Towards Biohydrogen Separation Using Poly(Ionic Liquid)/Ionic Liquid Composite Membranes

Considering the high potential of hydrogen (H2) as a clean energy carrier, the implementation of high performance and cost-effective biohydrogen (bioH2) purification techniques is of vital importance, particularly in fuel cell applications. As membrane technology is a potentially energy-saving solution to obtain high-quality biohydrogen, the most promising poly(ionic liquid) (PIL)–ionic liquid (IL) composite membranes that had previously been studied by our group for CO2/N2 separation, containing pyrrolidinium-based PILs with fluorinated or cyano-functionalized anions, were chosen as the starting point to explore the potential of PIL–IL membranes for CO2/H2 separation. The CO2 and H2 permeation properties at the typical conditions of biohydrogen production (T = 308 K and 100 kPa of feed pressure) were measured and discussed. PIL–IL composites prepared with the [C(CN)3]− anion showed higher CO2/H2 selectivity than those containing the [NTf2]− anion. All the membranes revealed CO2/H2 separation performances above the upper bound for this specific separation, highlighting the composite incorporating 60 wt % of [C2mim][C(CN)3] IL.


Introduction
Due to its outstanding intrinsic features, hydrogen (H 2 ) is considered to be the energy carrier of the future. Besides being the simplest element in the universe, the H 2 molecule has the highest energy content per unit weight of any known fuel. However, H 2 is not a primary fuel source, which means that it is not available in nature and thus needs to be produced [1]. Hydrogen has been produced mainly on an industrial scale from fossil fuels, through natural gas reforming or coal gasification, and from water, using electrolysis in which water (H 2 O) can be split into hydrogen and oxygen (O 2 ) [2]. Although water splitting is an ecologically clean process compared to the previously mentioned H 2 production processes, it is a highly energy-demanding technology [3].
Hydrogen production using biological processes has been attracting growing attention since it is more environmentally friendly and less energy intensive than the other described H 2 production systems because its conditions are close to room temperature (303-313 K) and ambient pressure (100 kPa) [2]. BioH 2 production can be divided into two main categories: light-dependent anions [35][36][37]. However, our group reported PIL-IL membranes made of pyrrolidinium-based PILs combining the same anions, which are particularly simple to prepare through a metathesis reaction of a commercially available polyelectrolyte. The PIL-IL membranes displayed CO 2 /N 2 separation performances near or even above the Robeson upper bound [38][39][40][41]. In fact, the CO 2 -phylic behavior of the [NTf 2 ] − anion and the CO 2 separation efficiency of the [C(CN) 3 ] − anion [42] motivated us to explore the most promising pyrrolidinium-based PIL-IL composites based on these two anions for CO 2 /N 2 separation, now for CO 2 /H 2 separation.
In this work, solvent casting method was used to prepare composite membranes composed of two pyrrolidinium-based PILs: poly ( Figure 1. The CO 2 and H 2 permeation properties (permeability, diffusivity and solubility) were determined at two different temperatures (T = 293 K and T = 308 K) under a transmembrane pressure differential of 100 kPa. A temperature of 293 K was used first so that the results obtained herein could be compared to those previously reported by our group, while T = 308 K was chosen to reproduce the hydrogen bioproduction conditions [13].  Figure 1. The CO2 and H2 permeation properties (permeability, diffusivity and solubility) were determined at two different temperatures (T = 293 K and T = 308 K) under a transmembrane pressure differential of 100 kPa. A temperature of 293 K was used first so that the results obtained herein could be compared to those previously reported by our group, while T = 308 K was chosen to reproduce the hydrogen bioproduction conditions [13].

Preparation of PIL-IL Membranes
Several free-standing membranes composed of the synthesized PILs and specific quantities of different ILs containing the same anions were produced by solvent casting. The first step was the preparation of 6 (w/v)% and 12 (w/v)% solutions of poly([Pyr 11 ][C(CN) 3 ]) and poly([Pyr 11 ][NTf 2 ]), respectively, in the most suitable solvents and the addition of the respective IL amounts ( Table 1). The solutions were mixed using a magnetic stirrer until complete dissolution of the PIL and IL components and then poured into Petri dishes for slow evaporation of the solvent at room temperature. With the aim of obtaining homogeneous membranes, the solvent evaporation took place slowly, for 2/3 days, depending on the solvent used (Table 1), and in a saturated solvent environment. The thicknesses of the prepared membranes (70-210 µm) were measured using a digital micrometer (Mitutoyo, model MDE-25PJ, Kanagawa, Japan). Average thickness was calculated from six measurements taken at different locations of each PIL-IL membrane. All the PIL-IL membranes studied were considered to have good stability since they were malleable and flexible enough to be used in the gas permeation measurements, even for the composites with 60 wt % of IL. Moreover, the evaluation of the mechanical stability of the PIL-IL composite membranes having the [C(CN) 3 ] − anion was recently reported by Tomé et al. [44] (Young's modulus (PIL-40IL)~14 MPa; Young's modulus (PIL-60IL)~5).

Gas Permeation Experiments
A time lag equipment described in detail elsewhere [38] was used to measure and determine the ideal CO 2 and H 2 permeabilities and diffusivities through the prepared PIL-IL composites. Initially, each membrane was degassed under vacuum inside the permeation cell for at least 12 h before testing. The gas permeation experiments were performed at 293 K and 308 K with an upstream pressure of 100 kPa (feed) and vacuum (<0.1 kPa) as the initial downstream pressure (permeate). Three separate CO 2 and H 2 experiments on a single composite membrane were measured. Between each run, the permeation cell and lines were evacuated until the pressure was below 0.1 kPa.
The gas transport through the PIL-IL membranes was assumed to occur according to the solution-diffusion mass transfer mechanism [45]. Thus, the permeability (P) is related to diffusivity (D) and solubility (S) as follows: The permeate flux of each studied gas (J i ) was experimentally determined using Equation (2) and assuming an ideal gas behavior and a homogeneous membrane [46]: where V p is the permeate volume, ∆p d is the variation of downstream pressure, A is the effective membrane surface area, t is the experimental time, R is the gas constant, and T is the temperature. Equation (3) was then used to calculate the ideal gas permeability (P i ) from the pressure driving force (∆p i ) and membrane thickness ( ): Gas diffusivity (D i ) was determined according to Equation (4). The time-lag parameter (θ) was deduced by extrapolating the slope of the linear portion of the p d vs. t curve back to the time axis, where the intercept is equal to θ [47]: (4) After defining both P i and D i , the gas solubility (S i ) was also calculated using Equation (1). The ideal permeability selectivity (or permselectivity), α i/j , which can also be expressed as the product of the diffusivity selectivity and the solubility selectivity, was obtained by dividing the permeability of the more permeable species i to the permeability of the less permeable species j, as expressed in Equation (5):

Gas Permeability (P)
The CO 2 and H 2 permeabilities through the PIL-IL composite membranes that were studied are presented in Figure 2. The CO 2 permeability was always higher than that of H 2 and both permeabilities increased with increasing temperature, although the increment was not the same between the studied composites, varying from 15 to 50% for CO 2 permeability values and from 39 to 77% for H 2 permeabilities. The CO 2 permeabilities at 293 K for all the membranes discussed here are in good agreement with those already reported [38,39,41], which emphasizes the high reproducibility of the method used. As expected, the incorporation of high amounts of IL led to enhanced CO 2 and H 2 permeabilities. Additionally, at 308 K, the temperature of bioH 2   Error bars represent standard deviations based on three experimental replicas. In some cases, the standard deviations are very small leading to error bars that cannot be clearly represented.

Gas Diffusivity (D)
The experimental gas diffusivity results at 293 K and 308 K through the prepared membranes are listed in Table 2. A high difference (one or, in some cases, two orders of magnitude) between CO2 and H2 diffusivity values, which corresponds to CO2/H2 diffusivity selectivities around 0.1, was observed. This difference in gas diffusivities was mainly due to the smaller size of H2 (2.89 Å) compared to CO2 kinetic diameter (3.30 Å) [48]. Moreover, both CO2 and H2 diffusivity increased with increasing temperature and with increasing IL content in the PIL-IL composite ( Table 2). The same behavior was also found for CO2 and H2 permeabilities (Figure 2). From Table 2, it can also be seen that CO2 and H2 diffusivities through the prepared membranes can be ordered as follows: PIL NTf2- 40 3], which means that the presence of the [C(CN)3] − anion in the composites, either in the PIL or IL's structure, leads to higher CO2 and H2 diffusivities compared to the [NTf2] − anion. The same trend was also observed for N2 diffusivities [38,39,41]. Thus, and as expected, it can be concluded that gas diffusivities follow the order of the kinetic diameters CO2 < N2 < H2. It can also be noted that the presence of imidazolium-based cation ([C2mim] + ) in the ILs led to superior gas diffusivities compared to the pyrrolidinium-based cation ([C4mpyr] + ).
Another interesting point is the comparison between gas permeability and diffusivity behaviors. Regardless of the anion, although the composite that comprised 60 wt% of IL had the highest H2 diffusivities (>1200 m 2 s −1 at 308 K), it did not present the highest H2 permeabilities (Figure 2). An equivalent behavior was also obtained for the PIL NTf2-40 [C4mpyr][NTf2] composite membrane, which displayed the lowest H2 diffusivities (546 m 2 s −1 at 308 K) but not the lowest H2 permeabilities. In the case of CO2, it can be seen from Table 2 and Figure 2 that CO2 permeability followed the same trend as CO2 diffusivity, with the exception of the PIL NTf2-60 [C4mpyr][NTf2] and PIL C(CN)3-40 [C2mim][C(CN)3] membranes. This behavior led us to conclude that the very high difference (three or, in some cases, four orders of magnitude) among H2 diffusivities is somehow compensated by a reverse behavior in H2 solubilities (as will be discussed in the next section), which has a significant impact on the H2 permeability results.

Gas Diffusivity (D)
The experimental gas diffusivity results at 293 K and 308 K through the prepared membranes are listed in Table 2. A high difference (one or, in some cases, two orders of magnitude) between CO 2 and H 2 diffusivity values, which corresponds to CO 2 /H 2 diffusivity selectivities around 0.1, was observed. This difference in gas diffusivities was mainly due to the smaller size of H 2 (2.89 Å) compared to CO 2 kinetic diameter (3.30 Å) [48]. Moreover, both CO 2 and H 2 diffusivity increased with increasing temperature and with increasing IL content in the PIL-IL composite ( Table 2). The same behavior was also found for CO 2 and H 2 permeabilities (Figure 2). From Table 2, it can also be seen that CO 2 and H 2 diffusivities through the prepared membranes can be ordered as follows: PIL NTf 2 -40 [C 4 3 ], which means that the presence of the [C(CN) 3 ] − anion in the composites, either in the PIL or IL's structure, leads to higher CO 2 and H 2 diffusivities compared to the [NTf 2 ] − anion. The same trend was also observed for N 2 diffusivities [38,39,41]. Thus, and as expected, it can be concluded that gas diffusivities follow the order of the kinetic diameters CO 2 < N 2 < H 2 . It can also be noted that the presence of imidazolium-based cation ([C 2 mim] + ) in the ILs led to superior gas diffusivities compared to the pyrrolidinium-based cation ([C 4 mpyr] + ).

mpyr][NTf 2 ] < PIL NTf 2 -60 [C 4 mpyr][NTf 2 ] < PIL NTf 2 -40 [C 2 mim][NTf 2 ] < PIL C(CN) 3 -40 [C 2 mim][C(CN) 3 ] < PIL C(CN) 3 -60 [C 2 mim][C(CN)
Another interesting point is the comparison between gas permeability and diffusivity behaviors. Regardless of the anion, although the composite that comprised 60 wt % of IL had the highest H 2 diffusivities (>1200 m 2 s −1 at 308 K), it did not present the highest H 2 permeabilities (Figure 2). An equivalent behavior was also obtained for the PIL NTf 2 -40 [C 4 mpyr][NTf 2 ] composite membrane, which displayed the lowest H 2 diffusivities (546 m 2 s −1 at 308 K) but not the lowest H 2 permeabilities. In the case of CO 2 , it can be seen from Table 2 and Figure 2 that CO 2 permeability followed the same trend as CO 2 diffusivity, with the exception of the PIL NTf 2 -60 [C 4 3 ] membranes. This behavior led us to conclude that the very high difference (three or, in some cases, four orders of magnitude) among H 2 diffusivities is somehow compensated by a reverse behavior in H 2 solubilities (as will be discussed in the next section), which has a significant impact on the H 2 permeability results. Table 2. Experimental gas diffusivities (D) through the studied PIL-IL membranes at T = 293 K and T = 308 K.

Gas Solubility (S)
Gas solubility (S) values were calculated using Equation (1) at temperatures of 293 K and 308 K and are displayed in Figure 3. The CO 2 solubility decreased with increasing temperature while H 2 solubility increased with increasing temperature for all the PIL-IL membranes studied. Analogous reverse H 2 solubility behavior with temperature was also found and discussed by Raeissi et al. [49] in imidazolium-based ILs, such as [C 4 mim][NTf 2 ], which means that hydrogen dissolves better at higher than at lower temperatures. This trend seems to be the general rule for H 2 solubility in ILs [49][50][51][52][53] and has been attributed to the extreme lightness and small intermolecular forces of hydrogen molecules [49].

CO2/H2 Separation Performance
The CO2 and H2 permeabilities and the ideal CO2/H2 permselectivities determined at 293 K and 308 K are summarized in Table 3. Amongst the PIL−IL membranes studied, those bearing the [C(CN)3] − anion revealed slightly higher CO2/H2 permselectivities than those containing the [NTf2] − anion. This behavior was also observed in our previous works concerning the use of PIL-IL membranes for CO2/N2 and CO2/CH4 separation [38,39,41]. Moreover, from Table 3, it can be seen that CO2/H2 permselectivities decreased with increasing temperature. This result can be explained by the variations in CO2/H2 solubility selectivity with temperature, which leads to a decrease in CO2/H2 permselectivity as the temperature increases [55]. In fact, CO2/H2 solubility selectivity through the studied composite membranes decreased from 78-145 at 293 K to 42-84 at 308 K. It can also be emphasized that CO2/H2 permselectivity seems to be controlled by a solubility mechanism, since CO2/H2 diffusivity selectivity (D CO2/H2) values were approximately equal to 0.1 at both temperatures while solubility selectivity (S CO2/H2) values ranged from 78-145 at 293 K and 42-84 at 308 K. Table 3. Single gas permeabilities (P) a and ideal permselectivities (α) of the PIL-IL membranes b Figure 3. Gas solubilities (S) for the studied PIL-IL membranes at 293 K and 308 K.
It can also be observed that, as expected, both CO 2 and H 2 solubilities were enhanced with the incorporation of high amounts of IL in the composite. For instance, when the amount of [C 2 mim][C(CN) 3 ] increased from 40 to 60 wt %, the CO 2 and H 2 solubilities at 293 K increased almost 50% and 15%, respectively, while at 308 K the increment in CO 2 and H 2 solubilities was around 39% and 5%, respectively. Similar behavior was found for the PIL-IL composites comprising the [C 4 mpyr][NTf 2 ] IL. Moreover, the large differences between CO 2 and H 2 solubilities can be explained by the high CO 2 critical temperature (CO 2 , 31 • C; H 2 , −240 • C), corresponding to the superior condensability of CO 2 (T ε/k = 195.2 K) compared to H 2 (T ε/k = 59.7 K) [48,54]. The fact that H 2 can almost be considered an ideal gas due to its small kinetic diameter and non-interacting nature, while CO 2 displays a higher kinetic diameter and a quadrupole moment, also plays a role in the difference in solubilities of the two gases. For the PIL-IL composites studied in this work at 308 K, the CO 2 solubility ranged from 14 to 28.5 (×10 −6 ) m 3 (STP) ·m −3 ·Pa −1 whereas the H 2 solubility values were two orders of magnitude lower, varying from 0.17 to 0.51 (×10 −6 ) m 3 (STP) ·m −3 ·Pa −1 . Among all the tested membranes, the PIL NTf 2 -60 [C 4 mpyr][NTf 2 ] composite presented the highest CO 2 and H 2 solubilities at both 293 and 308 K. Regarding the influence of the anion's structure and considering the same amount of free IL in the composite, it can be concluded that the presence of the [NTf 2 ] − anion in the PIL-IL membranes leads to higher CO 2 and H 2 solubilities compared to those membranes comprising the [C(CN) 3 ] − anion. As mentioned before, this behavior masks the higher H 2 diffusivities of composites with a cyano-functionalized anion, somehow explaining the low influence of H 2 diffusivity on the H 2 permeability results.

CO 2 /H 2 Separation Performance
The CO 2 and H 2 permeabilities and the ideal CO 2 /H 2 permselectivities determined at 293 K and 308 K are summarized in Table 3. Amongst the PIL−IL membranes studied, those bearing the [C(CN) 3 ] − anion revealed slightly higher CO 2 /H 2 permselectivities than those containing the [NTf 2 ] − anion. This behavior was also observed in our previous works concerning the use of PIL-IL membranes for CO 2 /N 2 and CO 2 /CH 4 separation [38,39,41]. Moreover, from Table 3, it can be seen that CO 2 /H 2 permselectivities decreased with increasing temperature. This result can be explained by the variations in CO 2 /H 2 solubility selectivity with temperature, which leads to a decrease in CO 2 /H 2 permselectivity as the temperature increases [55]. In fact, CO 2 /H 2 solubility selectivity through the studied composite membranes decreased from 78-145 at 293 K to 42-84 at 308 K. It can also be emphasized that CO 2 /H 2 permselectivity seems to be controlled by a solubility mechanism, since CO 2 /H 2 diffusivity selectivity (D CO 2 /H 2 ) values were approximately equal to 0.1 at both temperatures while solubility selectivity (S CO 2 /H 2 ) values ranged from 78-145 at 293 K and 42-84 at 308 K. Table 3. Single gas permeabilities (P) a and ideal permselectivities (α) of the PIL-IL membranes studied b .

Gas Permeability (Barrer) (T = 293 K)
Gas Permeability (Barrer) (T = 308 K) P CO 2 ± σ P H 2 ± σ α CO 2 /H 2 P CO 2 ± σ P H 2 ± σ α CO 2 /H 2 PIL C(CN) 3  The gas separation performance of the studied PIL-IL membranes is shown in Figure 4, where the CO 2 /H 2 permselectivity is plotted against the permeability of the more permeable gas species (CO 2 ). This graph displays a tradeoff (black line) between gas permeability and permselectivity. These upper bound limits for several gas pairs were first developed by Robeson [56] who correlated data obtained from measurements carried out with polymeric membranes at low temperatures (298-308 K). Later, Rowe et al. [55] studied the influence of temperature on the tradeoff between gas permeability and permselectivity for different gas pairs. Thus, the upper bound at 300 K developed by Rowe et al. [55] for the CO 2 /H 2 gas pair is represented in Figure 4 and was used to evaluate the performance of the studied PIL-IL membranes for biohydrogen purification (T = 308 K and 100 kPa). Figure 4 clearly shows that all the PIL-IL membranes that were studied displayed CO 2 /H 2 separation performances above the upper bound. The best CO 2 /H 2 separation performance was obtained for the membrane composed of poly([Pyr 11 ][C(CN) 3 ]) and 60 wt % of [C 2 mim][C(CN) 3 ] IL, which is in agreement with what has been observed in our recent works regarding the use of PIL-IL composites for CO 2 /N 2 separation [39]. Literature data points for other reported PIL-IL membranes are also plotted in Figure 4 for comparison. The gas permeation measurements reported by Carlisle et al. [33] were performed at room temperature with a transmembrane pressure differential of 200 kPa. Also, their PIL-IL membranes were not prepared by the solvent casting method but through UV polymerization by mixing different percentages of imidazolium-based IL monomers, a cross-linking monomer, and free IL [33]. From Figure 4, it can be seen that the PIL-IL membranes reported in the literature also present CO 2 /H 2 separation performances above the upper bound for the CO 2 /H 2 gas pair at 300 K, but the PIL C(CN) 3 -60 [C 2 mim][C(CN) 3 ] membrane studied in this work still revealed superior CO 2 /H 2 separation performance. Figure 4 clearly shows that all the PIL-IL membranes that were studied displayed CO2/H2 separation performances above the upper bound. The best CO2/H2 separation performance was obtained for the membrane composed of poly([Pyr11][C(CN)3]) and 60 wt% of [C2mim][C(CN)3] IL, which is in agreement with what has been observed in our recent works regarding the use of PIL-IL composites for CO2/N2 separation [39]. Literature data points for other reported PIL-IL membranes are also plotted in Figure 4 for comparison. The gas permeation measurements reported by Carlisle et al. [33] were performed at room temperature with a transmembrane pressure differential of 200 kPa. Also, their PIL-IL membranes were not prepared by the solvent casting method but through UV polymerization by mixing different percentages of imidazolium-based IL monomers, a cross-linking monomer, and free IL [33]. From Figure 4, it can be seen that the PIL-IL membranes reported in the literature also present CO2/H2 separation performances above the upper bound for the CO2/H2 gas pair at 300 K, but the PIL C(CN)3-60 [C2mim][C(CN)3] membrane studied in this work still revealed superior CO2/H2 separation performance.  [55]. Literature data points ( ) from other reported PIL-IL membranes are also displayed for comparison [33].

Conclusions
In this work, dense composite membranes made of pyrrolidinium-based PILs with [C (CN) incorporated were prepared by the solvent casting method. Their CO2 and H2 permeation properties (permeability, diffusivity, and solubility) were determined at biohydrogen production conditions (T = 308 K and 100 kPa of feed pressure) and discussed. The CO2 and H2 permeation properties were measured at T = 293 K and the effect of temperature on gas separation performance was evaluated. Data are plotted on a log-log scale and the upper bound at 300 K was adapted from Rowe et al. [55]. Literature data points (   Figure 4 clearly shows that all the PIL-IL membranes that were studied displayed CO2/H2 separation performances above the upper bound. The best CO2/H2 separation performance was obtained for the membrane composed of poly([Pyr11][C(CN)3]) and 60 wt% of [C2mim][C(CN)3] IL, which is in agreement with what has been observed in our recent works regarding the use of PIL-IL composites for CO2/N2 separation [39]. Literature data points for other reported PIL-IL membranes are also plotted in Figure 4 for comparison. The gas permeation measurements reported by Carlisle et al. [33] were performed at room temperature with a transmembrane pressure differential of 200 kPa. Also, their PIL-IL membranes were not prepared by the solvent casting method but through UV polymerization by mixing different percentages of imidazolium-based IL monomers, a cross-linking monomer, and free IL [33]. From Figure 4, it can be seen that the PIL-IL membranes reported in the literature also present CO2/H2 separation performances above the upper bound for the CO2/H2 gas pair at 300 K, but the PIL C(CN)3-60 [C2mim][C(CN)3] membrane studied in this work still revealed superior CO2/H2 separation performance.  [55]. Literature data points ( ) from other reported PIL-IL membranes are also displayed for comparison [33].

Conclusions
In this work, dense composite membranes made of pyrrolidinium-based PILs with [C (CN) incorporated were prepared by the solvent casting method. Their CO2 and H2 permeation properties (permeability, diffusivity, and solubility) were determined at biohydrogen production conditions (T = 308 K and 100 kPa of feed pressure) and discussed. The CO2 and H2 permeation properties were measured at T = 293 K and the effect of temperature on gas separation performance was evaluated.
) from other reported PIL-IL membranes are also displayed for comparison [33].

Conclusions
In this work, dense composite membranes made of pyrrolidinium-based PILs with [C(CN) 3  incorporated were prepared by the solvent casting method. Their CO 2 and H 2 permeation properties (permeability, diffusivity, and solubility) were determined at biohydrogen production conditions (T = 308 K and 100 kPa of feed pressure) and discussed. The CO 2 and H 2 permeation properties were measured at T = 293 K and the effect of temperature on gas separation performance was evaluated.
The PIL-IL membranes containing the [NTf 2 ] − anion presented the highest H 2 permeability and solubility values, while the PIL-IL composites having the [C(CN) 3 ] − anion showed the highest H 2 diffusivities and CO 2 /H 2 permselectivities. As previously reported, increments in gas permeabilities, diffusivities, and solubilities were observed with increasing temperature and amounts of IL, with the exception of H 2 solubility that showed the opposite behavior with temperature compared to what occurred in CO 2 solubility. Overall, all the PIL-IL membranes studied revealed similar or superior CO 2 /H 2 separation performance compared to the few PIL-IL composites reported so far in the literature. Particularly, at 308 K, the best result was obtained through the PIL C(CN) 3 -60 IL