Structure–Property Relationships of Inorganically Surface-Modified

Article pubs.acs.org/JPCC

Structure−Property Relationships of Inorganically Surface-Modified Zeolite Molecular Sieves for Nanocomposite Membrane Fabrication Megan E. Lydon,† Kinga A. Unocic,‡ Tae-Hyun Bae,§ Christopher W. Jones,*,†,§ and Sankar Nair*,§ †

School of Chemistry and Biochemistry, Georgia Institute of Technology, 901 Atlantic Dr., Atlanta, Georgia 30332-0400, United States ‡ Materials Science & Technology Division, Oak Ridge National Laboratory, 1 Bethel Valley Rd., Oak Ridge, Tennessee 37831-6064, United States § School of Chemical & Biomolecular Engineering, Georgia Institute of Technology, 311 Ferst Dr. NW, Atlanta, Georgia 30332-0100, United States S Supporting Information *

ABSTRACT: A multiscale experimental study of the structural, compositional, and morphological characteristics of aluminosilicate (LTA) and pure-silica (MFI) zeolite materials surface-modified with MgOxHy nanostructures is presented. These characteristics are correlated with the suitability of such materials in the fabrication of LTA/ Matrimid mixed-matrix membranes (MMMs) for CO2/CH4 separations. The four functionalization methods studied in this work produce surface nanostructures that may appear superficially similar under SEM observation but in fact differ considerably in shape, size, surface coverage, surface area/ roughness, degree of attachment to the zeolite surface, and degree of zeolite pore blocking. The evaluation of these characteristics by a combination of TEM, HRTEM, N2 physisorption, multiscale compositional analysis (XPS, EDX, and ICP-AES elemental analysis), and diffraction (ED and XRD) allows improved understanding of the origin of disparate gas permeation properties observed in MMMs made with four types of surface-modified zeolite LTA materials, as well as a rational selection of the method expected to result in the best enhancement of the desired properties (in the present case, CO2/CH4 selectivity increase without sacrificing permeability). A method based on ion exchange of the LTA with Mg2+, followed by base-induced precipitation and growth of MgOxHy nanostructures, deemed “ion exchange functionalization” here, offers modified particles with the best overall characteristics resulting in the most effective MMMs. LTA/ Matrimid MMMs containing ion exchange functionalized particles had a considerably higher CO2/CH4 selectivity (∼40) than could be obtained with the other functionalization techniques (∼30), while maintaining a CO2 permeability of ∼10 barrers. A parallel study on pure silica MFI surface nanostructures is also presented to compare and contrast with the zeolite LTA case.

■

supercapacitor devices,8 composite building materials,9,10 and photovoltaic devices.11 Previous work on MMMs has shown that void spaces often occur at the interface between inorganic (e.g., zeolite) particles and the polymer, wherein the polymer adheres poorly to the zeolite and a gap forms between the two phases.12,13 These voids create fast, nonselective gas permeation paths that bypass the zeolite. Several strategies for controlling the interface properties have been reported, such as surface-initiated polymerization,14,15 modification of the inorganic particle surfaces via silane-based covalent attachment to the polymer,2,16 in situ polymerization of the polymer on the inorganic particle,14,15 or the deposition of roughened inorganic nanostructures on the particle surface.5,7,17 The use of inorganic nanostructures to roughen

INTRODUCTION Membranes are an important emerging technology for gas separations because they offer lower energy consumption and capital costs compared to current, thermally driven methods such as distillation and adsorption.1 Although polymeric membranes have formed the mainstay of industrial membrane applications, inorganic and polymer/inorganic composite membranes are emerging as potentially higher-performance alternatives.1 Composite membranes containing a polymer bulk phase and a dispersed inorganic selective phase, also known as mixed matrix membranes (MMMs), allow the possibility of incorporating the selectivity of inorganic molecular sieving materials while maintaining the processability and low cost of polymer membranes.1−7 A critical design issue in the fabrication of MMMs is control over the polymer/inorganic interfaces, which often determines the membrane performance.2 Control over the interfacial properties of such composite systems may also find applications in other areas such as © 2012 American Chemical Society

Received: February 14, 2012 Revised: April 10, 2012 Published: April 11, 2012 9636

dx.doi.org/10.1021/jp301497d | J. Phys. Chem. C 2012, 116, 9636−9645

The Journal of Physical Chemistry C

Article

Figure 1. Synthesis of nanostructures by (A) Grignard, (B) solvothermal, (C) modified solvothermal, and (D) ion exchange functionalization methods.

the nanowhiskers on the zeolite surface. The surface nanostructures formed were previously suggested to be Mg(OH)2 when synthesized in the absence of zeolite.24 Membrane permeation data showed that the solvothermal methods lead to the formation of membranes with better gas separation properties than MMMs made without surface modification of the zeolite.6 In the ion exchange method, the extra-framework cations, Na+, are first exchanged with Mg2+ from a salt solution at a neutral pH. Subsequently, the zeolite is hydrothermally reacted with a source of Na+ ions at a slightly basic pH (∼9.5). During the hydrothermal treatment, reverse ion exchange occurs. Because of the low solubility products of Mg(OH)2 and related MgOxHy materials, the Mg2+ ions exiting the zeolite react immediately with the basic solution to form nanostructures at the zeolite surface.19 The previous work on surface-roughened zeolites has been mainly devoted to developing a range of surface nanostructuring and roughening methods and their use in the fabrication of MMMs for improved gas permeation properties. However, there is little knowledge of the microscopic structural properties of these surface nanostructures and their correlation to the properties of the membranes obtained using particles modified by each of the four methods described above. The purpose of the present investigation is to probe the structural properties of the nanowhiskers produced by each method and correlate these properties with its suitability for use in composite membranes. We focus on the surface modification of the aluminosilicate zeolite LTA and the pure-silica zeolite MFI, both of which have been used in the fabrication of MMMs for separation applications.4,6,7,13,25 In particular, we apply a range of physicochemical characterization techniques to study

molecular sieve particle surfaces has recently allowed enhancements in the properties of MMMs in gas separations.5,6,18 In particular, MgOxHy (1 ≤ x ≤ 2, 0 ≤ y ≤ 2) nanostructures have been grown on the surface of zeolites such as pure-silica MFI and aluminosilicate LTA through four techniques, namely Grignard decomposition reactions,17,18 solvothermal and modified solvothermal depositions,6,7 and ion exchange induced surface crystallization.19 The roughened surfaces provide a high surface area for noncovalent interactions and also allow for entanglement of the polymer chains leading to enhanced adhesion.9,10,17,20 Figure 1 summarizes the four current surface-modification techniques. In the original Grignard modification route, a Grignard (alkylmagnesium bromide) reagent is hydrolyzed in the presence of the zeolite material.17,18,21 There are currently two variants of the Grignard method that use different pretreatment procedures including the use of thionyl chloride for dealumination of LTA surfaces18,21 or the use of NaCl to seed zeolite surfaces.18,22 Additionally, a sol−gel variation of the Grignard method has recently been published.23 In previous work using the thionyl chloride dealumination route, the resulting nanostructures on the zeolite surface were shown by X-ray diffraction (XRD) to be composed of Mg(OH)2, and improved CO2/CH4 gas selectivity over unmodified composite and polymer membranes was observed.17,18 More recently, solvothermal methods were developed to produce nanostructures comparable in morphology to those produced by the Grignard reaction on both pure-silica MFI and aluminosilicate LTA zeolite surfaces via more benign chemistry amenable to scale-up.6,7 These methods depend on the use of a basic organic solvent complex with water and Mg2+ ions to nucleate and grow 9637

dx.doi.org/10.1021/jp301497d | J. Phys. Chem. C 2012, 116, 9636−9645

The Journal of Physical Chemistry C

Article

transferred to the reaction vessel by needle transfer from the secondary container. The Grignard reagent was added dropwise with stirring. The reaction mixture was sonicated for 4 h and stirred for another 12 h. The reaction was placed under N2 flow with stirring and an ice bath to dissipate extra heat. The reaction was quenched by slowly adding isopropanol dropwise and then to excess. Particles were washed and centrifuged twice with isopropanol and twice with 40 mL of DI H2O. Solvothermal Method. This method was carried out as described in previous work.6,7 All zeolites were washed with isopropanol prior to functionalization. A 23 mL Teflon linear, stir bar, and the isopropanol washed zeolite particles were dried at 80 °C before use. First, 0.2 g of zeolite was added to 10 mL of ethylenediamine (EDA) in a Teflon linear and sonicated for 30 s with a sonication horn. Then, 1 mL of a 1 M MgSO4 aqueous solution was added to the EDA mixture dropwise with stirring. The solution was stirred at room temperature for 1 h, autoclaved at 160 °C for 12 h with rotation, and washed with isopropanol and DI water. Modified Solvothermal Method. This method was carried out as described in previous work6,7 and differs from the solvothermal method in the substitution of the solvent. To prevent infiltration and trapping of EDA in the pores of zeolite LTA, a larger amine solvent molecule was used. Zeolites in this method were also washed prior to use in isopropanol. The washed particles and a 23 mL Teflon linear were dried prior to use in an 80 °C oven. First, 0.2 g of zeolite, 0.124 g of MgSO4, and 10 mL of diethylenetriamine (DETA) were combined in a linear and sonicated with a sonication horn for 30 s. Next, 1 mL of DI H2O was added dropwise to the solution and left at room temperature for 1 h. The solution was autoclaved at 180 °C for 12 h with rotation, and particles were washed with isopropanol and DI water. Ion Exchange Method. Before functionalization, as synthesized LTA (Na-LTA) was ion exchanged to Mg-LTA by dispersing 1 g of LTA in a solution of 1.0165 g of MgCl2 in 50 mL of H2O. The mixture was stirred for 1 h at room temperature, filtered with a 0.2 μm filter, and dried overnight. A NaNO3 solution was made by combining 0.851 g of NaNO3 in 100 mL of H2O, and the pH was adjusted to 9.55. 30 mL of the NaNO3 solution and 0.3 g of Mg-LTA are combined in a 45 mL Teflon linear, sonicated with sonication horn for 30 s, and autoclaved at 160 °C for 12 h with rotation. The particles were washed with water afterward. Membrane Fabrication. Bare or functionalized LTA and Matrimid were dried in an oven prior to use. First, 0.03 g of dried zeolite was dispersed in chloroform using a sonication bath. A dilute 10% w/w solution of 0.03 g of Matrimid in chloroform was added to the dispersed zeolite particles to introduce a thin layer of polymer around the particles, preventing agglomeration of the particles. The remaining polymer, 0.17 g, was added to the dilute solution with stirring. The solution was stirred slowly overnight, and chloroform was evaporated to produce a viscous but flowing polymer solution. The membranes were cast in a nitrogen−chloroform filled glovebag using a doctor’s knife. The membranes were kept in the bag overnight to dry. Characterization. Powder X-ray diffraction (PXRD) was carried out with a PAnalytical X’Pert PRO diffractometer operating with Cu Kα radiation and an X’celerator RTMS detector. A step size of 0.002° 2θ and a scan rate of 10 s per step were used. The morphology and composition of bare and surface-modified zeolites were initially characterized using a

the surface morphology, crystallinity, surface area, and micropore volume of the modified zeolite particles. All these factors are expected to strongly affect the control of polymer/zeolite adhesion properties. The new insights obtained also enable a rational selection and application of the appropriate surfacemodification technique. We illustrate this point in terms of structural and permeation properties of the corresponding MMMs.

■

EXPERIMENTAL SECTION Materials. The following chemicals were commercially available and were used as received: deionized water (DI water), sodium hydroxide (NaOH, 98%, Sigma-Aldrich), aluminum isopropoxide (Al(OiPr)3, 97%, Sigma-Aldrich), tetramethylammonium hydroxide (TMAOH, 25% w/w aqueous solution, Alfa Aesar), colloidal silica (Ludox HS-30, 30 wt % aqueous solution, Sigma-Aldrich), tetraethyl orthosilicate (TEOS, 98% Sigma-Aldrich), tetrapropylammonium hydroxide (TPAOH, 40% w/w aqueous solution, Alfa Aesar), magnesium chloride hexahydrate (MgCl2, 99−102.0%, Sigma-Aldrich), sodium nitrate (NaNO3, 99.0%, Sigma-Aldrich), sodium chloride (NaCl, Fisher Scientific), methylmagnesium bromide (Grignard reagent, 3.0 M in diethyl ether, Sigma-Aldrich), isopropanol (99.50%, BDH), magnesium sulfate heptahydrate (MgSO4, Acros), diethylenetriamine (DETA, 99%, SigmaAldrich), ethylenediamine (EDA, 99%, Sigma-Aldrich), calcium chloride dihydrate (CaCl2, 99%, Sigma-Aldrich), toluene (99.80%, Sigma-Aldrich), chloroform (99.8%, Sigma-Aldrich), and Matrimid (Ciba Specialty Chemicals). Synthesis of Zeolite LTA and MFI Particles. Zeolite LTA was prepared as described in previous work.6,7,26 First, 1.080 g of NaOH was combined with 246.86 g of DI H2O under stirring. A solution of molar composition 1 Na2O:10 SiO2:5 Al2O3:20 TMA2O:170 H2O was made by adding 28.435 g of aluminum isoproproxide, 196.62 tetramethylammonium hydroxide, and 28.43 g of Ludox silica solution (HS-30) in order with stirring. The solution was stirred 4 h and then heated in an autoclave at 100 °C for 24 h with rotation. The resulting zeolite LTA particles were washed with water, sonicated, and centrifuged four times to wash the particles. Zeolite MFI was also prepared as described in previous work.27,28 First, 25 g of TEOS was added dropwise into 21.53 g of TPAOH. This solution was stirred for 1 h until the solution turned clear. Next, 368.12 g of deionized water was added to the solution and stirred for an additional 24 h. The solution was then heated in an autoclave at 150 °C for 48 h with stirring. The zeolite was washed in the same manner as LTA. Surface Modification Techniques. Grignard Method. The Grignard reaction used in this work was completed using the NaCl seeding pretreatment method.22 This method produces nanostructures of a different morphology from those produced by the thionyl chloride method18,21 but does not change the surface properties of the zeolite. First, 0.805 g of zeolite LTA was added into a solution of 14.11 g of NaCl in 80.5 mL of DI H2O to “seed” the zeolite surface with NaCl. The zeolites were stirred for 2 h, filtered, and dried overnight. All glassware, stir bar, cannula needle, and the zeolite particles were dried at 80 °C before the reaction to eliminate moisture. Next, 0.75 g of NaCl seeded zeolite and 12 mL of toluene were added to round-bottom flask with stir bar. The stoppered flask was sonicated for 5 min to disperse particles and then purged with nitrogen. The Grignard reagent was transferred by cannula into a (sealed) secondary container, and 2 mL of reagent was 9638

dx.doi.org/10.1021/jp301497d | J. Phys. Chem. C 2012, 116, 9636−9645

The Journal of Physical Chemistry C

Article

Hitachi HF-2000 field-emission TEM in bright field mode at 200 keV equipped with a Thermo Scientific energy-dispersive spectroscopy (EDS) system. The resolution of the EDS data was determined by the beam focusing and generally included 3−10 particles. Each measurement was repeated on a second set of particles to confirm the composition. Noran System Six software was used to analyze the data. Samples were prepared by dispersing the particles in isopropanol, sonicating the dispersion, and dropping on a TEM grid. High-resolution images of the zeolite structure were recorded at Oak Ridge National Laboratory using a Hitachi HF-3300 TEM-STEM at 300 keV in TEM mode. The samples were embedded in epoxy, microtomed into 50−75 nm slices, and carbon coated to prevent charging. Elemental analysis by ICP-AES was carried out by Columbia Analytics (Tucson, AZ). X-ray photoelectron spectroscopy (XPS) was performed on a Thermo K-Alpha XPS instrument with Al Kα irradiation via irradiation of powder samples by a flood gun under vacuum. Micropore volume and surface area were determined by nitrogen physisorption using the t-plot and Brunauer−Emmett−Teller (BET) methods, respectively. Physisorption measurements were recorded using a Micromeritics TriStar II 3020. Samples were calcium exchanged prior to measurement (as explained later in this report) by combining 0.15 g of functionalized zeolite in 5 mL of a 1 M CaCl2 solution and vigorously stirring the mixture while it was heated in a 60 °C oven for 5 h to form Ca-LTA. Ca- LTA samples were filtered, dried, and degassed under vacuum at 120 °C for 2 h before ramping up and holding at 200 °C for 8 h. MFI particles were used as-prepared after being degassed using the same method as above. Mixed matrix membranes were annealed at 230 °C for 16 h in a vacuum oven prior to use and single-gas permeation properties were measured with an upstream pressure of 65 psia at 35 °C using a permeation apparatus described previously.29−31 A full description of the gas permeation measurements is included in the Supporting Information.

Figure 2. X-ray diffraction patterns of (A) unmodified 300 nm LTA particles and after modification by (B) Grignard, (C) solvothermal, (D) modified solvothermal, and (E) ion exchange methods. The reference peak positions of LTA are from ref 33. The arrow indicates the XRD peak corresponding to crystalline MgOxHy structures such as Mg(OH)2 or MgO.

■

RESULTS AND DISCUSSION Surface Nanostructures on Zeolite LTA. Powder XRD patterns of LTA zeolite particles (300 nm) in unmodified and modified forms are shown in Figure 2. All particles maintained the LTA structure after the four surface-modification treatments. A small peak broadening at ∼38° 2θ in the solvothermally modified LTA suggests a secondary crystalline phase. This peak suggests the presence of crystalline MgOxHy structures such as Mg(OH)2 or MgO, which both have an XRD peak in this region.32 The zeolite LTA treated by the modified solvothermal (Figure 2D), Grignard (Figure 2B), and ion exchange methods (Figure 2A) do not show additional crystalline phases observable by XRD. Crystalline nanostructures may still be present for these methods but are not prominent because they occupy only a small volume fraction of the sample.6,18 Lastly, the ratio between the [200] and [220] peak at 7° and 10° 2θ, respectively, varies due to the hydration of the zeolite with a ratio of 1 representing the dehydrated structure and higher values for the hydrated structure.33 TEM imaging of the modified particles shows the distinctive nanostructure morphology and approximate coverage of the surface structures produced by each functionalization method (Figure 3). All particles shown in the figure are representative of the samples for each method. The LTA particles modified by the NaCl-seeded Grignard (Figure 3A) and solvothermal (Figure 3B) methods show loosely attached surface structures

Figure 3. Low-magnification TEM images of 300 nm LTA particles modified by (A) Grignard, (B) solvothermal, (C) modified solvothermal, and (D) ion exchange functionalization methods.

with sparse and uneven coverage. The Grignard method produces straight, nanorod-shaped nanostructures, and the solvothermal method produces irregularly shaped nanostructures and a significant amount of unattached materials, thereby creating a mixture with a separate, secondary crystalline phase (Supporting Information Figure S1). In contrast, the modified solvothermal and ion exchange methods form nanostructures with a much more even coverage on the surface of the LTA particles. The modified solvothermal method produces a sheetlike material encapsulating the zeolite particle, whereas the ion exchange method produces finer nanostructures that 9639

dx.doi.org/10.1021/jp301497d | J. Phys. Chem. C 2012, 116, 9636−9645

The Journal of Physical Chemistry C

Article

quantity of these nanostructures makes them difficult to observe in powder XRD.34 The solvothermal method (Figures 4B and 5B) produces crystalline nanostructures that are less well ordered than those produced by the Grignard method but are observed to be a layered material with 0.72 and 1.9 nm spacing between the layers as determined by the Fourier transform of the lattice fringes (Figure S3). The nanostructures were imaged as quickly as possible to prevent dehydroxylation under the electron beam; however, some beam damage occurred. Grignard and solvothermal modification methods both lead to gaps of several nanometers at the zeolite/metal hydroxide interface. The gap between the nanostructures seen in the TEM images in Figure 5A and B and the zeolite surface strongly suggests that the nanostructures are nucleated in solution near the zeolite surface and subsequently propagate close to the surface of the zeolite, rather than growing directly from the surface. This will likely have an adverse impact on the strength of attachment of the nanostructures to the zeolite surface during subsequent membrane processing steps as well create nonselective paths for molecular diffusion in a membrane and lower its selectivity. The modified solvothermal method forms a relatively thick (10−40 nm) continuous, sheetlike nanostructure (Figures 4C and 5C) that appears mainly amorphous, with small crystalline areas also present. The crystalline areas were not significant enough to be observed clearly in XRD patterns. The ion exchange method produces a fine coverage of surface structures (Figure 4D). No lattice fringes could be detectedeven when imaged immediately to preclude any possibility of beam damagesuggesting that this method produced noncrystalline nanostructures. Both the modified solvothermal and the ion exchange methods exhibited excellent adhesion to the LTA with no gaps at the interface. In fact, nanostructures produced by the ion exchange method penetrated into the LTA (Figure S4). The above observations strongly indicate that these nanostructures were grown directly on the surface, leading to a higher expected strength of attachment between nanostructure and zeolite as well as the need for a substantially smaller quantity of roughening material than in the case of the Grignard and solvothermal methods. The Fourier transforms clearly demonstrated the crystalline nature of the surface materials produced by the solvothermal, modified solvothermal, and Grignard methods. The surface materials produced by the NaCl seeded Grignard and modified solvothermal methods both showed a d-spacing of 0.21 nm that is characteristic of MgO and is not found in Mg(OH)2 (Figure 5A,C and Table S1). The crystalline material formed by the solvothermal method indicated two distinct d-spacings near 0.15 nm that are characteristic of Mg(OH)2 (Figure 5B and Table S1). The HRTEM findings were, therefore, generally consistent with the powder XRD patterns but allow a much more detailed and conclusive understanding of the morphology, structure, and attachment of the surface materials produced by the four techniques. Next, the compositions of the surface nanostructures were studied using three different techniques that probe different length scales: elemental analysis (EA) for bulk analysis, energy dispersive X-ray spectroscopy (EDS) for microscale analysis (based on the micrometer−millimeter spread of the beam), and X-ray photoelectron spectroscopy (XPS) for nanoscale analysis (∼5−40 Å from the surface35). An abridged version of the composition analysis data is shown in Table 1, and the full set of data is presented in Table S2. All three analysis methods showed that the five types of LTA particles (bare and modified)

extend out from the zeolite surface. These results are also confirmed by SEM imaging included in Figure S2. HRTEM imaging was used to observe details of the zeolite− nanostructure interface and to obtain a better characterization of the nanostructure crystallinity. HRTEM images of the LTA− MgOxHy interface after treatment with each of the four surface modification methods are shown in Figures 4 and 5. The NaCl-

Figure 4. Intermediate-resolution transmission electron micrographs of nanostructures produced by (A) Grignard, (B) solvothermal, (C) modified solvothermal, and (D) ion exchange methods. The yellow line demarcates the border between zeolite and nanostructures.

seeded Grignard method (Figures 4A and 5A) produces a layered nanostructure displaying strong lattice fringes and thereby indicating high crystallinity; however, the small

Figure 5. HRTEM images of (A) Grignard, (B) solvothermal, and (C) modified solvothermal functionalized LTA and the Fourier transform of the functionalization highlighted by the white rectangle. The text is the corresponding d-spacing for each observable point. The yellow line demarcates the border between zeolite and nanostructures. 9640

dx.doi.org/10.1021/jp301497d | J. Phys. Chem. C 2012, 116, 9636−9645

The Journal of Physical Chemistry C

Article

Table 1. Value (atom %) of Ca-Exchanged LTA and Modified-LTA Samples Measured by Elemental Analysis (EA), Energy Dispersive X-ray Spectroscopy (EDS) at 200 keV, and X-ray Photoelectron Spectroscopy (XPS) and Normalized to Si Value To Be Compared across Methods (XPS Values are for Na-Exchanged Samples)

Table 2. Surface Area and Micropore Volume of LTA and Modified-LTA Materials, Obtained from Nitrogen Physisorption Isotherms and Expressed per gram of Si sample bare LTA Grignard solvothermal modified solvothermal ion exchange

atom % normalized to Si EA LTA

Grignard

solvothermal

modified solvothermal

ion exchange

Mg O Al Mg O Al Mg O Al Mg O Al Mg O Al

0.9 0.2 0.9 0.7 0.9 0.3 0.9 0.1 0.9

EDS

XPS

0.0 7.8 0.9 0.1 7.9 0.9 0.9 8.8 1.0 0.3 7.3 0.9 0.1 7.0 0.9

0.0 2.3 0.5 1.1 4.1 0.8 3.8 9.4 1.3 1.6 6.4 1.1 0.2 3.5 1.0

BET surface area (m2/g Si)

t-plot external surface area (m2/g Si)

t-plot micropore volume (cm3/g Si)

3610 3260 2230 958

109 102 185 133

1.40 1.26 0.82 0.33

3020

191

1.13

BET surface area is influenced by any blockage of the pores and has also been shown to be inaccurate for microporous materials,39 this analysis method is not representative of the actual surface roughening. The t-plot method is more useful in estimating the external surface area and hence the roughness of the surface.40 Using this method, the external surface area increases in the order of Grignard < bare LTA < modified solvothermal < solvothermal < ion exchange method. Since the ion exchange method produces fine, rodlike nanostructures, there is significantly more surface roughening than when the other methods are used. The solvothermal method creates the second highest external surface area. However, because of the existence of an impurity (byproduct) phase, the increase in surface area cannot be fully attributed to the roughening of the zeolite surface. In the literature, it is hypothesized that higher roughness leads to better adhesion of the particles with polymers in composite membranes,17 and therefore the ion exchange method should show the best adhesion properties. The micropore volumes of the modified LTA materials vary considerably. The solvothermal and modified solvothermal methods cause severe pore blocking effects (Table 2). The dense layers deposited on the surface, as well as the surfacemodification process, may block the diffusion of gases into the pores. The micropore volumes of the materials created by the ion exchange and Grignard methods are somewhat lower than that of bare LTA but do not indicate severe pore blocking. Pore blockage is clearly important in the use of the modified particles for MMM fabrication. Thus, the solvothermal-based methods for modifying LTA are less promising because of the large reduction in micropore accessibility. Surface Nanostructures on Pure-Silica Zeolite MFI. A parallel study was conducted on pure-silica MFI via the same functionalization methods as used for LTA with the exception of the ion exchange method, which cannot be carried out on pure-silica zeolites due to the lack of charge-balancing cations in the pores. X-ray diffraction showed that the MFI structure was also maintained after functionalization (Figure S5). XRD shows no evidence of a secondary crystal phase after any of the three modification processes. TEM investigations (summarized in Figure 7) indicated similar structures produced on MFI by the Grignard method as on aluminosilicate LTA. However, the nanostructures formed by the other two methods show slight differences in their appearance on MFI as compared to LTA. As described earlier, the solvothermal method on LTA generated layered nanostructures that are not attached to the zeolite surface. A somewhat different structure was produced on the surface of MFI (Figure 7B). Since the surface structure and surface potentials of the aluminosilicate LTA and pure-silica MFI are expected to be quite different,41,42 the above observations for the solvothermal method indicate that the

all have similar Al/Si ratios, suggesting that no significant change in the composition of the zeolite (e.g., dealumination) occurs during the surface modification process. The XPS measurements show more variation in Al/Si ratios because of the small integrated peak areas for these two components. The Mg content varies in the order solvothermal > modified solvothermal > Grignard > ion exchange. The XPS data show the same trend for Mg content as the EA and EDS data. Since XPS is a surface characterization technique, the amount of Mg measured is higher than by EA and EDS, thereby confirming the Mg-rich nature of the surface structures. XPS data also can be used to show the relative “material efficiency” of the modification methods and assess the economical use of reagents. The solvothermal method used the most Mg (Table 1) and demonstrated relatively poor nanostructure formation and a secondary phase (Figure 3B and Figure S1). The ion exchange method used the least amount of Mg, produced no magnesium-containing byproducts in the bulk solution, and formed highly adhered nanostructures (Figure 4d and Figure S4) showing that it is a highly efficient functionalization method. To investigate the effect of the surface nanostructures on the accessibility of the zeolite micropores to gas molecules and the surface roughness, nitrogen adsorption isotherms were measured for the bare and surface modified particles. The kinetics of N2 adsorption on Na-LTA were too slow at 77 K to reach equilibrium in a reasonable time scale because of the strong adsorbate−adsorbent polar interactions.36,37 However, Ca-exchanged LTA has been shown to adsorb nitrogen much faster at 77 K, and hence all samples were Ca-exchanged prior to physisorption measurements.26,38 Adsorption data and calculated physical properties were normalized to the Si content (obtained from elemental analysis) and are shown in Table 2 and Figure 6. The adsorption isotherms are of type I, representing LTA as a microporous material. LTA remained microporous after all the functionalization methods. The BET surface area increases in the order of modified solvothermal < solvothermal < ion exchange < Grignard < bare LTA. Since the 9641

dx.doi.org/10.1021/jp301497d | J. Phys. Chem. C 2012, 116, 9636−9645

The Journal of Physical Chemistry C

Article

Figure 6. Nitrogen adsorption isotherms for 300 nm Ca-exchanged (A) LTA, (B) Grignard, (C) solvothermal, (D) modified solvothermal, and (E) ion exchange functionalized LTA; all normalized to the Si content obtained from elemental analysis.

structures similar to those observed on LTA; however, an important difference is the emergence of a larger nanoscopic gap between the surface structures and the zeolite MFI surface as observed by TEM. The substitution of EDA with DETA caused fundamental changes in the nanostructure formation on the zeolite surfaces. Hence, we refer to the two methods separately throughout this work, although they are both carried out by solvothermal reactions. The elemental composition for MFI functionalization (Table S3) showed the same trends as LTA where the solvothermal method has the most Mg present in the sample and the Grignard has the least. Nitrogen physisorption analysis (Table 3 and Figure S7) displayed similar overall trends in the degree Table 3. Surface Area and Micropore Volume of MFI and Modified-MFI Materials, Obtained from Nitrogen Physisorption Isotherms and Expressed per gram of Sia sample

BET surface area (m2/g Si)

t-plot external surface area (m2/g Si)

t-plot micropore volume (cm3/g Si)

bare MFI Grignard solvothermal modified solvothermal

1430 1510 1100 966

306 344 340 447

0.45 0.47 0.31 0.21

Figure 7. TEM images of (A) Grignard, (B) solvothermal, and (C) modified-solvothermal functionalized MFI and (D) HRTEM of Grignard nanostructure on MFI with Fourier transform (inset) of the area in the white rectangle. The corresponding d-spacings are shown for each observable point in the Fourier transform.

a

zeolite surface properties may significantly impact the morphology of nanostructures formed. The different results obtained from the Grignard method after the thionyl chloride pretreatment method,18,21 and the NaCl seeding method may also be explained by the difference in LTA surface characteristics resulting from pretreatment conditions. The modified solvothermal method on MFI (Figure 7C) produced surface

of pore blocking using the various methods on MFI, in comparison to the modified-LTA samples. Both solvothermal methods show less nitrogen adsorption in cm3/g Si in the isotherms (Figure S7) and over 30% reduction in t-plot micropore volumes. The modified solvothermal method

These values can be converted to cm2/g zeolite via a conversion factor of 2.143 g zeolite/g Si.

9642

dx.doi.org/10.1021/jp301497d | J. Phys. Chem. C 2012, 116, 9636−9645

The Journal of Physical Chemistry C

Article

Figure 8. Polymer/zeolite mixed matrix membranes fabricated with Matrimid and LTA particles: (a) bare LTA, and LTA functionalized with the (b) Grignard, (c) solvothermal, (d) modified solvothermal, and (e) ion exchange methods.

created roughened MFI particles with the highest external surface area, as evaluated from the t-plot method. Mixed Matrix Membranes. Mixed matrix membranes (MMMs) were fabricated using bare and functionalized zeolite LTA particles and the glassy polyimide Matrimid as the matrix. The membranes contained ∼15 wt % of zeolite particles. SEM images of the membranes are shown in Figure 8. Poor adhesion between the zeolite and the polymer is seen in the membrane prepared with bare LTA; in particular, one can observe the presence of voids around the zeolite particles that act as nonselective gas transport pathways.13,43 On the other hand, all the membranes prepared by the surface-modified particles appear to show good adhesion between the zeolite and the polymer. The observation of a favorable membrane microstructure in SEM is not sufficient to ensure the desired permeation enhancements, as a number of other characteristics of the modified LTA particles must be taken into account before selection of the appropriate type of surface-modified LTA. The structural and morphological properties of the MgO xHy nanostructures on LTA are summarized in Table 4. Expected unfavorable characteristics for the MMM performance are marked in red. For example, the solvothermal methods exhibited severe pore blocking, a property that is detrimental to their use in separation applications because it blocks the more selective gas permeation path. On the other hand, the ion exchange method allowed the highest surface roughening (and surface area) without nucleation of a secondary phase, close attachment of the nanostructures to the zeolite, and limited pore blocking. The Grignard method showed similarly promising characteristics but leads to a much lower surface roughness. On the basis of the overall findings tabulated in Table 4, particles modified by the ion exchange method should show the best performance in enhancing the permeability and the selectivity in MMMs for gas separations because of its high surface area and open pores. A similar tabulation for MFI (Table S4) reveals a trade-off between the modified solvothermal and Grignard methods in terms of pore blockage and nanostructure attachment to the particle surfaces. CO2 and CH4 permeation measurements were conducted on the membranes imaged in Figure 8, and the results are shown

Table 4. Summary of Properties of Surface Nanostructures Produced by Four Different Surface Modification Methods on LTAa

a

All scale bars represent 100 nm except bare LTA (500 nm).

in Figure 9. Neither the bare LTA MMM nor the MMMs made with solvothermally modified and Grignard-modified LTA showed a consistent trend of improvement of permeation properties from the pure polymer membrane. On the other hand, the MMMs made with ion-exchange-modified LTA demonstrated a substantial CO2/CH4 selectivity increase without loss of CO2 permeability. Thus, the ion exchange modified particles show the greatest potential of all the presently investigated modification methods for use in the fabrication of mixed matrix membranes. This finding rationally follows from the results of Table 4 and clearly illustrates the importance of detailed structural, compositional, and morphological characterization of the modified zeolite materials at multiple length scales to reach reliable conclusions about their performance in MMMs. 9643

dx.doi.org/10.1021/jp301497d | J. Phys. Chem. C 2012, 116, 9636−9645

The Journal of Physical Chemistry C

Article

characteristics of modified MFI particles. This material is available free of charge via the Internet at http://pubs.acs.org.

■

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected] (S.N.); christopher. [email protected] (C.W.J.). Notes

The authors declare no competing financial interest.

ACKNOWLEDGMENTS

■

REFERENCES

This work was supported by King Abdullah University of Science and Technology under Award # KUS-I1-011-21. Microscopy research was supported in part by Oak Ridge National Laboratory’s ShaRE User Facility, which is sponsored by the Office of Basic Energy Sciences, U.S. Department of Energy. We acknowledge the following colleagues: (Georgia Tech) Y. Berta for assistance in microscopy, W. Long and P. Bollini for XPS data collection, J. Vaughn, J. Thompson, and W. J. Koros for assistance with gas permeation measurements; (ORNL): S. K. Reeves for assistance in TEM sample preparation, and L. F. Allard Jr. and D. N. Leonard for useful comments.

Figure 9. Single-gas CO2 and CH4 permeation characteristics of pure Matrimid and of zeolite/Matrimid MMMs made with 15 wt % LTA (either bare or functionalized by the four methods). The error ranges are calculated from measurements on two unique samples.

■

CONCLUSIONS A multiscale experimental study of the structural, compositional, and morphological characteristics of MgOxHy-modified zeolite materials has been presented, and the value of this characterization in assessing the suitability of such materials in the fabrication of mixed-matrix membranes (MMMs) has been shown via an example of CO2 and CH4 gas permeation properties. The four functionalization methods studied in this work produce surface nanostructures that may appear superficially similar under SEM observation but in fact differ considerably in shape, size, surface coverage, surface area/ roughness, degree of attachment to the zeolite surface, and degree of zeolite pore blocking. The evaluation of these characteristics by means of multiple techniques is summarizedboth quantitatively and qualitativelyin Table 4 (for LTA) and Table S4 (for MFI). The interpretation of these results in the context of MMM fabrication allowed a clear understanding of the origin of disparate gas permeation properties observed in MMMs made of the four types of modified zeolite materials as well as a rational selection of the method expected to result in the best enhancement of the desired properties (in the present case, CO2/CH4 selectivity increase without sacrificing the permeability). The ion exchange functionalization of LTA surfaces offered modified particles and MMMs with the best gas permeation and adhesion characteristics. For example, LTA/Matrimid MMMs containing ion exchange functionalized particles had a considerably high CO2/ CH4 selectivity (∼40) than with the other functionalization techniques (selectivity ∼30), while maintaining a CO2 permeability of ∼10 barrers.

■

■

(1) Chung, T. S.; Jiang, L. Y.; Li, Y.; Kulprathipanja, S. Prog. Polym. Sci. 2007, 32, 483−507. (2) Mahajan, R.; Koros, W. J. Polym. Eng. Sci. 2002, 42, 1420−1431. (3) Mahajan, R.; Koros, W. J. Polym. Eng. Sci. 2002, 42, 1432−1441. (4) Adams, R. T.; Lee, J. S.; Bae, T.-H.; Ward, J. K.; Johnson, J. R.; Jones, C. W.; Nair, S.; Koros, W. J. J. Membr. Sci. 2011, 367, 197−203. (5) Hillock, A. M. W.; Miller, S. J.; Koros, W. J. J. Membr. Sci. 2008, 314, 193−199. (6) Bae, T. H.; Liu, J. Q.; Lee, J. S.; Koros, W. J.; Jones, C. W.; Nair, S. J. Am. Chem. Soc. 2009, 131, 14662. (7) Bae, T.-H.; Liu, J. Q.; Thompson, J. A.; Koros, W. J.; Jones, C. W.; Nair, S. Microporous Mesoporous Mater. 2011, 139, 120−129. (8) Ye, J.-S.; Cui, H. F.; Liu, X.; Lim, T. M.; Zhang, W.-D.; Sheu, F.-S. Small 2005, 1, 560. (9) Thostenson, E. T.; Li, W. Z.; Wang, D. Z.; Ren, Z. F.; Chou, T. W. J. Appl. Phys. 2002, 91, 6034. (10) Ehlert, G. J.; Sodano, H. A. ACS Appl. Mater. Interfaces 2009, 1, 1827−33. (11) Lin, Y. Y.; Chen, C. W.; Chu, T. H.; Su, W. F.; Lin, C. C.; Ku, C. H.; Wu, J. J.; Chen, C. H. J. Mater. Chem. 2007, 17, 4571. (12) Mahajan, R.; Koros, W. J. Ind. Eng. Chem. 2000, 39, 2692−2696. (13) Mahajan, R.; Burns, R.; Schaeffer, M.; Koros, W. J. J. Appl. Polym. Sci. 2002, 86, 881−890. (14) Perruchot, C.; Khan, M. A.; Kamitsi, A.; Armes, S. P.; Werne, T.; vonPatten, T. E. Langmuir 2001, 17, 4479−4481. (15) Pyun, J.; Matyjaszewski, K. Chem. Mater. 2001, 13, 3436−3448. (16) Wall, J. S.; Hu, B.; Siddiqui, J. A.; Ottenbrite, R. M. Langmuir 2001, 17, 6027−6029. (17) Shu, S.; Husain, S.; Koros, W. J. J. Phys. Chem. C 2007, 111, 652−657. (18) Shu, S.; Husain, S.; Koros, W. J. Chem. Mater. 2007, 19, 4000− 4006. (19) Bae, T.-H. Engineering Nanoporous Materials for Application in Gas Separation Membranes. Ph.D. Thesis, Georgia Institute of Technology, 2010; pp 88−109. (20) Kowbel, W.; Bruce, C.; Withers, J.; Ransone, P. Composites, Part A 1997, 28, 993−1000. (21) Shu, S.; Husain, S.; Koros, W. J. Ind. Eng. Chem. 2007, 46, 767− 772.

ASSOCIATED CONTENT

S Supporting Information *

SEM images of all particles, HRTEM images and Fourier transform images of all crystalline nanostructures, table of MgO and Mg(OH)2 crystal spacings, detailed data on the elemental composition (EDX, EA, XPS) of all particles, XPS raw data for all particles, XRD patterns and N2 physisorption isotherms for modified MFI particles, and comparative summary table of 9644

dx.doi.org/10.1021/jp301497d | J. Phys. Chem. C 2012, 116, 9636−9645

The Journal of Physical Chemistry C

Article

(22) Shu, S. Engineering the performance of mixed matrix membranes for gas separations. Ph.D. Thesis, Georgia Institute of Technology, 2007. (23) Liu, J. Q.; Bae, T.-H.; Esekhile, O.; Nair, S.; Jones, C. W.; Koros, W. J. J. Sol-Gel Sci. Technol. 2011, 60, 189−197. (24) Li, Y. D.; Sui, M.; Ding, Y.; Zhang, G. H.; Zhuang, J.; Wang, C. Adv. Mater. 2000, 12, 818−821. (25) Huang, Z.; Guan, H.; Tan, W.; Qiao, X.; Kulprathipanja, S. J. Membr. Sci. 2006, 276, 260−271. (26) Breck, D. W.; Eversole, W. G.; Milton, R. M.; Reed, T. B.; Thomas, T. L. J. Am. Chem. Soc. 1956, 78, 5963−5971. (27) Cheng, C. H.; Bae, T. H.; McCool, B. A.; Chance, R. R.; Nair, S.; Jones, C. W. J. Phys. Chem. C 2008, 112, 3543−3551. (28) Schoeman, B. J. Microporous Mesoporous Mater. 1998, 22, 9−22. (29) Moore, T. T.; Damle, S.; Williams, P. J.; Koros, W. J. J. Membr. Sci. 2004, 245, 227−231. (30) Damle, S.; Koros, W. J. Ind. Eng. Chem. 2003, 42, 6389−6395. (31) Pye, D. G.; Hoehn, H. H.; Panar, M. J. Appl. Polym. Sci. 1976, 20, 1921−1931. (32) Ding, Y.; Zhang, G.; Wu, H.; Hai, B.; Wang, L.; Qian, Y. Chem. Mater. 2001, 13, 435−440. (33) Treacy, M. M. J., Higgens, J. B., Eds.; Collection of Simulated XRD Powder Patterns for Zeolites, 4th ed.; Elsevier: New York, 2001. (34) Self, P. G.; Bhadeshia, H. K. D. H.; Stobbs, W. M. Ultramicroscopy. 1981, 6, 29−40. (35) Fadley, C. S. J. Electron Spectrosc. Relat. Phenom. 1974, 5, 725− 754. (36) Du, X.; Wu, E. J. Phys. Chem. Solids 2007, 68, 1692−1699. (37) Breck, D. W. Zeolite Molecular Sieves; John Wiley & Sons Ltd.: New York, 1973; pp 596−604, 641−645. (38) Rakoczy, R.; Traa, Y. Microporous Mesoporous Mater. 2003, 60, 69−78. (39) Rouquerol, J.; Llewellyn, P.; Rouquerol, F. Stud. Surf. Sci. Catal. 2007, 160, 49−56. (40) deBoer, J. H.; Lippens, B. C.; Linsen, B. G.; Broekhoff, J. C. P.; van den Heuvel, A.; Osinga, T. J. J. Colloid Interface Sci. 1966, 21, 405− 414. (41) Davis, M. E. Nature 2002, 417, 813−821. (42) Turner, S.; Sieber, J. R.; Vetter, T. W.; Zeisler, R.; Marlow, A. F.; Moreno-Ramirez, M. G.; Davis, M. E.; Kennedy, G. J.; Borghard, W. G.; Yang, S.; Navrotsky, A.; Toby, B. H.; Kelly, J. F.; Fletcher, R. A.; Windsor, E. S.; Verkouteren, J. R.; Leigh, S. D. Microporous Mesoporous Mater. 2008, 107, 252−267. (43) Moore, T. T.; Koros, W. J. J. Mol. Struct. 2005, 739, 87−98.

9645

dx.doi.org/10.1021/jp301497d | J. Phys. Chem. C 2012, 116, 9636−9645