A review of the synthesis and characterization of anion exchange membranes

AEM head groups have traditionally been quaternary ammonium (QA) ions; however, current research is investigating other head groups such as tertiary diamines, phosphonium, sulphonium and metal cations [2, 5, 16]. Table 1 highlights common anion-conducting cations found in AEM head groups. AEMs were first synthesized with QA because of their relatively easy preparation and good stability [46]. AEMs with QA can be formed by reacting the polymer containing a benzyl halide (e.g. chlorine) with an amine (e.g. triethylamine) to add the ammonia group, and then treating with an alkaline (e.g. potassium hydroxide) to convert the ammonia group to the salt form which can participate in anion exchange [5, 24]. In terms of stability, QA has been shown to have higher thermal and chemical stability compared to quaternary phosphonium and tertiary sulphonium [46].

Additionally, the wide variety of tertiary amines permits the selection of diamines to act as both quaternization and cross-linking reagents when synthesizing AEMs. Notable tertiary diamine head groups include DABCO (1,4-diazabicyclo[2,2,2]octane) and TMHDA (N,N,N,N-tetramethylhexane diammonium) [49]. The ability to self-cross-link is important since it simplifies the membrane synthesis process and improves membrane mechanical stability. To compensate for the intrinsically lower electrochemical mobility of OH⁻ compared to H⁺, research has focused on increasing the ion exchange capacity (IEC) of AEM [16]. However, the trade-off is that higher IEC increases the membrane swelling and reduces mechanical stability [24]. By increasing cross-linking, it can mitigate, but not eliminate, these detrimental effects making these AEM good candidates for further development [64].

The main drawback of QA AEMs is poor chemical stability due to the ammonium group’s susceptibility to OH⁻ attack, leading to ammonium group degradation and reduced IEC [24]. The OH⁻ attack occurs via one of the following reaction pathways: Hoffman elimination, nucleophile substitution (S_N2) or ylide intermediate formation [38, 46, 65]. Figure 3a–d highlights the respective degradation reaction pathways. Given that all these reactions can be initiated by nucleophiles such as OH⁻, the high-pH environment in AEMFC/AEMWE makes it inevitable that the QA will be degraded over time [2, 13]. It has been postulated that the cation chemical stability could be improved by adding large functional groups or electron-donating groups. Large functional groups (e.g. phenyl groups) create a steric hindrance effect that blocks the OH⁻ from attacking the cation and electron-donating groups (e.g. methoxy groups) help protect the cation group from OH⁻ attack [5, 51].

Branching out from QA head groups, researchers have investigated other nitrogen-containing cations such as guanidinium, imidazolium and pyridinium [50‐52, 54‐57]. Of these, imidazolium-based head groups have shown the most promise due to their relatively easy synthesis method, adaptable structure which allows for the addition of various functional groups and selective solubility in water-miscible solvents [5, 51, 66]. With respect to alkaline stability, in addition to S_N2 and deprotonation degradation mechanisms, imidazolium-based head groups can also be degraded via a ring opening mechanism (Fig. 3e) [67, 68]. Multiple literature sources have reported that the electron-deficient C2 position of imidazolium-based head groups is highly susceptible to nucleophile attack, which could be mitigated through the addition of large functional groups to sterically hinder OH⁻ attack [67, 69, 70]. There is some conflicting information as to the importance steric hindrance plays in protecting the C2 carbon. Price et al. [69] commented that imidazolium cation stability can be increased primarily by competing reversible deprotonation reactions, followed by electronic stabilization of the C2 carbon through resonance and finally by steric hindrance of the C2 carbon. The proposed predominant stabilizing mechanism is from the presence of acidic protons which the OH⁻ attacks to deprotonate in a reversible reaction, therefore protecting the imidazolium nitrogen from being irreversibly degraded. Specifically, it was showed that imidazolium ions with a hydrogen at the C2 position was more stable than imidazolium ions with an isobutane group at the C2 position [69]. This conflicts with the theory that large electron dense functional groups at the C2 carbon would better stabilize the imidazolium cation, as shown by Wang et al. for imidazolium cations and by Thomas et al. for benzimidazolium cations [51, 70]. Additionally, Sun et al. summarized research done on large functional group substitutions for the N3, C4 and C5 positions of imidazolium cations, which all agreed with the trend of large electron dense functional groups improving imidazolium cation stability [71]. Of these substitutions, N3 substitutions are most promising as these imidazolium cations could be easily synthesized compared to C4- and C5-substituted imidazolium cations [71]. From the imidazolium-based cation head groups, benzimidazolium cations (benzene group bound to an imidazolium group) have been shown to have improved stability, due to benzene ring resonance structures, and improved anion conductivity, due to ion cluster formation, compared to similar QA and imidazolium-based AEM [50]. As such, these head groups are promising and worthy of additional research.

Through understanding the impacts of steric hindrance and electron-donating groups, researchers have revisited phosphonium and sulphonium cations with a focus on adding large electron-donating groups surrounding the cation to improve chemical stability [5]. Phosphonium-based AEM can be synthesized in similar methods to QA AEMs, except they use phosphine instead of amine for quaternization. Research has shown that stable phosphonium- and sulphonium-based AEM can be synthesized through the addition of phenyl and methoxy groups to the phosphorous and sulphur group to protect the cations [58, 60]. While this work is relatively recent, it has demonstrated that nitrogen-free AEM membranes can be synthesized, and suggests that further research is needed to improve phosphonium- and sulphonium-based AEM performance to match and/or exceed nitrogen-based AEM performance.

A final class of AEM head groups involves metal cations such as ruthenium, cobalt and nickel [63]. The first metal cation-based AEM was synthesized using ruthenium, which was significant as it is a divalent cation which can carry two anions per cation, as opposed to all previous AEM cations which are monovalent [62]. Given the lower electrochemical mobility of OH⁻ compared to H⁺, the ability to use multivalent cations can be a strategy to increase the IEC of AEM. Most recently, it has been found that nickel-based AEMs had the highest conductivity compared to ruthenium and cobalt-based AEMs, which suggests a new potential AEM head group and opportunity to explore other metals for AEM head groups [63].

Overall, there is no consensus on the “best” AEM head group as all head groups have inherent issues with chemical stability and limited IEC; however, there are promising head groups worthy of additional research. Imidazolium-based head groups, including benzimidazolium cations, are promising as stability and performance can be improved using large electron-donating functional groups. There has also been a focus on nitrogen-free AEM head groups, such as phosphonium, sulphonium, and metal cations, to investigate other materials that could be used in place of traditional QA cations in AEMs. While research into metal cation-based head groups is limited, this class of head groups shows great promise due to their high stability and high valency which can address AEM shortcomings related to chemical stability and low IEC.

In an ideal situation, the “best” AEM head group or membrane, is one that is both functional and practical. Functional in that it accomplishes the purpose of the given AEM application. This may include ensuring suitable ion exchange capacity and hydroxide conductivity, stable long-term operation, chemical stability and adequate mechanical properties for routine operation (continuous and/or intermittent). Functionality relates to the material, whereas practicality refers to the synthesis procedure for the head group/membrane. If this membrane is to be used in a commercial application, it will likely be manufactured at a large scale. Overly, complex membrane chemistries using multi-step synthesis with harsh chemicals and operating conditions requiring specialized equipment are not practical. Therefore, in designing the “best” AEM head group or membrane it is important to keep the end goal of the application in mind to engineer a cost-effective solution that will be functional and practical to use.

In parallel with enhancing AEM head groups, research also focuses on polymer structure to improve IEC and chemical stability [28, 46]. Since AEM have traditionally used QA groups, most work on AEM polymer structure involves polymers with QA, with benzyltrimethylammonium being considered the benchmark for AEM head groups [2]. Recently, it has been suggested that benzyl-N-methylpyrrolidinium should be considered the new QA benchmark in AEM research as it exhibits improved alkali stability, conductivity and in situ fuel cell performance compared to benzyltrimethylammonium [47]. As previously mentioned, there is a trade-off between increasing IEC, through the number of ion exchange sites, and decreasing mechanical stability due to water update and membrane swelling [24]. Therefore, AEM polymer research focuses on increasing polymer cross-linking and the formation of ion channels in polymers with distinct hydrophilic and hydrophobic regions [5]. This is driven in part by the success of Nafion^® as a PEMFC membrane since it exhibits a “comb like” structure with a PTFE backbone and regularly spaced perfluorovinyl ether side chain terminated with a sulphonate group for ion exchange [72].

Cross-linking is done to impart more favourable thermal, mechanical and physiochemical properties on a polymer. It can be done as a cross-linking step in a polymerization reaction using high molecular weight or directly cross-linkable oligomers (one-step synthesis) or as a post-cross-linking step after polymerization (multi-step synthesis) [28, 46]. Most (post)cross-linking steps involve covalent bonding and the use of heat, radiation and/or chemicals to facilitate a cross-linking chemical reaction [73]. Given the variety of monomers used to synthesize AEMs, there is no universal cross-linking mechanism, but rather a variety of cross-linking and post-cross-linking reactions. To achieve easier and “greener” AEM synthesis, it is logical to expect that the one-step synthesis method is preferred compared to the post-cross-linking route.

Polymer backbones are commonly polysulfones or fluorinated polymers [e.g. poly(vinylidene fluoride)] [28, 46]. Figure 4 highlights common polymer backbone degradation pathways for polysulfones and fluorinated polymers. Polysulfones are susceptible to ether hydrolysis and quaternary carbon hydrolysis due to hydroxide attack, while fluorinated polymers are susceptible to dehydrofluorination [74‐78]. Therefore, in addition to AEM head group alkaline degradation, AEM chemical stability is also affected by the polymer backbone design. Within these classes of polymers, chemical modifications have allowed more thermal and chemically stable polymer backbones to be designed and/or selected [79‐81].

Inspired by Nafion^®, rather than just having the QA attached to the polymer backbone, polymers were created with regularly spaced flexible side chains containing one or multiple QA groups [5, 82]. Small improvements in stability were seen by changing the polymer backbone to less polar polymers; however, the greatest stability improvements were achieved by attaching the QA groups by a long aliphatic side chain [83]. By grafting multiple QA groups on the side chains, regions of hydrophobicity (polymer backbone) and hydrophilicity (polymer side chains) developed which has been shown to improve IEC and chemical stability [84, 85]. It has been reported that AEM with 99 mS/cm OH⁻ conductivity at room temperature have been synthesized, which is greater than conductivity values reported for Nafion^® [82].

Strategies to obtain AEM with ion channels include locating ion-conducting groups at the ends of polymer side chains, synthesizing polymer main chains using multiblock co-polymers containing regions of ion-conducting groups [86, 87], monomers with densely functionalized ion-conducting regions on the main chains [88] or separately attaching the hydrophobic side change and ion-conducting group to the polymer backbone [89]. Work by Pan et al. and Weiber et al. shows that increasing the number of QA groups in the block copolymer or the hydrophobic side chain length improved the membrane’s IEC to a certain point, after which the IEC decreased with additional QA groups or chain length [86, 89]. This was likely due to the QA group proximity (limited ion-dissociating ability) and the over-assembly of ion clusters that resulted in separate ion-conducting regions.

This demonstrates the need for “rational polymer architecture” to optimize the location, type and concentration of anion-conducting groups and hydrophobic side chains to achieve optimal AEM performance through effective hydrophobic/hydrophilic region interactions. In Fig. 5, the B scenario is what has been shown to be most effective as it creates ion channels to facilitate higher anion conductivity while providing improved alkaline stability since the polymer backbone is protected in the hydrophobic region [89].

Another factor to consider when synthesizing IEMs is Manning’s counterion condensation theory, which suggests that counterions can condense on polyelectrolytes if the linear charge density of the polyelectrolyte chain is greater than one [90, 91]. Due to counterion condensation, reduced effective charge is seen compared to expected values from elemental analysis since the counterion is effectively “screening” the polyelectrolyte charge [92]. While minimal research has investigated this effect in AEMs, multiple sources have demonstrated and modelled the significance of this effect for sulphonated CEMs [92‐94]. For example, Nafion^® 117 has been reported to have approximately 80% of protons in the condensed state [94]. As research targets improvements in ion exchange capacity, understanding and mitigating the effect of counterion condensation can provide an opportunity for optimized IEM polymer and membrane structures.

In terms of membrane synthesis, most AEMs are homogenous membranes prepared by (a) direct polymerization and cross-linking, (b) chemical modification of polymers by irradiation or grafting or (c) chemical reactions to modify polymers [46]. This usually involves phase inversion methods where solutions of membrane precursors are dissolved in polar solvents and casts on a plate after which the solvent is evaporated producing an IEM [5]. Typically, there are multiple steps with harsh solvents (e.g. chloromethyl methyl ether which is carcinogenic for chloromethylation) or radiation sources (e.g. UV, gamma or X-ray for grafting various head groups) [46, 95].

Alternatively, heterogeneous AEM can be prepared using (a) a pore filling or pore immersion technique which synthesizes polymeric membranes on a porous support or (b) mixed matrix membranes that fix inorganic nanoparticles in organic polymers [5]. Pore filling and pore immersion are similar techniques in that a polymer solution is either poured over or immersed in a porous substrate allowing the polymer matrix to fill the porous substrate pores creating a membrane [96‐99]. The porous substrate is selected to be chemically inert and mechanically stable (e.g. high-density polyethylene, polypropylene, polystyrene, polyimide or similar porous polyolefin) [5]. This technique combines the beneficial characteristics of the polymer (high ion conductivity) and porous support (mechanical strength and reduced membrane swelling) to produce membranes with improved performance [100]. While this method may involve repeated pouring and immersion steps, the literature sources have reported the ability to obtain both cation and anion exchange membranes with high IEC [97, 98, 101]. Work by Lee et al. [98] relating to anion exchange membranes is significant as AEM with high IEC were achieved without significant membrane swelling, which was attributed to the use of the porous substrate. This produced membranes with improved mechanical strength compared to similar AEMs synthesized without porous supports. Additionally, by using a porous support, AEMs could be synthesized with multiple narrow ion channels that allowed for the high OH⁻ conductivity [98].

Mixed matrix membranes are another promising type of heterogeneous membranes due to the variety of inorganic nanoparticles and organic polymers that can be blended to achieve desired membrane characteristics [102]. Examples of inorganic nanoparticles that have been used include metal ions, metal oxides, silica, functionalized nanoparticles (e.g. imidazolium-functionalized silsesquioxane), graphene oxide and carbon nanotubes [48, 53, 102‐104]. The inorganic phase is selected to provide improved ion conductivity and thermal, chemical and mechanical stability, while the organic phase is selected to provide the flexibility to the membrane [5, 102]. Sol–gel techniques are typically used to prepare mixed matrix membranes, which stresses the importance of well-dispersed inorganic nanoparticles in the organic phase to produce uniform membranes [105]. Heterogeneous membranes synthesized using porous supports or inorganic nanoparticles are promising methodologies to achieve AEM with high IEC without compromising mechanical strength. This methodology still uses volatile organic solvents, which suggests further research is needed to develop IEM synthesis pathways that can minimize harsh solvent usage. With the variety of porous supports and inorganic nanoparticles available, ample research opportunities are available to tune membrane properties for various applications.

The IEC is a measure of the number of exchangeable ions per membrane dry weight (meqiv/g or mmol/g) [106]. It can be measured via different methods including titration, spectroscopy to determine NO₃⁻ ion concentrations and ion selective electrodes (e.g. pH probe) to determine the presence of H⁺/OH⁻ ions in solution [109]. Titration methods, either through acid/base titration or the Mohr method, are the most common methods to determine IEC. From a safety perspective, the acid/base titration method may be preferred since the Mohr method involves hexavalent chromium (CrO₄²⁻) which is a known carcinogen [110, 111]; however, it has been postulated that there are inherent shortcomings with the acid/base titration method related to CO₂ poisoning of the OH⁻ groups. When the AEM in OH⁻ form is exposed to CO₂-containing environments (e.g. ambient air or air-saturated water), these groups can convert to HCO₃⁻ form and alter the calculated IEC [109]. This influence may be minimal given the short exposure time to air, yet Karas et al. [109] demonstrated IEC decreases of 3.5 and 2.0% per minute for homogeneous and heterogeneous AEMS, respectively, when exposed to air for 5 min. Therefore, efficient procedures when performing acid/base titrations and rinsing AEMs with degassed DI water could help mitigate, but not eliminate, the risk of CO₂ poisoning when measuring IEC [109, 112].

Furthermore, it has been suggested to measure the membrane IEC in the Cl⁻ form, which is the form the membrane is typically synthesized in, to eliminate any deviations in IEC measurements due to pH swings during acid/base titrations [2]. To measure the AEM IEC in Cl⁻ form, the AEM is initially equilibrated in a NaNO₃ solution and then acidified using HNO₃. The resulting solution containing the displaced Cl⁻ ions is then titrated with AgNO₃ using Ag-titrodes to the endpoint, which is when all Cl⁻ has been converted to AgCl. Using the following equation, where m_d is the membrane dry mass, which is the membrane mass after drying at 80 °C for 48 h until there is no change in membrane mass, the IEC (Cl⁻ form) can be determined [47, 113]:

For the acid/base titration method, various procedures are reported depending on acid/base strengths used and soaking times. The general premise is to soak the AEM in a strong base solution (e.g. 1 M NaOH) to convert the AEM to the OH⁻ form and then soaking in a strong acid solution of known volume and concentration to convert the AEM to the Cl⁻ form. Then, the AEM is removed and rinsed with DI water so the resulting diluted HCl solution is titrated with standardized NaOH to the phenolphthalein endpoint. To calculate the number of exchangeable ions (OH⁻) present, the moles of NaOH added are subtracted from the moles of HCl added. This value is then divided by the membrane dry mass, and the resulting IEC calculated by the acid/base titration method is [66, 70]:

With the Mohr method, an AEM is converted to the Cl⁻ form by soaking in a salt solution (e.g. 1 M NaCl). The AEM is then rinsed and equilibrated in a 0.5 M Na₂SO₄ solution to facilitate the release of Cl⁻. Using a AgNO₃ solution with K₂CrO₄ indicator, the AEM/Na₂SO₄ solution is titrated until the K₂CrO₄ endpoint, which indicates when all the chlorides have been precipitated and now Ag₂CrO₄ forms. The resulting IEC calculated by the Mohr method is [110, 114]:

As with any titration, there are inherent human errors in determining the colour change at the endpoint, which ultimately affects the calculated IEC. A further complication for titrations is the dilute nature of the ion of interest targeted in the titrations. To ensure complete conversion of the AEM to the given form, strong bulk solutions (e.g. HCl, NaSO₄) are needed for the titrations. For both the acid/base titration and Mohr method, the concentration of the ion participating in the ion exchange is low relative to the bulk solution, resulting in challenges to accurately determine the endpoint when titrating. To improve accuracy and reduce variability when performing titrations, it is possible to use an ISE (e.g. pH probe) to determine the endpoint rather than relying on the visual colour change [109].

Unlike determining IEC for CEM, IEC procedures for AEM are less well defined. While the most common IEC procedures involve titrations, this may not be the most accurate method. Karas et al. [109] demonstrated that using UV–Vis spectroscopy to determine the NO₃⁻ ions that exchange with Cl⁻ ions in an AEM produced IEC results in greatest agreement with theoretical IEC determined from elemental analysis of AEM composition. By understanding and mitigating the shortcomings with the different IEC procedures, it is possible to reduce the resulting errors in IEC values obtained. More importantly, this demonstrates a need to develop a robust and universal IEC measurement procedure for AEMs to allow accurate comparisons between different AEMs.

The swelling ratio is a measure of the linear expansion of the membranes when exposed to water [115]. It is calculated as a per cent difference between wet and dry membrane lengths. The “dry” membrane state is defined the same as above for IEC.

The water uptake is a measure of how the membrane mass changes when exposed to water [116]. It is calculated as a per cent difference between wet and dry membrane masses. The “dry” membrane state is defined the same as above for IEC.

The membrane water content is a measure of the number of water molecules per mobile anion and is calculated by dividing the water uptake by the molecular weight of water and IEC [117]. Note that the WU is multiplied by 10 to account for the WU being reported in per cent and the IEC being reported in mmol/g.

The water contact angle (θ) is a measure of the wettability of a membrane surface with large contact angles indicating highly hydrophobic surfaces. This can be measured using the sessile-drop technique [101].

Hydroxide conductivity can be calculated from electrochemical impedance spectroscopy (EIS) and a two- or four-electrode testing cell [118]. After soaking an AEM in DI water overnight, the membrane is secured in the testing cell and varying AC current is applied to collect impedance data. Using nonlinear least squares regression analysis, the membrane ionic resistance (R_m) can be obtained and from that the conductivity (σ) can be calculated from the following [70, 119]:

Given that the aforementioned procedure is performed in ambient atmosphere, the presence of CO₂ presents challenges when measuring the true OH⁻ conductivity due to the rapid formation of carbonates and bicarbonates (refer to Eq. 1) [31, 120]. This effect was previously believed to be minimal; however, Ziv et al. [31] have shown that CO₂ can significantly impact true OH⁻ conductivity measurements (~ 50 mS/cm (conventional procedure) versus 103 mS/cm (CO₂-free environment). Ziv et al. proposes modifying conventional hydroxide conductivity testing procedure to ensure a CO₂-free environment by subjecting the AEM to a nitrogen sweep gas flow in the testing cell. Then, a current is applied to generate OH⁻ at the cathode and convert (bi)carbonates to CO₂ which are released at the anode. Once all the CO₂ is released, the AEM would be in the pure OH⁻ form allowing for true OH⁻ conductivities to be measured and thus providing a standardized platform to compare hydroxide conductivity measurements between various AEMs [31].

The alkaline stability is a measure of how the AEM performance changes over time when exposed to high-pH environments [116]. Testing conditions vary; however, the general premise is to soak the AEM in a high-pH solution (e.g. 1–10 M KOH) at a given temperature (room temperature or elevated temperature) for extended periods of time and periodically testing the membrane IEC to see how it changes with time [70]. Inconsistencies in alkaline stability testing conditions may be problematic, as it’s been shown that alkaline stability is influenced by the hydration level of the nucleophile (OH⁻); specifically, reducing hydration levels reduces alkaline stability [121]. At higher hydration levels, the water molecules fill the solvation sphere surrounding the OH⁻, in effect shielding it and reducing its nucleophilic character, resulting in improved alkaline stability. Ex situ alkaline stability has been tested using KOH or NaOH solutions up to 10 M, which corresponds to a water content of approximately 5 (γ = 5) [66, 121, 122]. Higher KOH concentrations have lower water contents; however, the higher viscosities may adversely affect OH⁻ diffusivity and resulting measured alkaline stability. In AEMFCs, as Fig. 1 shows, the cathode can become water-depleted, especially at higher current densities, thus exposing the AEM to ultralow hydration levels (γ = 0). Work by Dekel et al. demonstrated that QA groups had excellent stability at γ = 4; however, this stability was significantly reduced at γ = 0, which was attributed to the change in S_N2 reaction energies, which was the predominant degradation mechanism for the QA group studied [121]. As the hydration level increased, the OH⁻ nucleophilicity decreased, resulting in higher activation energies and reaction energies. This demonstrates that current ex situ alkalinity stability testing using aqueous solutions may produce artificially high alkalinity stability values that would not be representative of in situ alkaline stability in AEMFC. Therefore, Dekel et al. [121] proposed an alternative ex situ alkalinity stability testing procedure using NMR and water-free hydroxide (crown ether/KOH) solution where the water/OH⁻ ratio (γ) could be controlled to assess alkaline stability at different hydration levels.

To analyse AEM mechanical properties, properties such as thermal stability and tensile strength are measured. Knowing the elevated operating temperatures of AEMFC (up to 200 °C) and AEMWE (typically 50–70 °C), thermal stability of the membrane is important and can be determined using thermogravimetric analysis (TGA) and differential scanning calorimetry (DSC) [13, 101, 123]. TGA is used to assess thermal stability by monitoring the temperature at which membrane changes occurs due to water loss, head group decomposition and/or polymer decomposition [30, 123, 124]. DSC can be used to evaluate the glass transition temperature, the effects of thermal cycling and changes in polymer crystallinity and cross-linking [112, 123‐125]. By stretching membrane samples in a universal testing machine, various physical properties like tensile strength, stress–strain curves and elongation at break can be determined [103, 116, 125].