Report

Interfacial microfluidic processing of metal-organic framework hollow fiber membranes

See allHide authors and affiliations

Science  04 Jul 2014:
Vol. 345, Issue 6192, pp. 72-75
DOI: 10.1126/science.1251181

High-surface-area gas separation membranes

Membranes for gas separation require a combination of high surface area and selective transport pathways. Brown et al. present a potentially scalable route for making high-quality gas separation membranes in a high-surface-area configuration. Using two different solvents flowing in opposite directions, a metal-organic framework material was selectively deposited within hollow polymer fibers. The membranes showed high-performance separation capabilities when tested with mixtures of hydrocarbon gases.

Science, this issue p. 72

Abstract

Molecular sieving metal-organic framework (MOF) membranes have great potential for energy-efficient chemical separations, but a major hurdle is the lack of a scalable and inexpensive membrane fabrication mechanism. We describe a route for processing MOF membranes in polymeric hollow fibers, combining a two-solvent interfacial approach for positional control over membrane formation (at inner and outer surfaces, or in the bulk, of the fibers), a microfluidic approach to replenishment or recycling of reactants, and an in situ module for membrane fabrication and permeation. We fabricated continuous molecular sieving ZIF-8 membranes in single and multiple poly(amide-imide) hollow fibers, with H2/C3H8 and C3H6/C3H8 separation factors as high as 370 and 12, respectively. We also demonstrate positional control of the ZIF-8 films and characterize the contributions of membrane defects and lumen bypass.

Molecular sieving membranes have created interest as high-performance separation systems for production of petro-based and renewable fuels and chemicals. Compared to thermodynamically driven separation methods such as distillation, membrane-based processes can substantially reduce the energy and capital costs of separating molecules on a large scale. Membranes composed of molecular sieving materials such as zeolites (1), layered zeolites (2), or metal-organic frameworks (MOFs) (3) have intrinsic advantages over polymeric membranes, such as a simultaneously high permeability and selectivity. Despite their performance limitations, polymeric membranes have continued to dominate industrial membrane separations owing to their relative ease of processing into morphologies such as hollow fibers (4). One challenge facing molecular sieving membranes is the lack of an easily scalable, reliable, and benign fabrication process (57). Zeolite membranes are further hampered by the need for hydrothermal synthesis on high-cost support materials. MOFs consist of metal centers connected by coordination bonds to organic linker molecules. They have been used to grow crystalline membranes on disk and tubular substrates through techniques similar to those developed for zeolitic membranes (8). The zeolitic imidazolite framework (ZIF) subclass of MOFs is of particular interest for membrane fabrication because of its tunable pore size and chemistry (9) and relatively good thermal and chemical stability (10, 11). In an early demonstration of scalable ZIF membrane processing (12), we synthesized ZIF-90 membranes by seeded growth on the outer surfaces of porous polymeric poly(amide-imide) (Torlon) hollow fibers of ~250-μm outer diameter by immersion in a methanolic precursor solution at mild conditions (65°C). Free-standing MOF films can also be synthesized at the interfaces between two immiscible solvents (13). However, molecular sieving membranes on the inner surfaces of hollow fibers also have advantages for rapid, scalable fabrication due to the ability to be bundled in close proximity while avoiding membrane-membrane contact points and interfaces that lead to defects during synthesis. Synthesis of selective membranes in microscopic confined spaces faces a number of challenges: reactant availability and transport, positional control of the membrane, and scalability. As the bore size (and hence volume) is decreased to microscopic dimensions, film formation becomes limited by reactant availability and local inhomogeneities (14).

We report a methodology for fabricating molecular sieving MOF membranes (specifically, ZIF-8), which we refer to as interfacial microfluidic membrane processing (IMMP) (Fig. 1). IMMP thus combines three key concepts: (i) in situ ZIF-8 film synthesis in the membrane module (Fig. 1A); (ii) a two-solvent interfacial approach (Fig. 1, B and C) that can be tuned to achieve positional control over membrane formation (at inner and outer surfaces, as well as inside the bulk, of the porous fiber); and (iii) the controlled supply, replenishment, and recycling of reactants at microfluidic conditions in the hollow fiber bore. Our approach can be applied more generally to other MOF materials, but we demonstrate our key findings here with the example of ZIF-8, which has been identified as a promising candidate for important separations such as H2 from hydrocarbons and propylene from propane (3, 15). To study the IMMP concept, we designed and fabricated a reusable flow module that serves as both a membrane fabrication reactor as well as a gas permeation module (fig. S1). Torlon hollow fibers with a pore size of ~100 nm and room-temperature N2 permeance of 53,000 gas permeation units (GPUs; 1 GPU = 3.348 × 10−10 mol m−2 s−1 Pa−1), produced in-house and characterized as previously described in detail (12, 16), were used as membrane fabrication supports. A dilute zinc nitrate hexahydrate/1-octanol solution was flowed at a typical rate of 10 μl/hour through the bore of a horizontally mounted Torlon fiber, and a concentrated aqueous 2-methylimidazole (2-mIm) solution was present in the reactor chamber on the shell side (17). Membrane growth was controlled (fig. S2) by the use of continuous bore solution flow conditions, static bore solution conditions, or intermittent conditions (initial continuous bore solution flow followed by static conditions and then flow replenishment of bore reactant). Static growth conditions produced dense, noncontinuous coatings of ZIF-8 particles in the bore of the fiber (fig. S2). This is due to the lack of sufficient Zn2+ ions available in the fiber bore (volume 1.5 μl) to sustain film growth after the initial nucleation and growth of ZIF-8 crystals at the fiber surface. By contrast, a thin membrane (~2 μm, fig. S2) was formed via growth under continuous flow without static growth phases. This is due to the relatively rapid transport of reactants to the interface under continuous flow, leading to rapid formation of a ZIF-8 layer. The growing membrane itself becomes a barrier between the two immiscible solvents and confines the liquid-liquid interface to the gaps and interstices between the ZIF-8 crystals. Crystal growth becomes more favorable in these void regions owing to higher reagent accessibility, leading to a more continuous film (13). When the flow profile was altered to include an initial continuous flow followed by static growth phases interrupted only by short reactant replenishment steps, highly intergrown ZIF-8 membranes of different thicknesses (e.g., ~9 μm under the present conditions) were formed (fig. S2). Figure 2A is a lower-magnification image of a ZIF-8 membrane formed under intermittent conditions. Zinc elemental mapping (Fig. 2B) confirms the localization of the membrane to the inner surface of the fiber. To determine the uniformity of membrane formation, we obtained cross sections of the fiber at different locations along the fiber length and measured the membrane thickness at these locations. Figure 2C depicts the membrane thickness measured along the length of the fiber and shows that a continuous membrane of average thickness 8.8 ± 1.4 μm was formed throughout the fiber length (also see fig. S3).

Fig. 1 Scheme depicting the interfacial microfluidic membrane processing (IMMP) approach for MOF membranes in hollow fibers.

(A) Side view of a series of fibers mounted in the IMMP reactor. (B) The Zn2+ ions are supplied in a 1-octanol solution (light red) flowing through the bore of the fiber, whereas the methylimidazole linkers are supplied on the outer (shell) side of the fiber in an aqueous solution (light blue). (C) Magnified view of fiber support during synthesis. In this example, the membrane forms on the inner surface of the fiber by reaction of the two precursors to form a polycrystalline ZIF-8 layer (dark blue).

Fig. 2 Cross-sectional SEM and EDX characterization of ZIF-8/hollow fiber membrane morphology and thickness.

(A) Cross section SEM showing the ZIF-8 membrane on the inner surface of the hollow fiber. (B) EDX elemental maps of carbon (red) and superimposed zinc (green) showing the localization of the ZIF-8 membrane to the inner surface of the fiber. (C) Membrane thickness measured from cross sections obtained at different locations along the 5-cm length of the hollow fiber, showing the formation of a continuous membrane of essentially uniform thickness. The average membrane thickness along the entire fiber (red dashed line) is 8.8 ± 1.4 μm. The thickness at each location was measured from five randomly chosen points along the circumference of the hollow fiber cross section. Figure S3 shows several examples of the membrane cross sections taken at different distances along the fiber.

We then hypothesized that controlling the precursor and solvent locations would allow control over the membrane position (inner surface, outer surface, or in the bulk of the fiber). Although we have studied only inner-surface membranes in detail here, demonstration of outer-surface and in-fiber membranes is also of interest. For example, zeolite DDR membranes formed inside (rather than on) porous ceramic supports were previously shown to have better CO2/CH4 separation, and this effect was attributed to greater mechanical and thermal stability (18). In separations limited by low permeability, outer-surface membranes may be desired because of the higher surface area. The two ZIF precursors (here, Zn2+ ions and 2-mIm) can be supplied in two different solvents (here, water and 1-octanol) on either the inner (bore) or outer (shell) sides of the hollow fiber. This creates eight different reactor operation possibilities. Table S1 lists five of these cases studied so far, with case 1 shown in Fig. 2, A and B, and cases 2 to 5 shown in fig. S4. X-ray diffraction confirmed the ZIF-8 crystal structure of the films formed in cases 1 to 5 (fig. S5). We found that the primary factor influencing the position of the film is the location of the Zn2+ reactant. The ZIF-8 film forms at (or near) the location of the Zn2+ reactant. The Zn2+ concentration is 0.018 mol/liter in all cases, whereas the 2-mIm concentration is 1.37 mol/liter and hence in large excess (table S1). Assuming that the diffusion rate of reactants to the film surface limits the growth of the membrane (i.e., the reactants are quickly consumed by chemical reaction upon encountering each other), and considering that the diffusivities of the two precursors in the liquid solvents are of the same order of magnitude, the ratio of distances of the membrane from the two bulk solutions will be proportional to the ratio of the bulk concentrations of the two reactants (table S1). This is in agreement with the results of detailed multiscale modeling of film formation in opposing-reactant geometries (19), wherein the ratio of reactant concentrations is identified as the most important parameter. In all five cases, the bore side reactant is first introduced, filling the fiber bore and also infiltrating the porous support, before the shell side reactant is introduced. In cases 1, 4, and 5, the two solvents are immiscible and form a sharp interface (e.g., Fig. 1C), but the precursors are soluble in both solvents and can diffuse between them. In cases 4 and 5, the limiting Zn2+ reactant is introduced on the shell side and hence the membrane forms on the outer surface. In case 1, although Zn2+ ions are initially present throughout the porous fiber and can react with the 2-mIm linker molecules diffusing from the shell side, the successful nucleation and growth of ZIF-8 only occurs when there is a sufficiently replenished supply of Zn2+, i.e., on the bore side. This is again in agreement with the modeling predictions of the role of initial conditions in determining the location of solid nucleation (19). In cases 2 and 3, where the two solvents are identical and Zn2+ is introduced on the bore side, the film again forms at the bore side, but about 10 μm into the porous fiber. The miscibility of the two solvents likely alters the transport rate of the limiting Zn2+ reactant toward the shell side, thus moving the location of ZIF-8 nucleation a short distance into the porous support.

We characterized the hydrogen (H2) and propylene (C3H6) separation properties of the case-1 ZIF-8 membrane formed under intermittent flow conditions by H2/C3H8 and C3H6/C3H8 equimolar mixture permeation as a function of temperature, with the IMMP reactor directly acting as a permeation module (fig. S6). The as-made ZIF-8 membranes exhibited clear molecular sieving properties: high H2 permeances, sharp H2/C3H8 separation factors as high as 130 ± 8 at 120°C, and strong temperature dependence of H2 permeance, indicating activated molecular transport through the ZIF-8 pores (Fig. 3, A and B). However, the C3H8 permeance was much larger than that expected from previous studies (8, 2023) and prevented a high C3H6/C3H8 separation factor (2.2 ± 0.4 at 25°C). This was hypothesized to result from both molecular transport through the ZIF-8 membrane and from bypassing (24) of the ZIF-8 membrane by the feed molecules through the ends of the fiber (fig. S7). To suppress this membrane bypass, we included a lumen-capping step to the IMMP, accomplished by applying a controlled amount of a solution containing poly(dimethylsiloxane) (PDMS) to the ends of the mounted module. The PDMS solution was readily absorbed by capillary action at the fiber ends, blocking the 100-nm pores of the fiber and leading to bypass suppression. Scanning electron micrograph (SEM) imaging and energy-dispersive x-ray (EDX) analysis showed that the porous fiber was infiltrated by PDMS, whereas the bore remained unblocked (fig. S8). PDMS infiltration into the fiber was also profiled with EDX mapping of the Si/Zn ratio along the hollow fiber length (fig. S9), showing that the infiltration depth was 8 mm from the fiber ends. Because the permeances of the gases through PDMS and ZIF-8 (2527) are three orders of magnitude lower than through the bare porous fiber, the C3H8 flux should decrease substantially after capping. The PDMS-sealed fibers showed much higher H2/C3H8 and C3H6/C3H8 separation factors (370 ± 60 at 120°C and 12 ± 3 at 25°C, respectively) (Fig. 3, A and B, and table S2). This is consistent with previously reported ZIF-8 membranes with low defect densities (8, 2023). Notably, the C3H8 permeance decreased by a factor of 10 after capping, showing that most of the earlier observed C3H8 flux was due to bypassing and that the capping step largely shuts down this nonselective path. The permeate streams after capping contain essentially pure H2, or 92% C3H6/8% C3H8, which are substantial upgrades from the equimolar feeds. Longer-term H2/C3H8 and C3H6/C3H8 permeation measurements carried out over a period of 35 days showed the high stability of the ZIF-8 membrane (figs. S10 and S11).

Fig. 3 Gas permeation properties of ZIF-8/hollow fiber membranes fabricated by IMMP in binary H2/C3H8 and C3H6/C3H8 separations with equimolar feed mixtures.

(A) Binary H2/C3H8 and (B) binary C3H6/C3H8 separation characteristics (permeances and separation factors) of single-fiber membranes as a function of temperature, before and after lumen sealing with PDMS. Lumen sealing leads to a large increase in separation factor in both cases. (C) and (D) depict the same performance characteristics obtained from a module of three fibers processed simultaneously with IMMP, showing separation factors similar to those of the single-fiber membranes. The error bars in (A) and (B) were estimated from three independently synthesized membrane samples, and the numerical data for all of the lumen-sealed membranes are shown in table S2.

To reliably quantify the efficacy of the lumen-capping step and the contribution of membrane defects such as grain boundaries or microcracks, we conducted single-component measurements on four gases (C3H6, C3H8, n-C4H10, i-C4H10) at 25°C using as-made and sealed membranes (table S3). The results were fitted to a permeation model (17) that yields the relative molar flow contributions of the ZIF-8 membrane, membrane defects, and bypass for each gas. The PDMS capping step is seen to be highly effective in closing the bypass path (table S4). In the as-made membranes, about 60% of the molar flow of C3H8 occurs through bypass, about 37% through membrane defects, and 3% through ZIF-8. After PDMS capping, the contribution of bypass is essentially eliminated, and about 92% of the small remaining C3H8 flow (~2 GPU of permeance) is through membrane defects. This indicates that for the C3H6/C3H8 separation case, optimization of ZIF-8 IMMP could focus on further reduction of defect permeance by variation of growth conditions or by postsynthesis treatments (28, 29). By contrast, separations such as H2/C3H8 that involve a fast-permeating species are not appreciably affected by membrane defects. IMMP is also inherently a modular and parallel approach that should allow independent and simultaneous processing of membranes in multiple fibers. To test this hypothesis, we applied IMMP to the simultaneous processing of three hollow fibers. The total bore flow rate was increased by a factor of 3 so that the flow rate through individual fibers was maintained. The ends of the module were capped with PDMS, as described earlier. Figure 3, C and D, shows that the H2/C3H8 and C3H6/C3H8 separation behavior is essentially identical to the single-fiber case, demonstrating the potential for scalability of IMMP. Given the overall importance of tunable ZIF materials for a range of hydrocarbon and light-gas separations, the membrane-processing approach reported here overcomes many limitations of current processes and is a notable step toward realizing scalable molecular sieving MOF membranes.

Supplementary Materials

www.sciencemag.org/content/345/6192/72/suppl/DC1

Materials and Methods

Supplementary Text

Figs. S1 to S11

Tables S1 to S4

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: This work was supported by Phillips 66 Company. S.N., A.J.B., and C.W.J. conceived the research. A.J.B. and N.A.B. designed the synthesis reactor. Hollow-fiber fabrication was carried out by J.R.J. and W.J.K. Membrane synthesis, characterization, and permeation measurements were carried out by A.J.B., K.E., and F.R. Permeation modeling was carried out by S.N. and A.J.B. All authors contributed to manuscript writing and editing. We thank W. Qiu, R. P. Lively, and A. Rownaghi (all at Georgia Institute of Technology) for helpful discussions. The Supplementary Materials includes a detailed description of materials and methods, details of the IMMP reactor, time-dependent flow profiles and synthesis cases, SEM images of ZIF-8 membranes, XRD patterns of membranes, schematics of permeation apparatus and gas bypass effects, EDX mapping of the ZIF-8 membrane, permeation modeling equations, and gas permeation data. A patent application related to this work has been filed [U.S. patent application 61/820,489, filed 7 May  2013;  S. Nair et al., Flow processing and characterization of metal-organic framework (MOF) membranes in tubular and hollow fiber modules].
View Abstract

Navigate This Article