Crosslinked PEG and PEBAX Membranes for Concurrent Permeation of Water and Carbon Dioxide

Membrane technology can be used for both post combustion carbon dioxide capture and acidic gas sweetening and dehydration of natural gas. These processes are especially suited for polymeric membranes with polyether functionality, because of the high affinity of this species for both H2O and CO2. Here, both crosslinked polyethylene glycol diacrylate and a polyether-polyamide block copolymer (PEBAX 2533©) are studied for their ability to separate CO2 from CH4 and N2 under single and mixed gas conditions, for both dry and wet feeds, as well as when 500 ppm H2S is present. The solubility of gases within these polymers is shown to be better correlated with the Lennard Jones well depth than with critical temperature. Under dry mixed gas conditions, CO2 permeability is reduced compared to the single gas measurement because of competitive sorption from CH4 or N2. However, selectivity for CO2 is retained in both polymers. The presence of water in the feed is observed to swell the PEG membrane resulting in a significant increase in CO2 permeability relative to the dry gas scenario. Importantly, the selectivity is again retained under wet feed gas conditions. The presence of H2S is observed to only slightly reduce CO2 permeability through both membranes.


Introduction
Post combustion carbon dioxide capture requires the removal of carbon dioxide from flue gas streams saturated with water. Membrane technology is considered of potential for this application because of its smaller lighter footprint, relative to solvent absorption, but comparable costs [1][2][3]. In this application, co-permeation of water vapour with the CO 2 can be advantageous, as it dilutes the permeate stream, reducing the CO 2 partial pressure and thus increasing the driving force for CO 2 permeation [2].
Similarly, two major unit operations in natural gas processing are gas sweetening, which removes carbon dioxide and hydrogen sulphide; and gas dehydration which removes water [4]. This ensures the natural gas composition is standardised for transport and limits the risk of water hydrates and pipeline corrosion. Membrane gas separation is increasingly being used for natural gas sweetening [5][6][7][8][9][10]. Dehydration of natural gas is traditionally undertaken through an ethylene glycol process [11]; however polysulfone membranes have been used for this application [12]. There is the potential to combine both natural gas sweetening and dehydration in a single membrane process, although care would be required to ensure pressure conditions are suitable to avoid both methane and carbon dioxide hydrates; and to avoid corrosion in the permeate stream piping. To undertake either of these operations, the membrane needs to have high permeability for water, and CO 2 , while also having good selectivity against other gases. Water permeability through any polymeric membrane is usually orders of magnitude greater than other gases and vapors, due to the small size and high critical temperature of this molecule. Membranes containing polyether groups are of particular interest for CO 2 removal, due to the strong interaction between the quadripolar CO 2 and the polar ether bonds [13][14][15]. However, crystalline regions within the membrane can reduce the permeability of such gas species, due to the reduction in fractional free volume within these regions. Pure polyethylene glycol (PEG) membranes suffer significantly from such crystallinity, with values reported of up to 71 vol% [16]. The use of cross linking, or the incorporation of polyamide blocks within the structure, as commercialised through the PEBAX © series of block copolymers, reduces the overall crystallinity, with values of 14% to 51% reported for PEBAX systems [17].
In this investigation, the performance of cross-linked PEG and PEBAX 2533 in single and mixed gas feeds of CO 2 and water are reported. In particular, the effect of competitive sorption on the permeability of CO 2 , water, N 2 , CH 4 and H 2 S is studied, to identify the potential of these two polymeric membranes for post combustion carbon capture and for simultaneous removal of both acid gases and water from natural gas.
Sorption measurements were conducted on a Gravimetric Sorption Analyzer (GHP-FS, VTI Scientific Instruments, Irvine, CA, USA), with a Cahn D-200 microbalance, described elsewhere [19]. For CO 2 , CH 4 and N 2 isotherms the analyzer was operated in static mode. The membranes were initially exposed to vacuum overnight and heated to 80˝C to remove air and water vapour. The penetrant gas was introduced into the chamber in pressure steps at 35˝C and the equilibrium mass of membrane plus sorbed gas was measured. In an incremental manner, penetrant isotherms as a function of penetrant pressure were determined. Sorption mass equilibrium for all gases was reached within a maximum of 6 h at each pressure. A comparable experiment using helium was completed to determine the buoyancy correction. For water sorption, the Gravimetric Sorption Analyzer was operated in a flow mode. The standard degassing procedure as above was undertaken. Helium, the carrier gas, was then introduced into the sample chamber at a set pressure,~1 atm, in a flow through arrangement. The helium source was divided into two streams, one dry, the other passed through a water entrainer at a set temperature. Relative humidity within the sample chamber was achieved by varying the flowrate ratio of these two streams. Sorption mass equilibrium was achieved within 3 h at each relative humidity stage.
Single gas permeabilities were measured on a constant volume, variable pressure gas permeation apparatus described elsewhere [20]. Mixed gas permeabilities were tested on a mixed gas instrument, also reported elsewhere [21]. The effect of humidity was measured on a modified mixed gas permeability instrument reported elsewhere [13]. Pure gas or a gas mixture (10% CO 2 in CH 4 ) was fed into a saturator vessel filled with water, then a demister vessel to generate the humid gas stream. Both the saturator and demister were partially filled with steel wool and immersed in a temperature Membranes 2016, 6, 0001 3 of 10 controlled bath. The wet gas stream was then passed through a humidity and temperature transmitter (HMT, Probe type 334 Vaisala Oyj, Vantaa, Finland), which was fitted within a fan forced oven. The oven also contained the permeation cell, another humidity sensor (HMT, Vantaa, Finland) on the permeate side and associated tubing. Stainless steel wool packing was present in both sides of the membrane within the permeation cell, to enhance mixing and minimize concentration polarization. Before each series of experiments, the permeation cell was pre-heated to the operating temperature for at least an hour, with nitrogen and argon flowing through both sides of the membrane, to avoid vapor condensation during the experiment. The humidity of the feed stream was controlled by the temperature of the saturator and the demister. The permeate side of the membrane used argon as the sweep gas and after the humidity sensor passed through an iced cold trap to remove water. All permeabilities were measured at 35˝C for a total upstream pressure of 600 kPa gauge. For mixed gas measures, the concentrations of gases in the permeate stream were determined by gas chromatography (Varian CP-3800, column PORAPAK Q in a bypass series with a Molecular Sieve 5A) (Varian, Melbourne, Australia).

Results and Discussion
The sorption isotherms for CO 2 , N 2 , CH 4 in both crosslinked PEG and PEBAX membranes are provided in Figure 1. The isotherms are linear and can be described by Henry's Law, consistent with the literature for small gases in rubbery polymers at low pressure [14]. The Henry's law constant for CO 2 , N 2 and CH 4 can be determined from the concentration isotherms based on fugacity and are compared to literature values in Table 1 and Figure 2. The data for crosslinked PEG is consistent with the literature, while that for PEBAX 2533 is slightly higher.

Results and Discussion
The sorption isotherms for CO2, N2, CH4 in both crosslinked PEG and PEBAX membranes are provided in Figure 1. The isotherms are linear and can be described by Henry's Law, consistent with the literature for small gases in rubbery polymers at low pressure [14]. The Henry's law constant for CO2, N2 and CH4 can be determined from the concentration isotherms based on fugacity and are compared to literature values in Table 1 and Figure 2. The data for crosslinked PEG is consistent with the literature, while that for PEBAX 2533 is slightly higher.     The H2O isotherm for both membranes (Figure 3) is different from the other gases in that it is strongly concave at high H2O activity, especially for the crosslinked PEG system. This behaviour is typical of highly condensable vapors [24] and the PEG result at high activity suggests extensive swelling of the polymer and the formation of a hydrogel, as commonly observed in the literature [25]. This represents a significant morphology change from the dry membrane, as the hydrogel membrane swells and elasticity increases dramatically. The water solubility in PEBAX is lower due to the presence of the "hard" polyamide domains, which do not swell as readily in water [26]. Sophisticated models such as PC-SAFT [27,28] or the ENSIC model [29] are required to describe the swelling effects that are evident at higher water activity, but this is outside the scope of this paper. However, the Henry's Law constant can be estimated from the data at low water activity (the infinite dilution solubility) and these values are also provided in Table 1 and Figure 2.
Generally, it is found that such Henry's Law coefficients can be correlated to either the Lennard-Jones well depth [17] of the gas or to its critical temperature [18]. Either approach is able to fit the data here for CO2, N2 and CH4, but the Lennard Jones approach is significantly more consistent with the results for water. The use of the critical temperature as the correlating parameter leads to a substantial The H 2 O isotherm for both membranes (Figure 3) is different from the other gases in that it is strongly concave at high H 2 O activity, especially for the crosslinked PEG system. This behaviour is typical of highly condensable vapors [24] and the PEG result at high activity suggests extensive swelling of the polymer and the formation of a hydrogel, as commonly observed in the literature [25]. This represents a significant morphology change from the dry membrane, as the hydrogel membrane swells and elasticity increases dramatically. The water solubility in PEBAX is lower due to the presence of the "hard" polyamide domains, which do not swell as readily in water [26]. Sophisticated models such as PC-SAFT [27,28] or the ENSIC model [29] are required to describe the swelling effects that are evident at higher water activity, but this is outside the scope of this paper. However, the Henry's Law constant can be estimated from the data at low water activity (the infinite dilution solubility) and these values are also provided in Table 1 and Figure 2.
Generally, it is found that such Henry's Law coefficients can be correlated to either the Lennard-Jones well depth [17] of the gas or to its critical temperature [18]. Either approach is able to Membranes 2016, 6, 0001 5 of 10 fit the data here for CO 2 , N 2 and CH 4 , but the Lennard Jones approach is significantly more consistent with the results for water. The use of the critical temperature as the correlating parameter leads to a substantial under-estimation of this parameter (Figure 2). The pure gas permeability in crosslinked PEG and PEBAX 2533 for CO2, N2 and CH4 at 35 °C are provided in Table 2. Both PEG and PEBAX clearly favour CO2 over N2 and CH4. For PEG, the N2 and CH4 permeability is reduced compared to literature, while the CO2 permeability is half that reported in the literature (130 Barrer) [23,30]. This suggests that the level of crystallinity has not been reduced to the same extent as that observed by Lin et al. [23]. This is associated with the level of cross-linking and the difference in the acrylic length of the original polymer influencing membrane density and fractional free volume. However, the CO2 permeability is around five times greater than that of semicrystalline (uncrosslinked) PEG of 12 Barrer [16]. For PEBAX 2533, the CO2 permeability is comparable to Bondar et al. [17] who obtained a value of 221 Barrer for membranes prepared by melt extrusion. Tocci et al. [31] report a slightly lower CO2 permeability of 200 Barrer and a slightly higher CH4 value of 40 Barrer for a membrane formed from a n-butanol/2-propanol solution and annealed at 70 °C for 24 h. Barbi et al. [26] show that the permeability value is indeed strongly dependent upon the casting solvent and annealing conditions. They record values of 276 Barrer (CO2) and 40 Barrer (CH4) when the membrane is cast in 1-butanol and 241 Barrer (CO2) and 35 Barrer (CH4) when it is cast in cyclohexanol; with annealing in both The pure gas permeability in crosslinked PEG and PEBAX 2533 for CO 2 , N 2 and CH 4 at 35˝C are provided in Table 2. Both PEG and PEBAX clearly favour CO 2 over N 2 and CH 4 . For PEG, the N 2 and CH 4 permeability is reduced compared to literature, while the CO 2 permeability is half that reported in the literature (130 Barrer) [23,30]. This suggests that the level of crystallinity has not been reduced to the same extent as that observed by Lin et al. [23]. This is associated with the level of cross-linking and the difference in the acrylic length of the original polymer influencing membrane density and fractional free volume. However, the CO 2 permeability is around five times greater than that of semicrystalline (uncrosslinked) PEG of 12 Barrer [16]. For PEBAX 2533, the CO 2 permeability is comparable to Bondar et al. [17] who obtained a value of 221 Barrer for membranes prepared by melt extrusion. Tocci et al. [31] report a slightly lower CO 2 permeability of 200 Barrer and a slightly higher CH 4 value of 40 Barrer for a membrane formed from a n-butanol/2-propanol solution and annealed at 70˝C for 24 h. Barbi et al. [26] show that the permeability value is indeed strongly dependent upon the casting solvent and annealing conditions. They record values of 276 Barrer (CO 2 ) and 40 Barrer (CH 4 ) when the membrane is cast in 1-butanol and 241 Barrer (CO 2 ) and 35 Barrer (CH 4 ) when it is cast in cyclohexanol; with annealing in both cases at 70˝C for 12 h. They argue that membranes cast from cyclohexanol exhibit a decreased "hard" (polyamide) domain fraction due to an imperfect phase separation.
In terms of selectivity, PEBAX has lower CO 2 /N 2 and CO 2 /CH 4 selectivity compared to PEG. These differences are believed to be due to the underlying morphology of the polymers. PEBAX 2533 has a lower density (1.00 g/cm 3 ) than the cross-linked PEG (1.31 g/cm 3 ), and as such the more open morphology of PEBAX enabled higher gas permeabilities. However, the more open morphology is at the expense of selectivity, which is a common trade-off for polymeric membranes [32].
The H 2 O permeability through PEG and PEBAX membranes as a function of feed humidity is provided in Figure 4, with N 2 carrier gas. The PEBAX H 2 O permeabilities are of the same order of magnitude as reported in the literature; Sijbesma et al. [33] reported a H 2 O permeability of~25,000 Barrer, Gugliuzza and Drioli [34] reported 25,030 Barrer, while Rezac et al. [35] reported 25,600 for a water activity of 0.53. Our own values are slightly higher, possibly as a result of the care we have taken to eliminate concentration polarization. Conversely, the H 2 O permeabilities for PEG reported here are lower than those in the literature for copolymer series based on PEG [36,37]. This difference is attributed to the cross-linking in this PEG membrane, decreasing water diffusion within the resulting membrane and hence reducing water permeability. For both membranes the H 2 O permeability increases as a function of relative humidity, which is observed for other polymeric membranes and is indicative of water swelling. For PEG, a significant increase in permeability occurs above an activity of 0.6, suggestive of a transition in the membrane morphology to a hydrogel. This observation has also been observed for other PEG based membranes [36,37]. The average H 2 O/N 2 ideal selectivity of PEG is 28,000 and for PEBAX is 5400, which is comparable to other rubbery polymeric membranes [33].
cases at 70 °C for 12 h. They argue that membranes cast from cyclohexanol exhibit a decreased "hard" (polyamide) domain fraction due to an imperfect phase separation.
In terms of selectivity, PEBAX has lower CO2/N2 and CO2/CH4 selectivity compared to PEG. These differences are believed to be due to the underlying morphology of the polymers. PEBAX 2533 has a lower density (1.00 g/cm 3 ) than the cross-linked PEG (1.31 g/cm 3 ), and as such the more open morphology of PEBAX enabled higher gas permeabilities. However, the more open morphology is at the expense of selectivity, which is a common trade-off for polymeric membranes [32].
The H2O permeability through PEG and PEBAX membranes as a function of feed humidity is provided in Figure 4, with N2 carrier gas. The PEBAX H2O permeabilities are of the same order of magnitude as reported in the literature; Sijbesma et al. [33] reported a H2O permeability of ~25,000 Barrer, Gugliuzza and Drioli [34] reported 25,030 Barrer, while Rezac et al. [35] reported 25,600 for a water activity of 0.53. Our own values are slightly higher, possibly as a result of the care we have taken to eliminate concentration polarization. Conversely, the H2O permeabilities for PEG reported here are lower than those in the literature for copolymer series based on PEG [36,37]. This difference is attributed to the cross-linking in this PEG membrane, decreasing water diffusion within the resulting membrane and hence reducing water permeability. For both membranes the H2O permeability increases as a function of relative humidity, which is observed for other polymeric membranes and is indicative of water swelling. For PEG, a significant increase in permeability occurs above an activity of 0.6, suggestive of a transition in the membrane morphology to a hydrogel. This observation has also been observed for other PEG based membranes [36,37]. The average H2O/N2 ideal selectivity of PEG is 28,000 and for PEBAX is 5400, which is comparable to other rubbery polymeric membranes [33]. The permeability of CO2 for PEG and PEBAX in dry N2-CO2 and CH4-CO2 mixed gas conditions are shown in Table 3. In both membranes and both mixed gas systems the observed permeability of The permeability of CO 2 for PEG and PEBAX in dry N 2 -CO 2 and CH 4 -CO 2 mixed gas conditions are shown in Table 3. In both membranes and both mixed gas systems the observed permeability of CO 2 is reduced compared to the pure gas permeability, because of competition from the other gas. For both N 2 -CO 2 and CH 4 -CO 2 feeds, the CO 2 permeability in PEG is reduced by 11% and for PEBAX by 10%, compared to the single gas case. For N 2 and CH 4 , the average permeability through both membranes is also reduced, due to competitive sorption from CO 2 . As such, the CO 2 /N 2 and CO 2 /CH 4 selectivities of both membranes are slightly lower compared to the single gas measurements, but both still retain significant selectivity for CO 2 .  14 7 A water activity of 0.2 is then applied to the CH 4 -CO 2 feed gas to simulate wet conditions in a standard natural gas process [6]. For both PEG and PEBAX the CO 2 permeability changed little compared to the dry mixed gas (Table 4). Similarly, the CH 4 permeability increased only slightly for both PEG and PEBAX compared to the dry mixed gas feed. As a result, the ideal CO 2 /CH 4 selectivity of both membranes decreased only slightly. Hence, CO 2 separation performance is retained under wet feed conditions, indicating both membranes could process a low RH feed gas without a loss in performance. The water permeability through PEBAX also appears to increase slightly under the mixed gas conditions; compared to the pure N 2 feed gas measurement (Figure 3) at the same activity, although the increase is possibly within experimental error. This increase is attributed to CO 2 inducing plasticization in the PEBAX structure and thus increasing the water diffusivity through the membranes. The water permeability through PEG is within error of the N 2 feed gas result at the same activity ( Figure 3).
Under dry feed gas conditions adding 500 ppm H 2 S to the N 2 -CO 2 mixed gas system enables the impact of H 2 S on separation performance to be measured, with the change in CO 2 and N 2 permeability reported in Table 4. For PEG, the presence of H 2 S reduces CO 2 permeability by 1 Barrer for PEG and 2 Barrer for PEBAX. These are only minor decreases in performance and signify that H 2 S competition for sorption in both of these membranes is small. When compared to that previously reported for PDMS exposed to H 2 S (~8% reduction observed [38]), it highlights the different chemical interactions between the gases and moieties in the two polymers studied here. Similarly, only a small change in N 2 permeability upon exposure to H 2 S for both PEG and PEBAX was observed. Hence, the CO 2 /N 2 selectivity of both membranes in the presence of 500 ppm H 2 S is retained.

Conclusions
Membranes consisting of cross-linked PEG and PEBAX 2533 where studied for CO 2 and H 2 O separation from CH 4 and N 2 , to simulate applications including post-combustion carbon capture and natural gas sweetening and dehydration. Sorption measurements indicated that the simple gases followed Henry's law within both membranes while water sorption increased substantially at higher water activities, particularly for the PEG membrane. The Henry's Law coefficients were more readily fitted to a correlation with the Lennard Jones well depth than to the critical temperature. Single gas measurements indicated that both membranes were selective for CO 2 against CH 4 and N 2 , while water permeability increased strongly associated with water activity in the feed gas. CO 2 permeability fell in mixtures with CH 4 and N 2 , compared to the single gas measurement, because of competitive sorption. Under wet feed gas conditions, the presence of water swelled both membranes and the gas permeability increased, with only a small decrease in selectivity. It was also observed that H 2 S had very little impact on both PEG and PEBAX membranes separation of CO 2 from N 2 .