Tunable Anion Exchange Membrane Conductivity and Permselectivity via Non-Covalent, Hydrogen Bond Cross-Linking

Abstract

Ion exchange membranes (IEMs) are a key component of electrochemical processes that purify water, generate clean energy, and treat waste. Most conventional polymer IEMs are covalently cross-linked, which results in a challenging tradeoff relationship between two desirable properties—high permselectivity and high conductivity—in which one property cannot be changed without negatively affecting the other. In an attempt to overcome this limitation, in this work we synthesized a series of anion exchange membranes containing non-covalent cross-links formed by a hydrogen bond donor (methacrylic acid) and a hydrogen bond acceptor (dimethylacrylamide). We show that these monomers act synergistically to improve both membrane permselectivity and conductivity relative to a control membrane without non-covalent cross-links. Furthermore, we show that the hydrogen bond donor and acceptor loading can be used to tune permselectivity and conductivity relatively independently of one another, escaping the tradeoff observed in conventional membranes.

Graphical Abstract

1. INTRODUCTION

Ion exchange membranes (IEMs) are a key component of electrochemical processes that purify water, generate clean energy, and treat waste.^1–6 It is desirable for IEMs to be both selective and conductive to certain ions (counter-ions), and these properties are achieved through a high concentration of charged sites covalently bound to a polymer backbone.^7–9 However, the charged hydrophilic sites cause the polymer to swell excessively in water, thus necessitating incorporation of covalent cross-links to ensure mechanical stability.^2,9–13 The introduction of covalent cross-links reduces the conductivity of the membrane by introducing obstacles to ion permeation,^10,11 thus imposing a challenging tradeoff between high selectivity (favored by high charge concentration) and high conductivity (favored by lower cross-linking).^14–16

Polymer networks can also be formed using non-covalent cross-links. Non-covalent cross-links mediated by exchangeable bonds (“dynamic cross-links”),¹⁷ ionic interactions (“ionic cross-links”),^18,19 or hydrogen bonding^20,21 can provide an alternate means of enhancing mechanical strength. Non-covalent cross-linking mechanisms have also been recognized in the renowned Nafion IEM: physical cross-links mediated by non-covalent interactions between its crystalline domains and ionic cross-links between charged sites have a strong influence on its thermal and mechanical properties.^22,23 Generally speaking, non-covalent cross-links increase the strength and toughness of polymer networks because they can break and reform or “self-heal” in response to stress.^17–21 For example, Hu et al.²¹ used non-covalent cross-links formed by hydrogen bond donor–acceptor pairs to create a hydrogel with an impressive combination of strength, stiffness, and toughness that was unprecedented in covalently cross-linked hydrogels. The remarkable strength of this hydrogel was attributed to the continual breaking and re-forming of hydrogen bond cross-links within the network. Furthermore, the hydrogels synthesized by Hu et al.²¹ maintained structural integrity in aqueous solutions, which is the environment in which most IEMs operate.

Inspired by the work of Hu et al.,²¹ we sought to investigate whether non-covalent (hydrogen bond) cross-linking could benefit the performance of IEMs. We hypothesized that adding non-covalent (hydrogen bond) cross-links could result in novel combinations of conductivity, selectivity, and strength that are inaccessible to conventional, covalently cross-linked membranes. For example, a transient hydrogen bond cross-link in the “broken” state might present less of an obstacle to ion conduction than a covalent cross-link, thus improving membrane conductivity without sacrificing physical integrity. In addition, hydrogen bond cross-links may affect other relevant phenomena mediated by hydrogen bonding, such as counter-ion hydration and the balance of free versus bound water within the membrane, which are known to play a central role in determining ion transport properties.^23–31

Accordingly, our objective in this work was to study the effects of non-covalent cross-linking on the performance of IEMs. We synthesized anion exchange membranes incorporating methacrylic acid (MAAc) as a hydrogen bond donor and dimethylacrylamide (DMAA) as a hydrogen bond acceptor to promote non-covalent interactions within the membranes. MAAc is known to function as a proton donor, forming hydrogen bonds with acceptor moieties,^21,32 and has also been shown to exhibit strong hydrophobic interactions with ether oxygen groups (which are also present in our membranes).³² DMAA is known to be a potent hydrogen bond acceptor.^21,33 To enable a systematic study of the influence of non-covalent cross-links, we varied the amounts of hydrogen bond donor groups and hydrogen bond acceptor groups while maintaining an identical polymer backbone with constant charge concentration and cross-link density. We quantified membrane performance via conductivity and permselectivity measurements in sodium chloride solutions to reflect application conditions relevant to environmental separations (e.g., electrodialysis). The effects of the hydrogen bond donor and acceptor content on water dynamics and phase-separation within the membranes were investigated via differential scanning calorimetry (DSC), atomic force microscopy (AFM), and small-angle X-ray scattering (SAXS). We show that varying the loadings of MAAc and DMAA in the membranes allows conductivity and permselectivity to be partially decoupled, suggesting that engineering non-covalent interactions may be a promising strategy for designing membranes that escape the permselectivity–conductivity tradeoff.

2. RESULTS AND DISCUSSION

2.1. Anion Exchange Membranes with Non-Covalent Cross-Linking Moieties.

Membranes in this study contained three groups of monomers (Figure 1): (1) covalent cross-linker poly(ethylene glycol)diacrylate or PEGDA (at a molar ratio X); (2) neutral ester (methyl methacrylate or MMA), hydrogen bond donor (MAAc), and/or hydrogen bond acceptor (DMAA), all of which are monoacrylates (at total molar ratio Y); and (3) cationic charged group [(methacryloyloxy)ethyl]trimethylammonium chloride or MAOTMAC (at a molar ratio Z). The molar ratios of these three components were kept fixed in all membranes (X/Y/Z = 1.2: 1: 1). Only the relative amounts of the hydrogen bond donor (MAAc), hydrogen bond acceptor (DMAA), and neutral ester (MMA) varied, keeping the total molar ratio Y fixed. Note that by design, all membranes share an identical C–C backbone with C=O groups immediately adjacent to the backbone. By fixing the X and Z values, we ensured that all membranes also had the same covalent cross-linking density and the same molar ratio of fixed charge groups to other monomers. The X, Y, and Z values we adopted resulted in a very narrow range of theoretical ion exchange capacity (IEC) of 2.18–2.25 mequiv g⁻¹, which varies slightly with the membrane composition because the respective monomers have different molecular weights, even though the molar ratio of charged sites to monomers was constant for all membranes. By keeping the same covalent cross-linking density and IEC across membranes, we ensured that differences in the performance were primarily a result of non-covalent cross-linking (hydrogen bond) behavior.

Figure 1.

Synthesis of anion exchange membranes with physical cross-links mediated by hydrogen bonding. (a) Photocross-linkable monomers; (b) membrane naming scheme; and (c) functions of the various monomers. PEGDA is the covalent cross-linker, MMA is the neutral ester, MAAc is the hydrogen bond donor, DMAA is the hydrogen bond acceptor, and MAOTMAC is the cationic monomer. A molar ratio X/Y/Z = 1.2:1:1 was used in the synthesis of all membranes.

The molar compositions of all membranes prepared in this study are listed in Table 1. Each membrane was designated D#–A#–E#, where the #s following D, A, and E represent the molar percentages of Y occupied by the hydrogen bond donor, acceptor, and neutral ester, respectively. Thus, in membrane D25–A25–E50, MAAc (donor) and DMAA (acceptor) each comprises 25% of the total monoacrylates (Y), and MMA (neutral ester) comprises the remaining 50%. After UV-initiated photopolymerization, all compositions formed flexible, freestanding membranes that were optically clear immediately after synthesis and became translucent and somewhat cloudy upon immersion in water. Photographs illustrating this behavior are provided in the Supporting Information. The cloudiness that we observed may indicate nanoscale phase separation within the polymer^34,35 and is a topic we explore further in a later section. The gel fraction of the membranes we prepared ranged from 81 to 86%, with an average value of 84% (Table 1), indicating that the majority of the monomers in our mixtures were incorporated into the cross-linked polymer network and that all compositions we prepared were similarly reactive. The unreacted monomer was removed from the membranes prior to further testing (see Section 4.2).

Table 1.

Molar Composition of Pre-Polymerization Mixtures Used to Synthesize Membranes in This Work

sample	theoretical IEC	% of Y that is donor or acceptor	PEGDAa (X)	MMAa (Y)	MAAca (Y)	DMAAa (Y)	MAOTMACa (Z)	total monomer concentrationb before polymerization	gel fraction
	(mequiv. g−1)		(mol × 2)	(E, mol)	(D, mol)	(A, mol)	(mol)	(wt %)	(%)
Control Membrane
D0–A0–E100	2.18	0	1.2	1	0	0	1	73	86 ± <1%
Membranes Containing Donor Without Acceptor
D50–A0–E50	2.22	50	1.2	0.5	0.5	0	1	73	82 ± <1%
D100–A0–E0	2.25	100	1.2	0	1.0	0	1	73	83 ± <1%
Membranes Containing Acceptor Without Donor
D0–A50–E50	2.19	50	1.2	0.5	0	0.5	1	73	85 ± <1%
D0–A100–E0	2.19	100	1.2	0	0	1.0	1	73	85 ± 1%
Membranes Containing Both Donor And Acceptor
D13–A13–E74	2.19	25	1.2	0.75	0.125	0.125	1	73	84 ± 1%
D25–A25–E50	2.20	50	1.2	0.5	0.25	0.25	1	73	82 ± 3%
D38–A38–E24	2.21	75	1.2	0.25	0.375	0.375	1	73	84 ± <1%
D50–A50–E0	2.22	100	1.2	0	0.5	0.5	1	73	84 ± 1%
D25–A75–E0	2.20	100	1.2	0	0.25	0.75	1	73	83 ± <1%
D75–A25–E0	2.24	100	1.2	0	0.75	0.25	1	73	81 ± <1%

^aPEGDA is the covalent cross-linker, MMA is the neutral ester, MAAc is the hydrogen bond donor, DMAA is the hydrogen bond acceptor, and MAOTMAC is the cationic monomer. The moles of PEGDA listed in the table represent moles of polymerizable groups. Since PEGDA is difunctional, the values listed are 2× the number of moles of PEGDA molecules in the pre-polymerization mixture.

^b“Total monomer concentration” refers to the combined concentration of PEGDA, MMA, MAAc, DMAA, and MAOTMAC. The remaining 27% is accounted for by the solvent.

Fourier transform-infrared (FTIR) spectra of the dry membranes (Figure 2) further confirmed that hydrogen bond donor (MAAc) and hydrogen bond acceptor (DMAA) monomers were incorporated into the polymer structure. Individual monomer spectra, as shown in the Supporting Information, revealed a distinctive band at 1690 cm⁻¹ for MAAc, characteristic of C=O stretching in unsaturated carboxylic acids.³⁶ DMAA exhibited two distinctive absorption bands at 1647 and 1610 cm⁻¹. The first is characteristic of C=O stretching in tertiary amides,³⁶ while the origin of the second band is not clear. PEGDA (cross-linker), MAOTMAC (charged monomer), and MMA (neutral ester) all displayed a strong absorption band at ~1720 cm⁻¹, which corresponds to C=O stretching in unsaturated esters.^36–40 In addition, MAOTMAC showed a strong, sharp band at 956 cm⁻¹, which is characteristic of the quaternary ammonium headgroup.⁴¹

Figure 2.

FTIR spectra of synthesized anion exchange membranes. (a) Complete spectra of all membranes. Vertical lines are placed at 1720 cm⁻¹ (C=O stretch in PEGDA, MMA, and MAOTMAC), 1690 cm⁻¹ (C=O stretch in MAAc), 1647 cm⁻¹ (C=O stretch in DMAA), 1610 cm⁻¹ (unidentified band associated with DMAA), and 956 cm⁻¹ (quaternary ammonium headgroups in MAOTMAC). (b) Enlargement of the C=O stretching region for a subset of the membranes. In this subset, the proportion of MAAc/DMAA varied from 0 to 1 (these membranes contain no MMA). The incorporation of MAAc (hydrogen bond donor) into the membranes can be seen in the broadening of the band at ~1720 cm⁻¹ toward lower wavenumbers; incorporation of DMAA is shown by the emergence of a band between 1647 and 1610 cm⁻¹. Band positions were identified via spectra of individual monomers and binary mixtures containing only two monomers (see the Supporting Information).

The incorporation of MAAc into membranes was indicated by the broadening of the ~1720 cm⁻¹ band toward lower frequencies, reflecting the combination of C=O stretching from the unsaturated esters (1720 cm⁻¹) and the carboxylic acid (1690 cm⁻¹). The presence of DMAA in the membranes resulted in the emergence of an additional absorption band between 1647 and 1610 cm⁻¹ that was not visible in membranes without DMAA. This band is associated with the amide group.⁴² All membranes exhibited a strong, sharp band at 956 cm⁻¹ with a similar intensity, confirming the incorporation of the MAOTMAC monomer,⁴¹ and weak, broad absorption bands in the O–H stretching (3200–3400 cm⁻¹) region normally associated with hydrogen bonding.³⁶ The lack of a stronger signal in this region may be a consequence of the fact that the membranes were scanned in the dry state.

2.2. Permselectivity, Conductivity, and Water Uptake Properties.

2.2.1. Performance of Membranes Containing Only Donor or Acceptor.

To evaluate the effects of non-covalent cross-linking by the hydrogen bond donor (MAAc) and hydrogen bond acceptor (DMAA) on membrane performance, we first consider membranes in which MAAc and DMAA were added individually, relative to the control membrane D0–A0–E100 which contained neither donor nor acceptor. Neutral ester (MMA) was substituted with MAAc or DMAA to various degrees, such that the total monoacrylate content of every membrane remained constant (Y = 1; see Table 1). As noted in the introduction section, in addition to donating protons to form hydrogen bonds, MAAc is also known to form interpolymer complexes with ether oxygen groups (e.g., on PEGDA)^21,32 and can potentially complex with the oppositely charged MAOTMAC⁴³ which is another mechanism by which MAAC may promote non-covalent cross-linking. The water uptake, conductivity, and permselectivity of these membranes are summarized in Figure 3. Focusing first on water uptake, Figures 3a,b shows clearly that adding MAAc to the membranes reduced their water uptake relative to the control membrane, while adding DMAA increased it. Since the carboxyl and acrylamide groups on MAAc and DMAA, respectively, are both more hydrophilic than the methyl group on MMA, which is the only neutral monoacrylate present in the control membrane, one would expect addition of either MAAc or DMAA to increase the water uptake. The fact that adding MAAc decreased water uptake suggests that MAAc was not preferentially hydrogen bonding with water molecules, which would presumably cause additional water uptake. The fact that adding MAAc decreased water uptake despite its hydrophilicity⁴⁴ suggests that ionic and hydrogen bond interactions between MAAc and the polymer^{18–21,32,45} reduced the number of sites available for hydrogen bonding with water. Instead, MAAc may have formed hydrogen bonds with itself (see Figure 1c) or with ether oxygen groups on the PEGDA cross-linker. MAAc–O complexes, arising from hydrogen bonding and further stabilized by strong hydrophobic effects (i.e., structuring of water around the COOH–O moieties), have been shown to form between the polymer analogue of MAAc—poly(methacrylic acid) or PMAA—and PEG.³² Either type of bonding would have the effect of pulling adjacent polymer chains closer together since the O–H–O hydrogen bond would be much shorter than the length of the PEGDA cross-linker.

Figure 3.

Membrane water uptake (a,b), conductivity (c,d), and permselectivity (e,f) vs hydrogen bond donor (i.e., MAAc) or acceptor (i.e., DMAA) content for membranes containing only MAAc (left column) or DMAA (right column). Refer to the legend and Table 1 for the full composition of the membrane represented by each data point. For all membranes shown, the total mol of hydrogen bond donor (MAAc) + acceptor (DMAA) + neutral ester (MMA) was fixed at Y = 1 mol; (see Figure 1 and Table 1). Red color corresponds to the hydrogen bond donor (MAAc), and the lighter peach color corresponds to the hydrogen bond acceptor (DMAA). Error bars represent the standard error of at least five replicate measurements.

Figures 3c and 3d illustrates the effects of the hydrogen bond donor and acceptor, respectively, on conductivity. It is striking that adding donor (MAAc) to the membrane did not result in any substantial change in conductivity (Figure 3c), even though water uptake was reduced. In most IEMs, decreasing the water uptake is expected to reduce the conductivity.^16,46 In contrast to MAAc, the hydrogen bond acceptor (DMAA) increased conductivity relative to the control membrane (Figure 3d), which is what would be expected from the observed increase in the water uptake (Figure 3b).

A contrast in behavior between MAAc and DMAA is also evident in the plots of membrane permselectivity. Figure 3e shows that adding MAAc to the membranes substantially increased permselectivity, and this effect was not linear with the MAAc loading: the membrane containing 0.5 mol MAAc had almost the same permselectivity as the membrane containing 1.0 mol MAAc, and both were substantially higher than that of the control membrane containing only MMA. In contrast, DMAA had only a minor effect on permselectivity (Figure 3f). At 0.5 mol loading, the permselectivity was roughly equal to that of the control membrane, while at 1.0 mol loading, the permselectivity was slightly lower. The higher permselectivity of MAAc-containing membranes may be related to the reduction in the water uptake presumably caused by non-covalent interactions between MAAc and carbonyl or ether oxygen groups, as discussed above.

Overall, addition of a hydrogen bond donor (MAAc) resulted in a decrease in the water uptake and increase in permselectivity, but no change in conductivity. Addition of a hydrogen bond acceptor (DMAA) resulted in an increase in the water uptake and substantial increase in conductivity but only a slight decrease in permselectivity. Therefore, our results indicate that addition of hydrogen bond donors and/or acceptors may enable tuning conductivity and permselectivity relatively independently from one another and therefore escaping the conductivity–permselectivity tradeoff typically observed in IEMs.

2.2.2. Performance of Membranes Containing Both Donor and Acceptor.

We next consider the water uptake, conductivity, and permselectivity of the membranes containing both donor and acceptor, which are summarized in Figure 4. Two series of membranes are shown. In the first series (left column), hydrogen bond donor and acceptor (MAAc and DMAA, respectively) were introduced in a 1:1 ratio under the assumption that each donor–acceptor pair would form a single non-covalent cross-link (see Figure 1c). In this series, the neutral ester (MMA) was progressively replaced by equal amounts of donor and acceptor, such that the total monoacrylate content of every membrane remained constant (Y = 1; see Table 1). As such, the last membrane in the series (D50–A50–E0) contained 0.5 mol each of donor and acceptor and 0 mol MMA. In the second series (right column), we held the total loading of donor + acceptor constant at Y = 1 mol (with 0 mol MMA) and varied the donor/acceptor ratio. Thus, the monoacrylate content of this subset of membranes varied from 0 mol acceptor: 1 mol donor to 1 mol donor: 0 mol acceptor, and none contained neutral ester (MMA).

Figure 4.

Membrane water uptake (a,b), conductivity (c,d), and permselectivity (e,f) of membranes containing variable amounts of hydrogen bond donor (i.e., MAAc) and acceptor (i.e., DMAA). The left column (a,c,e) shows membranes in which donor and acceptor were added in a 1:1 ratio, while the right column (b,d,f) shows membranes in which the ratio of donor to acceptor varied. Refer to the legend and Table 1 for the composition of the membrane represented by each data point. For all membranes shown, the total mol of cross-linker PEGDA (X), monoacrylates (Y), and charged monomer MAOTMAC (Z) are fixed (X = 1.2 mol, Y = 1 mol, Z = 1 mol; see Figure 1 and Table 1). Some membranes appear in both columns. Red color corresponds to the hydrogen bond donor (MAAc), and peach color corresponds to the hydrogen bond acceptor (DMAA). Error bars represent the standard error of at least five replicate measurements.

Focusing first on the water uptake, we see in Figure 4a that when donor and acceptor were added in a 1:1 ratio, the water uptake increased slightly compared to the control membrane, before decreasing substantially as the total donor + acceptor loading approached 1 mol. With a variable donor/acceptor ratio (Figure 4b), water uptake generally increased relative to the control membrane as the proportion of the acceptor increased, although membrane D50–A50–E0 was an exception. Confirming what was observed when the donor (MAAc) was added by itself (Figure 3a), Figure 4b suggests that MAAc reduced the water uptake. This result could be explained by hydrogen bonding between MAAc–DMAA or by non-covalent interactions between MAAc and other carbonyl or ether oxygen groups (see Figure 1c), any of which may create non-covalent cross-links that help counteract the swelling pressure in the membrane. In contrast, hydrogen bond acceptor (DMAA) appears to increase the water uptake, as indicated by the fact that membrane D0–A100–E0 (containing DMAA without MAAc or MMA) had the highest water uptake of any membrane shown.

This finding can be explained by considering that the DMAA molecule is more hydrophilic than MMA but (unlike MAAc) lacks the ability to form hydrogen bonds in the absence of a proton donor. Thus, the non-monotonic dependence of water uptake on donor/acceptor loading in Figure 4a can be interpreted as the result of competitive effects of the donor and acceptor: at higher donor loadings, the tendency of MAAc to reduce water uptake outweighs the tendency of DMAA to increase it.

It is interesting to compare the individual effects of the donor and acceptor to their combined effect on water uptake. By itself, the donor (MAAc) had a smaller effect on water uptake than the acceptor (DMAA): the decrease in water uptake between the control membrane and D100–A0–E0 (0.116 g H₂O per g, corresponding to the maximum MAAc content) was smaller than the increase in water uptake between the control membrane D0–A0–E100 and D0–A100–E0 (0.167 g H₂O per g, corresponding to the maximum DMAA content). Nevertheless, adding MAAc and DMAA in a 1:1 ratio (D50–A50–E0) had as net effect a reduction of the water uptake by 0.121 g H₂O per g compared to the control membrane (Figure 4a). Thus, the combined effects of MAAc and DMAA on the water uptake were not additive. This observation is consistent with the presence of non-covalent, hydrogen-bonded cross-links between the two monomers: acrylamide moieties on DMAA that are bonded to MAAc are not free to interact with water, so their ability to increase the membrane water uptake is suppressed. If the carboxyl and acrylamide moieties in the –D50–A50–E0 membrane were unbonded, the combined effects of MAAc and DMAA should have been roughly additive, and the water uptake of the D50–A50–E0 membrane would have increased.

Membrane conductivity increased in a roughly linear fashion relative to the control membrane as hydrogen bond donor and acceptor were added to the membrane in a 1:1 ratio (Figure 4c). As was the case when MAAc was added by itself (Figure 3a,c), this result was contrary to expectations considering that adding donor and acceptor also reduced the water uptake (Figure 4a). Normally, reducing the water uptake would be expected to reduce the conductivity, and vice versa.^16,46 On the other hand, the expected relationship between the water uptake and conductivity is exhibited by the series of membranes in Figures 4b and 4d, where increasing the proportion of hydrogen bond acceptor (DMAA) increased both the water uptake and conductivity. A possible explanation for these observations is that when MAAc is present by itself or in excess (i.e., without a sufficient number of DMAA hydrogen bond acceptors to pair with), its –COOH group ionizes to –COO⁻ and repels chloride counter-ions. Previous studies^47–50 have shown that some –COOH groups inside of polyamide reverse osmosis membranes dissociate with pK_a ~ 5.4, lower than the neutral pH solutions used here. Repelling counter-ions from the polymer may reduce intrapore energy barriers and lead to higher conductivity.⁵¹

The combined effects of the donor and acceptor on permselectivity are shown in Figure 4e,f. Adding donor and acceptor in a 1:1 ratio always improved membrane permselectivity relative to the control membrane (0.854–0.866 vs 0.831; Figure 4e); however, permselectivity did not increase monotonically with donor and acceptor concentration. Surprisingly, the membrane with the highest permselectivity (D25–A25–E0) had a water uptake that was slightly higher than that of the control membrane (0.871 vs 0.849 g H₂O g⁻¹; Figure 3a). In general, permselectivity is expected to decrease with increasing water uptake,¹⁶ contrary to observations in Figure 4e, so once again introduction of donor and acceptor in a 1:1 ratio resulted in novel behavior in these membranes. In Figure 4f, we see that membrane D100–A0–E0, which contained no hydrogen bond acceptor (DMAA), had the highest permselectivity of all (0.869) and that increasing the proportion of the acceptor generally reduced the permselectivity, consistent with Figure 3f.

Altogether, Figures 3 and 4 illustrate that certain combinations of donor and acceptor resulted in novel changes in membrane properties (e.g., decreases in water uptake accompanied by increases in conductivity) that defied expectations based on known tradeoff relationships. The hydrogen bond donor (MAAc) and acceptor (DMAA) exhibited contrasting behavior with respect to their effects on water uptake, conductivity, and permselectivity. Specifically, MAAc reduced water uptake and increased permselectivity without affecting conductivity (Figure 3c,e), while DMAA increased water uptake and conductivity without substantially decreasing permselectivity (Figure 3d,f). Furthermore, the effects of MAAc and DMAA when added together (Figure 4) were not an additive combination of their individual effects (Figure 3).

The combined effects of donor, acceptor, and neutral ester on membrane conductivity and permselectivity are summarized in Figure 5, which presents contour plots of each property as a function of the fractional monoacrylate composition (Y). Figure 5 illustrates clearly that MAAc-containing membranes (left edge) exhibit high permselectivity, while DMAA-containing membranes (lower right vertex) exhibit higher conductivity.

Figure 5.

Fractional monoacrylate (MMA, MAAc, or DMAA) composition vs permselectivity (left) and conductivity (right) for all membranes in this work. The fractional composition at a particular location within the plot is determined by reading the axis value in the direction indicated by the arrow next to each component. For example, horizontal lines correspond to different fractional amounts of MMA (neutral ester), which are labeled on the left axis. Downward-sloping lines at −60° from the horizontal correspond to different amounts of MAAc (donor), which are labeled on the bottom axis. Upward-sloping lines at 60° from the horizontal correspond to different amounts of DMAA (acceptor), which are labeled on the right axis. The vertices of each plot represent membranes that contain only MMA, MAAc, or DMAA in the monoacrylate block (e.g., membrane D100–A0–E0 is the lower left vertex).

2.2.3. Permselectivity–Conductivity Tradeoff in Membranes Containing Donor and Acceptor.

The unusual effects of donor and/or acceptor addition, as described in Sections 2.2.1 (Figure 3) and 2.2.2 (Figure 4), appear to allow MAAc and DMAA to be used to tune conductivity and permselectivity somewhat independently of the water uptake. For example, consider membranes D100–A0–E0 (solid red triangle) and D50–A50–E0 (half-filled circle). The difference between the two membranes is that the membrane D100–A0–E0 contained no hydrogen bond acceptor, while the membrane D50–A50–E0 contained 0.5 mol each of donor and acceptor. Both had essentially the same water uptake (~0.73 g H₂O per g; Figure 4a,b), but the D50–A50–E0 membrane had roughly twice the conductivity of the D100–A0–E0 membrane (5.78 mS cm⁻¹ vs 2.92 mS cm⁻¹), while the D100–A0–E0 membrane had higher permselectivity (0.869 vs 0.845). Thus, replacing some of the hydrogen bond donor (MAAc) with hydrogen bond acceptor (DMAA) substantially increased the conductivity of the membrane without changing its water uptake and only slightly decreasing its permselectivity. Plots of conductivity and permselectivity versus water uptake for all membranes are provided in the Supporting Information to further illustrate this tunability.

The permselectivity and conductivity tradeoff exhibited by all membranes is summarized in Figure 6, which shows that altogether, the membranes exhibited a tradeoff between high permselectivity and high conductivity, similar to that observed in conventional IEMs.^14–16 However, all membranes containing hydrogen bond donor (MAAc) or acceptor (DMAA) had higher performance (i.e., were closer to the upper right corner of the plot, corresponding to high selectivity and conductivity), than the control membrane which contained neither. Moreover, there is a general trend, visible by examining the symbol colors, that membranes containing more MAAc tended to have higher permselectivity/lower conductivity, while membranes containing more DMAA tended to have higher conductivity/lower permselectivity, as illustrated previously in Figure 5. Thus, manipulating the MAAc and DMAA content in these membranes provides a means of navigating along the tradeoff frontier to optimize for one property or the other, and their combined inclusion can be used to improve the permselectivity and conductivity simultaneously compared with the control membrane.

Figure 6.

Membrane permselectivity vs conductivity for all membranes in this work. Refer to the legend and Table 1 for the composition of the membrane represented by each data point. Red color corresponds to the hydrogen bond donor (MAAc), and peach color corresponds to the hydrogen bond acceptor (DMAA). Error bars represent the standard error of at least five replicate measurements.

2.2.4. State-of-Water Analysis.

To gain a more fundamental understanding of how MAAc and DMAA affected the membrane water uptake, we examined the state of water in the membranes. Three populations of water molecules are generally observed inside charged polymer membranes— “free” water that behaves similarly to bulk water, “bound” or “non-freezable” water that is strongly associated with the polymer or ions, and “weakly bound” water in an intermediate state.^24,52–54 For the membranes studied here, the fractions of the total water uptake attributable to bound water ranged from 70 to 98% and were in a similar range to the bound water fractions of other methacrylate-based membranes reported in the literature.^24,55 Hence, the majority of the water absorbed by our membranes was strongly associated with the polymer.

Figure 7 presents the relationship between the bound water fraction and the total water uptake for all membranes studied. Overall, the plot resembles a sigmoid shape in which higher fractions of bound water (relative to the control membrane) are associated with lower total water uptake, while lower fractions of bound water are associated with higher total water uptake. There are four membranes with a higher bound water fraction than the control membrane (~0.95 vs 0.90), and all of these contain either (1) donor without acceptor (e.g., D100–A0–E0, red triangle) or (2) donor and acceptor in a 1:1 ratio. The remaining six membranes have a substantially lower bound water fraction than the control (0.70–0.80). Among these six, four contain either (1) acceptor without donor (e.g., D0–A100–E0, peach triangle) or (2) donor and acceptor in an unequal ratio.

Figure 7.

Membrane bound (non-freezable) water fraction vs total water uptake. Refer to the legend and Table 1 for the composition of the membrane represented by each data point. Red color corresponds to the hydrogen bond donor (MAAc), and peach color corresponds to the hydrogen bond acceptor (DMAA). The dashed curve is shown to guide the eye.

Based on this rough classification, we may hypothesize that hydrogen bond donor (MAAc) generally increases the bound water fraction by (1) reducing the space between adjacent polymer chains via non-covalent cross-linking with various moieties on the membranes, as illustrated in Figure 1c, (2) forming complexes with ether oxygen groups on the PEGDA cross-linker that bind water via hydrophobic effects,³² or (3) by directly bonding with water molecules since, by definition, bound water interacts strongly via electrostatic or hydrogen bonding with the polymer structure.⁵³ However, the fact that MAAc caused a decrease in the total water uptake (see Section 2.2.1) suggests that direct MAAc–water hydrogen bonding is unlikely since MAAc–water hydrogen bonding would likely have increased water uptake. In contrast, hydrophilic DMAA lowers the bound water fraction by attracting additional free water molecules, increasing the overall water content of the membrane. Furthermore, the effects of donor and acceptor on the bound water fraction are clearly non-additive, as was the case with the total water uptake (see Section 2.2.2).

To further evaluate whether the MAAc interacted directly with water in the membranes, we carried out pH-dependent swelling tests on selected membranes (see the Supporting Information). We did not observe any dimensional change when the membranes were exposed to pH 3, 9, or 11, indicating either that the –COOH groups on MAAc were entirely inaccessible to the aqueous phase (which would prevent them from deprotonating at high pH), or that the covalent cross-links in the membranes were sufficiently strong to resist any increase in the swelling pressure caused by the breaking of non-covalent cross-links due to –COOH deprotonation.

Together, the total water uptake (Sections 2.2.1 and 2.2.2) and bound water uptake (this section) results provide evidence that DMAA interacts directly with water molecules in the membrane, thereby increasing the water uptake in linear fashion with DMAA loading (Figure 3b), while MAAc generally does not (Figure 3a) presumably because its hydrogen bond capacity is used for non-covalent cross-linking with DMAA or other moieties.

2.3. Phase Separation in the Membrane.

The hypothesis that MAAc does not interact directly with sorbed water molecules is consistent with an observation of Hu et al.²¹ that hydrogels containing MAAc and DMAA formed hydrophobic clusters which allowed the MAAc and DMAA to remain hydrogen bonded (and therefore to not interact with water) even when the gel was immersed in aqueous solution. The lack of dimensional changes with pH and the cloudy appearance of our membranes when hydrated (see Section 2.1) are additional signs consistent with the presence of micro- or nanoscopic phase separation.^34,35 To investigate this possibility further, we performed phase-contrast AFM on a subset of the membranes studied. Although the images showed contrasts indicative of sub-micron phase separation, the presence of hydrogen bond donor or acceptor did not appear to cause systematic changes in phase separation compared to the control membrane (see Figure S7). We also obtained SAXS profiles of membranes D0–A0–E100, D100–A0–E0, D0–A100–E0, and D-50–A50–E0 to provide additional insights into their structural features. As with the AFM results, the membranes containing hydrogen bond donor and acceptor showed similar features to the control membrane (see Figure S8). All membranes imaged with SAXS exhibited broad scattering peaks, indicating an amorphous structure consistent with that observed in other polymer anion exchange membranes.⁴⁶

Thus, although we found evidence of phase separation within our membranes, their micro- and nanostructure did not appear to change significantly or systematically compared to the control membrane when hydrogen bond donor and/or acceptor were present.

3. CONCLUSIONS

We introduced non-covalent cross-links comprising hydrogen bond donor–acceptor pairs into anion exchange membrane polymers and evaluated their effects on membrane properties. We found that non-covalent cross-linking enables a degree of independent control over permselectivity and conductivity that is usually not available in conventional, covalently cross-linked membranes. Specifically, we found that

• Hydrogen bond donor (MAAc) reduced water uptake and increased permselectivity relative to the control membrane, without affecting conductivity.

• Hydrogen bond acceptor (DMAA) increased water uptake and conductivity without substantially reducing permselectivity.

• The combined inclusion of donor (MAAc) and acceptor (DMAA) improved both permselectivity and conductivity simultaneously, escaping the tradeoff relationship typically observed in covalently cross-linked membranes.

• The combined effects of MAAc and DMAA were not additive (i.e., not easily predictable based on their individual effects), suggesting that they interact with one another inside the membrane.

• The exact mechanism by which hydrogen bond donors and acceptors affect membrane permselectivity and conductivity remains unclear. However, it does not appear that MAAc or DMAA caused significant changes to the membrane’s micro- or nanostructure.

• Based on the exciting prospect of independently tunable permselectivity and conductivity, we believe the behavior of non-covalent cross-links in IEMs merits further study.

4. EXPERIMENTAL SECTION

4.1. Materials.

Membranes in this study were synthesized from commercially available materials. MAOTMAC (cationic monomer, 75% solution in water), PEGDA (covalent cross-linker, average M_n = 250 g mol⁻¹), MMA (neutral ester), and 2-hydroxy-4′-(2-hydroxyethoxy)-2-methylpropiophenone (Irgacure 2959, photoinitiator) were purchased from Sigma-Aldrich. MAAc (hydrogen bond donor) and N,N-DMAA (hydrogen bond acceptor) were purchased from Fisher Scientific. Laboratory grade water (LGW, >18 MΩ cm), ethanol, and 1-propanol were used as solvents. All materials were used without further purification.

4.2. Membrane Synthesis.

Freestanding membranes were prepared by photoinitiated free-radical polymerization of mixtures containing solvent (ethanol), positively charged monomer (MAOTMAC), covalent cross-linker (PEGDA), neutral ester (MMA), non-covalent cross-linkers (MAAc as the hydrogen bond donor and/or DMAA as the hydrogen bond acceptor), and photoinitiator (IrgaCure 2959). The chosen amounts of each component were combined into a single glass vial and stirred at 500 rpm on a hot plate set to 50 °C. To facilitate dispensing small quantities, IrgaCure was added to the mixture from a 50 mg mL⁻¹ stock solution in 1-propanol. The initiator concentration in all cases was 0.05 wt %. After the pre-polymerization mixture became optically transparent (i.e., all components were fully dissolved), we dispensed 1.3 mL onto the surface of a clear borosilicate glass plate coated with a hydrophobic silane coupling agent (Rain-X, Illinois Tool Works, Houston, TX). Stainless-steel spacers (nominal thickness = 127 μm) were placed on the edges of the glass, and another glass plate identical to the first was placed on top. This arrangement sandwiched the liquid pre-polymerization mixture between the glass plates, creating a film with a thickness defined by the stainless-steel spacers. The plates were fastened together with clips and exposed to medium wavelength ultraviolet light delivered by five UVP 34004201 bulbs (nominal peak output at 302 nm). We cured the membranes for a total of 30 min, flipping the plates over every 7.5 min to ensure equal exposure of the top and bottom surfaces to the UV light. After curing, we carefully separated the glass plates and removed the membrane. Samples from each film were cut for analysis of the gel fraction (see below), and the remainder of the membrane was immediately placed in LGW. Prior to any further testing, membranes were equilibrated in LGW for at least 48 h, during which the water was replaced twice to remove any unreacted monomer and residual solvent.

4.3. Gel Fraction.

To confirm that the monomers reacted to form a cross-linked network, we estimated the gel fraction of the membranes, after Tran et al.⁵³ The unhydrated sample of each film cut immediately after curing was dried in an incubator at 65 °C for 48 h to remove residual solvent and then weighed. Next, it was equilibrated in DI water for at least 48 h (replacing the water twice) to extract any unreacted monomer. The sample was dried at 65 °C for 48 h and weighed again. We calculated the gel fraction as the weight after extracting unreacted monomer divided by the weight of the sample before extraction.

4.4. Water Uptake.

Water uptake (i.e., swelling degree) of the membranes was measured gravimetrically.⁵⁶ After removing unreacted monomers by soaking in LGW, as described in Section 4.3, membrane coupons were equilibrated in 0.5 M sodium chloride for 24 h, then blotted dry with a laboratory wipe^16,57 and weighed using an analytical balance. Membrane coupons were subsequently dried in an incubator at 65 °C for 48 h⁵⁸ and then weighed again. The water uptake (w_u, g H₂O per g dry polymer) was calculated from the wet and dry masses (m_wet and m_dry, respectively) as^56,59

$$w_{\text{u}}=\frac{m_{\text{wet}}-m_{\text{dry}}}{m_{\text{dry}}}$$ (1)

We report the average and standard error of at least five replicate SD measurements.

4.5. Permselectivity.

The preferential transport of counter-ions was quantified by the membrane permselectivity, which we measured using a technique similar to that described in our previous work.⁶⁰ Briefly, membrane coupons were installed in a two-compartment cell (volume of each compartment = 17 mL, open area = 7.065 cm²) containing Ag/AgCl wire reference electrodes. Prior to each experiment, reservoirs connected to the respective cell compartments were filled with 500 mL of either 0.5 or 0.1 M sodium chloride solution. These solutions were circulated through the cell at a flow rate of approximately 4 mL min⁻¹ using a peristaltic pump. A potentiostat (VMP3, Bio-Logic) was used to record the potential difference between the reference electrodes until a stable reading was obtained. The measured potential difference was corrected for electrode offset potentials, as described previously,⁶⁰ to obtain the membrane potential, E_mem (mV). E_mem was used to calculate the apparent permselectivity, α (dimensionless), as

$$\alpha =\frac{\frac{E_{\text{mem}}}{E_{\text{mem},\text{ideal}}}+1-2t_{\text{ct}}}{2t_{\text{co}}}$$ (2)

where Emem,ideal (37.9 mV at room temperature) is the potential of an ideally selective membrane separating 0.5 and 0.1 M sodium chloride solution calculated according to the Nernst equation and t_ct (0.604, dimensionless) and t_co (0.396, dimensionless) are the transport numbers of the counter-ion (Cl⁻) and co-ion (Na⁺), respectively. We report the average and standard error of replicate measurements conducted on at least five different membrane coupons.

4.6. Conductivity.

Membrane conductivity was obtained from membrane resistance and thickness measurements. Membrane coupons were installed in the same two-compartment cell as described in Section 4.5, but this time the cell contained platinized titanium working and counter electrodes and single-junction Ag/AgCl reference electrodes (BaSi, Inc. RE-5B) installed in fine-tipped Luggin capillaries. Cell compartments were filled with 0.5 M sodium chloride, which was circulated through both compartments for the duration of the measurement. A potentiostat (VMP3, Bio-logic) was used to perform galvanostatic electrochemical impedance spectroscopy (EIS) over a frequency range of 1–1000 Hz with an amplitude of 1 mA. The combined ohmic resistance of the membrane and solution was determined from the high-frequency intercept of the Nyquist plot. The resistance of a “blank” cell (i.e., the measurement cell without a membrane) was subtracted from this value to obtain the membrane resistance, R_mem (Ω). Further details of the technique can be found in our previous work.²⁸ We then used a digital micrometer (L.S. Starrett Co., Athol, MA, model 3732XFL-1) to measure the membrane thickness L (μm) in the hydrated state. Membrane resistance was normalized by the membrane area, A (7.065 cm²), and thickness to obtain membrane conductivity, κ_mem (mS cm⁻¹), according to

$${\kappa }_{\text{mem }}=\frac{L}{A{R}_{\text{mem }}}\frac{1\text{cm}}{{10}^4\mu \text{m}}$$ (3)

We report the average and standard error of at least five replicate measurements, conducted on different membrane coupons.

4.7. Bound Water Analysis.

We estimated the amount of non-freezable or “bound” water in the membranes (i.e., water strongly associated with the polymer)^24,53,54 using DSC. DSC was performed on hydrated membrane samples (~5 mg) in Tzero aluminum pans with hermetic lids using a TA Instruments DSC2500. Two cool–heat cycles between −100 and 30 °C were performed at a rate of 10 °C/min under a nitrogen atmosphere. Samples were loaded at 40 °C. Melt enthalpies of water in the membrane $\left(\Delta H_{\text{w}}^{\text{m}},\text{ J }{\text{g}}^{-1}\right)$ were calculated by integrating the entire melting enthalpy in the heating curve (including multiple peaks, when present). We calculated the free water content (w_f, g free water per g dry membrane) and bound (non-freezable) water content (w_nf, g bound water per g dry membrane) according to^24,54

$$w_{\text{f}}=\frac{\Delta H_{\text{w}}^{\text{m}}}{\Delta H_{\text{w}}^{\text{o}}}\left(w_{\text{u}}+1\right)$$ (4)

and

$$w_{\text{nf}}=w_{\text{u}}-w_{\text{nf}}$$ (5)

where $\Delta H_{\text{w}}^{\circ }\left(333.5\text{ J }{\text{g}}^{-1}\right)$ is the enthalpy of melting for pure water and w_u is the water uptake defined in Section 4.4. We calculated the bound (non-freezable) fraction of total water in the membranes as w_nf/w_u.

4.8. FTIR Analysis.

Fourier transform infrared (FTIR) measurements were collected using a PerkinElmer Frontier spectrometer in attenuated total reflection (ATR) mode at a resolution of 1 cm⁻¹. Membrane samples were scanned in the dry state between 4000 and 650 cm⁻¹. Liquid monomers, comonomer mixtures, and pre-polymerization solutions were also scanned to aid in band identification. For each sample, we report the average of 16 scans. The background signal was measured prior to any sample analysis and was subtracted from each scan automatically. All spectra were baseline corrected and normalized to the range 0–1.

4.9. Atomic Force Microscopy.

The microscale morphology of the membrane was probed via phase contrast AFM using an Asylum Research MFP3D Atomic Force Microscope. This tapping mode technique responds to differences in the density of the sample associated with differences in the composition^61–65 and hence can detect microscale phase separation. Membranes were analyzed in the wet state after being equilibrated in 0.5 M NaCl solution for at least 24 h.

4.10. Small-Angle X-Ray Scattering.

SAXS was performed at beamline 7.3.3 of the Advanced Light Source (ALS) at Lawrence Berkeley National Laboratory (LBNL). The scattering vector q ranged from 0.005 to 0.5 Å⁻¹, corresponding to nanoscopic feature sizes of approximately 1–100 nm. Membrane samples were imaged in the wet state while immersed in aqueous salt solutions using cells with X-ray transparent Kapton windows and also in the dry state. Wet membranes were equilibrated in 0.5 M NaCl solutions for at least 24 h prior to imaging, and all images were collected at 25 °C. The background scattering signal due to the Kapton windows was obtained from an empty cell containing ambient air and was subtracted from each sample scattering signal prior to analysis.