Article pubs.acs.org/Langmuir
Effect of Hydrophilic Defects on Water Transport in MFI Zeolites Thomas Humplik,† Rishi Raj,† Shalabh C. Maroo,‡ Tahar Laoui,§ and Evelyn N. Wang*,† †
Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, United States Department of Mechanical & Aerospace Engineering, Syracuse University, Syracuse, New York 13244, United States § Department of Mechanical Engineering, King Fahd University of Petroleum and Minerals, Dhahran, Saudi Arabia ‡
S Supporting Information *
ABSTRACT: The subnanometer pore structure of zeolites and other microporous materials has been proposed to act as a molecular sieve for various water separation technologies. However, due to the increased interaction between the solid and water in these nanoconfined spaces, it is unclear which type of interface, be it hydrophilic or hydrophobic, offers an advantageous medium for enhancing transport properties. In this work, we probe the role of hydrophilic defects on the transport of water inside the microporous hydrophobic MFI zeolite pore structure via combined sorption and highpressure infiltration experiments. While the inclusion of defects was observed to increase the amount of water within the zeolite pore network by up to 7 times at the saturation pressure, the diffusivity of this infiltrated water was lowered by up to 2 orders of magnitude in comparison to that of water within the nearly defectfree hydrophobic MFI zeolite. Subsequently, the permeability of water within the more defective MFI zeolite was an order of magnitude lower than that of the nearly defect-free zeolite. The results from these experiments suggest that the intrinsic hydrophobic pore structure of MFI zeolites can facilitate faster water transport due to the decreased attraction between the water and the defect-free surface. While the strong attraction of water to the defects allows for water to infiltrate the porous network at lower pressures, the results suggest that this strong attraction decreases the mobility of the infiltrated water. The insights gained from this study can be utilized to improve the design of future membranes for water desalination and other separation techniques.
1. INTRODUCTION The physical and transport properties of water are usually altered due to the increased effect of solid−liquid molecular interactions when water is confined to nanometer length scales.1−6 Depending on the strength of interaction between a water molecule and the surface, the local interface can be defined either as hydrophobic (i.e., the ratio of the water−water interaction energy to water−solid interaction energy is high) or hydrophilic (i.e., the ratio of the water−water interaction energy to the water−solid interaction energy is low).52 While a hydrophilic surface is by definition water-loving and has long been expected to facilitate better water transport,8−11 some recent studies have shown the opposite behavior where a significant increase in the flow rate of water confined within a nanometer-sized hydrophobic pore (e.g., a carbon nanotube) was observed.12−14 Such an interface could potentially be exploited to improve the performance of membranes used in seawater reverse osmosis desalination and other water-based separation applications. However, additional research is needed to improve our understanding of the physical mechanism behind the water transport within a generic hydrophobic pore. Zeolites, which are microporous (1000) MFI (Mobil Five) zeolites require pressures in excess of 100 MPa to be applied to completely saturate the porous network with water.18,20−24 This large “infiltration pressure” is due to the lack of hydrogenbinding sites within the zeolite, thus creating both enthalpic and entropic barriers for entry.21,25 However, when a large number of hydrophilic defects (either acidic defects or extra-framework cations) are introduced within the zeolite, such as with FAU (Faujasite) or LTA (Linde type A) zeolites, water can completely saturate the porous network at pressures of 1 kPa or less.26,27 The low-pressure pore filling is caused by the large attraction energy of water to these defects, the magnitude of which surpasses even the heat of sublimation for water, thereby Received: March 10, 2014 Revised: May 5, 2014
A
dx.doi.org/10.1021/la500939t | Langmuir XXXX, XXX, XXX−XXX
Langmuir
Article
Figure 1. SEM (A) and XRD (B) analysis of MFI zeolites. Although there is some variation in size among the different Si/Al ratio zeolites, the zeolites all have an internal to external surface area ratio that is greater than 1000 (see Table S1) such that the sorption/infiltration analysis was limited to the internal pore structure. The scale bar for all of the SEM images is 5 μm. The XRD patterns for each zeolite sample matched well with that of known MFI zeolites.
cell (N/UC).24 The increasing defect density decreased the measured diffusivity by up to 2 orders of magnitude and the permeability by upwards of an order of magnitude compared to water within the nearly defect-free Silicalite-1 MFI zeolite. The results from these experiments highlight the strong attraction of water to the hydrophilic defect sites within the zeolite, which, although increases the solubility of the zeolites, has a pronounced detrimental effect on the diffusivity and consequently on the permeability. The experimental results from this study can be utilized to provide detailed physical insights into the transport mechanisms, which can then help guide the design of high-permeability membranes in various water-based separation applications.
overcoming any entry barriers into the subnanometer pores.28−31 It is due to these competing interactions that the mechanism of water transport within these nanoscale pores is not well-understood and that the design of an optimal interface to enhance water transport within these pores is complicated. For transport processes that are limited by diffusion, the permeability of the fluid through the material is based on the product of the diffusivity and the solubility (i.e., sorption coefficient), indicating that both the rate of transport and the amount of infiltrated water are important in controlling the permeability.32 While it is well-known that increasing the defect density increases the solubility of the zeolite,31,33,34 both molecular dynamics simulations and nuclear magnetic resonance experiments have demonstrated that the inclusion of defects decreases the water diffusivity.19,35,36 These trends suggest that the influence of the defect density on the interplay between the solubility and diffusivity, and ultimately the permeability, is not straightforward. To understand these effects better, increased control over experimental techniques and analysis is needed to systematically investigate the competing parametric effects of the diffusivity and the solubility within zeolite crystals. In this work, we experimentally investigated the role of the concentration of hydrophilic defects in MFI (i.e., ZSM-5) zeolites on the permeability of water by independently quantifying both the solubility and diffusivity. We synthesized MFI zeolites with a compositional silicon/aluminum (Si/Al) ratio varying from 100 to infinite (i.e., Silicalite-1) and confirmed the structure and morphology using X-ray diffraction (XRD) and scanning electron microscopy (SEM). Through combined sorption and high-pressure infiltration experiments, we investigated the solubility of water within the zeolite pores as a function of pressure and chemical potential. The diffusivity and subsequently, the permeability of water within the pores were evaluated by analyzing the transient adsorption and desorption behavior. For a given pressure, we found that the amount of water that infiltrated the zeolite porous network increased as the internal defect density increased. However, none of the zeolites studied were found to be completely filled at the saturation pressure of water (3.14 kPa at 298 K), and each zeolite sample required upwards of 40 MPa to reach the measured framework capacity of 35 water molecules per unit
2. EXPERIMENTAL MATERIALS AND METHODS Five different MFI zeolites were synthesized for this study with a varying Si/Al ratio (Figure 1A). For all but the purely siliceous MFI zeolite (which does not contain aluminum), aluminum nitrate was used as the aluminum source (ACS reagent >98%, Sigma-Aldrich). First (if needed), the appropriate amount of aluminum nitrate was dissolved in tetrapropylammonium (TPA) hydroxide (TPAOH, 1 M in H2O, Sigma-Aldrich). Once the solution became clear after agitation, the proper amounts of tetraethyl orthosilicate (TEOS, >98%, Sigma-Aldrich) and deionized (DI) water (class 2, VWR) were added to the solution, after which the solution was left to age overnight under agitation to fully hydrolyze the TEOS. The molar concentration of the solutions was as follows, where X is the appropriate molar concentration of aluminum nitrate for each respective zeolite. 35:10:X:1750 TEOS/TPAOH/AlNit/H 2O The solution was then transferred into a PTFE-lined stainless steel autoclave (45 mL, Parr, Inc.), placed into a preheated furnace at 175 °C (BlueM, Lindberg), and heated under rotation for 5 h. After synthesis, the resulting solution was centrifuged, decanted, and washed with DI water at least three times to lower the pH of the solution to ∼9. The crystals were then dried at 60 °C and calcined at 550 °C (with 3 °C/min ramp rates) in air to remove the organic template and to condense any silanol groups that may have formed during synthesis.37,38 The zeolites were imaged using SEM (Ultra-55, Zeiss) (Figure 1A). For SEM analysis, the samples were coated with ∼5 nm of Pt/Pd to prevent charging during imaging. The volume and exterior surface area of all zeolites were determined by the statistical sampling of >20 crystals and are provided in Table S1 (Supporting Information S.1). B
dx.doi.org/10.1021/la500939t | Langmuir XXXX, XXX, XXX−XXX
Langmuir
Article
XRD (PANalytical X’Pert Pro, Phillips) was performed using a Cu Kα target and a nickel filter to collect the diffraction patterns (Figure 1B) and was used to confirm the MFI structure of the synthesized zeolites. To quantify the amount of water that adsorbed in the zeolites up to approximately 98% of the saturation pressure (3.14 kPa at 25 °C), physiosorption of water vapor was carried out at 25 °C using a gravimetric vapor sorption analyzer (Q5000SA, TA Instruments). Before the tests, the samples were dried in a furnace at 500 °C under air for at least 10 h. The calculated adsorbed water was converted into units of water molecules per unit cell of zeolite. Following the sorption experiments, the zeolites were immersed in 10 mL of DI water and the solution was introduced into a custom-built pressure vessel. Specific details on the pressure vessel can be found in our previous paper.24 A mechanical testing apparatus (5582, Instron) was used to examine additional water infiltration within the zeolite framework by compressing the solution to a significantly high pressure beyond the saturation pressure limit of the gravimetric vapor sorption analyzer. A volumetric compression rate of ∼254 mm3/min (corresponding to a displacement rate of 2 mm/min) was applied, and the displacement and load were recorded. The compression rate was varied to ensure that transient effects did not affect the behavior. The vessel was compressed to a load of 20 kN corresponding to a pressure of ∼150 MPa, and the displacement data was corrected for water compressibility at this high pressure. The water capacity was calculated by equating the displaced volume to an equivalent number of water molecules entering per unit cell of the zeolites.
Figure 2. Combined adsorption and infiltration isotherms for varying Si/Al ratio MFI zeolites. Symbols indicate recorded data while dashed lines are provided to guide the eye. The pressure, P, is normalized by the saturation pressure at 298 K (3.14 kPa), Po. At lower relative pressures, the water uptake increased as the Si/Al ratio decreased. None of the zeolites were filled at the saturation pressure, and all zeolites studied exhibited some amount of high-pressure infiltration. The error associated with the water uptake is ∼0.5 N/UC, while the error associated with the pressure infiltration is ∼±2 N/UC.
3. RESULTS AND DISCUSSION We first confirmed the morphology and crystallinity of the zeolite samples using SEM and XRD, respectively. The SEM images for the five synthesized zeolite samples are shown in Figure 1A. The typical prismatic morphology associated with MFI-type zeolites synthesized using TPA ions as the structuredirecting agent was observed for all zeolites.39 In addition, the size of the crystal was controlled such that each zeolite crystal had an internal to external surface area ratio of at least 1000 so that the role of the textural porosity (external surface) on the water sorption analysis could be neglected.24 To make a fair comparison of the sorption behavior between the different zeolites, a significant effort was made to approximately maintain the same crystal volume (±2.5 μm3) among all zeolite samples.31 The XRD patterns of the calcined samples are shown in Figure 1B. All of the samples had well-defined peaks that correspond to the known MFI zeolite diffraction pattern. Furthermore, the micropore volume for all zeolites quantified via nitrogen sorption was ∼0.18 cm3/g (typical for MFI zeolites), which demonstrated that the inclusion of the defects did not alter the internal pore structure of the zeolites. Details regarding this characterization are presented in the Supporting Information (S.5). 3.1. Sorption and Infiltration Behavior. Using the combined sorption and high-pressure infiltration experimental approach introduced in our previous work,24 we investigated and quantified the effect of the defect density on the pressure at which water entered the zeolite pores (Figure 2). The uptake behavior (recorded up to a relative pressure of 0.98 at 298 K) in Figure 2 showed an increase in total water uptake as the Si/Al ratio decreased. As seen in Table 1, under near saturation conditions, the purely siliceous MFI INF zeolite adsorbed only ∼4 water molecules per unit cell (N/UC), while the most defective MFI 100 zeolite adsorbed ∼27 N/UC, indicating an ∼7-fold increase in sorption capacity. It should be noted that since the overall capacity of the zeolites is ∼35 N/UC,18,20,24 none of these zeolites were completely filled with water at the saturation pressure. The low-pressure adsorption behavior
Table 1. Experimental Results from the Sorption/Infiltration Analysisa zeolite MFI 100 MFI 200 MFI 300 MFI 1000 MFI INF
adsorption @ P/Po = 0.98 (N/UC)
total infiltration @ P = 150 MPa (N/UC)
estimated defect density (N/UC)
26.9
36.5
1.33
18.5
36.6
1.14
14.9
35.4
0.97
6.9
34.3
0.42
3.8
33.5
0.1
a
The defect density was estimated by the low-pressure linear extrapolation method previously used by Olson et al.31 The method is described in detail in the Supporting Information (S.4).
provided information to better estimate the internal defect density of the zeolites, since the compositional Si/Al ratio does not necessarily reflect the actual defect density due to synthesis uncertainties.31 The defect density was estimated via a low relative pressure linear extrapolation of the adsorption quantity (N/UC in Figure 2) to zero relative pressure. This adsorption quantity at “zero pressure” corresponds to water that is specifically adsorbed to the defect sites, thereby giving a better approximation of the defect density than the composition alone.24,31 The estimated defect density is provided in Table 1, while further details of the procedure used to quantify the defect density are provided in the Supporting Information (S.4). Following the sorption experiments, the zeolites were immersed in water and placed within the pressure vessel for high-pressure testing. The total adsorption amount at a relative pressure of 0.98 for each zeolite was used as the starting point for the high-pressure experiments.53 For all zeolites, water C
dx.doi.org/10.1021/la500939t | Langmuir XXXX, XXX, XXX−XXX
Langmuir
Article
Figure 3. Comparison of the experimental data with Cailliez’s weaker MFI zeolite defect model20 (A) and stronger defect model20 (B). The solid symbols correspond with the experimental data from this work while the solid lines represent the modeling results. Note that the vapor to liquid phase transformation (saturation pressure) occurs at a chemical potential of −0.1 kJ/mol and is indicated by the black dashed line in each plot. Magnified plots highlighting the vapor to liquid transition region are provided in the Supporting Information (S.2). The stronger defect model showed better agreement with the experimental data since increasing the defect density not only decreased the infiltration pressure but also substantially increased the amount of adsorbed water prior to the saturation pressure. However, the strength of the modeled defects was underestimated as the experimental defects contain an aluminum cation which provides an additional attractive force compared to the modeled silanol groups.
was stronger than previously estimated values.44 Additionally, using this larger prescribed partial charge, it was predicted that the full infiltration isotherm (i.e., an isotherm from 0 Pa to hundreds of MPa) would shift from type V to type IV45 as the defect density increased.20 For an appropriate comparison with the experiments reported in this work, the relative pressure (abscissa in Figure 2) was converted to a chemical potential using the NIST properties database for water46 (with the assumption that the temperature remained at 298 K throughout the experiments47). Both the strong (+0.65 au) and weak (+0.35 au) defect models were compared to our experimental results and are shown in Figure 3A and B, respectively. The symbols in both figures represent our experimental data while the lines correspond to the respective defect models. Magnified plots of the vapor to liquid transition region are also provided in the Supporting Information (S.2). For the simulated weak defect model (Figure 3A), nearly all of the pore filling was estimated to occur over a small range of pressure (this behavior is more commonly referred to as pore condensation and is classified by a type V isotherm20,45), with a sharp onset of infiltration with respect to the chemical potential. However, this type of pore filling was not observed from our experimental data, as all of the zeolites studied exhibited a combination of low-pressure adsorption and high-pressure infiltration (which is classified by a type IV isotherm). In contrast, the simulated strong defect model (Figure 3B) showed this mix of adsorption and high-pressure infiltration, particularly for the 1 D/UC (defect per unit cell) and 4 D/UC cases. A complete filling of the pores below the saturation pressure (as shown for the modeled 12 strong defects per unit cell) was not experimentally observed; however, constraints on the crystal morphology and size limited the defect density of this study.57 Nonetheless, even for the strong defect model, the simulated defect density underestimated the strength of the Al-substituted defects investigated in this work, i.e., the experimentally calculated defect density of 1 D/UC exhibited nearly the same behavior as the simulated 4 D/UC. Since the heat treatment processes used should have removed the majority of the silanol functional
saturated the remaining pore network between the pressures of 40 and 120 MPa. This high-pressure infiltration behavior of the hydrophobic MFI pores was first reported by Eroshenko et al.16 and was later confirmed by a number of experimental and simulation-based studies including our recent work.18,20,21,23−25,40−43 The mechanism of water infiltration into the hydrophobic pores (and subsequent pore evacuation upon pressure release) has been attributed to the decrease in both the number of available water−water hydrogen bonds within the subnanometer pore network (from ∼5 to 6 outside the zeolite to ∼2 within the pores) and to a lack of zeolite− water hydrogen bonding sites within the zeolite. As the defect density increased, the sharp onset that is associated with the high-pressure water infiltration significantly decreased, and ultimately, the infiltration followed a linear trend as a function of pressure. It should be noted that for all zeolites studied in this work, the high-pressure infiltration was repeatable56 and a total number of ∼35 ± 2 N/UC entered the zeolite framework up to a pressure of 150 MPa, which corroborates our previous work.24 To better understand the mechanistic effect of an increasing defect density on the infiltration behavior, we compared the experimental data with the defective zeolite modeling work of Cailliez et al.20 Cailliez et al. probed the effect of the prescribed partial charge of silanol defect groups within MFI zeolites on the water infiltration behavior. Although Cailliez et al. investigated a different type of acidic defect group within MFI zeolites (silanol groups rather than the Al-substituted defects in this work), the comparison with the modeling provides qualitative insights into the effect of the quantity of defects on the infiltration pressure and an investigation of the relative strength of the attraction of water to the Al-substituted defects. Cailliez et al. reported that a higher assigned partial charge than is typically prescribed (+0.65 atomic units (au) on the hydrogen atom versus +0.35 au on the hydrogen atom) better matched their own experimental infiltration data, which indicated that the attraction of water to the silanol defects D
dx.doi.org/10.1021/la500939t | Langmuir XXXX, XXX, XXX−XXX
Langmuir
Article
Figure 4. (A) Water diffusivity (black squares, left axis) and solubility (red triangles, right axis) plotted against the estimated internal defect density taken at a relative pressure of 0.4. As the defect density increased, the water diffusivity decreased by up to 2 orders of magnitude while the solubility increased only by ∼7-fold. (B) Estimated average water permeability of MFI zeolites as a function of the defect density at 298 K. The decreasing trend as the defect density increased highlights the detrimental effect of hydrophilic defects on water transport within the MFI structure. The complete data sets of the diffusivity, solubility, and permeability as a function of the relative pressure are shown in the Supporting Information (S.3). Note that 1 Barrer is 3.06 × 10−16 (mol m)(m2 s Pa)−1 at STP.
linear uptake behavior (as a function of t1/2) at all relative pressures investigated. Accordingly, we modeled both the adsorption and desorption behavior as a Fickian diffusion process. The micrographs from the SEM analysis were used to estimate the external surface area (A) and volume (V) of the crystal (Table S1). Since the Fickian transport diffusivity accounts for gradients in concentration, it is more appropriate to utilize the thermodynamically corrected diffusivity (which takes into account gradients in the chemical potential) for transport-related processes. Following Zhang et al., the transport diffusivity was converted to the corrected diffusivity by utilizing eq 2,
groups within the pore structure, the bulk of the defects within the lower (
