Mixed matrix hollow fiber membranes made with modified HSSZ-13 zeolite in polyetherimide polymer matrix for gas separation

Journal of Membrane Science 288 (2007) 195–207

Mixed matrix hollow fiber membranes made with modified HSSZ-13 zeolite in polyetherimide polymer matrix for gas separation Shabbir Husain ∗ , William J. Koros School of Chemical & Biomolecular Engineering, Georgia Institute of Technology, 778 Atlantic Drive, Atlanta, GA 30332-0100, United States Received 4 May 2006; received in revised form 20 October 2006; accepted 9 November 2006 Available online 21 November 2006

Abstract Organic–inorganic hybrid (mixed matrix) asymmetric hollow fiber membranes were spun via a dry jet-wet quench procedure using surface modified inorganic small pore size zeolite incorporated in an Ultem® 1000 polyetherimide matrix. The zeolites were modified via two separate techniques and termed as (1) Ultem® “sized” and (2) Grignard treated. The first technique failed to achieve successful mixed matrix performance, but the second approach gave very attractive results. The Ultem® “sized” zeolites were prepared by treating the zeolites with a silane coupling agent to allow Ultem® polymer chains to be grafted to the surface. Poor adhesion was observed between the bulk polymer and most of the zeolite particles in the final membrane using Ultem® “sized” particles. The post-treated fibers did not display enhanced selectivity over neat polymer with pure gas nitrogen, oxygen, helium, methane and carbon dioxide testing or mixed gas carbon dioxide and methane gas pair. The absence of the mixed matrix effect is hypothesized to be due to the nucleation of hydrophilic solvent and non-solvent around the Ultem® “sized” zeolite particles during phase separation in the quench bath forming a so called sieve-in-a-cage defect. Although such defects have been reported in dense mixed matrix films, they have not yet been investigated in hollow fibers format which are formed via a phase separation process and thus remain prone to the effects of the non-solvent quenching media. To test the nucleation hypothesis, fibers incorporating 10.3 vol% (with respect to polymer) of zeolites modified with a Grignard reagent were spun. These fibers, after post-treatment, showed significant selectivity enhancement of 10% for O2 /N2 , 29% for He/N2 , 17% for CO2 /CH4 pure gases and 25% for mixed gas CO2 /CH4 pairs over neat polymer results. The selectivity properties in these fibers are similar to or exceed the Maxwell model predictions for these hybrid materials. © 2006 Elsevier B.V. All rights reserved. Keywords: Mixed matrix; Hollow fibers; Zeolite surface modification; Grignard reagent; Phase separation

1. Introduction Current use of membranes in separation processes is growing at a slow but steady rate [1]. However, as a separation process for gases, membrane usage has the potential to grow tremendously if selectivity of the membranes can be enhanced, especially in high volume areas of natural gas sweetening and low purity nitrogen production. Existing challenges of low selectivity and permeability in polymer membranes are being addressed with some degree of success with advanced engineering polymers. Yet, even though such new polymers are available is no guarantee that the polymer can be processed into asymmetric hollow fiber

∗

Corresponding author. Tel.: +1 404 385 4717; fax: +1 404 385 2683. E-mail address: [email protected] (S. Husain).

0376-7388/$ – see front matter © 2006 Elsevier B.V. All rights reserved. doi:10.1016/j.memsci.2006.11.016

membranes. This polymer characteristic, ambiguously called “spinnability”, is often undermined by polymer crystallinity under shear flow and lack of capillary stability of the spinning solution (dope) which prevents polymer solutions from being extruded and drawn as required in the formation of hollow fiber membranes [2]. Notwithstanding hollow fiber spinning challenges, selectivity enhancement via new polymers offers slow progress as is evidenced by the marginal increases in selectivity and permeability over the previous decades [1]. At the other end of the spectrum, inorganic and carbon based molecular sieves offer very high selectivities and permeabilities, yet are prohibitively expensive to fashion into working membranes. A promising route to enhanced transport properties thus involves the formation of mixed matrix membranes combining the processibility of spinnable polymers with the excellent transport properties of molecular sieves.

196

S. Husain, W.J. Koros / Journal of Membrane Science 288 (2007) 195–207

1.1. Mixed matrix background

1.2. Zeolites and phase separation kinetics in asymmetric membranes

skin can be formed in the prior step via solvent evaporation in the air gap. The final morphology of the hollow fibers is controlled by the phase separation kinetics. While dope additives are used to control the phase separation kinetics of the membrane; however, their presence could potentially obstruct the formation of successful mixed matrix membranes since zeolites are highly susceptible to contaminants which can block the molecular sieving function of the zeolite [17]. Addressing this factor is critical if any enhancement in selectivity is to be achieved in the membrane. There are two aspects of selecting additives for mixed matrix dopes. The first aspect is that the components of the dope must not interfere or block the molecular sieving attributes of the zeolite. The second aspect is that the additive must not interfere in the adhesion between the zeolite and the polymer. While the first aspect can be dealt with using large or very small components in the dope mixture, the effect on polymer–sieve adhesion is more difficult to predict especially when dealing with the considerable variation in chemical functionality on the surface of the zeolite. Initial research carried out in dense film format has identified many material issues leading to the optimization in the selection of molecular sieves and polymers for use in mixed matrix membranes [18]. These challenges include agglomeration of molecular sieves, defective sieve–polymer interfaces and surface pattern formation in the mixed matrix films [4,10,11,19,20]. Yet transitioning from a dense film to asymmetric hollow fiber morphology introduces considerably more components and parameters into the membrane formation process. These challenges are compounded by the use of phase separation to form asymmetric membranes, thereby necessitating reconsideration of material selection and process parameters used for dense mixed matrix films. This reconsideration derives from the influence of zeolites on phase separation kinetics, and subsequently on the final membrane morphology. Although the effects of phase separation on membrane morphology have been widely studied for polymer-only membranes, the effect of zeolites on phase separation equilibrium and kinetics in mixed matrix fibers has not yet been explored. This remains a critical aspect, as the desired asymmetric morphology is formed via a control of the phase separation kinetics using a number of dope components including non-solvents, viscosity modifiers, inorganic salts and quenching media [21–23].

For high rates of production, the dense separating layer of the membrane must be as thin as possible, yet strong enough to withstand considerable transmembrane pressure differential driving forces. Such an arrangement is ideally achieved with asymmetric hollow fibers which consist of a thin dense layer (skin) and a porous support layer. Such membranes are typically formed in a single step via a dry jet-wet quench spinning process where phase separation within the quenched membrane is initiated as rapidly as possible. The rapid phase separation traps in significant porosity for the support layer of the membrane, while a thin

1.2.1. Zeolite surface characteristics In the spinning dope, the zeolites interact with the remaining dope components via the surface functionalities. The first approximation equates zeolites surfaces with the surface characteristics of silica and alumina materials. As such, the zeolite surface, by comparison to silica, is believed to contain as many as four to five hydroxyl groups per square nanometer of the surface attached to silicon and, if present, aluminum [17]. This estimate of the number of hydroxyl groups on the zeolite used in this work, HSSZ-13, has been validated by the reaction of surface

Drawing considerably from composite literature, the incorporation of zeolites into a polymer matrix was initially viewed as the formation of a composite material where adhesion of the polymer phase with the zeolite was expected to be of extreme importance [3–5]. Incorporation of molecular sieves in polymer matrices (usually rubbery polymers) for gas separations followed two hypotheses. In the first case, the adsorptive properties of molecular sieve could be used to enhance selectivity for a given gas mixture by increasing the sorption of the desired gas component within the mixed matrix membrane. This approach was pioneered by Paul and Kemp [6] and further expanded by Kulprathipanja et al. [7]. Initial success in incorporating molecular sieves was achieved with zeolites in highly flexible rubbers such as polydimethylsiloxane and ethylene–propylene diene rubber (EPDM) [8], and in glassy flexible polymers, such as cellulose acetate [7] and polyvinyl acetate (PVAc) [3]. The flexibility of the rubbery polymers and low glass transition temperature (Tg ) glassy polymers was cited as the primary cause of good polymer–zeolite interaction and the polymer deemed to be flexible enough to form stress free interfaces with the zeolite [3]. Formation of successful mixed matrix membranes was then extended to rigid, higher Tg polymers using a variety of material modification and process parameters, which included the use of silane coupling agents, polymers with pendant acidic groups, high temperature casting and removal of dense films from the casting surface at solution Tg [9–11]. However, few discussions exist in the open literature on asymmetric mixed matrix hollow fiber spinning for gas separations. A recent publication by Jiang et al. [12] describes the formation of zeolite–polymer mixed matrix hollow fiber membranes which unfortunately were highly defective, capable only of Knudsen selectivities for oxygen and nitrogen gas pair. While mention has also been made of ion exchange polymer–molecular sieve composite hollow fiber membranes in the public domain [13], to this date, all of the existing research for successful mixed matrix gas separation hollow fiber membranes has been reported in patents focusing on product development [14–16]. This publication, to the author’s best knowledge, is the first report of successful mixed matrix hollow fibers for gas separation outside of patent literature.

S. Husain, W.J. Koros / Journal of Membrane Science 288 (2007) 195–207

hydroxyls with vanadium oxytrichloride (VOCl3 ) followed by an elemental analysis of vanadium on the surface [17]. With these surface hydroxyl groups, zeolites added to the dope mixture cannot be considered inert filler. The zeolites in the spinning dope affect long term dope stability and phase separation kinetics of the membrane. The stability of the dope refers to the characteristic of the zeolite particles to remain homogeneously suspended in the spinning dope. Such stability, firstly, depends on the rate of settling of the non-colloidal particles, and secondly, on the compatibility of the zeolite surface with the remaining components of the dope. When sub-micron particles are used, dope stability only remains a function of the zeolite surface interactions with the dope components dictating whether zeolite aggregate formation and subsequent aggregate settling is prevented. The zeolites can have favorable, unfavorable or neutral interaction with the dope components. These interactions involve zeolite-solvent/non-solvent and zeolite–polymer interactions. Earlier work by Mahajan and Koros [3], employing Hildebrand solubility parameters to select solvents, though feasible for the limited production of dense films, is difficult to implement in large scale production of hollow fibers where safety concerns limit the choice of solvents and non-solvents. The industrial preference for the non-solvent for the quench bath is water due to safety and environmental reasons. Further, among the limited number of solvents for Ultem® 1000, N-methyl-2-pyrrolidione (NMP) is selected due to its relatively benign nature. Mahajan and Koros [3] suggest the use of solvents that interact less or poorly with the molecular sieve compared to the polymer; thus preventing competition of the solvent molecules with the polymer for the sieve surface. Using correlations from Barton [24], Table 1 outlines the solubility parameters of selected solvents and non-solvents for Ultem® along with the estimated liquid–solid parameters for the same solvents with silica and alumina. Similar values of the solubility parameter indicate higher compatibility between the two components (neglecting polar and hydrogen bonding interactions), while a higher liquid–solid parameter indicates a more favorable interaction between the liquid and the solid. As can be seen in Table 1, solvents for Ultem® have a high affinity for the zeolite surface through a variety of interactions. Further, even though NMP and tetrahydrofuran (THF) have very similar solubility parameters,

197

NMP is a strong solvent for Ultem® , while THF is only a weak swelling agent for the polymer. Likewise, even though the solubility parameter for dichloromethane (DCM) of 20.3 (MJ/m3 )1/2 is considerably different from that calculated for Ultem® of 26.3 (MJ/m3 )1/2 , DCM is a strong solvent for Ultem® . These discrepancies highlight the significant contribution of polarity and acid base type interactions prevalent in the components used in this research. While comparisons to silica and alumina surfaces are made as first approximations for the zeolite surface, the reactivity of zeolites having multiple acid sites may be poorly modeled by silica and/or alumina surfaces. Additionally, the synthesis of sub-micron zeolite particles is still a relatively new field and zeolite surfaces and particle characteristics (including the formation of multi-crystal particles) can vary from batch to batch during synthesis. Of particular concern are multi-crystal particles which contain multiple grain boundaries. These grain boundaries are believed to have significantly higher reactivities because of the high free energy of the defects [25]. The estimation of surface chemistry is further complicated by differences in surface and bulk elemental compositions within a zeolite particle, which is again strongly dependent on the batch processing parameters used in the zeolite synthesis. These factors have made the estimation of the interactions between the zeolite and dope components difficult to quantify and only qualitative assessments can be made currently. This understanding has been one of the drivers for the modification of the zeolite to obtain a uniform characterizable surface. 1.2.2. Zeolite modification Although predicting zeolite-solvent and zeolite–polymer interactions accurately can be difficult, by analogy to organic/inorganic composites, silanes can be used to modify the zeolite surface to improve compatibility with the polymer [9,19]. Due to the reactivity with the silica surface, silanes could be attached to the surface of the zeolites via the surface hydroxyls. Further, silanes with a second reactive end group could be used to bond polymer chains to the zeolite thus promoting adhesion between the zeolite and the bulk polymer phase in the membrane. Often, ␥-aminopropyldimethylethoxy silane (APDMES) has been used since it is believed to be large enough to be unable to enter into and block internal pores of small pore

Table 1 Hildebrand solubility and liquid–solid interactions parameters for silica and alumina for selected solvents and non-solvents of Ultem® Solvent/polymer

Hildebrand solubility parameter, δt (MJ/m3 )1/2

ε0 silica (liquid–solid interaction parameter for silica)

ε0 alumina (liquid–solid interaction parameter for alumina)

Solvent for Ultem®

N-Methyl-2-pyrrolidione Dimethyl sulfoxide N,N-Dimethylformamide Dichloromethane Tetrahydrofuran n-Hexane Methanol Water Ultem® 1000 polyetherimide

22.93 26.4 24.8 20.3 22.5 14.8 29.1 47.9 26.3

0.467 0.616 0.547 0.354 0.449 0.117 0.732 1.541

0.615 0.827 0.729 0.454 0.589 0.118 0.992 2.141

Yes Yes Yes Yes No No No No

198

S. Husain, W.J. Koros / Journal of Membrane Science 288 (2007) 195–207

Fig. 1. Schematic of envisioned Ultem® “sizing” of the zeolite surface using a silane coupling agent (APDMES).

˚ and yet able to form a tether between the zeozeolites (3.8–4 A), lite and the polymer that would not allow non-selective flow of gases between the zeolite and the attached polymer [10]. Earlier work with APDMES indicated that the silane reacted well with zeolite 4A and did not impede the molecular sieving function of the small pore zeolite [10]. Fig. 1 shows a schematic of the envisioned coupling reaction. A second modification of the zeolite surface was performed to minimize zeolite-solvent/non-solvent interaction especially for use of the modified zeolite in asymmetric membranes. The as-received zeolite particles were treated with a vigorous surface modification in an attempt to hydrophobize the surface by replacing the hydroxyl groups with methyl groups. Guidelines for the components and reaction procedure for this process were obtained from earlier work performed on the methylation of silicas [26]. A schematic of the envisioned modification is outlined in Fig. 2. This report outlines two procedures to modify the surface of a particular aluminosilicate zeolite, HSSZ-13, with the objectives of: (1) transitioning the success of dense film work with Ultem® “sized” zeolites to asymmetric hollow fiber geometry and (2) describing a second zeolite modification procedure to address challenges observed in the first objective above.

2. Experimental 2.1. Materials The HSSZ-13 zeolites used in this research were graciously provided by Chevron Energy Technology Company (ETC) (Richmond, CA). Ultem® 1000 polymer was purchased from GE Plastics (Pittsfield, MA) and dried at 120 ◦ C for 12–24 h before use. Anhydrous N-methyl-2pyrrolidione (NMP), tetrahydrofuran (THF), iso-propanol (IPA), methanol (MeOH), hexanes (Hx), anhydrous toluene, trimesoylchloride (TMC), lithium nitrate (LiNO3 ), thionyl chloride (TC) and methyl magnesium bromide (MMB) were purchased from Aldrich (Milwaukee, WI) and were of reagent grade. ␥-Aminopropyldimethylethoxysilane (APDMES) was obtained from Gelest (Morrisville, PA). Diethyltoluenediamine (Ethacure® -100) and Sylgard® 184 were graciously donated by Albemarle Corporation (Baton Rouge, LA) and Dow Chemical (Midland, MI), respectively. All chemicals and polymers were used as received without any further purification. 2.2. Silanation and Ultem® sizing treatment Ten grams of HSSZ-13 zeolites were dried overnight in a vacuum oven at 120 ◦ C. The drying was carried out based on

Fig. 2. Schematic of envisioned hydrophobizing reaction on the zeolite surface.

S. Husain, W.J. Koros / Journal of Membrane Science 288 (2007) 195–207

results from dense film work which indicated that better membranes were obtained with the drying of the zeolites before they were added to the aqueous alcohol solution for silanation [17]. The dried zeolites were added to 200 ml of a 95:5 vol% solution of IPA and deionized (DI) water (18 M; Model: D4521, Barnstead International, Dubuque, IA). The mixture was sonicated at high intensity using a 1000 W (maximum power output) sonication horn (Dukane, Leesburg, VA) to disperse the zeolites. After adding 5.0 ml of fresh APDMES, the mixture was sonicated with the 1000 W sonication horn for a total of 30 min in 10 min installments with 10 min “rest” periods in between to prevent excess heating of the mixture. The maximum temperature of the mixture attained during sonication was 50 ◦ C. The zeolite dispersion in IPA was then left undisturbed overnight to allow larger (greater than 1–2 ␮m) particles to settle out. Thereafter, the supernatant containing the dispersed zeolites was pipetted off. The zeolites were recovered using a high pressure filtration setup (#4280, Pall Gelman, East Hills, NY) and 0.2 ␮m PTFE filters with 160 psig of nitrogen back pressure to aid filtration. The zeolites were washed with a 150 ml aliquot of IPA with sonication used to disperse the zeolites in IPA followed by the above mentioned filtration. The washing, sonication and filtration was repeated twice more. The collected zeolites were dried in a vacuum oven at 140 ◦ C for 12 h. The dried zeolites were dispersed in NMP to form a 10 wt% dispersion using sonication (1000 W sonication horn). The dispersion was heated in an oil bath at 145–150 ◦ C. Dried Ultem® 1000 polymer, enough to form a 0.5 wt% solution in NMP, was added to the dispersion. The mixture was stirred with a Teflon coated stir bar for 4 h in the oil bath until a thick paste was formed due to the evaporation of the NMP. The mixture was then removed from heat and a further 25 ml of NMP added to dissolve the paste which dispersed rapidly in the added solvent. 2.3. Grignard treatment Molecular sieves were dried in a vacuum oven at 150 ◦ C for 24 h to remove adsorbed water on the surface. Under anhydrous conditions in a sealed flask, 8.0 g of the sieves were sonicated at low intensity in 80 ml of anhydrous toluene and 10 ml of thionyl chloride for 4 h in a sonication bath (Model: 1510R-MTH, maximum output 70 W, Branson Ultrasonics Corp., Danbury, CT). Low intensity sonication was used to minimize undesirable side reactions from taking place in the presence of aggressive reagents (SOCl2 and MMB) and wherever possible the

199

lower sonication energy sufficient to disperse the sieve particles was used. The dispersion was allowed to stir overnight with a dry nitrogen sweep at room temperature and then sonicated in the sonication bath for 4 h followed by heating at 110 ◦ C with stirring until the zeolites formed a dry cake. Any remaining volatile solvent/reactant was removed under vacuum. The zeolites were re-dispersed in anhydrous toluene using a 50 W (maximum power output) sonication horn (Vibracell Model VC50, Sonics & Materials Inc., Danbury, CT) for a total sonication time of 30 min. The sonication horn, in direct contact with the mixture, was used as the sonication bath was not able to break up and disperse the zeolite cake in the solvent. To this zeolite dispersion, 20 ml of MMB was gradually added with stirring under anhydrous conditions. Thereafter the dispersion was sonicated in the sonication bath for 3 h. A thin bore needle was used as a vent to prevent pressure build-up within the flask. After sonication, the flask was stirred overnight at room temperature. The excess MMB was quenched by slowly adding IPA while cooled by an ice bath. The zeolites were collected using a high pressure filtration setup and 0.2 ␮m PTFE filter paper. The zeolites were washed with three aliquots of 150 ml of isopropanol followed by DI water until the conductivity of the filtrate was reduced to 40–50 ␮S. The zeolites were then dried at 150 ◦ C for 24 h. 2.4. Mixed matrix dope preparation The modified zeolites (silanated or Grignard treated variants) in NMP were sonicated using the 1000 W horn in 30 s bursts until well dispersed to form a 10 wt% (solids) dispersion. This dispersion was added to a sealed glass reaction vessel and heated to 45 ◦ C. A high torque motor (TalBoys Laboratory Stirrers, Model: 409, Troemner LLC, Thorofare, NJ) with a Teflon impeller was used to stir the mixture. THF (as per the dope composition; see Table 2) was added to the dispersion followed by the polymer solution (26–30 wt% Ultem® ) containing lithium nitrate (LiNO3 ) in NMP. The addition of the polymer solution adds about 10–14 wt% polymer with respect to solvents (NMP and THF) into the dope. The mixture was stirred for 15–30 min, until a uniform mixture was observed. When the dispersion appeared well blended, 27.8 g of dried polymer powder was added to the mixture to increase dope viscosity. The mix was then stirred for 5.5 h at 45 ◦ C to completely dissolve the polymer. The prepared dope was then poured into a syringe pump (Model 100DM, Isco, Lincoln, NE) and allowed to degas overnight.

Table 2 Core and sheath dope compositions Component

Core dope (wt%)

Ultem® “sized” sieve sheath dope (wt%)

Grignard treated sieve sheath dope (wt%)

Ultem® 1000 N-Methyl-2-pyrrolidione Tetrahydrofuran Lithium nitrate HSSZ-13

32.0 55.7–57.7 8–10 2.25 –

29.8 51.4 13.2 1.2 4.4

28.0 49.6 17.6 1.0 3.8

200

S. Husain, W.J. Koros / Journal of Membrane Science 288 (2007) 195–207

Table 3 Dual layer mixed matrix hollow fiber spinning parameters Spinning parameter

Range

Core flow rate (ml/h) Sheath flow rate (ml/h) Bore flow rate (ml/h) Air gap (cm) Draw ratio

150–180 15–18 50–60 1–20 1.7–3.5

The fibers were removed from the drum by cutting cleanly using a sharp blade and placed in DI water for 3 days, with the water being changed everyday. At the end of 3 days, the fibers (approximately 75 g) were solvent exchanged by immersing for 20 min each in three successive aliquots (400 ml) of methanol, followed by 20 min each in three aliquots (400 ml) of hexane. The fibers were removed from the last hexane bath and allowed to dry in the hood for 1 h. The fibers were then dried in a vacuum oven at 75 ◦ C for 2 h.

2.5. Core dope preparation

2.7. Characterization

Dried LiNO3 was dissolved in NMP using sonication. After the salt had dissolved, THF and dried polymer powder were added and the mixture placed on a roll mill until completely dissolved. Heated air (50 ◦ C) was used to aid dissolution of the polymer solution on the roll mill. The core and sheath dopes compositions are listed in Table 2.

2.7.1. Scanning electron microscopy Dried fibers were soaked in hexane for a couple of minutes, gently patted dry and placed in liquid nitrogen for at least 1 min before they were shear fractured using two fine point tweezers. The fractured fibers were sputter coated with a 10–20 nm thick coating of gold (Model P-S1, ISI, Mountain View, CA). Images were captured using a high resolution scanning electron microscope (Leo 1530, Leo Electron Microscopy, Cambridge, UK).

2.6. Fiber spinning The bore, core, and sheath fluids were co-extruded using a dual layer spinneret graciously provided by Medal L.P. (Newport, DE) with various air gaps, extrusion rates and draw ratios in a quench bath of tap water maintained at 25 ◦ C (Table 3). The Isco syringe pumps provide excellent flow stability and are ideal for use in laboratory spinning of hollow fibers. The bore fluid consisted of 10 wt% DI water in NMP. Filters (Swagelock, OH) were attached upstream of the spinneret to trap large particles which could potentially block the spinneret ports. The spinning was carried out at 50 ◦ C by heating the spinneret and filter blocks using multiple heating tapes (BriskHeatTM , Barnstead International, Dubuque, IA) regulated by temperature controllers (Model: CN9111A, Omega Engineering Inc., Stamford, CT). The nascent membrane was extruded through an adjustable air gap into the quench bath, passed under a Teflon guide in the quench bath and collected on a controlled speed rotating drum partially immersed in tap water. Fig. 3 displays a schematic of the spinning apparatus.

2.7.2. Permeation Modules of 30–50 fibers of 20 cm lengths with total membrane surface areas between 50 and 75 cm2 were made. The fibers were post-treated with a reactive post-treatment described by Ekiner and Kulkarni [14]. Transport properties of the fibers were tested with pure gas (oxygen, nitrogen, helium, methane, carbon dioxide) and mixed gas mixtures (methane/carbon dioxide) obtained from the local vendor of air products/airgas. Feed pressures of 114.7 psia were used for the pure gas O2 , N2 , He, CH4 and mixed gas CO2 :CH4 testing. Pure gas CO2 was measured at 23 psia. All permeation testing was performed at 35 ◦ C. Mixed gas testing for methane/carbon dioxide gas pair was also performed to more accurately represent testing under “real” feed conditions. Gas compositions for mixed gas testing were determined by injecting equilibrated gas samples into a gas chromatograph (EG&G Chandler Engineering, Carle Series 100 AGC, model: 72000-00). Stage cut of less than 1% were used.

Fig. 3. Mixed matrix dual layer hollow fiber spinning setup.

S. Husain, W.J. Koros / Journal of Membrane Science 288 (2007) 195–207

201

Details of module making and permeation testing are provided elsewhere [27]. 3. Results and discussion 3.1. Mixed matrix hollow fiber membranes based on Ultem® “sized” HSSZ-13 zeolites Macroscopic views of the hollow fibers made from Ultem® “sized” zeolites show circular and concentric bores. No interface between the sheath and core layers is observable indicating good adhesion and intermixing between the layers. Fig. 4 displays SEM micrograph of a representative fiber. High permeability and low selectivity, though above Knudsen selectivity for the O2 /N2 gas pair, obtained in the permeation data of unpost-treated fibers indicated that defects exist in the skin of the fiber through which non-selective gas flow can take place. The skin thickness has a strong role to play in determining whether Knudsen or above Knudsen selectivity can be obtained for unpost-treated fibers. As the selectivity was above that predicted by Knudsen diffusivity, very few such defects exist presumably due to a thick skin; however, these defects

Fig. 4. SEM microphotograph of dual layer mixed matrix hollow fiber membrane incorporating Ultem® “sized” HSSZ-13 zeolites.

Fig. 5. SEM microphotographs of the skin region showing sieve-in-a-cage morphology in mixed matrix hollow fiber membranes incorporating Ultem® “sized” HSSZ-13 zeolites.

Permeation measurements using feed pressures: pure gas O2 , N2 , He, CH4 (114.7 psia), mixed gas CO2 /CH4 (114.7 psia) and pure gas CO2 (23 psia). Measurements made at 35 ◦ C. Error represents the standard deviation in the measurements. Neat Ultem® dense film data is obtained from Ref. [28]. GPU = 10−6 cm3 (STP)/(cm2 s cmHg). a Dense film and Maxwell model permeance (P/l) calculation for a thickness of 100 nm. b Helium permeability of HSSZ-13 zeolite not currently available.

13.8 ± 0.3 – – 39.6 ± 0.2 – – 11.69 ± 0.07 13 20 36.0 ± 0.2 37.4 42 65.4 ± 0.2 94 N/Ab 3.00 ± 0.01 4.0 5.6 7.26 ± 0.03 7.6 8.4 No PT N/A N/A

158.9 ± 0.4 178.6 N/Ab

12.4 ± 0.1 40.1 ± 0.1 11.3 ± 0.1 39.6 ± 0.7 69.3 ± 0.2 3.26 ± 0.02 7.7 ± 0.1

SH31 fibers (Ultem® sized sieves) SH21 fibers (neat Ultem® ) Neat Ultem® dense filma Maxwell model based on 11.2 vol% SSZ-13a

Reactive PT

170 ± 3

CO2 /CH4 selectivity mixed gas (20:80) (P/ l)CO2 (GPU) CO2 /CH4 selectivity pure gas (P/ l)He (GPU) He/N2 selectivity pure gas (P/ l)O2 (GPU) O2 /N2 selectivity pure gas Post-treatment (PT) Membrane

were enough to reduce selectivity significantly below intrinsic polymer levels. As a similar dope composition without zeolites was earlier spun with a defect-free skin, it is postulated that these defects most likely exist in the region surrounding zeolite particles protruding through the skin region of the hollow fiber. The defects/gaps may be formed due to poor zeolite polymer contact and their effect is magnified due to the presence of a thin skin region. After post-treatment, the fibers attained only intrinsic polymer selectivity demonstrating that although, the skin defects were caulked, any enhancement in selectivity (mixed matrix effect) for O2 /N2 , He/N2 and CO2 /CH4 gas pairs was absent (see Table 4). These results indicate that most of the zeolites were being bypassed by the gas molecules, an observation that is confirmed by the sieve-in-a-cage structures seen in the skin region of the fibers as shown in Fig. 5. Although some particles appear close together, they do not appear to be agglomerates. The data for the post-treated fibers are shown in Table 4; the permeation results of mixed matrix fiber spun with a draw ratio (ratio of take-up velocity to extrusion velocity) (DR) of 3.5 and air gap (AG) of 20 cm are compared to those of earlier spun Ultem® defect-free fibers (DR = 2.8; AG = 20 cm), dense neat polymer film values [28] and Maxwell model predicted values. A film thickness of 100 nanometers (nm) was used to generate permeances for dense film and Maxwell model predictions. Experimental permeance values for fibers lower than predicted film values indicate the presence of a skin thickness greater than 100 nm. A variety of explanations have been considered to explain sieve-in-a-cage morphology seen in mixed matrix dense films. Foremost theories among them suggest that these defects are formed as a result of polymer and sieve surface incompatibility. Although, an attempt was made by Mahajan and Koros [3] to estimate such incompatibilities using Hildebrand solubility parameters by approximating the zeolite surface as a combination of silica and alumina surfaces, they still remain theories that need to be tested in practice with each polymer, zeolite and solvent combination. Recently a second hypothesis has been put forward to account for the formation of sieve-in-a-cage defects seen in mixed matrix dense films made using Ultem® “sized” HSSZ-13 zeolites in Ultem® 1000 matrix [17]. This hypothesis claims that sievein-a-cage defects originate due to stresses that develop in a shrinking film on a constrained surface. Arguments are made that as the solvent evaporates, the film tries to shrink in both the thickness and the plane of the film; however, due to the rigid substrate this shrinkage is constrained in the planar direction resulting in the development of large stresses. If the polymer is not flexible, the polymer chains are unable to dissipate stresses that are built up. For Ultem® polymer with 10 wt% residual solvent these stresses have been calculated to be as high as 170 MPa [17]. These stresses are significantly higher than the expected interfacial strength of the silanated zeolite and the bulk polymer and are believed to result in the debonding of the interface [17]. However, in asymmetric hollow fiber membranes, as the nascent membrane encounters a free surface in the air gap before coagulating in the quench bath, the presence of

(P/ l)CO2 (GPU)

S. Husain, W.J. Koros / Journal of Membrane Science 288 (2007) 195–207

Table 4 Comparison of data for Ultem® “sized” mixed matrix fibers with neat Ultem® fibers, neat Ultem® film and Maxwell model predictions

202

S. Husain, W.J. Koros / Journal of Membrane Science 288 (2007) 195–207

203

Fig. 6. Cartoon depiction of nucleation around a sieve particle. Molecules shown in the insets are hydroxyls attached to the sieve surface, polymer molecules, n-methyl-2-pyrrolidione, tetrahydrofuran and water.

sieve-in-a-cage defects indicates a different mechanism for their formation. It is proposed that these defects in asymmetric hollow fibers are a product of the phase separation phenomena and can be attributed to nucleation initiated around a particle. Even though the silanation procedure followed by the grafting of Ultem® onto the surface of the zeolite has been proven to occur, complete silane coverage of the surface is not observed [17]. Based on thermogravametric analysis (TGA) and VOCl3 titration, Moore estimates a maximum of two silane molecules per square nanometer of the zeolite surface after the silanization reaction [5]. Thus, the silanated zeolite particles still retain a high number

of hydroxyl groups which can act as excellent adsorption sites for polar hydrophilic molecules in the dope mixture, including traces of water. If comparisons drawn with adsorbed water layers on silica surfaces are applicable [29,30], the adsorbed layer on the zeolite surface may be as thick as 10 nm. This favorable interaction with the solvents and non-solvents of the dope over and above the polymer molecules can lead to the formation of a locally phase separated polymer lean phase or a lower polymer concentration region around the zeolite particle. Such regions of local phase separation are believed to lead to agglomeration of the zeolite particles which can be observed by the poor stability of mixed matrix dopes a few days after preparation.

Fig. 7. SEM microphotograph of dual layer hollow fiber membranes incorporating Grignard treated HSSZ-13 zeolites.

204

S. Husain, W.J. Koros / Journal of Membrane Science 288 (2007) 195–207

Although poor stability of the dope is visually observed within a week, an adsorbed layer of hydrophilic molecules around the particle can be envisioned to form or exist during the formation of the dope mixture. The high liquid–solid interaction parameters, shown in Table 1 between water and silica/alumina surfaces, give some indication of the preferential adsorption of water and hydrophilic solvents around the zeolite. When this dope is extruded into the quench bath, non-solvent, usually water, enters the membrane and initiates phase separation. As the phase separation is mass transfer initiated, it is believed to proceed via a nucleation and growth mechanism. The ingressing water molecules are attracted to the surface of the zeolite particle where the particles with pre-absorbed layers of hydrophilic molecules act as natural nuclei. Here, the decrease in free energy from phase separation does not have to contend with an increase in free energy from making new interfaces as the interface already exists. The ingressing water molecules expand this adsorbed layer thereby increasing the non-solvent concentration around the zeolite particle. When this non-solvent concentration reaches a critical value, phase separation occurs around the zeolite, with the polymer lean phase surrounding the zeolite and the polymer rich phase forming a cage around it. Fig. 6 presents a cartoon depiction of the proposed hypothesis. A secondary aspect of the zeolite–polymer adhesion is the effect of extension of the fiber in the air gap on the interphase. These stresses are induced during free fall of the nascent membrane in the air gap or if the fiber is subjected to drawing. When the draw ratio, that is, the ratio of the velocity of the fiber just

before it enters the quench bath to the velocity of the extruded fiber, reaches a critical value, polymer chains may be pulled away from the surface of the zeolite. For this to occur, the tensile yield strength of the polymer–zeolite interphase must be less than the tensile yield strength of the polymer solution. Although the interfacial strength of silane modified surfaces is at most expected to be around 30 MPa in a poly(vinyl butyral) matrix [31], the interfacial strength of Ultem® “sized” HSSZ-13 in Ultem® matrix might be much lower because of the inherent incompatibility between the residual hydrophilic hydroxyl groups and hydrophobic Ultem® . The stresses induced in the dope and nascent membrane also contribute to the phase instability of the dope. During the spinning process, the dope undergoes shear stress within the spinneret. The alignment of the chains within the shear field is believed to lead to a decrease in the entropy of the chains within the dope solution, resulting in a less negative free energy of mixing and thus a decrease in the stability of the dope [2,32]. Further, the alignment of polymer chains is greatly enhanced if the nascent membrane undergoes extensional strain in the air gap [33]. Mechanistically, the imposed stress, both shear and extensional, on the dope solution leads to an increase in the energy that must be stored (via the alignment of the polymer chains) or dissipated as heat energy. Wolf [34] has described the effects of shear on the phase behavior of polymer solutions leading to either increase or decease in the homogeneous envelope based on energy contribution from flow to the Gibbs free energy of

Fig. 8. SEM micrographs of the skin region of fibers made with Grignard treated HSSZ-13 zeolites.

1.70 ± 0.02 5.4 8.2 ± 0.09 8.4 Reactive PT N/A SH38 fibers (Grignard treated zeolites) Maxwell model based on 10.3 vol% HSSZ-13a

Permeation measurements using feed pressures: pure gas O2 , N2 , He, CH4 (114.7 psia), mixed gas CO2 /CH4 (114.7 psia) and pure gas CO2 (23 psia). Measurements made at 35 ◦ C. Error represents the standard deviation in the measurements. GPU = 10−6 cm3 (STP)/(cm2 s cmHg). a Maxwell model permeance (P/l) calculation for a thickness of 100 nm. b Helium permeability of HSSZ-13 zeolite not currently available.

46.9 ± 0.1 – 6.23 ± 0.02 20 43.9 ± 0.3 42 48 ± 0.8 N/Ab

(P/ l)He (GPU) He/N2 selectivity pure gas (P/ l)O2 (GPU) O2 /N2 selectivity pure gas Post-treatment (PT) Membrane

Table 5 Comparison of data for Grignard treated zeolite based mixed matrix fibers with Maxwell model predictions

As hypothesized above, the formation of the sieve-in-a-cage defects around zeolites is believed to be due to the hydrophilic nature of the zeolite surface even after silanation and Ultem® “sizing”. The Grignard treated zeolites were developed and tested with the hypothesis, that the hydrophobized surface would suppress the zeolite particles from acting as nucleating agents. The dual layer fibers incorporating the Grignard treated zeolites did not display an interface between the sheath and core regions indicating good adhesion between the two layers. Circular fibers with concentric bores were observed. Some macrovoids were observed initiating in both the sheath and core region of the fiber as can be seen in SEM micrographs of selected fibers in Fig. 7. These features do not affect the discussion here and hypotheses on the formation of these macrovoids will be presented in a forthcoming publication. The SEM micrographs of the skin region of the fibers show excellent adhesion between the zeolite and the polymer matrix as can be seen in Fig. 8. In the images, both the skin region (dense layer) and the beginning of the porous support can be seen. It is expected and also observed that the bonding between the polymer and Grignard treated particles is poor in the porous region of the asymmetric hollow fiber membrane. As the porous region of the fiber only performs the function of mechanical support, poor zeolite–polymer bonding within the region is not of importance. However, it is essential that good zeolite–polymer bonding occur in the skin region of the fiber to achieve the mixed matrix effect. The fibers incorporating Grignard treated zeolites, spun at DR = 3.5 and AG = 10 cm, show only Knudsen selectivity for the O2 /N2 gas pair, indicating that defects exist in the fibers. The defects in the above fibers can be attributed to the thin skin thickness of the hollow fiber membranes due to a lower air gap (10 cm) when compared to the 20 cm air gaps used for fibers incorporating Ultem® “sized” zeolites. It is expected that with a thicker skin region, several particle diameters thick, defect free membranes would be attained. Unfortunately, a considerably thick skin would severely retard gas flux through the hollow fibers, resulting in loss of productivity. A more feasible option of forming thin but defective skins that can be post-treated with a “caulking” layer was pursued in this work, which does not sacrifice gas flux. A thicker skin, encapsulating mostly sieve-in-

CO2 /CH4 selectivity pure gas

3.2. Mixed matrix hollow fiber membranes based on Grignard treated HSSZ-13 zeolites

231 ± 4 N/Ab

(P/ l)CO2 (GPU)

CO2 /CH4 selectivity mixed gas (20:80)

(P/ l)CO2 (GPU)

the polymer solution. Thus, the shear stress in the spinneret and the extensional stress imposed on the nascent membrane during drawing in the air gap can result in demixing or an increase in the nucleation rate, thereby inducing a more rapid phase separation within the membrane. If the nucleation hypothesis in mixed matrix hollow fiber membranes is valid, successful mixed matrix membranes can be achieved if nucleation at the sieve surface can be suppressed. One such method to test this suggestion could be via the use of zeolites with hydrophobic surfaces which would energetically not support the nucleation of the hydrophilic components of the dope.

205 6.8 ± 0.1 –

S. Husain, W.J. Koros / Journal of Membrane Science 288 (2007) 195–207

206

S. Husain, W.J. Koros / Journal of Membrane Science 288 (2007) 195–207

a-cage defects would display selectivity higher than Knudsen but would not achieve a mixed matrix effect after post-treatment as the sieve-in-a-cage defects can not be “caulked”. After the reactive post-treatment, the fibers showed enhanced selectivity of 10%, 29% and 17% for O2 /N2 , He/N2 , CO2 /CH4 pure gas pairs, respectively, and 25% for mixed gas CO2 /CH4 . The permeation results of the post-treated fibers are shown in Table 5. The results for CO2 /CH4 gas pair exceed the enhancement in selectivity as predicted by the Maxwell model. The discrepancy in the actual results and those predicted by the Maxwell model can be most likely traced to the underestimation of the input transport properties of the zeolite in the model. Further work is being performed to identify the changes induced in the zeolite by the Grignard treatment along with tests to estimate the level of hydrophobicity of the surface after the modification. 4. Conclusions Successful mixed matrix membranes are proposed to form only if good bonding of the zeolite particle can be achieved with the polymer matrix. Generally, zeolites with hydrophilic surfaces do not interact well with hydrophobic polymers used in fiber spinning. This requires that the surface of the zeolite particles be modified to change the level of interaction between polymer and the zeolite. The first method of increasing zeolite–polymer compatibility via the use of silane coupling agents and subsequent polymer “sizing” did not result in a mixed matrix enhancement of selectivity. Sieve-in-a-cage defects were observed in the fibers and identified as the cause for the absence of the mixed matrix effect. The defects are hypothesized to form as a result of nucleation of non-solvent and/or polymer lean phase around the zeolite during the phase separation process. For successful mixed matrix asymmetric hollow fiber membranes it appears necessary that nucleation of solvents and non-solvents at the zeolite surface be restricted. One such approach is via increasing the hydrophobicity of the zeolite surface by capping surface hydroxyls with hydrophobic organic molecules. Hollow fibers incorporating Grignard treated zeolites showed a selectivity enhancement of 10%, 29% and 17% for O2 /N2 , He/N2 and CO2 /CH4 pure gas pairs, respectively, and 25% for mixed gas CO2 /CH4 . The success of the Grignard treated zeolites in the polymer matrix also highlights that coupling of the polymer to the sieve surface is not a prerequisite for successful mixed matrix membrane formation. Acknowledgements The authors would like to acknowledge the financial support of Medal L.P./Advanced Technology Program (ATP), Project No. 00-00-39, and fruitful discussions with Dr. Kulkarni at Medal L.P. Also, Chevron Corp. is acknowledged for the supply of submicron HSSZ-13 zeolites used in this work. References [1] R.W. Baker, Future directions of membrane gas separation technology, Ind. Eng. Chem. Res. 41 (6) (2002) 1393–1411.

[2] A. Ziabicki, Fundamentals of Fibre Formation, John Wiley & Sons, London, 1976, p. 488. [3] R. Mahajan, W.J. Koros, Factors controlling successful formation of mixedmatrix gas separation materials, Ind. Eng. Chem. Res. 39 (8) (2000) 2692–2696. [4] T.T. Moore, R. Mahajan, D.Q. Vu, W.J. Koros, Hybrid membrane mate s, AIChE J. 50 (2) (2004) 311–321. [5] T.T. Moore, Effects of materials, processing, and operating conditions on the morphology and gas transport properties of mixed matrix membranes, 2004, 311 pp. [6] D.R. Paul, D.R. Kemp, Diffusion time lag in polymer membranes containing adsorptive fillers, J. Polym. Sci. Polym. Symp. (41) (1973) 79–93. [7] S. Kulprathipanja, R.W. Neuzil, N.N. Li, Separation of fluids by means of mixed matrix membranes, Cont. of U.S. Ser. No. 697,990, Allied-Signal, Inc., USA, 1988, p. 7. [8] J.M. Duval, B. Folkers, M.H.V. Mulder, G. Desgrandchamps, C.A. Smolders, Adsorbent-filled membranes for gas separation. Part 1. Improvement of the gas separation properties of polymeric membranes by incorporation of microporous adsorbents, J. Membr. Sci. 80 (1993) 189–198. [9] J.M. Duval, A.J.B. Kemperman, B. Folkers, M.H.V. Mulder, G. Desgrandchamps, C.A. Smolders, Preparation of zeolite filled glassy polymer membranes, J. Appl. Polym. Sci. 54 (4) (1994) 409–418. [10] R. Mahajan, W.J. Koros, Mixed matrix membrane materials with glassy polymers. Part 2, Polym. Eng. Sci. 42 (7) (2002) 1432–1441. [11] R. Mahajan, R. Burns, M. Schaefer, W. Koros, Challenges in forming successful mixed matrix membranes with rigid polymeric materials, J. Appl. Polym. Sci. 86 (2002) 881–890. [12] L.Y. Jiang, T.S. Chung, C. Cao, Z. Huang, S. Kulprathipanja, Fundamental understanding of nano-sized zeolite distribution in the formation of the mixed matrix single- and dual-layer asymmetric hollow fiber membranes, J. Membr. Sci. 252 (1/2) (2005) 89–100. [13] R. Kiyono, G.H. Koops, M. Wessling, H. Strathmann, Mixed matrix microporous hollow fibers with ion-exchange functionality, J. Membr. Sci. 231 (1/2) (2004) 109–115. [14] O.M. Ekiner, S.S. Kulkarni, Process for making mixed matrix hollow fiber membranes for gas separation, in U.S. L’Air Liquide Societe Anonyme A Directoire Et Conseil De Surveillance Pour L’etude Et L’exploitation Des Procedes Georges Claude, Fr, 2003, 10 pp. [15] W.J. Koros, D. Wallace, J.D. Wind, S.J. Miller, C. Staudt-Bickel, D.Q. Vu, Crosslinked and crosslinkable hollow fiber mixed matrix membrane and method of making same, Cont-in-part of U.S. Ser. No. 32,315, Chevron U.S.A., Inc., USA; University of Texas System, 2003, 24 pp. [16] S.J. Miller, Purification of p-Xylene using Composite Mixed Matrix Membranes, Chevron U.S.A. Inc./Medal, L.P., USA, 2002, 12 pp. [17] T.T. Moore, Effects of Materials, Processing, and Operating Conditions on the Morphology and Gas Transport Properties of Mixed Matrix Membranes, Department of Chemical Engineering, University of Texas, Austin, TX, USA, 2004, p. 312. [18] C.M. Zimmerman, A. Singh, W.J. Koros, Tailoring mixed matrix composite membranes for gas separations, J. Membr. Sci. 137 (1/2) (1997) 145–154. [19] R. Mahajan, W.J. Koros, Mixed matrix membrane materials with glassy polymers. Part 1, Polym. Eng. Sci. 42 (7) (2002) 1420–1431. [20] C.M. Zimmerman, R. Mahajan, W.J. Koros, Fundamental and practical aspects of mixed matrix gas separation membranes, Polym. Mater. Sci. Eng. 77 (1997) 328–329. [21] S.C. Pesek, Aqueous Quenched Asymmetric Polysulfone Flat Sheet and Hollow Fiber Membranes Prepared by Dry/Wet Phase Separation, Department of Chemical Engineering, University of Texas, Austin, TX, USA, 1993, 260 pp. [22] O.M. Ekiner, G. Vassilatos, Polyaramide hollow fibers for hydrogen/methane separation—spinning and properties, J. Membr. Sci. 53 (3) (1990) 259–273. [23] R.M. Boom, Membrane Formation by Immersion Precipitation: The Role of the Polymeric Additive, University of Twente, The Netherlands, 1992. [24] A.F.M. Barton, CRC Handbook of Solubility Parameters and Other Cohesion Parameters, CRC Press Inc., Boca Raton, FL, 1983. [25] W.D. Callister, Materials Science and Engineering: An Introduction, 4th ed., John Wiley & Sons, 1996, 864 pp.

S. Husain, W.J. Koros / Journal of Membrane Science 288 (2007) 195–207 [26] T. Tao, G.E. Maciel, Reactivities of silicas with organometallic methylating agents, J. Am. Chem. Soc. 122 (13) (2000) 3118–3126. [27] S.B. Carruthers, Integral-skin Formation in Hollow Fiber Membranes for Gas Separations, University of Texas, Austin, TX, USA, 2001, 233 pp. [28] T.A. Barbari, W.J. Koros, D.R. Paul, Polymeric membranes based on bisphenol A for gas separations, J. Membr. Sci. 42 (1/2) (1989) 69–86. [29] P.G. Nutting, The adsorptive force of silica for water, J. Phys. Chem. 31 (1927) 531–534. [30] G.F. Ershova, Z.M. Zorin, N.V. Churaev, Temperature dependence of the thickness of polymolecular adsorption films of water on quartz surfaces, Kolloidnyi Zh. 37 (1) (1975) 208–210.

207

[31] P.H. Harding, J.C. Berg, The adhesion promotion mechanism of organofunctional silanes, J. Appl. Polym. Sci. 67 (6) (1998) 1025–1033. [32] D. Chopra, C. Haynes, S.G. Hatzikiriakos, D. Vlassopoulos, Modeling the shear-induced structural changes in polymeric fluids, J. Non-Newton. Fluid Mechan. 82 (2/3) (1999) 367–385. [33] O.M. Ekiner, G. Vassilatos, Polyaramide hollow fibers for H2 /CH4 separation II. Spinning and properties, J. Membr. Sci. 186 (1) (2001) 71– 84. [34] B.A. Wolf, Thermodynamic theory of flowing polymer solutions and its application to phase separation, Macromolecules 17 (4) (1984) 615– 618.