Report

An Efficient Polymer Molecular Sieve for Membrane Gas Separations

See allHide authors and affiliations

Science  18 Jan 2013:
Vol. 339, Issue 6117, pp. 303-307
DOI: 10.1126/science.1228032

Abstract

Microporous polymers of extreme rigidity are required for gas-separation membranes that combine high permeability with selectivity. We report a shape-persistent ladder polymer consisting of benzene rings fused together by inflexible bridged bicyclic units. The polymer’s contorted shape ensures both microporosity—with an internal surface area greater than 1000 square meters per gram—and solubility so that it is readily cast from solution into robust films. These films demonstrate exceptional performance as molecular sieves with high gas permeabilities and good selectivities for smaller gas molecules, such as hydrogen and oxygen, over larger molecules, such as nitrogen and methane. Hence, this polymer has excellent potential for making membranes suitable for large-scale gas separations of commercial and environmental relevance.

Commercially important membrane-based gas separations include O2 and N2 enrichment of air, hydrogen recovery from ammonia production or hydrocarbon processing, and the upgrading of natural gas (1). They also have potential for both post-combustion and pre-combustion CO2 capture during electricity generation from fossil fuels (2, 3). Polymer membranes provide an energy-efficient method for gas separations because they do not require thermal regeneration, a phase change, or active moving parts in their operation and as such are predicted to play a growing role in an energy-constrained and low-carbon future (4). However, polymers suffer from a well-defined trade-off between the desirable properties of permeability and selectivity for the required gas component. Presently, most commercial gas-separation membranes are based on a few polymers with low permeability and high selectivity; therefore, large membrane areas are required to compensate for lack of permeance, which increases costs and space requirements for large-scale applications. Unfortunately, highly permeable microporous polymers demonstrate insufficient selectivity for practical applications because, unlike classical molecular sieves such as zeolites, they possess ill-defined voids that because of chain flexibility fluctuate in size and so have limited size-selectivity (5). However, microporous polymers have the great advantage over classical inorganic molecular sieve materials of being easily processed into membranes (such as thin coatings or hollow fibers). Therefore, it is an important challenge to develop microporous polymers that behave as efficient molecular sieves so that they can provide both the permeability and selectivity to support large-scale gas separations.

Robeson used data from a very large number of polymers to quantify the trade-off between single gas permeability (Px) and ideal selectivity (αxy = Px/Py) for a number of gas pairs (6, 7). Empirical upper bounds in plots of log αxy versus log Px were established, and the position of data points of new polymers relative to these upper bounds is used routinely as an indicator of their potential performance for gas separations. In effect, the positions of the upper bounds represent the state of the art for approaching true molecular sieve behavior in polymers. Hence, rigid glassy polymers that facilitate size-selective gas diffusivity through reduced chain mobility, especially those composed of fused-ring ladder-like structures such as polymers of intrinsic microporosity (PIMs; PIM-1, for example) (Fig. 1C), help to define the present upper bounds (6, 8). Freeman’s theoretical analysis of the position of the upper bounds suggests that further enhancement in gas selectivities could be achieved by designing polymers with even greater shape persistence (9). Recently, we tested this prediction by making a PIM in which the relatively flexible spirobisindane (SBI) component of PIM-1 was replaced with the more rigid spirobifluorene unit (PIM-SBF). The enhanced gas selectivities demonstrated by PIM-SBF resulted in data with modest advances above the current Robeson upper bounds for most gas pairs of interest (10). However, because of the inherent flexibility of the spiro-centers and dioxan linking groups used to assemble these PIMs (11), a fundamental structural redesign is required to provide microporous polymers that offer substantial further increases in rigidity and performance.

Fig. 1

The synthesis and molecular structures of (A) PIM-EA-TB and (B) PIM-SBI-TB (DMM, dimethoxymethane; TFA, trifluoroacetic acid). (C) The structure of PIM-1, the archetypal PIM. (D) A plot showing the increase in energy associated with the deviation in the marked dihedral angle within the bridged bicyclic units of EA (red) and TB (purple) as compared with typical components of PIMs, such as the spiro-center of SBI (blue) and the dioxan linking unit (green). The narrower energy wells of the bridged bicyclic units TB and EA demonstrate their greater rigidity. (E) A molecular model of PIM-EA-TB showing its contorted shape, which combined with its rigidity generates intrinsic microporosity due to an inability to pack efficiently in the solid state. (F) A solvent-cast film (10 cm in diameter) of PIM-EA-TB, through which is visible its molecular structure printed on a piece of paper.

Molecular modeling of potential structural components for assembling polymers with greater shape-persistence suggests that bridged bicyclic ring systems, such as ethanoanthracene (EA), are highly inflexible when compared with the spiro-centers and dioxan rings used commonly within PIMs (Fig. 1D). However, devising an efficient polymerization reaction on the basis of forming bridged bicyclic linking groups is a difficult synthetic challenge because of the need for the simultaneous formation of multiple covalent bonds. With this challenge in mind, we noted that the bridged bicyclic amine 2,8-dimethyl-6H,12H-5,11-methanodibenzo[b,f][1,5]diazocine, commonly called Tröger’s base (TB), can be prepared in excellent yield, despite requiring the participation of five precursor molecules and the net formation of six covalent bonds for its construction (12). TB was originally reported in 1887 after its serendipitous isolation from the reaction between p-toluidine and dimethoxymethane (13) and has since been the subject of numerous studies because of its interesting stereochemistry (14), supramolecular chemistry (15, 16), and strongly basic nature for organocatalysis (17). However, TB has rarely been used as a component for polymer construction (18, 19), and these previous studies involved preformed TB monomers rather than in situ formation of the TB unit during the polymerization reaction. The rigidity conferred on TB by its bridged bicyclic structure was confirmed through modeling (Fig. 1D); therefore, it was anticipated that polymers prepared by using TB polymerization from suitably rigid aromatic diamine monomers would be highly shape-persistent. Hence, we designed two monomers, 2,6(7)-diamino-9,10-dimethylethanoanthracene and 5,5′,(6),(6’)-diamino-3,3,3′,3′-tetramethyl-1,1’-spirobisindane, to assess the efficiency of the proposed TB polymerization reaction and, if successful, to provide PIM-like fused-ring macromolecular structures.

Once synthesized, each monomer was subsequently reacted with five equivalents of dimethoxymethane in trifluoroacetic acid (TFA) at ambient temperature (Fig. 1, A and B) until the solutions became too viscous to stir (20, 21). Each of the resulting colorless polymers proved highly soluble in chloroform, which can be attributed to their highly contorted macromolecular structures (Fig. 1E). This allowed for analysis by means of gel permeation chromatography (GPC), which indicated that the TB polymerization reaction yields polymers of high average molecular mass—typically, number-average molecular weight (Mn) > 40,000 and weight-average molecular weight (Mw) > 100, 000—relative to polystyrene standards (fig. S1). The polymers as reprecipitated powders demonstrate intrinsic microporosity by nitrogen adsorption at low pressure, at 77 K (fig. S2). Apparent BET (Brunauer, Emmett, Teller) surface areas of 1028 m2 g−1 for the ethanoanthracene-based TB polymer (PIM-EA-TB) and 745 m2 g−1 for the spirobisindane-based TB polymer (PIM-SBI-TB) were calculated from the isotherms. The apparent surface area for PIM-EA-TB is greater than for previously reported PIMs (22) or other solution-processable, amorphous microporous materials, such as those derived from molecular cages (23, 24). Thermal gravimetric analysis (TGA) indicated that neither PIM-EA-TB nor PIM-SBI-TB lose mass below 260°C except for the loss of adsorbates below 100°C. They are also stable toward hydrolysis in mild aqueous acids and bases. Casting of chloroform solutions gave optically transparent films (Fig. 1F), of good mechanical strength (fig. S3), that proved suitable for gas permeability measurements.

The single gas permeabilities and ideal selectivities of PIM-EA-TB films of different thickness (91 and 180 μm), which had first been soaked in methanol and then dried in air at room temperature, are given in Table 1. Methanol (or ethanol) treatment has been shown previously to reverse the effects of physical aging for highly permeable glassy polymers and also to remove the last residues of casting solvent (8, 25). Therefore, this treatment allows a direct comparison between the gas permeabilities of different polymers. The equivalent data for films of PIM-SBI-TB (128 and 157 μm), which had been treated under identical conditions to those of PIM-EA-TB, are also provided in Table 1 for comparison. The gas permeabilities of PIM-EA-TB are particularly high, which is consistent with its enhanced microporosity. As is generally the case for thicker films of glassy polymers (26, 27), the gas permeabilities are higher, and selectivities are lower, and this is attributed to the smaller relative contribution of a more densely packed surface region of the film relative to thinner films. Unusually for a PIM, the order of gas permeabilities of PIM-EA-TB is H2 > CO2 > He > O2 > CH4 > N2, whereas normally CO2 permeates more than H2, and O2 permeates more than He (5), suggesting a greater preference for the transport of smaller gas molecules. PIM-SBI-TB has a much lower permeability and shows the usual trend for PIMs.

Table 1

Single gas permeability Px, diffusivity Dx, solubility coefficient Sx, and their corresponding selectivities with respect to N2 for methanol-treated films of PIM-EA-TB (181 μm) and PIM-SBI-TB (157 μm). Values for a thinner film (95 and 128 μm, respectively) are given between parentheses. Values for the 181-μm PIM-EA-TB film aged for 24 hours are given in square brackets. Each value is an average of two separate measurements on the same film. O2/N2 mixed gas permeability data through a 3-day-old 129-μm PIM-EA-TB film are given in curly brackets; these are an average of eight measurements over the pressure range up to 7 bar (fig. S4). Dashes indicate no units or not applicable.

View this table:

The gas selectivities of PIM-EA-TB are remarkably high for such a permeable polymer, so that its data lie well above the 2008 Robeson upper bounds (6) for O2/N2, H2/N2, H2/CH4, and H2/CO2, each of which is of technological relevance (Fig. 2, A to D). This performance can be directly ascribed to the molecular sieving characteristics of PIM-EA-TB facilitating enhanced diffusivity selectivity (Dx/Dy) (Table 1) for molecules with smaller kinetic diameters (He = 2.69, H2 = 2.89, and O2 = 3.46 Å) over those of larger diameter (CO2 = 3.3, N2 = 3.64, and CH4 = 3.87 Å). The improved performance relative to the upper bound is particularly evident for gas pairs that include H2 (Fig. 1, B, C, and D). In addition, for H2/N2 and H2/CH4 the H2 permeation is faster (7760 Barrer), and selectivities are higher (αH2/N2 = 15 and αH2/CH4 = 11) than for PIM-SBF (PH2 = 6320 Barrer, αH2/N2 = 8, and αH2/CH4 = 6), which is the only other solution-processable polymer that provides data above the 2008 upper bounds for these gas pairs (10). Therefore, PIM-EA-TB has unrivalled potential for making highly permeable membranes for separations involving H2, such as its recovery from ammonia production, hydrocarbon processing, or precombustion carbon capture. For the latter, it is likely that the modest H2/CO2 selectivity demonstrated at 25°C will be enhanced at the required elevated temperatures (150 to 200°C) because of a large reduction in CO2 solubility that facilitates CO2 transport at lower temperatures (28).

Fig. 2

Portions of the Robeson plots (log αxy versus log Px) relevant to highly permeable polymers for (A) O2/N2; (B) H2/N2; (C) H2/CH4; (D) H2/CO2; (E) CO2/CH4, and (F) CO2/N2 gas pairs showing the data for methanol-treated PIM-EA-TB, with data points 1 (solid red square) for a 181-μm film, 2 (solid red square) for a 95-μm film, and 3 (solid red square) for the same 181-μm film after aging for 24 hours. Data point 4 (open red square) represents an average value for O2 permeabilities and O2/N2 selectivities obtained for air at variable feed pressures on a 129-μm sample (fig. S4). The black and red lines represent the 1991 (7) and 2008 (6) upper bounds, respectively. Data points 5 (solid blue squares) are from 157- and 128-μm films of PIM-SBI-TB. Other data points (solid black triangles) represent PIMs (10, 3038) and other highly permeable polymers (3942) that have been reported since the upper bounds were updated in 2008. For PIM data points above the 2008 upper bounds, see (10, 30, 34, 37).

PIM-EA-TB also shows promise for the separation of O2 from N2 as permeability data surpass the 2008 upper bound (Fig. 2A). Hence, the performance of PIM-EA-TB for the O2 enrichment of air was assessed over a range of feed pressures up to 7 bar by using a slightly aged (3-day-old) methanol-treated film of 181 μm thickness (fig. S4). The observed permeabilities of O2 (900 to 1100 Barrer) and N2 (160 to 205 Barrer) were almost independent of pressure and gave excellent O2/N2 selectivity (4.5 to 6.1). The rapid aging of PIM-EA-TB is a general feature of ultra-permeable polymers, but in this case, it is accompanied by a commensurate increase in selectivity so that a data point derived from the average of the O2 and N2 permeabilities also lies well above the upper bound (Fig. 2A, data point 4). This result suggests that PIM-EA-TB has excellent potential for the important industrial applications of nitrogen and oxygen enrichment from air because further physical aging and the use of thinner films are likely to produce a membrane with commercially desirable O2/N2 selectivity while maintaining high permeability relative to polymers presently used for these processes.

The performance of PIM-EA-TB for the gas pairs CO2/N2 and CO2/CH4 falls below the 2008 upper bounds (Fig. 2, E and F). However, for these gas pairs the selectivity is primarily due to the much higher solubility of CO2 [solubility selectivity (Sx/Sy)] (Table 1) rather than selectivity based on diffusivity. Very high CO2 solubility is a characteristic feature of all PIMs. PIM-EA-TB has complementary behavior to the promising thermally rearranged (TR) polymers and tetrazole-substituted PIMs (TZ-PIMs) that lie above the 2008 upper bound for CO2/N2 and CO2/CH4 (29, 30).

The single gas permeabilities and selectivities of PIM-SBI-TB films (from 128- and 157-μm films) are within the range that is typical for PIMs for all gas pairs (Fig. 2 and Table 1). Most PIMs contain the SBI component, and molecular modeling shows that it is relatively flexible as compared with the EA component of PIM-EA-TB (Fig. 2D) (11). Therefore, it can be deduced that the enhanced molecular sieve behavior of PIM-EA-TB (fig. S5) is attributable to the combined rigidity of the bridged bicyclic TB and EA units rather than originating from the TB unit alone. In addition to providing an inflexible structural unit, TB also offers in-built amines suitable for quaternization via reaction with alkyl dihalides, which may be useful for simple cross-linking reactions to aid membrane stability and reduce physical aging. Furthermore, given the range of available aromatic diamine monomers, the TB polymerization reaction could be used to prepare polymers with diverse applications beyond gas separation membranes, which take advantage of the functionality of TB.

Supplementary Materials

www.sciencemag.org/cgi/content/full/339/6117/303/DC1

Materials and Methods

Figs. S1 to S5

Table S1

References (4351)

References and Notes

  1. Materials and methods are available as supplementary materials on Science Online.
  2. Acknowledgments: Part of the work leading to these results has received funding from the European Community’s Seventh Framework Programme (FP7/2007-2013) under grant agreement NMP3-SL-2009-228631, project DoubleNanoMem. We also thank the Engineering and Physical Sciences Research Council for funding (grants EP/G01244X and EP/G062129/1). K. Pilnácek and G. Clarizia are gratefully acknowledged for their help in the data elaboration and the performance of mixed-gas permeation measurements. We thank G. Hutchings (Cardiff University) for reading the manuscript and making helpful suggestions. A patent has been filed through Cardiff University on the polymer synthesis presented in this report (UK patent application PCT/GB2011/051703; WO 2012/035327 (2010).
View Abstract

Navigate This Article