto the ORTHO structure is shifted to somewhat lower concentrations and yields a different isotherm shape for C6. The initial onset of diol adsorption from aqueous solution was found to decrease by about one order of magnitude with increasing number of methylene segments, and this was found to result primarily from the more favorable interactions in the adsorbed phase (as opposed to in the aqueous phase). The diol isotherms were found to exhibit steep changes in loading with respect to the solution concentration up to loadings of about 7–10 molec/uc, analogous to a first-order phase transition. The steep changes in loading nearly coincide with minima in the free energy of transfer from vapor to zeolite. The associated water loadings exhibit a maximum with a height and width that depends on how efficiently the diols can pack in the pore channels.

The C3 and C4 isotherms follow the initial step with a “knee” until maximum loading is reached. Their channel-specific adsorbed locations and conformations are linear with total loading up until the end of the initial step. The C5 and C6 loadings, however, are observed to be nearly a step function, and the channel-specific adsorbed locations and conformations correspond to a reversal in preferences below and above the step. At low loadings, the C5 and C6 diols adsorb with COMs occupying the channel intersections with oxygens orienting along the straight channels and forming intermolecular hydrogen bonds. At saturation loadings, however, the C5 and C6 diol COMs occupy the straight and zig-zag channels.

The number of diol–diol hydrogen bonds per adsorbed diol molecule as a function of solution concentration was found to be significantly similar in shape to the diol loading as a function of solution concentration. This suggests that diol–diol intermolecular hydrogen bonds play a large role in the steep change in loading with solution concentration. Cluster analysis of diol hydrogen-bonding networks revealed large diol clusters to form at maximum loading. The maximum cluster size found for C5 and C6 corresponds to almost all diols forming an entire network.

With increasing temperature, the C5 isotherm does not shift significantly in concentration of the initial onset with increasing temperature because the free energy of adsorption and of solvation increase by similar and compensating amounts. At the same time, the C5 isotherm flattens as the free energy of transfer from vapor to zeolite decreases in depth, and the water adsorption isotherm increases in height and width. Compared to 323 K, the maximum of C5 COM in the intersection at 383 K was associated with a lower proportion of diol oxygens in the straight channels, and the fraction of all trans conformers in the zig-zag channels at high loading decreased by a factor of 2. With increasing temperature, the diol cluster size at maximum loading decreases, with the peak at the high cluster size decreasing at 383 K. Therefore, the decrease in depth of the free energy of adsorption (and of the decrease in steepness) at increased temperature is attributed to result from the increased water adsorption, decreased order, and decreased population of hydrogen bonds at higher temperatures.

Simulations indicate that the step in the C5 isotherm corresponds to adsorbed phase coexistence at 323 K and 348 K, with distinct low and high-density regions found in a large supercell. Analysis of structure factors reveals that the high-density phase for C5 is positionally ordered in both channels at 323 and 343 K, whereas less order is found in the straight channels for C5 at 383 K and C6 at 323 K. At 383 K, there is no steep step in the isotherm, and the distribution of C5 molecules at half of saturation is uniform over a large supercell. This unique behavior could play a large role in reactions and separations and is expected to be quite general for the adsorption of diols in zeolites.

See supplementary material for tabular data for simulated equilibria, an examination of the sensitivity of simulated C5 and C6 isotherms from aqueous solution to the experimental structure adopted, unary and binary isotherms from the vapor phase, sensitivity of simulated C5 and C6 selectivities and compositions adsorbed from aqueous solution to the experimental structure adopted, selectivity and composition adsorbed for C5 as a function of temperature, unary and binary adsorption of C5 as a function of temperature from the vapor phase, enthalpic and entropic terms for the free energies of transfer for unary adsorption of diols at 323 K, the free energy of solvation and of adsorption for water associated with the diols at 323 K, the free energy of solvation and of adsorption for water associated with C5 at the various temperatures investigated, two-dimensional histograms of angles and distances relevant to hydrogen bonding criteria, the sensitivity of hydrogen bonding criteria adopted, hydrogen bonding in zeolites for C5 as a function of temperature, hydrogen bonding of diols in the vapor phase, hydrogen bonding of C5 as a function of temperature, hydrogen bonding of diols in solution, additional channel-specific adsorption information for the adsorption of C5 and C6 mixtures from aqueous solution, additional channel-specific adsorption information for the adsorption of C3 and C4 mixtures from aqueous solution, probability distributions of diol COMs at the highest loading investigated, hydrogen bonding cluster distributions as a function of temperature, and simulation snapshots of the phase coexistence simulations for the massively large near-cubic box.

This material is based upon work supported by the Department of Energy (DOE), Office of Energy Efficiency and Renewable Energy (EERE), under Award No. DE-EE0006878. This research used resources of the Argonne Leadership Computing Facility, which is a DOE Office of Science User Facility supported under Contract No. DE-AC02-06CH11357. Additional computer resources were provided by the Minnesota Supercomputing Institute at the University of Minnesota. We thank Kefeng Huang and George Huber for helpful discussions on process conditions for the catalytic production of diols.

1.
A.
Primo
and
H.
Garcia
,
Chem. Soc. Rev.
43
,
7548
(
2014
).
2.
E. T. C.
Vogt
and
B.
Weckhuysen
,
Chem. Soc. Rev.
44
,
7342
(
2015
).
3.
H.
Nguyen
,
R. F.
DeJaco
,
N.
Mittal
,
J. I.
Siepmann
,
M.
Tsapatsis
,
M. A.
Snyder
,
W.
Fan
,
B.
Saha
, and
D. G.
Vlachos
,
Annu. Rev. Chem. Biomol. Eng.
8
,
115
(
2017
).
4.
E. E.
Mallon
,
M. Y.
Jeon
,
M.
Navarro
,
A.
Bhan
, and
M.
Tsapatsis
,
Langmuir
29
,
6546
(
2013
).
5.
M.
Leon
,
T. D.
Swift
,
V.
Nikolakis
, and
D. G.
Vlachos
,
Langmuir
29
,
6597
(
2013
).
6.
A.
Farzaneh
,
R. F.
DeJaco
,
L.
Ohlin
,
A.
Holmgren
,
J. I.
Siepmann
, and
M.
Grahn
,
Langmuir
33
,
8420
(
2017
).
7.
L. M.
Vane
,
Biofuels, Bioprod. Biorefin.
2
,
553
(
2008
).
8.
R.
Xiong
,
S. I.
Sandler
, and
D. G.
Vlachos
,
Langmuir
28
,
4491
(
2012
).
9.
N. B.
Milestone
and
D. M.
Bibby
,
J. Chem. Technol. Biotechnol.
31
,
732
(
1981
).
10.
P.
Bai
,
M.
Tsapatsis
, and
J. I.
Siepmann
,
Langmuir
28
,
15566
(
2012
).
11.
B.
Cekova
,
D.
Kocev
,
E.
Kolcakovska
, and
D.
Stojanova
,
Acta Period. Technol.
(
37
),
83
(
2006
).
12.
N. B.
Milestone
and
D. M.
Bibby
,
J. Chem. Technol. Biotechnol., Chem. Technol.
34
,
73
(
1983
).
13.
R. F.
DeJaco
,
P.
Bai
,
M.
Tsapatsis
, and
J. I.
Siepmann
,
Langmuir
32
,
2093
(
2016
).
14.
T.
Remy
,
J.
Cousin Saint Remi
,
R.
Singh
,
P. A.
Webley
,
G. V.
Baron
, and
J. F. M.
Denayer
,
J. Phys. Chem. C
115
,
8117
(
2011
).
15.
J.
Cousin Saint Remi
,
T.
Remy
,
V.
Van Hunskerken
,
S.
van de Perre
,
T.
Duerinck
,
M.
Maes
,
D.
De Vos
,
E.
Gobechiya
,
C. E. A.
Kirschhock
,
G. V.
Baron
, and
J. F. M.
Denayer
,
ChemSusChem
4
,
1074
(
2011
).
16.
J.
Cousin-Saint-Remi
and
J. F. M.
Denayer
,
Chem. Eng. J.
324
,
313
(
2017
).
17.
T. C.
Bowen
and
L. M.
Vane
,
Langmuir
22
,
3721
(
2006
).
18.
R.
Xiong
,
S. I.
Sandler
, and
D. G.
Vlachos
,
J. Phys. Chem. C
115
,
18659
(
2011
).
19.
F. A.
Farhadpour
and
A.
Bono
,
J. Colloid Interface Sci.
124
,
209
(
1988
).
20.
M.
Yu
,
J. L.
Falconer
, and
R. D.
Noble
,
Langmuir
21
,
7390
(
2005
).
21.
P.
Bai
,
J. I.
Siepmann
, and
M. W.
Deem
,
AIChE J.
59
,
3523
(
2013
).
22.
T.
Ennaert
,
J.
Van Aelst
,
J.
Dijkmans
,
R.
De Clercq
,
W.
Schutyser
,
M.
Dusselier
,
D.
Verboekend
, and
B. F.
Sels
,
Chem. Soc. Rev.
45
,
584
(
2016
).
23.
P.
Werle
,
M.
Morawietz
,
S.
Lundmark
,
K.
Sörensen
,
E.
Karvinen
, and
J.
Lehtonen
,
Ullmann’s Enclyclopedia of Industrial Chemistry
(
Wiley-VCH
,
2008
), pp.
263
284
.
24.
M.
Dashtban
,
A.
Gilbert
, and
P.
Fatehi
,
J. Sci. Technol. For. Prod. Processes
2
,
44
(
2012
).
25.
M.
Chia
,
Y. J.
Pagán-Torres
,
D.
Hibbitts
,
Q.
Tan
,
H. N.
Pham
,
A. K.
Datye
,
M.
Neurock
,
R. J.
Davis
, and
J. A.
Dumesic
,
J. Am. Chem. Soc.
133
,
12675
(
2011
).
26.
Y.
Nakagawa
,
M.
Tamura
, and
K.
Tomishige
,
Catal. Surv. Asia
19
,
249
(
2015
).
27.
S.
Koso
,
I.
Furikado
,
A.
Shimao
,
T.
Miyazawa
,
K.
Kunimori
, and
K.
Tomishige
,
Chem. Commun.
0
,
2035
(
2009
).
28.
P.
Bai
and
J. I.
Siepmann
,
J. Chem. Theory Comput.
13
,
431
(
2017
).
29.
K.
Huang
,
Z. J.
Brentzel
,
K. J.
Barnett
,
J. A.
Dumesic
,
G. W.
Huber
, and
C. T.
Maravelias
,
ACS Sustainable Chem. Eng.
5
,
4699
(
2017
).
30.
Z. J.
Brentzel
,
K. J.
Barnett
,
C. T.
Huang
,
K.
Maravelias
,
J. A.
Dumesic
, and
G. W.
Huber
,
ChemSusChem
10
,
1351
(
2017
).
31.
S. P.
Burt
,
K. J.
Barnett
,
D. J.
McClelland
,
P.
Wolf
,
J. A.
Dumesic
,
G. W.
Huber
, and
I.
Hermans
,
Green Chem.
19
,
1390
(
2017
).
32.
K.
Huang
,
W.
Won
,
K. J.
Barnett
,
Z. J.
Brentzel
,
D. M.
Alonso
,
G. W.
Huber
,
J. A.
Dumesic
, and
C. T.
Maravelias
,
Appl. Energy
213
,
585
(
2018
).
33.
S.
Li
,
V. A.
Tuan
,
J. L.
Falconer
, and
R. D.
Noble
,
Chem. Mater.
13
,
1865
(
2001
).
34.
S.
Li
,
V. A.
Tuan
,
J. L.
Falconer
, and
R. D.
Noble
,
Ind. Eng. Chem. Res.
40
,
1952
(
2001
).
35.
S.
Li
,
V. A.
Tuan
,
J. L.
Falconer
, and
R. D.
Noble
,
J. Membr. Sci.
191
,
53
(
2001
).
36.
V. R.
Rani
,
N.
Srinivas
,
S. J.
Kulkarni
, and
K. V.
Raghavan
,
J. Mol. Catal. A: Chem.
187
,
237
(
2002
).
37.
C. P.
Bezouhanova
and
F. A.
Jabur
,
React. Kinet. Catal. Lett.
51
,
177
(
1993
).
38.
M.
Aghaziarati
,
M.
Kazemeini
,
M.
Soltanieh
, and
S.
Sahebdelfar
,
Ind. Eng. Chem. Res.
46
,
726
(
2007
).
39.
A.
Araya
and
B. M.
Lowe
,
Zeolites
6
,
111
(
1986
).
40.
A.
Chatterjee
and
R.
Vetrivel
,
J. Chem. Soc., Faraday Trans.
91
,
4313
(
1995
).
41.
C.
Falamaki
,
M.
Edrissi
, and
M.
Sohrabi
,
Zeolites
19
,
2
(
1997
).
42.
C.
Falamaki
,
M.
Sohrabi
, and
G.
Talebi
,
Chem. Eng. Technol.
24
,
501
(
2001
).
43.
S. G.
Fegan
and
B. M.
Lowe
,
J. Inclusion Phenom. Mol. Recognit. Chem.
7
,
461
(
1989
).
44.
R. E.
Richards
and
L. V. C.
Rees
,
Langmuir
3
,
335
(
1987
).
45.
M. M.
Dubinin
,
G. U.
Rakhmatkariev
, and
A. A.
Isirikyan
,
Russ. Chem. Bull.
38
,
2146
(
1989
).
46.
F.
Eder
and
J. A.
Lercher
,
Zeolites
18
,
75
(
1997
).
47.
M. S.
Sun
,
O.
Talu
, and
D. B.
Shah
,
J. Phys. Chem.
100
,
17276
(
1996
).
48.
B.
Smit
and
T. L. M.
Maesen
,
Nature
374
,
42
(
1995
).
49.
W. J. M.
van Well
,
J. P.
Wolthuizen
,
B.
Smit
,
J. H. C.
van Hooff
, and
R. A.
van Santen
,
Angew. Chem., Int. Ed. Engl.
34
,
2543
(
1995
).
50.
D. H.
Olson
and
P. T.
Reischman
,
Zeolites
17
,
434
(
1996
).
51.
N.
Floquet
,
J. P.
Coulomb
,
J. P.
Bellat
,
J. M.
Simon
,
G.
Weber
, and
G.
Andre
,
J. Phys. Chem. C
111
,
18182
(
2007
).
52.
N.
Desbiens
,
I.
Demachy
,
A. H.
Fuchs
,
H.
Kirsh-Rodeschini
,
M.
Soulard
, and
J.
Patariin
,
Angew. Chem., Int. Ed.
44
,
5310
(
2005
).
53.
N.
Desbiens
,
A.
Boutin
, and
I.
Demachy
,
J. Phys. Chem. B
109
,
24071
(
2005
).
54.
F.
Cailliez
,
M.
Trzpit
,
M.
Soulard
,
I.
Demachy
,
A.
Boutin
,
J.
Patarin
, and
A. H.
Fuchs
,
Phys. Chem. Chem. Phys.
10
,
4817
(
2008
).
55.
G.
Fraux
,
F.-X.
Coudert
,
A.
Boutin
, and
A. H.
Fuchs
,
Chem. Soc. Rev.
46
,
7421
(
2017
).
56.
Q.
Yang
and
C.
Zhong
,
Langmuir
25
,
2302
(
2009
).
57.
R.
Krishna
and
J. M.
van Baten
,
Langmuir
26
,
3981
(
2009
).
58.
R.
Krishna
and
J. M.
van Baten
,
Langmuir
26
,
10854
(
2010
).
59.
J. K.
Kim
,
G.
Natarajan
, and
P. C.
Wankat
,
Adsorption
9
,
117
(
2003
).
60.
R. A.
Klein
,
J. Comput. Chem.
23
,
585
(
2002
).
61.
R. A.
Klein
,
J. Am. Chem. Soc.
124
,
13931
(
2002
).
62.
F. J.
Luque
,
J. M.
López
,
M. L.
de la Paz
,
C.
Vicent
, and
M.
Orozco
,
J. Phys. Chem. A
102
,
6690
(
1998
).
63.
W. L.
Jorgensen
,
J.
Chandrasekhar
,
J. D.
Madura
,
R. W.
Impey
, and
M. L.
Klien
,
J. Chem. Phys.
79
,
926
(
1983
).
64.
M. G.
Martin
and
J. I.
Siepmann
,
J. Phys. Chem. B
102
,
2569
(
1998
).
65.
B.
Chen
,
J. J.
Potoff
, and
J. I.
Siepmann
,
J. Phys. Chem. B
105
,
3093
(
2001
).
66.
J. M.
Stubbs
,
J. J.
Potoff
, and
J. I.
Siepmann
,
J. Phys. Chem. B
108
,
17596
(
2004
).
67.
M. S.
Kelkar
,
J. L.
Rafferty
,
E. J.
Maginn
, and
J. I.
Siepmann
,
Fluid Phase Equilib.
260
,
218
(
2007
).
68.
K. A.
Maerzke
and
J. I.
Siepmann
,
J. Phys. Chem. B
114
,
4261
(
2010
).
69.
S. Y.
Mun
and
H.
Lee
,
J. Chem. Eng. Data
44
,
1231
(
1999
).
70.
K.
Jelínek
,
F.
Lešek
, and
M.
Sivoková
,
Collect. Czech. Chem. Commun.
41
,
2650
(
1976
).
71.
H.
van Koningsveld
,
H.
van Bekkum
, and
J. C.
Jansen
,
Acta Crystallogr., Sect. B: Struct. Sci.
43
,
127
(
1987
).
72.
C.
Baerlocher
and
L. B.
McCusker
, Database of Zeolite Structures,
2017
.
73.
P.
Bai
,
M.
Tsapatsis
, and
J. I.
Siepmann
,
J. Phys. Chem. C
117
,
24375
(
2013
).
74.
R. L.
June
,
A. T.
Bell
, and
D. N.
Theodorou
,
J. Phys. Chem.
94
,
1508
(
1990
).
75.
H.
van Koningsveld
,
H.
van Bekkum
, and
J. C.
Jansen
,
Zeolites
10
,
235
(
1990
).
76.
H.
van Koningsveld
,
F.
Tuinstra
,
H.
van Bekkum
, and
J. C.
Jansen
,
Acta Crystallogr., Sect. B: Struct. Sci.
45
,
423
(
1989
).
77.
A. Z.
Panagiotopoulos
,
Mol. Phys.
61
,
813
(
1987
).
78.
A. Z.
Panagiotopoulos
,
N.
Quirke
,
M.
Stapleton
, and
D. J.
Tildesley
,
Mol. Phys.
63
,
527
(
1988
).
79.
B.
Smit
,
P.
Desmedt
, and
D.
Frenkel
,
Mol. Phys.
68
,
931
(
1989
).
80.
S. M.
Melnikov
,
A.
Höltzel
,
A.
Siedel-Morgenstern
, and
U.
Tallarek
,
J. Phys. Chem. C
117
,
6620
(
2013
).
81.
M. G.
Martin
and
J. I.
Siepmann
,
J. Phys. Chem. B
103
,
4508
(
1999
).
82.
J. I.
Siepmann
and
D.
Frenkel
,
Mol. Phys.
75
,
59
(
1992
).
83.
D.
Frenkel
,
G. C. A. M.
Mooij
, and
B.
Smit
,
J. Phys.: Condens. Matter
4
,
3053
(
1992
).
84.
J. J.
de Pablo
,
M.
Laso
,
J. I.
Siepmann
, and
U. W.
Suter
,
Mol. Phys.
80
,
55
(
1993
).
85.
G. C. A. M.
Mooij
,
D.
Frenkel
, and
B.
Smit
,
J. Phys.: Condens. Matter
4
,
L225
(
1992
).
86.
M.
Laso
,
J. J.
de Pablo
, and
U. W.
Suter
,
J. Chem. Phys.
97
,
2817
(
1992
).
87.
J. N. C.
Lopes
and
D. J.
Tildesley
,
Mol. Phys.
92
,
187
(
1997
).
88.
A. V. A. V.
Neimark
,
Phys. Rev. E
62
,
4611
(
2000
).
89.
A. V.
Neimark
and
A.
Vishnyakov
,
J. Chem. Phys.
122
,
234108
(
2005
).
90.
M. G.
Martin
and
J. I.
Siepmann
,
J. Am. Chem. Soc.
119
,
8921
(
1997
).
91.
M. P.
Allen
and
D. J.
Tildesley
,
Computer Simulation of Liquids
(
Oxford University Press
,
New York
,
1987
).
92.
E.
Braun
,
J. J.
Chen
,
S. K.
Schnell
,
L.-C.
Lin
,
J. A.
Reimer
, and
B.
Smit
,
Angew. Chem.
127
,
14557
(
2015
).
93.
M.
Moosavi
and
A. A.
Rostami
,
J. Chem. Eng. Data
62
,
156
(
2016
).
94.
L.
Hnĕdkovský
and
I.
Cibulka
,
J. Chem. Eng. Data
58
,
1724
(
2013
).
95.
M.
Trzpit
,
M.
Soulard
,
J.
Patarin
,
N.
Desbiens
,
F.
Cailliez
,
A.
Boutin
,
I.
Demachy
, and
A. H.
Fuchs
,
Langmuir
23
,
10131
(
2007
).
96.
B. D.
Smith
and
R.
Srivastava
,
Thermodynamics Data for Pure Compounds
(
Elsevier
,
Amsterdam
,
1986
), p.
284
.
97.
C. H.
Wang
,
P.
Bai
,
J. I.
Siepmann
, and
A. E.
Clark
,
J. Phys. Chem. C
118
,
19723
(
2014
).
98.
H.
Morell
,
K.
Angermund
,
A. R.
Lewis
,
D. H.
Brouwer
,
C. A.
Fyfe
, and
H.
Gies
,
Chem. Mater.
14
,
2192
(
2002
).
99.
R. Q.
Snurr
,
A. T.
Bell
, and
D. N.
Theodorou
,
J. Phys. Chem.
97
,
13742
(
1993
).
100.
Y.
Huang
,
J. Am. Chem. Soc.
118
,
7233
(
1996
).
101.
C. A.
Fyfe
,
H.
Strobl
,
G. T.
Kokotailo
,
G. J.
Kennedy
, and
G. E.
BArlow
,
J. Am. Chem. Soc.
110
,
3373
(
1988
).
102.
T. J. H.
Vlugt
,
R.
Krishna
, and
B.
Smit
,
J. Phys. Chem. B
103
,
1102
(
1999
).
103.
M. G.
Martin
and
J. I.
Siepmann
,
Theor. Chem. Acc.
99
,
347
(
1998
).
104.
A.
Ben-Naim
,
Statistical Thermodynamics for Chemists and Biochemists
(
Plenum Press
,
New York
,
1992
).
105.
R. P. J.
Ranatunga
and
P. W.
Carr
,
Anal. Chem.
72
,
5679
(
2000
).
106.
J. L.
Rafferty
,
L.
Sun
,
J. I.
Siepmann
, and
M. R.
Schure
,
Fluid Phase Equilib.
290
,
25
(
2010
).
107.
D. L.
Howard
and
H. G.
Kjaergaard
,
J. Phys. Chem. A
110
,
10245
(
2006
).
108.
P.
Das
,
P. K.
Das
, and
E.
Arunan
,
J. Phys. Chem. A
119
,
3710
(
2015
).
109.
Y.-L.
Cheng
,
H.-Y.
Chen
, and
K.
Takahashi
,
J. Phys. Chem. A
115
,
5641
(
2011
).
110.
H.-Y.
Chen
,
Y. L.
Cheng
, and
K.
Takahashi
,
J. Phys. Chem. A
115
,
14315
(
2011
).
111.
L.
Evangelisti
,
Q.
Gou
,
L.
Spada
,
G.
Feng
, and
W.
Caminati
,
Chem. Phys. Lett.
556
,
55
(
2013
).
112.
B.
Smit
and
J. I.
Siepmann
,
Science
264
,
1118
(
1994
).
113.
B.
Smit
and
J. I.
Siepmann
,
J. Phys. Chem.
98
,
8442
(
1994
).
114.
E. J.
Maginn
,
A. T.
Bell
, and
D. N.
Theodorou
,
J. Phys. Chem.
99
,
2057
(
1995
).
115.
K. M. A.
De Meyer
,
S.
Chempath
,
J. F. M.
Denayer
,
J. A.
Martens
,
R. Q.
Snurr
, and
G. V.
Baron
,
J. Phys. Chem. B
107
,
10760
(
2003
).
116.
A. A.
Hagberg
,
D. A.
Schult
, and
P. J.
Swart
, in
Proceedings of the 7th Python in Science Conference (SciPy2008)
(
SciPy Conference
,
Pasadena, CA, USA
,
2008
), pp.
11
15
.
117.
A. J.
Schultz
,
C. K.
Hall
, and
J.
Genzer
,
Macromolecules
40
,
2629
(
2007
).

Supplementary Material