The synthesis, chain-packing simulation and long-term gas permeability of highly selective spirobifluorene-based polymers of intrinsic microporosity

Membranes composed of Polymers of Intrinsic Microporosity (SBF-PIMs) have potential for commercial gas separation. Here we report a combined simulation and experimental study to investigate the effect on polymer microporosity and gas permeability by placing simple substituents such as methyl, t-butyl and fused benzo groups onto PIMs derived from spirobifluorene (PIM-SBFs). It is shown that methyl or t-butyl substituents both cause a large increase in gas permeabilities with four methyl groups enhancing the concentration of ultramicropores ( 1.0 nm). Long-term ageing studies (>3.5 years) demonstrate the potential of PIM-SBFs as high-performance membrane materials for gas separations. In particular, the data for the PIM derived from tetramethyl substituted SBF reaches the proposed 2015 Robeson upper bound for O2/N2 and, hence, hold promise for the oxygen or nitrogen enrichment of air. Mixed gas permeation measurements for CO2/CH4 of the aged PIM-SBFs also demonstrate their potential for natural gas or biogas upgrading.


Introduction
Polymer membranes are used commercially for the efficient separation of gases such as oxygen or nitrogen enrichment of air (i.e.O 2 from N 2 ), natural gas and biogas upgrading (i.e.2][3][4] Recently, their potential for post-combustion carbon-capture, requiring the separation of CO 2 , predominantly from N 2 but on a massive scale, has also been suggested. 5,6For membranes it is desirable to have good selectivity for one gas over another and for large-scale separations it is necessary to have very high ux (i.e.permeance or throughput) to minimise the required size of the membrane system.Unfortunately, known highly permeable polymers possess poor selectivity and conversely selective polymers, such as those used presently for commercial applications, possess low permeability and are hence uncompetitive for large-scale gas separations with other technologies such as cryogenic distillation or solvent extraction. 7,8The empirical trade-off between the polymer properties of permeability (P x ) and ideal selectivity (P x /P y ) for a given gas pair (x and y) was quantied by Robeson in 1991, 9 and updated in 2008, 10 using the gas permeability data of a large number of polymers.For each gas pair, Robeson identied the upper bound in plots of log P x versus log P x /P y and the position of the gas permeability data for a new polymer relative to the 1991 and 2008 upper bounds is used as a universal performance indicator.Theoretical analysis by Freeman suggests that the position of the Robeson upper bound is a direct consequence of the sizesieving nature (i.e.diffusivity selectivity) of the highly rigid glassy polymers, which were used to dene it. 11,12Consequently, it was suggested that more selective polymers would be obtained by increasing polymer chain rigidity.In addition, ensuring high free volume through increasing inter-chain separation would increase gas diffusivity.Polymers of intrinsic microporosity, such as the archetypal PIM-1 (Fig. 1), follow this design concept by using a wholly fused-ring structure to provide rigidity and spiro-centres to ensure a contorted chain structure that cannot pack space efficiently. 13,14Gas permeability data for PIMs generally lie over the 1991 upper bounds for most important gas pairs, [15][16][17] with data for PIM-1 and PIM-7 being used by Robeson to position 2008 upper bounds for several important gas pairs. 10In recent years permeability data for an increasing number of PIMs, particularly those which incorporate rigid bridged bicyclic units such as ethanoanthracene, [18][19][20] triptycene, 21-25 methanopentacene 26 and Trögers base, 18,21 have been shown to exceed the 2008 upper bounds due to their enhanced selectivity.Indeed, in 2015 Pinnau et al. proposed that the O 2 /N 2 , H 2 /N 2 and H 2 /CH 4 upper bounds should be updated using permeability data from aged lms of triptycene-based PIMs. 27n 2012, we reported a preliminary account of a novel PIM in which the spirobisindane (SBI) unit of PIM-1 was replaced with the spirobiuorene (SBF) unit (PIM-SBF-1; Fig. 1). 28PIM-SBF-1 demonstrated greater selectivity as compared to PIM-1, without loss of permeability, which placed the data above the 2008 upper bound for the O 2 /N 2 , H 2 /N 2 , H 2 /CH 4 and CO 2 /CH 4 gas pairs.The improved performance of PIM-SBF-1 compared to PIM-1 was attributed to the enhanced rigidity of the SBF unit 29 over that of SBI as demonstrated by molecular dynamics modelling.[32] Following the original report on PIM-SBF-1, a macromolecular packing simulation study predicted that adding four methyl substituents to the SBF unit would result in greater intrinsic microporosity. 33Here we report a combined chainpacking simulation and experimental study to conrm this prediction and to investigate further the effect on microporosity and gas permeability by placing simple substituents such as methyl, t-butyl and fused benzo units onto the SBF unit of PIM-SBF.In addition, data from lms aged over prolonged periods of time (>3.5 years) are reported, which suggest that SBF-based PIMs have potential for gas separations involving enhanced diffusivity selectivity via a molecular sieving effect.

Synthesis
A series of seven PIMs, designated PIM-SBF-1 to PIM-SBF-7, was prepared by the aromatic nucleophilic substitution reaction between their respective 2,2 0 ,3,3 0 -tetrahydroxy-9,9-spirobiuorene precursors (1a-g) and 2,3,5,6-tetrauoroterephthalonitrile as summarised in Scheme 1. 13,28 The synthesis of the SBF monomers was based on simple adaptations of the well-established route to spirobiuorenes, 34 involving the addition of the 2-lithio-anion of biphenyl to uorenone followed by acid-mediated cyclisation.Hence, using a series of substituted 2-bromobiphenyl derivatives (2a-d) as precursors, several 2,3-dimethoxyuorenones (4a-e) were prepared via a three-step procedure involving treatment with n-butyl lithium and addition of CO 2 to give the respective carboxylic acids (3a-d), followed by an acid mediated ring closure (i.e.intramolecular acylation).The 2-lithio-anions derived from the same 2-bromobiphenyl derivatives (2a-e) were combined with uorenones 4a-e to give, aer acid treatment, a series of tetra-methoxyspirobiuorenes (5a-g).Demethylation using boron tribromide gave the SBF monomers 1a-g, which were used in the polymerisation reactions immediately due to their tendency to oxidise on standing.
With the exception of those containing fused benzosubstituents (i.e.PIM-SBFs 6 and 7), the resulting PIM-SBFs all proved soluble in chloroform, facilitating purication by reprecipitation into a non-solvent (methanol) to remove oligomeric impurities.Gel Permeation Chromatography (GPC) of the soluble PIM-SBFs 1-5 conrmed that high molecular mass polymer was achieved in each case with a range of weightaverage molecular mass (M w ) of 75-110 Â 10 3 g mol À1 relative to polystyrene standards (Table 1).The ESI † gives full experimental details and spectroscopic data for all precursors, monomers and polymers.

Simulation of solid-state packing and prediction of porosity
Molecular dynamics (MD) simulations were conducted using the LAMMPS soware package. 35PIM-SBF samples were constructed using the Polymatic simulated polymerization algorithm 36 and the compression/decompression scheme. 37The PIM-SBF structure generation consists of three steps: (1) random packing of respective PIM-SBF 100 repeat monomers into a simulation box under periodic boundary conditions at

Gas adsorption and gas transport properties
The powdered form of all of the PIM-SBFs, obtained from precipitation, adsorb a large amount of nitrogen (N 2 ) at 77 K. Analysis of the N 2 adsorption isotherms (Fig. S5-10 †) gives apparent Brunauer-Emmett-Teller (BET) surface areas (SA BET ) within the range of 727-938 m 2 g À1 (Table 1), with the value for PIM-SBF-2 being similar to that from PIMs possessing the highest SA BET for solution processable polymers. 18,24It is notable that the BET value for PIM-SBF-2 calculated from the simulation study is similar to the apparent value measured directly from N 2 isotherms and supports the prediction that methyl substituents will enhance intrinsic microporosity by increasing the inter-chain distance. 33Both simulation and experimental analysis of PIM-SBF-5 conrm that microporosity is also enhanced by the use of t-butyl substituents.Overall, there is good correlation between the predicted and experimentally Scheme 1 The synthesis of PIM-SBFs 1-7 (see Section 2 of the ESI † for details).
Table 1 Molecular mass and gas adsorption properties of PIM-SBFs 1-7 determined BET surface areas for all PIM-SBFs (i.e.within 100 m 2 g À1 ), which is within the range of reported experimental values of the BET surface area of PIM-1 quoted in the literature (720-870 m 2 g À1 ) caused by the variation in N 2 adsorption isotherms at 77 K. 38 Micropore size distributions calculated from CO 2 adsorption at 273 K for PIM-SBF-1, PIM-SBF-2 and PIM-SBF-5 (Fig. S11 †) using DFT correlate well with those calculated from the chain packing simulations (Fig. 2 and S12 †).Signicantly, both PIM-SBF-2 and PIM-SBF-5 possess greater microporosity than PIM-SBF-1 but with a clear difference in pore size distribution so that PIM-SBF-2 has a greater contribution from ultramicropores (<0.7 nm) whereas PIM-SBF-5 has greater pore volume from larger micropores (>1.0 nm).

Gas permeability
Simple solvent casting from chloroform solution allowed the fabrication of optically clear and robust lms of PIM-SBFs 1-5, of $100 mm thickness, suitable for gas permeability studies.The pure gas permeabilities of PIM-SBFs 1-5, which had rst been soaked in methanol and then dried in air at room temperature for 24 h, are given in Table 2. Methanol treatment has been shown previously to reverse the effects of physical ageing for glassy ultra-permeable polymers and it also removes the last residues of casting solvent. 16,39,40In addition, methanol treatment allows a direct comparison between the gas permeabilities of different polymers prior to ageing.The equivalent data for a freshly methanol treated lm of PIM-1 of similar thickness, measured under identical conditions to those of the PIM-SBFs, are also provided in Table 2 for comparison.
For each methanol treated PIM-SBF the order of gas permeabilities is CO 2 > H 2 > O 2 > He > CH 4 > N 2 , which is the same as that for PIM-1.As noted previously, 28 whilst the values for gas permeabilities are broadly similar for PIM-1 and PIM-SBF-1 for gases with smaller kinetic diameters (i.e.He ¼ 2.69; H 2 ¼ 2.8; CO 2 ¼ 3.3; O 2 ¼ 3.46 Å), those for larger molecules (i.e.N 2 ¼ 3.64; CH 4 ¼ 3.87 Å) are smaller (>30%) resulting in a signicant enhancement of ideal selectivities.As noted previously, 28 this places the data for PIM-SBF-1 above the 2008 upper bound for the O 2 /N 2 , H 2 /N 2 , N 2 /CH 4 and CO 2 /CH 4 gas pairs and can be directly ascribed to the greater rigidity of the SBF units as compared to SBI units of PIM-1.It is evident from the data for PIM-SBFs 2-5 that placing methyl or t-butyl substituents onto the SBF unit causes a moderate to large increase in gas permeabilities relative to those of both PIM-1 and PIM-SBF-1.This is consistent with the conclusion derived from predictive packing simulations 33 and experimental gas adsorption, discussed above, that methyl groups help to increase the inter-chain separation distance and generate greater intrinsic microporosity.The order of increasing gas permeability is PIM-SBF-1 (0 Me) < PIM-SBF-4 z PIM-SBF-3 (both 2 Me) < PIM-SBF-5 (6 Me) < PIM-SBF-2 (4 Me).The gas permeabilities for PIM-SBF-2 and PIM-SBF-5 are signicantly higher than for all reported PIMs with the exception of PIM-1, 42 which is slightly less permeable than PIM-SBF-2 but more so than PIM-SBF-5, and the recently reported PIM-TMN-Trip, 24 the latter deriving its exceptionally high permeability from its Table 2 Ideal gas permeabilities of freshly methanol treated and aged films of PIM-SBFs 1-5 measured at 25 C with data for PIM-1 unique 2D chain structure.In addition, there are several polyacetylenes with high gas permeabilities, such as the ultrapermeable poly(trimethylsilylpropyne) (PTMSP), when freshly treated with methanol, although with much lower selectivities relative to the PIMs. 43For PIM-SBFs 2-4, in which the methyl groups are attached directly to the SBF unit, the selectivity enhancement noted for PIM-SBF-1 over PIM-1 is maintained providing data points, which are of a similar distance above the 2008 Robeson upper bounds for the O 2 /N 2 , H 2 /N 2 , CO 2 /N 2 and CO 2 /CH 4 gas pairs (Fig. 3).This is consistent with the high concentration of ultramicropores (<0.7 nm) suggested by chain-packing simulation and analysis of gas adsorption data for PIM-SBF-2 (Fig. 2).Ultramicropores, being of similar size to the kinetic diameter of gas molecules, are responsible for selective gas transport.In contrast, the relatively low selectivity of PIM-SBF-5 can be ascribed to a broader pore size distribution with a greater concentration of larger, less selective, micropores (>1.0 nm) (Fig. 2).In addition, the t-butyl substituents of PIM-SBF-5 will be relatively free to undergo thermal motions so that they may act as molecular turnstiles facilitating the transport of larger gas molecules such as N 2 and CH 4 .

Extended ageing studies
Physical ageing (i.e.loss of free volume over time) is a general feature of glassy polymers.For example, data for PIM-1 shows that aer an extended period of 1360 days the value for PO 2 falls to 317 Barrer but with a corresponding increase in selectivity of PO 2 /PN 2 ¼ 5.0. 41Ageing studies were conducted on lms of all of the PIM-SBFs with gas permeability measurements taken aer at least 1200 days to allow for a direct comparison with the previous PIM-1 ageing study 41 (Table 2).On extended ageing of PIM-SBF-1-4 the order of gas permeability becomes H 2 > CO 2 > He > O 2 > CH 4 > N 2 , which is more conventional for glassy polymers.When the data from aged lms are placed on the appropriate Robeson plot (Fig. 3), it can be observed that PIM-SBF-2 ages at an approximately similar rate relative to both PIM-1 and PIM-SBF-1 with loss of permeability combined with a signicant increase in selectivity.Interestingly, for the O 2 /N 2 and H 2 /N 2 gas pairs these trend lines are approximately parallel and are steeper in gradient than the Robeson upper bounds.For these two gas pairs selectivity is based on differences in gas diffusivity, therefore, size-selectivity is being enhanced beyond that expected by the standard polymer permeability-selectivity trade-off.The pore structure of PIM-SBF-1-4 translates into a very steep correlation of the diffusion coefficient with the gas diameter, highlighting the strong size-selectivity of these polymers particularly aer ageing (Fig. 4a).The slope is highest for the less permeable and more selective PIM-SBF-1 and lowest for most permeable and less selective PIM-SBF-5, showing that the permselectivity is closely related to the diffusion selectivity.The steeper slope, and thus increased size-sieving behaviour aer aging, is an indication for a denser packing and lower free volume of the aged sample with respect to the MeOH treated sample (Fig. 4b).
Hence, ageing is preferentially reducing larger micropores within the PIM-SBFs.Notably, the data for PIM-SBF-2 aer ageing for 1295 days (PO 2 ¼ 910 Barrer; PO 2 /PN 2 ¼ 5.5) is very close to the proposed 2015 upper bound for O 2 /N 2 (Fig. 3a). 27n contrast, for the CO 2 /N 2 gas pair the ageing trend line for each of PIM-SBF-1-4 are at a much shallower gradient than the 2008 Robeson upper bound (Fig. 3c), which can be explained by CO 2 /N 2 separation being dominated by solubility selectivity.CO 2 solubility is correlated to the amount of free volume within the polymer which is lost on ageing.For the CO 2 /CH 4 gas pair there is a contribution to selectivity from both gas solubility and diffusivity and the ageing trend lines for PIM-SBF-1-4 are well above and roughly parallel to the 2008 Robeson upper bound (Fig. 3d).For this gas pair the data for aged lm of PIM-SBF-1 is particularly impressive (PCO 2 ¼ 2410 Barrer; PCO 2 /PCH 4 ¼ 24).
The ageing of PIM-SBF-5 is anomalously slow relative to that of the other PIM-SBFs with large values of gas permeability demonstrated almost four years aer methanol treatment (e.g.PCO 2 > 10 000 Barrer) and the original CO 2 > H 2 > O 2 > He > CH 4 > N 2 order of gas permeabilities is maintained (Table 2; Fig. 3).Remarkably, the values of gas permeabilities for this aged lm of PIM-SBF-5 are greater than those of freshly methanol treated PIM-SBF-1.Therefore, the bulky t-butyl groups of PIM-SBF-5 appear better than the methyl substituents of PIM-SBFs 2-4 at maintaining the distance between polymer chains during ageing.
Ultimately, for practical membrane application in gas separations, the PIM selective layer would be the thin skin on a hollow bre or a thin lm within a composite membrane.It is well-established that physical ageing of polymers, particularly those of high free volume, is much more rapid in thin lms. 45,46 recent study on PIM-1 hollow bres 47 estimates that the gas permeability and selectivity of the thin selective layer ($3 mm; PO 2 ¼ 200 Barrer; PO 2 /PN 2 ¼ 4.6) are similar to those of a thick lm (102 mm) that had been aged for 1360 days. 41The long-term ageing study of the PIM-SBFs suggest that, with the exception of PIM-SBF-5, they age similarly to PIM-1 with commensurate increases in selectivities (Fig. 3).Therefore, for thin lms of PIM-SBF-2, it can be estimated that the separation of important gas pairs based on diffusivity selectivity (e.g.O 2 /N 2 ) would approach those of commercial polymer membranes (e.g.PO 2 / PN 2 > 5.5) but with much higher permeance ($three times that expected for PIM-1).

Mixed gas permeation
Mixed gas permeation measurements with CO 2 /CH 4 were carried out both across a total pressure range of 1-6 bar using a 35/65 vol% mixture and a CO 2 feed concentration of 10-50% in CH 4 at 3 bar for two representative aged samples: the most selective sample, PIM-SBF-1 aged over 2088 days, and the most permeable sample, PIM-SBF-5 aged for 1439 days (Fig. 5).
For the mixed gas measurements PIM-SBF-1 and PIM-SBF-5 both maintain excellent gas permeability to give values that are Fig. 4 Correlation of the diffusion coefficient with the square of the effective gas diameter for aged samples (A) 44 PIM-SBF-1 ( ), PIM-SBF-similar to the ideal values measured with pure gases.In both lms the CO 2 permeability and CO 2 /CH 4 selectivity decrease about 20% upon a pressure increase from 1 to 6 bar, while the CH 4 permeability remains virtually constant.On the other hand, at a feed pressure of 3 bar the transport parameters of PIM-SBF-5 are independent of the gas composition in the range of 10-50% CO 2 , which is good for a stable operation of a membrane separation plant under variable feed conditions, for instance in the case of biogas separation.At a pressure of 1 bar, the permselectivity is substantially higher than the ideal value and the points lie in a more favourable region of the Robeson diagram (Fig. 3d).At higher pressures, they converge close to the ideal values measured from pure gases.

Conclusions
Both methyl (PIM-SBF-2-4) and t-butyl substituents (PIM-SBF-5) attached to the SBF unit increase intrinsic microporosity as demonstrated by chain packing simulations, and by analysis of both gas adsorption data and gas permeability of the PIM-SBF series.However, these two substituents generate a signicant difference in pore size distribution, as predicted by packing simulation, that has a marked effect on the ideal gas selectivity of these polymers.Furthermore, these differences appear to be accentuated on ageing over an extended period with good selectivities being obtained.Indeed, the excellent performance of aged PIM-SBF-1 in mixed gas CO 2 /CH 4 separation experiments (PCO 2 >2000 Barrer; PCO 2 /PCH 4 > 25) suggests its use might allow signicant reductions to the size of current membrane systems for industrially relevant gas separations, such as natural gas or biogas upgrading.

Fig. 3
Fig. 3 Robeson plots for the (a) O 2 /N 2 , (b) H 2 /N 2 , (c) CO 2 /N 2 and (d) CO 2 /CH 4 gas pairs showing the position of the gas permeability data for PIM-SBFs 1-5 and that of PIM-1.Data from films aged for >3.5 years (e.g.2a) are joined by an arrow to those of the freshly methanol treated films to form ageing trend-lines.Upper bounds are represented by black (1991), red (2008) and dotted (2015) lines.Mixed CO 2 /CH 4 data at pressures of 1-6 bar for the film of PIM-SBF-1 aged for 2088 days are also plotted with red arrow indicating increasing pressure.

Fig. 5
Fig. 5 Mixed gas separation performance of aged PIM-SBF-1 (A) and aged PIM-SBF-5 (C) as a function of the feed pressure for the CO 2 /CH 4 (35/65 vol%) mixture and performance of PIM-SBF-1 (B) and PIM-SBF-5 (D) as a function of the feed composition for the CO 2 /CH 4 mixture at a pressure of 3 bar(a).Filled symbols for increasing, and empty symbols for decreasing pressure and CO 2 concentration steps.