3.1 Mechanisms of proton conduction in Nafion

Proton conduction in Nafion occurs through the channels formed by sulfonic groups and filled with water. Therefore, λ has a crucial effect on the proton conductivity, σ. For example, in Nafion 112 at 80°C, it decreases from 9 to 2 S m⁻¹ when RH is lowered from 90% to 50% (c). The effect of λ on the conductivity originates from three mechanisms that are responsible for proton transfer in Nafion (Kraytsberg and Ein-Eli 2014), as explained below.

At low hydration (λ<5), the state of water in Nafion is defined as “non-freezable” (“bound water”, i.e. tightly bound by the electrostatic interactions to the sulfonic groups forming the walls of ionic channels) (Devanathan 2008, Wu 2008). The bound water region exists within 3–4Å near the pore surface, which is roughly the size of a water molecule (Sahu et al. 2009). Only slow hopping of protons between the sulfonic groups occurs in this regime, referred to as “surface diffusion mechanism.” As hydration increases, the ionic channels expand, reaching an average diameter of 4–6nm. The water accumulating in the central part of the ionic channels is defined as “freezable” (“bulk water”). Protons travel through the bulk water the same way as through aqueous acid solutions, i.e. by “en masse” (“vehicle”) diffusion of H₃O⁺ ions and by “structural diffusion” (“Grotthuss mechanism”) (Choi 2004, Choi et al. 2005). The structural diffusion means that a proton passes from a H₃O⁺ ion to a neighboring H₂O molecule through a series of spatial rotations involving breaking of hydrogen and covalent bonds of a hydrogen atom with adjacent water oxygen. At full hydration, the contributions to overall conductivity are roughly 80% Grotthuss and 20% vehicle, while the surface hopping contribution is negligible. As the total number of protons is constant and equal to the number of sulfonic groups, the increase in conductivity with hydration occurs through the increase of proton fraction found in the bulk water. In addition, the width and tortuosity of the proton path through the 3D network decrease as the ion channels expand, thus the resistance to the proton transport decreases.

The conductivity contributions of these three types of protons and their diffusion in the water channels to the conductivity can be described by the modified Nernst-Einstein equation (Choi et al. 2005):

where F, R and T are the Faraday constant, the universal gas constant and the absolute temperature; D_H+ and C_H+ are the proton diffusion coefficient and concentration in the hydrophilic phase; ϕ_h and τ are the water volume fraction and the tortuosity factor, both dependent on λ; and the indices B, S, G and E stand for “bulk”, “surface”, “Grotthuss” and “en masse”, respectively.

Recent research has revealed that the Grotthuss mechanism can be subdivided into two mechanisms. One is the structural diffusion described above, for which water fluctuation is indispensable. The second is a “packed-acid mechanism,” which occurs in materials containing highly concentrated (packed) acids (Ogawa et al. 2014). In contrast to structural diffusion, the packed-acid mechanism can lead to conduction of protons without water movement or fluctuation. This type of conductivity is used in solid-acid PEMs, which demonstrate high conductivity under anhydrous conditions (Goni-Urtiaga et al. 2012).

3.2 Methanol transport in Nafion

In general, the transport of small molecules across a dense (nonporous) polymer membrane follows a solution-diffusion mechanism, i.e. sorption of solutes into the membrane (on the high concentration side), diffusion across the membrane driven by electrochemical potential gradient and desorption of solutes out of the membrane (on the low concentration side) (DeLuca 2004). The resulting profile is represented schematically in Figure 4.

Figure 4:

A schematic description of a solution-diffusion mechanism in a polymer membrane separating two volumes of solution with high and low solute concentrations.

The transport rate of the solute in the membrane J [mol m⁻² s⁻¹] under concentration gradient and without additional external fields is defined by the following expression:

where D [m² s⁻¹] is the diffusion coefficient, ΔC_m =C_m1–C_m2 is the concentration drop in the membrane [mol m⁻³], and d is the membrane thickness [m]. As the solute concentrations in the membrane surface layers C_m1, C_m2 are not easily measurable, the transport rate is usually expressed through the specific permeability P [m² s⁻¹]:

where ΔC_s=C_s1−C_s2 is the concentration drop between the high concentration and the low concentration solutions, and the partitioning coefficient K=C_m /C_s is the ratio of solute concentrations in the membrane and in the solution at the membrane-solution interface.

When phase-separated ionomers, such as Nafion, and water-miscible solutes, e.g. methanol (MeOH) or ethanol, are considered, K depends on their solubility in the hydrophilic and hydrophobic phases of the ionomer. As was shown by Zhao et al. (2012), sorption of pure water, methanol and ethanol in Nafion occurs essentially by the same mechanism, i.e. solvation of the sulfonic groups, so the contribution of the hydrophobic phase is minor. The first solvation shell consists of four strongly bound molecules; the second solvation shell and beyond are less strongly bound. However, with water-alcohol solutions, the first solvation shell, i.e. the bound water, is formed exclusively by water molecules due to their higher affinity to the sulfonic groups. As a result, alcohol molecules are excluded from the bound water region. The effect is known as “salting out” by fixed ionic groups and obeys the Setschenow equation (Freger et al. 1997a):

where β is the salting-out coefficient and CAm is the concentration of the fixed anion groups in the swollen ionomer membrane.

In the simplest case, methanol permeation through Nafion can be viewed as diffusion through the bulk water region, where MeOH concentration and diffusion coefficient are essentially the same as in the external MeOH solution (Lin et al. 2008a,b). The overall permeability, therefore, is controlled by the hydrophilic volume fraction ϕ_h, or more specifically, by the fraction of bulk water in the membrane, as K increases with ϕ_h (i.e. decreasing CAm) (Wu 2008). As the proton and the methanol transport in Nafion and other sulfonated ionomer systems follow essentially the same routes, σ and P_MeOH are coupled and change simultaneously with ϕ_h, which makes improving selectivity difficult (DeLuca and Elabd 2006). ϕ_h, however, is not the only parameter determining permeability. The morphology and microstructure of the hydrophilic phase may also be very important (Freger et al. 1997b, Freger et al. 1999). For example, Schmidt-Rohr and Chen (2008) concluded that the protonic channels in Nafion, in comparison to stiffer sulfonated aromatic polymers, are wider and less branched due to flexible side chains. This results in higher water diffusion coefficients and higher methanol permeation in Nafion. MeOH concentration can influence the permeability to some extent, too, e.g. an increase in P_MeOH from 1.5×10⁻¹⁰ to 2.5×10⁻¹⁰m² s⁻¹ was reported for a Nafion N117 membrane when the methanol concentration increased from 1 m to 10 m (Yildirim et al. 2008).

Methanol crossover poses a serious problem to the operation of DMFCs. While in hydrogen-operated PEMFCs, thin Nafion membranes (N112) are used to minimize Ohmic losses, in DMFCs, the losses due to methanol crossover are even more significant than the Ohmic ones, requiring thicker membranes (N117) (Hickner et al. 2004). Still, even using thicker membranes, MeOH concentrations in the DMFCs have to be low (1–2 m; 4–8vol.%) to mitigate the crossover problem. If methanol crossover in DMFCs was not an issue, a higher methanol concentration could be used, which would yield a significantly higher cell voltage. Current DMFCs have an open circuit voltage of 0.7V, which is approximately half of the reversible “no-loss” cell voltage (1.2V). However, a 66% reduction in voltage at an optimal operating current density and a decrease in open circuit voltage were observed when the MeOH concentration in the feed was increased from 2 to 6 m (DeLuca and Elabd 2006).

3.3 Salt transport in Nafion

Nafion membranes were originally developed as ion-selective barriers in chlor-alkali applications (Schlick 1996). Nafion selectivity, i.e. exclusion of free anions, results from the effect of fixed anionic groups, which is explained as follows. When uncharged polymer film is immersed in a monovalent salt solution (M⁺X⁻), salt sorption occurs by a simple partitioning mechanism, whereby for each equivalent of uptaken cations, an equivalent of anions has to be uptaken to maintain electro-neutrality within the film. In a polymer with fixed anions A⁻, however, salt sorption follows both a simple partitioning and an ion-exchange mechanism, whereby cations M⁺ act as counter ions for both the free (X⁻) and the fixed (A⁻) anions. As electro-neutrality must be maintained, the resulting concentration of X⁻ in the membrane is lower than that of M⁺. This phenomenon whereby the difference in free anion and cation concentrations in the polymer and in the solution is balanced electrostatically by the polymer’s fixed charge groups is referred to as Donnan exclusion (Geise 2012).

For a phase-separated ionomer like Nafion, salt is almost exclusively limited to the hydrophilic phase; the partition coefficient for the free salt in the membrane can be given by the following expression (Geise 2012):

where the superscript indices s, m and h represent the solution, the membrane and the hydrophilic phase of the membrane, respectively. The γ_± terms are the mean activities of the ions, and their ratio K0≡γ±s/γ±h represents all the non-Donnan contributions to the partitioning (Bason et al. 2010). At very low salt concentrations CMXs≪CAh2K0, the partition coefficient is nearly proportional to the inverse of the fixed anion concentration: K≈ϕhK02CMXsCAh_. At very high salt concentrations CMXs≫CAh/2K0, the Donnan contribution becomes negligible and the partition coefficient is determined only by the non-Donnan contributions: K≈ϕ_hK₀.

Prediction of K is, however, complicated. Even in the simplest case of Nafion membrane immersed in NaCl solution, the presence of salt affects K in several non-linear ways. First, the exchange of protons for sodium ions lowers λ and thus ϕ_h due to the lower hydration number of Na⁺. In addition, λ further decreases with increasing CNaCls (so-called osmotic de-swelling), resulting in higher CAh, narrower ion channels, increased tortuosity and stiffening of the polymer chains (Narebska et al. 1984, Cwirco and Carbonell 1992a,b, Lehmani et al. 1997). The dielectric constant in the membrane will change, too, affecting K₀.

When considering a more complicated system, e.g. Nafion membrane separating two compartments, one filled with concentrated NaCl solution and the other, with pure water, λ should increase across the membrane from the solution side to the pure water side, consequently affecting K. From a practical point of view, however, the permeability of NaCl in Nafion and other sulfonated polymers under such conditions follows a simple empirical log-log linear correlation with the bulk NaCl concentration (Geise et al. 2014). For a Nafion N115 membrane, for example, P_NaCl=5×10⁻¹²C_NaCl^0.427 when a membrane is pre-treated by soaking in water at room temperature and P_NaCl=2×10⁻¹¹C_NaCl^0.452 when a membrane is pre-treated by boiling in water (the units of P_NaCl and C_NaCl are m² s⁻¹ and mol l⁻¹, respectively). P_NaCl in both cases increases with C_NaCl concentration due to the suppression of Donnan exclusion, and the higher P_NaCl of a boiled membrane results from its higher λ.

4.1 In situ PEMFC characterization by electrochemical impedance spectroscopy

PEM transport properties can be studied in situ and ex situ, i.e. as a MEA performance in a PEMFC or as isolated membrane properties. In the former case, the contributions of the catalyst kinetics, membrane resistance and mass transfer to the overall overpotential are studied at different current densities. The MEA permeability to the reactants can be assessed as a half-cell limiting current density under external potential, e.g. when the anode side is fed with H₂ (Tang et al. 2007a) or MeOH (Choi et al. 2008) solution and the cathode side is fed with N₂. The contributions mentioned above can be separated and quantified by means of electrochemical impedance spectroscopy (EIS) (Scribner Associates – Tutorial). In this technique, a small AC voltage or current bias is applied to the FC over a broad range of frequencies (f) to acquire the system impedance response Z (spectrum). The response at the high end of the frequency range is generally associated with the dielectric effects and resistance to charge transfer in the bulk electrolyte, while the electrode effects (reaction, adsorption, double layer response) are typically observed at very low frequencies (Macdonald 1992).

The system impedance can be analyzed by building an equivalent circuit (EC) out of several elements connected in parallel and/or in series, so the frequency response of the EC fits as close as possible to the system response (Latham 2004). The advantage of the EC approach is being relatively fast and simple; however, one of the problems is that ECs are not unique. Several different ECs can result in the same spectrum, leading to completely different physical interpretations (Baltianski and Tsur 2003). Thus, other modern analysis techniques, notably Impedance Spectroscopy Genetic Programming (ISGP) (Hershkovitz et al. 2011, Hershkovitz et al. 2012, Drach et al. 2016, Oz et al. 2016, Oz et al. 2017), analyzing distribution of relaxation times in the system, and mechanistic models (Amphlett et al. 1995, Springer et al. 1996, Franco et al. 2007, Reshetenko and Kulikovsky 2017), are being also employed.

4.2 Ex situ techniques for measurement of PEM transport properties

While an in situ membrane testing provides the most adequate information on the membrane’s performance, this method requires considerable time and skill, as well as expensive specialized equipment (Cooper 2009). Moreover, the hot-pressing procedure during the MEA preparation can alter the PEM properties. As a result, PEM transport characteristics are often measured ex situ in electrochemical and/or diffusion cells. For example, H₂ permeability was measured when a membrane was placed between two vessels, one filled with pressurized H₂ and the other one initially kept under vacuum (Lin et al. 2004). In the setup employed, the pressure change in the second vessel was monitored over time (nearly 60 h). In a similar way, methanol permeability is commonly measured in diffusion cells, in which the membrane separates two stirred compartments. One compartment is filled with MeOH, and the other, with deionized water (dH₂O) (Lin et al. 2005, Yildirim et al. 2008, Zhang et al. 2009, Lin et al. 2010, Gloukhovski et al. 2016). The change in MeOH concentration in the second compartment is measured with gas chromatograph (GC) or density meter using a suitable calibration curve.

The literature suggests that, while the permeability measurements are relatively standardized and straightforward, ex situ conductivity measurements appear to be more controversial. The measurement of the through-plane membrane conductivity under controlled temperature and RH conditions, a mode that best represents the actual mode of operation in PEMFC, requires long equilibration (approximately 2 h) prior to testing (Cooper 2010, Cooper 2011). In addition, poor membrane-electrode contact due to the surface roughness of the experimental membranes can impair the measurements (Cho et al. 2004). Consequently, numerous methods for conductivity measurement in liquid at ambient conditions were proposed, resulting in a wide scatter of reported conductivity values, often significantly different from 10 S m⁻¹ (Slade et al. 2002, Silva et al. 2004). As a result, comparing conductivity and, correspondingly, selectivity values of different PEMs can be problematic due to limitations and uncertainties of the measurement techniques. Gloukhovski et al. (2016) presented a detailed discussion of the subject, which may be summarized as follows.

The first very common approach is to measure in-plane conductivity of a PEM, when the resistance of a long wet membrane stripe is measured in the tangential direction by ac impedance (typically below 100 kHz) with two or four electrodes placed along the stripe. This method is simple and reliable, as the contact and the electrode resistances are negligible relative to the membrane resistance and results in σ≈6–9 S m⁻¹ for Nafion, depending on λ. However, as the actual “working” direction in a PEM is through-plane (normal to the surface), this method is adequate only for isotropic membranes. An anisotropic experimental PEM where some alignment is expected in either in-plane or through-plane direction requires measuring conductivity in the through-plane direction.

In an often used second approach, the through-plane resistance is deduced from ac impedance measurements when the membrane is equilibrated with dH₂O. High frequencies up to 10 MHz are typically required to eliminate effects other than the bulk electrical conductivity. Yet the resistance of the thin water layer in the region of membrane-electrode contact still makes a significant relative contribution to the overall resistance, especially with thin membranes. For Nafion membranes, the conductivity measured by this method was reported to be σ≈1–3 S m⁻¹, typically with lower values for thinner membranes and higher ones for thicker films, strongly suggesting a significant contribution of contact and solution film resistances.

The contact resistance can be eliminated by equilibrating the membrane with strong acids of concentration above 1 m rather than with dH₂O (Gloukhovski et al. 2016). Again, the proton conduction in pore-filling composite PEMs discussed later can result from Nafion presumably deposited in the pores (specific transport) as well as from free transport of acid ions through the unblocked pores (non-specific transport). The distinction can be made with a complementary measurement of P_NaCl, as Cl⁻ transport will be much slower in the Nafion-filled pores than in the empty pores due to Donnan exclusion. The advantage of this method is that it requires only a simple real-time monitoring of cell conductivity or impedance, and thus, it is faster and simpler compared to P_H2 and P_MeOH measurements. Thereby, it can be used in combination with σ measurements for the initial screening of numerous experimental membranes, before more technically demanding characterization techniques are applied (Gloukhovski et al. 2017).

6.1 Perfluorinated and partially fluorinated PEMs

Short side-chain (SSC) PFSA membranes seem to be a promising strategy in search for PEMs with higher conductivity, as shorter side chains allow for a reduction in the EW of the membrane and an increase in the density of acid groups. On the other hand, they are known for problems of durability and mechanical integrity due to increased swelling. Dow Chemical Company demonstrated an SSC PEM with EW of 800–900, which achieved four times the power of Nafion membrane (EW 1100) at the same operating voltage (Savadogo 1998) due to higher conductivity provided by higher density of sulfonic groups. However, the Dow membrane advantage in conductivity over Nafion diminishes with decreasing hydration, and at λ<5, the conductivity of Nafion is superior, probably due to longer and thus more flexible side chains. Other SSC membranes like Flemion (Asahi Glass Co. Ltd.), Aciplex (Asahi Chemical Co. Ltd.) and 3M membrane (3M Inc.) are known to allow higher operational temperatures due to higher crystallinity at a given EW and thus a higher glass transition temperature, as compared to Nafion. Aquivion (Solvay Solexis) membranes are also currently marketed and appear to outperform Nafion with their lower methanol crossover and superior water management (Devanathan 2008). However, in hydrogen-fed FC applications, they are less effective, as their conductivity at 100°C decreases from 11 S m⁻¹ at RH 90% to 7 S m⁻¹ at RH 50%.

An interesting approach to overcome excessive swelling was proposed by 3M Company (Wycisk et al. 2014). Their membranes with EW 580 demonstrated conductivity of 10 S/m at 80°C and RH 50%, yet, below EW 800, polymer crystallinity disappeared, resulting in membrane excessive swelling and loss of mechanical strength in a water vapor environment. In liquid water, the polymer became soluble at elevated temperatures. Later, 3M Company synthesized perfluorinated and partially fluorinated ionomers with side chains containing multiple protogenic groups. This allowed longer runs of backbone CF₂ groups between side chains for the same EW, rendering the membrane with crystallinity sufficient to be insoluble in liquid water. For example, such membranes with EW 625 demonstrated proton conductivity larger than 10 S m⁻¹ at 80°C and RH 50%, similar to a PFSA film of EW 700, yet the water uptake was only half that of PFSA.

Ballard Power Systems developed a partially fluorinated PEM with aromatic side-chains based on trifluorostyrene compositions (Ballard Advanced Materials 3rd Generation, BAM3G) (Tian 2004, Devanathan 2008). It was claimed that in FC applications, BAM3G demonstrated a superior performance compared with Nafion and Dow membranes, as well as excellent durability. EW of 509 appears to be optimal from the conductivity standpoint (17 S m⁻¹ at room temperature and λ=33). Significantly lower conductivities were found for lower EW because of acid dilution by an excessive water uptake (λ≈100) and for higher EW because of poor water uptake (λ≈10). As the styrenic side-chains of Ballard membranes are much more rigid than Nafion side-chains, phase separation is not well developed and sulfonic acid groups are distributed uniformly throughout the membrane. The aggregation of ionic sites is inhibited, and a large fraction of water in the membrane is not associated with ionic aggregates, so the proton conductivity mechanism in BAM membranes may be different from Nafion.

Sulfonated perfluorocyclobutane (sPFCB) ionomer developed by General Motors Company demonstrated good proton conductivity and gas crossover lower than that of PFSA films of similar thickness. The insufficient mechanical stability of this membrane material could be overcome by blending it with polyvinylidene fluoride (PVDF) or using a porous reinforcing mat of expanded polytetrafluoroethylene (ePTFE) (Kraytsberg and Ein-Eli 2014). The latter approach will be elaborated in Section 7.1.

6.2 Non-fluorinated PEMs

Ionomers based on sulfonated or phosphonated hydrocarbon polymers appear to be an attractive replacement to Nafion due to their low production cost and convenience of recycling (Peighambardoust et al. 2010). The main disadvantage of hydrocarbon PEMs is that very high acid group concentrations are required to achieve proton conductivities comparable with Nafion. This leads to very high water content, resulting in poor mechanical properties and even the risk of membrane dissolution in water. Phosphonic groups, being weaker acids than sulfonic groups, induce lower water uptake; however, their conductivity is lower, too. Another undesirable consequence is that loss of water and conductivity at low RH in these membranes is more severe than in Nafion, which makes them less attractive for PEMFC applications. As was demonstrated with sulfonated poly(ether ether ketone) (SPEEK), this difference arises from the backbone rigidity and lack of side-chains, as the sulfonic groups are attached directly to the backbone (Hickner et al. 2004, Dupuis 2011, Park et al. 2011a). The distances between the sulfonic groups are then greater and water channels are narrower, with more branches and dead-ends, leading to poorer connectivity. Also, the acidity of the sulfonic groups in hydrocarbon ionomers is lower due to lower electronegativity of the backbone, as compared with Nafion. Yet even given that, SPEEK and sulfonated poly(ether ether ketone ketone) (SPEEKK) demonstrated selectivity, expressed as the ratio of conductivity to methanol permeability(η_MeOH), seven times higher than Nafion.

Another limitation of hydrocarbon polymers is their inherently lower resistance to oxidative attack as compared with PFSA. In sulfonated polyimides (SPI), for example, this lower resistance arises from greater susceptibility of the sulfonated polymer backbones to chemical attack, as compared with non-sulfonated PI (DeLuca and Elabd 2006), or from their low molecular weight due to the low reactivity of sulfonated monomers. Sulfonated polysulfones, on the other hand, demonstrate excellent oxidative stability in an operating FC due to exceptionally low H₂ and O₂ gas permeability, resulting in low concentration of oxidizing species in the membrane (Wycisk et al. 2014).

Many attempts were made to overcome the limitations of conductivity and oxidative stability via synthesis of copolymers with desirable polymer architectures (e.g. block copolymers, high-free volume copolymers, grafted/branched copolymers etc.) (Zhang and Shen 2012b, Awang et al. 2015, Kim et al. 2015). The aforementioned PSS-b-PMB demonstrated a high degree of phase separation, stability of the ion-conducting channels and effective maintenance of hydration and conductivity at elevated temperatures and low RH (Park et al. 2007). Increasing system order results in faster percolation for proton transport due to lower percolation threshold ϕ_h,c and critical exponent n values, as defined by Hsu et al. (1980): σ=σ₀(ϕ_h−ϕ_h,c )ⁿ. For example, for the triblock copolymer ionomer, sulfonated poly(styrene-isobutylene-styrene) (S-SIBS), ϕ_h,c =0.077 and n=0.76 were reported for proton conductivity (Elabd et al. 2003), lower than Nafion (0.1 and 1.5, respectively), which is a random copolymer. Interestingly, the same research reported higher values of ϕ_h,c and n for methanol permeability, 0.095 and 1.15, respectively. This indicates that increasing the ordering of the system can improve selectivity, which was about five times higher for SSIBS as compared to Nafion. The good connectivity between hydrophilic domains, however, also provides an effective transportation channel for free radicals, which can decrease the durability of these multi-block copolymer membranes (Zhang and Shen 2012a).

6.3 Incorporation of hygroscopic and proton-conductive fillers into PEMs

Water retention at low RH in Nafion and other sulfonated ionomers can be improved by incorporation of hygroscopic inorganic fillers (oxides of aluminum, silicon, titanium, zirconium, zirconium and cesium phosphates, zeolites, clays) (Li et al. 2003, Cele and Ray 2009, Ahmad et al. 2010, Peighambardoust et al. 2010, Patil et al. 2011, Thiam et al. 2011, Tripathi and Shahi 2011, Kim et al. 2015). These particles also improve thermal and mechanical stability of the parent polymer. Despite the increased water retention, the conductivity of the composites is generally on par with the unfilled polymer or lower, even when high filler content is used. The probable reason is that significant amount of water retained by the filler is not available for deprotonation of the sulfonic groups, and filler particles partially block ion-conducting pathways, increasing tortuosity. The composite membranes also become brittle at high filler concentration (Viswanathan and Helen 2007). Methanol permeability, however, also decreases due to pathway blockage and reduced water mobility in the composite membranes. Significant reduction of methanol permeability (>10 times) was reported for hygroscopic particles [e.g. zirconium phosphates (Alberti and Casciola 2003) and clays like laponite and montmorillonite (DeLuca and Elabd 2006)] that undergo exfoliation into thin sheets oriented in parallel to the membrane surface. The effect can be attributed to the orientation of the exfoliated sheets, which is normal to the direction of methanol transport, so the methanol diffusion pathways are significantly extended. An additional advantage of that geometry is a high aspect ratio of the filler particles, resulting in large organic/inorganic interfaces, which help avoid generation of cavities (Zhang and Shen 2012b). Curiously, only a minor decrease in proton conductivity was measured in the composites. This result, however, can be misleading as conductivity is often measured along the membrane surface, while methanol permeability is measured in the direction normal to the surface.

A feasible solution for the problem of poor conductivity mentioned above is using bifunctional fillers, which are both hydrophilic and conductive. They can replace the bulk water in the ionomer channels, thus significantly reducing methanol transport, while still participating in the proton transport. Nafion membranes filled with sulfonated silanes demonstrated η_MeOH three times higher than the pristine Nafion N117 membrane (Ladewig et al. 2007). A hybrid membrane based on SPI and sulfonated molecular sieve Si-MCM-41 showed significantly higher PEMFC performance than that of pure SPI or Nafion membranes (Zhang and Shen 2012b).

7.1 Isotropic pore-filling membranes

The material most widely used for support membranes is porous ePTFE due to its high porosity, mechanical and chemical stability and good compatibility with Nafion. The basic technique for preparation of an ePTFE-Nafion composite membrane is pouring Nafion solution on a pre-cleaned ePTFE film, followed by drying and annealing. The annealing of recast Nafion at temperatures above Nafion T_g=105°C is essential to allow the backbone chain to reorganize and render the film insoluble (Zook and Leddy 1996). It should also improve adherence between the PTFE domains in Nafion backbone and porous matrix due to Van der Waals forces. Liu et al. (2003) used this technique with a≈4.5wt.% in-house made Nafion solution in three kinds of alcohol+N,N-dimethylformamide (DMF) and a PTFE film (0.3–0.5-μm pore size) pre-cleaned with ethanol. The composite membrane demonstrated very effective pore impregnation with Nafion and, similarly to the Gore Select membrane, smaller loss of tensile strength upon hydration and smaller water uptake, which was attributed to higher hydrophobicity due to the ePTFE matrix. The composite membrane performance in H₂/O₂ FC (80°C) was slightly better than the N115 membrane, especially at current density above 1000 mA cm⁻². The oxygen permeability coefficient, however, was 15% higher for the composite, which could be attributed to higher intrinsic PTFE permeability to O₂ as well as to non-ideal pore plugging.

The group of Huang investigated impregnation of ePTFE film (pore size 0.5 μm, porosity 85%) pre-cleaned by boiling in acetone with 5 wt.% Nafion solution in a mixture of water, propanol, ethanol, methanol and unspecified ethers (DuPont) (Lin et al. 2004, Yu et al. 2004, Lin et al. 2005). They demonstrated that pore impregnation is poor with the original DuPont solution but becomes better with the addition of Triton X-100 surfactant (up to 5 wt.%), which both improves compatibility between the solvent mixture and the PTFE matrix and reduces the size of Nafion aggregates in the solution (Lin et al. 2004). The composite membrane demonstrated σ and P_H2, respectively, by 3 and 10 times lower than Nafion and outperformed the N112 membrane in H₂/O₂ FC (70°C) at current density above 1000 mA cm⁻². The aggregation in Nafion solutions in alcohol-water mixtures (4:1 wt. ratio) also decreased in the following order: methanol→ethanol→isopropanol (Lin et al. 2005), as the solubility parameter δ of isopropanol (IPA) is the closest to that of Nafion. Also, ePTFE shrinks under evaporation-induced stress, which is highest for IPA (Zhengbang et al. 2011). This property might be beneficial for Nafion/ePTFE membrane preparation, leading to decrease in pore size during the impregnation and thus tighter pore-filling. The composite prepared from this solution demonstrated properties similar to those found in the previous research, i.e. effective pore impregnation, σ and P_N2 lower than those of Nafion and better performance in H₂/O₂ FC (70°C) (Yu et al. 2004). A similar trend was observed for P_MeOH and performance in a DMFC; moreover, performance deterioration with increasing MeOH concentration was less significant for the composite membrane than for the N112 membrane (Lin et al. 2005).

Ramya et al. (2006) showed that adding ethylene glycol, a high-boiling-point solvent, could probably further improve impregnation and tighten the composite membrane, while adding n-hexane, a low-boiling-point solvent, results in fissure formation. Teng et al. (2013) repeated the methods of Liu et al. (2003) and presented a membrane prepared from a 5 wt.% Nafion/DMF solution, which looked essentially the same as a Nafion 212 membrane both to the naked eye and in scanning electron microscopy (SEM) pictures. The composite was completely transparent, in contrast to the milky-white original PTFE film, which was taken as evidence of tight and complete pore-filling.

Wang et al. (2011) concluded that T_g of PFSA films increased upon exchange of H⁺ to alkali ions Li⁺, Na⁺, K⁺ reaching 290°C for Na⁺, which is the highest among the four and the closest to the T_g of ePTFE (340°C). Consequently, the recast PFSA films and composite PFSA/ePTFE prepared from Nafion 5 wt.% DuPont solution (original or alkali-exchanged), annealed at the corresponding T_g and finally converted to H⁺ form demonstrated a change in various properties. For the membranes prepared from the H⁺-PFSA solution, the crystallinity, the degree of PTFE pore-filling, σ and the PEMFC performance were the lowest, while those prepared from the Na⁺-PFSA solution were the best in terms of all the above properties. In contrast, H₂ crossover and solubility in alcohol were the highest for H⁺-PFSA and the lowest for Na⁺-PFSA. It can be concluded that Na⁺-PFSA solutions are more suitable for preparation of PTFE pore-filling membranes; however, this study did not rule out the possibility that all the differences were caused mainly by the increase in the annealing temperature.

The aggregation of Nafion in pore-filling solutions appears to become the dominant problem with decreasing pore size, requiring special precautions to achieve effective impregnation. For example, Yang et al. (2013) demonstrated that pore-filling using ~4wt.% Nafion solution in IPA-water mixture (4:1 wt. ratio), which was good for 0.3–0.5-μm pore size (Lin et al. 2005), was poor for 0.1–0.2-μm pores. The filling was, however, successful using a similar solution of Nafion precursor, perfluorosulfonyl fluoride (PFSF), followed by alkali hydrolysis and conversion to sulfonyl chlorides SO₂F to sulfonic acid groups SO₃H. The advantage of the method is that unhydrolyzed -SO₂F groups in PFSF, in contrast to SO₃H groups in Nafion, do not induce microphase separation and aggregation in water-alcohol solutions. The resulting performance of the composite membrane PEMFC was very similar to that of the Nafion 211 membrane.

Tang and co-workers investigated impregnation of chemically modified ePTFE film (pore size, 0.1–0.2 μm; 85% porosity) with 5 wt.% Nafion (DuPont) (Tang et al. 2007a,b,c, 2008). The impregnation technique was based on soaking the ePTFE film in the ionomer solution followed by drying, which was repeated three times. The pore-filling of the as-received ePTFE film with Nafion solution and 5% Triton X-100 added was incomplete, again, in contrast to the effective filling of larger pores with the same solution (Yu et al. 2004). However, it improved significantly with pre-hydrophilization of the PTFE film via treatment with sodium naphthalene and N-methyl acrylamide (Tang et al. 2007a). Comparing with a Nafion 211 membrane, the composite demonstrated lower water uptake and conductivity, similar H₂ crossover and slightly better PEMFC performance. It must be noted, however, that the composite membranes were only briefly annealed (5 min at 120°C) and probably could be partially dissolved upon immersion in IPA, which was required to remove Triton X100 after Nafion deposition. It was claimed that lowering the initial pressure to 500Pa before applying the Nafion+Triton X100 solution to the as-received ePTFE film significantly improved pore impregnation (Tang et al. 2007a). It was also demonstrated that the use of ePTFE treated with sodium naphthalene for impregnation with Nafion solution in the Na⁺ form followed by annealing at 270°C (Tang et al. 2007c) resulted in a superior resistance to Fenton oxidation compared with Nafion 111 membrane. The DMFC performance of such a membrane, however, was very poor (Tang et al. 2008). A possible explanation is that the annealing at 270°C for 2 min was still too brief to induce sufficient degree of crystallinity and thus reduce P_MeOH.

Improper choice of the solvent, Nafion concentration, matrix material, pore size and pore-filling method frequently lead to poor pore impregnation. For instance, the following four examples proved to be unsuccessful combinations:

• porous polyethylene, polypropylene and PTFE supports (pore size, 0.1–0.5 μm) and 18wt.% Nafion solution in ethanol (Nouel and Fedkiw 1998);

• polysulfone and microfiber glass fleeces (pore size, 7–15 μm) and Aldrich 5 wt.% Nafion solution (Haufe and Stimming 2001);

• porous polyethylene, polypropylene and PTFE supports (pore size, 0.1–10 μm) and 20–25 wt.% Nafion solution in DMF (Rodgers et al. 2008); and

• ePTFE support (pore size 0.075 μm) and DuPont 5 wt.% Nafion solution applied by solution spraying (Xing et al. 2013).

Apart from Nafion, some other ion-conductive polymers were used to impregnate ePTFE films. Shin et al. (2005) prepared sulfonated polystyrene (SPS)/PTFE membranes by direct polymerization of styrene and DVB (divinylbenzene, the crosslinker) in the ePTFE pores, followed by sulfonation. The degree of sulfonation decreased with the crosslinker concentration, as it became more difficult for the reactants to access the polymer chains. As expected, water content, methanol permeability and proton conductivity dropped, as the degree of sulfonation decreased. In comparison with Nafion 117, however, ion exchange capacity (IEC, measured in milliequivalents of acid groups per gram of dry membrane), water content and conductivity were higher while methanol permeability was lower at all degrees of crosslinking. As a result, η_MeOH increased with the crosslinking degree up to a value of about eight times higher than η_MeOH in Nafion for the styrene/DVB ratio of 85/15. It is unclear, though, how the ePTFE matrix influenced the membrane properties, as no comparison with bulk SPS film was made.

In another study, SPEEK/PTFE membranes (pore size, 0.3–0.5 μm; porosity, 85%) prepared from SPEEK solutions in N-methyl pyrrolidone (NMP) (Xing et al. 2005) and N,N-dimethylacetamide (DMAC) (Wei et al. 2012) demonstrated lower σ and water uptake and higher oxygen permeability, as compared with recast SPEEK films. The PEMFC performance was below that of both recast SPEEK and Nafion. Moreover, the composite membranes included a thick layer of bulk SPEEK above the SPEEK/PTFE region, thus concealing the “true composite” properties.

A similar result was obtained with Na⁺-Aquivion solution in IPA+Triton X-100 (Xiao et al. 2013). Composite membrane conductivity and PEMFC performance, due to thick surface layers and apparently poor pore-filling, were nearly identical to the recast Aquivion film. Using the same ePTFE support, Zhu et al. (2007) prepared composite membranes from 5 wt.% solutions of disulfonated poly(arylene ether sulfone) (SPSU) and Nafion in a mixture of dimethyl sulfoxide (DMSO) and n-butanol (1:1 v/v). Interestingly, the SPSU/PTFE membrane prepared from the solution in DMSO only was brownish and opaque, as compared with the transparent and almost colorless SPSU/PTFE membrane prepared from the DMSO/n-butanol solvent. This indicated that addition of n-butanol increased compatibility between the SPSU and PTFE. The SPSU/PTFE water uptake was smaller than for the recast SPSU, while in Nafion/PTFE and recast Nafion, it was about the same. The conductivities of the composites were about 50% of the corresponding bulk polymer. Curiously, the PEMFC performance increased in the order SPSU→Nafion→Nafion/PTFE→SPSU/PTFE. This order can be explained by higher intrinsic conductivity of SPSU and lower thickness of the composite membranes. However, it is also possible that the choice of solvents was more compatible with SPSU than with Nafion, resulting in better pore impregnation and higher relative increase in performance.

Some research was dedicated to Nafion filling into electrospun nanofiber mats, which provide relatively large and open pores and thus are easily impregnated. The conductivity of Nafion-impregnated polyvinylidene fluoride (PVDF) electrospun mats demonstrated an increase in σ with Nafion content (Choi et al. 2008). It was still, however, an order of magnitude lower than for bulk Nafion due to the insulating properties of PVDF. The DMFC performance, though, was better than for Nafion 115 membrane of commensurate thickness due to lower methanol crossover. The problem of immiscibility of Nafion and PVDF was addressed by preparation of Nafion-modified PVDF nanofiber mats, impregnated then with Nafion (Li and Liu 2014). PEMFC performance of the resulting composite membrane was better than for Nafion 212 and Nafion/non-modified PVDF composite. DMFC performance was better than for Nafion 117 membrane, due to higher σ and lower P_MeOH. It must be noted, however, that the conductivity was measured in parallel to the surface, which is also the prevalent orientation of the nanofibers, while the direction of proton conduction in FCs is normal to the surface.

Similar results were obtained for Nafion-filled mats made of PVDF grafted with poly(styrene sulfonic acid) (PSSA) (Li et al. 2014). The composite demonstrated higher water uptake and σ due to the additional sulfonic groups in the nanofibers, and lower P_MeOH, as compared to Nafion 212 membrane. PEMFC performance of the composite was notably better than for Nafion, although DMFC performance was only slightly better, as the increase in selectivity to methanol was moderate. An interesting result was obtained with Nafion-filled SPEEK nanofibers mats, in which conductivity was higher than in both Nafion and recast SPEEK over a wide range of temperatures and RH values (Xu et al. 2015). However, similar to the above report by Li and Liu (2014), conductivity was measured in parallel to the surface and may not represent correctly the conductivity in the direction normal to the surface. Moreover, this modification resulted in only moderate increase in η_MeOH, as the intrinsic methanol permeability of SPEEK is of the same order of magnitude as that of Nafion.

Wang and Lin (2014) studied Nafion-filled mats made from poly(vinylidene fluoride-co-hexafluoropropylene) (PVDP) or PVDP blended with polybenzimidazole (BI) to improve the compatibility with Nafion. Nafion content was 90 wt.%. The conductivity of the Nafion/PVDP-BI composite was on par with Nafion, but P_MeOH was about 200 times lower. Both conductivity and permeability decreased with the BI content due to BI interaction with the sulfonic groups, thereby the resulting selectivity for all the composite membranes was about 200 times higher than for Nafion. Addition of the minimal fraction of BI (5 wt.%) was found to be optimal, resulting in the highest selectivity and best DMFC performance.

Several studies dealt with Nafion/poly(vinyl alcohol) (Zhang et al. 2009, Lin et al. 2010, Molla and Compan 2011, Molla et al. 2011, Lin and Wang 2014) and poly(lactic-co-glycolic acid) (Wang et al. 2013) electro-spun fiber composite membranes. Due to the high hydrophilicity of PVA and PLGA fibers, the composite membranes showed only moderate decrease in conductivity and methanol permeability, while the water uptake could be on par with Nafion or even higher (Molla and Compan 2011, Lin and Wang 2014). DMFC and PEMFC performance of such membranes was also on par with Nafion or only slightly better, indicating that PVA and PLGA nanofibers induce little change in membrane properties, except for significant increase in tensile strength for PVA (Molla et al. 2011). Similar results were obtained with triple-layer composite membranes, in which a sulfonated polyether sulfone (SPES) fiber mat filled with Nafion was sandwiched between two layers of plain Nafion (Hasani-Sadrabadi et al. 2011, Shabani et al. 2011). The composite DMFC performance was on par with N117 membrane at methanol concentration 1 m and notably better at 5 m.

Apart from Nafion, sulfonated poly(arylene ether sulfone) (SPAES) was filled into nonwoven polyacrylonitrile (PAN) nanofiber mat (Yu et al. 2013). The recast SPAES membranes demonstrated higher water uptake, conductivity and PEMFC performance at RH 100%, as compared with Nafion 212 membrane. However, typically for hydrocarbon PEMs (Section 6.2), conductivity and performance deteriorated faster for SPAES with decreasing RH. The SPAES-PAN composite membrane characteristics were below those of both Nafion and recast SPAES, except for tensile strength, which increased significantly due to the PAN reinforcement. Membranes made from SPEEK-PVA blends, recast or impregnated into mats of electrospun SPEEK-PVB (polyvinyl butyral) demonstrated significant decrease in water uptake, σ and P_MeOH with increasing annealing temperature (from 110 to 140°C) (Molla and Compan 2015). As compared with recast SPEEK-PVA films, water uptake and σ were lower in the composite membranes while P_MeOH remained almost the same. However, both conductivity and selectivity were significantly lower for all the experimental membranes than for Nafion.

Yamaguchi et al. (2003) noted that swelling suppression, especially in the in-plane dimension, is important in PEMs for DMFCs not only to reduce methanol crossover but also to prevent detachment of carbon electrodes, which do not swell. Using porous supports filled with poly(acrylamide-tert-butylsulfonic acid) (PTABS) by direct polymerization in the pores, they showed that composites based on cross-linked high-density polyethylene (CLPE) and polyimide (PI) demonstrate significantly lower areal swelling in the surface dimension in water than those based on ePTFE supports. In comparison to Nafion 117, methanol permeability of PTABS/CLPE composites was many times lower and increased to a significantly lesser degree with methanol concentration, while proton conductivity was only about 50% lower (Yamaguchi et al. 2005). As a result, the composite DMFC performance was only slightly better than N117 at methanol concentration of 8 wt.% but significantly better at 32wt.%.

Yildirim et al. (2008) reported on composite membranes made by filling of ultra-high molecular weight polyethylene (UHMWPE, pore size 0.7 of μm) with Nafion (Yildirim et al. 2008) and sulfonated poly(phthalazinone ether ketone) (SPPEK) (Yildirim et al. 2009). Proton conductivity and methanol permeability of the recast SPPEK film were about 20 and 10 times lower than for the N117 membrane. Nafion/UHMWPE swelling in area dimension was negligible and independent on methanol concentration, while N117 swelling increased. The total swelling by weight, however, was significantly higher for the composite and increased much stronger with methanol concentration, approaching that of recast Nafion. Proton conductivity and methanol permeability of the composite were about six and two times lower, respectively, than for the N117. In SPPEK/UHMWPE composite, both types of swelling decreased by about half while conductivity and methanol permeability decreased about three times, as compared to the recast SPPEK film. Therefore, in both the Nafion and SPPEK cases, selectivity of the composite membrane was notably lower than in N117. The most probable reason could be the incomplete filling of the porous support (≈60% out of ≈90% pore volume fraction) and effect of the bulk ionomer layers deposited on the composite membrane surfaces.

Navarro et al. (2008) filled the UHMWPE supports (pore size, 0.3 μm) from a 5% solution of sulfonated and hydrogenated styrene/butadiene block copolymer (SHSBS) in chloroform/ethanol mixture. Several plasma treatments were tested on the supports to improve the pore-filling. All the composite membranes demonstrated significantly lower methanol permeability as compared with N117, while conductivity only slightly decreased. Electron cyclotron resonance (ECR) treatment in argon provided the best results, increasing selectivity 1.6 and 10 times, as compared to untreated SHSBS/UHMWPE and N117, respectively. Interestingly, liquid uptake in water-methanol solutions decreased with methanol concentration for the composite membranes, in contrast with N117, where it increased.

Porous polyimide (PI) is probably the most effective material for suppressing PEM swelling, because it is rigid and mechanically very strong, as was shown by Yamaguchi et al. (2007) using a PI support (pore size, 0.1 μm; porosity, 55%), filled with SPAES by direct polymerization in the pores. As compared with N117, composite conductivity was three times lower while methanol permeability was 300 times lower and did not increase with methanol concentration or temperature. Later, the same group presented a PI support (pore size, 0.1–0.3 μm; porosity, 50%) filled with SPAES from 10 wt.% solution in NMP (Hara et al. 2009). The increase in water content with IEC in the composite was significantly lower than in the recast SPAES film. Conductivity in SPAES/PI was an order of magnitude lower than in SPAES for low IEC values but closed the gap significantly as IEC increased. This is because conductivity in SPAES reached a plateau due to sulfonic group dilution with bulk water, while restricted swelling in SPAES/PI composite resulted in the absence of bulk (“free”) water, so conductivity increased through the entire IEC range. The authors attributed the extremely low methanol permeability of SPAES/PI to the absence of bulk water and suggested that the proton conduction mechanisms in pore-filling PEMs with high swelling restriction may be quite different from bulk PEMs. It can be related to the “packed-acid mechanism” (Ogawa et al. 2014), which does not require water movement and thus can possibly operate through closely located acid groups and bound water in the PEMs with suppressed swelling.

Similar results were obtained by Nguyen for sulfonated poly(styrene-ran-ethylene) (SPSE)/PI (Nguyen et al. 2009) and Nafion/PI (Nguyen and Wang 2010) composite membranes (support porosity, 80%). In both cases, water uptake in the composite membrane was about one-third of the bulk ionomer, while proton conductivity remained almost the same, and methanol permeability decreased by two orders of magnitude. With the sulfonated polyimide (SPI)/PI membranes, however, the increase in selectivity, as compared with recast SPI, was moderate, although swelling was reduced by half (Wang et al. 2012).

7.2 Anisotropic pore-filling membranes

The results presented in the previous section demonstrate that confinement of ionomers in isotropic rigid and porous supports results in proton conductivity at best on par and generally lower than of the bulk ionomer, even for highly porous supports (80–85% porosity). The reason is that supports of that type increase the tortuosity of proton-conducting paths and, apparently, induce no preferable orientation in the pore-filling ionomer phase. The only exception was demonstrated with ionomer-filled mats of nanofibers bearing sulfonic groups (Li and Liu 2014, Xu et al. 2015) when conductivity was measured in parallel to the surface. As this is also the prevalent orientation of the nanofibers and significant conductivity enhancement along the nanofiber was demonstrated (Dong et al. 2010), the “true” through-plane conductivity of these PEMs might be substantially lower.

Therefore, there is significant evidence (see also Section 5) that conductivity of Nafion and other proton conducting polymers can be increased by alignment. However, to increase “trough-plane” conductivity, alignment normal to the membrane surface is required. This could be achieved by filling an array of highly ordered parallel pores with Nafion or other proton-conducting polymer, i.e. to use a support membrane that would provide a properly aligned solid interface, as shown in Figure 5B.

Anodized aluminum oxide (AAO) membranes and polycarbonate track-etched (PCTE) membranes are the most popular architectures for making anisotropic structures aligned normal to the surface. For instance, an increased electric conductivity of nanowires prepared within their pores is reported, as compared with the same bulk polymerized materials (Malinauskas et al. 2005). However, the examples of PEM preparation with AAO and PCTE membranes are even scarcer than with the isotropic supports reviewed earlier. Probably, the first publications by Leddy and Vanderborgh (1987) and Fang and Leddy (1995) presented PCTE membranes filled with Nafion from an in-house-made ~3wt.% water-ethanol solution (50:50) by soaking and drying. A 20-fold increase in cation flux with pore diameter decreasing from 600 to 30 nm was reported, and increasing ordering of Nafion within the pores was suggested as an explanation for the phenomenon. However, no SEM visualization or structural study results were provided.

Later, Raghav (2005) investigated pore-filling of PCTE (pore diameters are 10, 50, 100nm) and AAO (pore diameters are 20, 100nm) membranes with Nafion and sulfonated polystyrene (SPS). Applying the methods of Leddy and Vanderborgh (1987) to commercial Nafion solutions (Sigma-Aldrich, Ion Power) failed entirely to show any filling of the PCTE membrane pores. Several other impregnation methods, including vacuum filtration, also failed to fill the PCTE and AAO membranes with Nafion. Vacuum filtration of 5 wt.% SPS solution in acetone, however, resulted in the formation of well-defined SPS nanorods in the PCTE pores and incomplete, although still noticeable filling of the AAO pores. These results were attributed to poor substrate wetting by the Nafion water-alcohol solutions, as opposed to very good wetting by the SPS/acetone solution.

Bocchetta et al. (2007) used prolonged soaking of AAO membranes (pore size, 200nm) in a commercial Nafion solution; however, pore-filling was mostly next to the AAO membrane surface without significant penetration into the pores. A work based on the same support filled with Nafion solutions (5 and 20wt.%) under low vacuum (0.1–0.3atm) offered some interesting results (Rayon et al. 2006). While no pore-filling was achieved with the as-received materials, widening the AAO pores to 500nm resulted in some Nafion deposition, and neutralization of Nafion into Na⁺ form allowed the formation of nanorods in the 200-nm pores, as verified by SEM. No conductivity/selectivity results were provided, though. The only example of successful pore-filling with Nafion combined with electrochemical measurements could be found with straight-hole PI support (Guo et al. 2012). However, very large pore diameters (50–200 μm) were employed. As a result, composite PEM conductivity increased with pore size, mostly due to the increasing Nafion fraction and with RH, similarly to the Nafion 211 membrane. PEMFC performance under various conditions was, at best, on par with Nafion 211.

Available data suggest that the main obstacle for Nafion-H⁺ impregnation into pores of nanometric size was the aggregation of Nafion in the solutions. Indeed, recent reports indicate that there is a threshold concentration C*, depending on the solvent, below which Nafion micelles break up to single molecules of estimated ~5-nm width and ~100-nm length (Aldebert et al. 1988) and hydrodynamic radius (R_h) of 20–40 nm (Lee et al. 2004, Zhang et al. 2008). C* was found to be about 1 g l⁻¹ in water and water-methanol mixtures (Jiang et al. 2001, Lee et al. 2004). On the other hand, in 10:9 v/v IPA/water mixture, a better solvent for the perfluorinated backbone, C* increases to ~4.5 g l⁻¹ (Zhang et al. 2008). As previous attempts to fill PCTE or AAO pores with Nafion (Leddy and Vanderborgh 1987, Fang and Leddy 1995, Raghav 2005, Bocchetta et al. 2007) were carried out with solutions containing ≥30 g l⁻¹ Nafion in solvents containing ≥50% water and <15% IPA, Nafion molecules must have aggregated to form large particles (>500 nm) (Jiang et al. 2001, Zhang et al. 2008) and were unlikely to enter 100–200-nm pores. This is also consistent with the aggregation problems during the pore-filling of PTFE membranes described in the previous section.

Recently, Gloukhovski et al. (2016) demonstrated successful filling of the AAO membrane (pore diameter, 200 nm) by through-the-membrane evaporation of Nafion solution (~1.2 g l⁻¹ in IPA) under vacuum. Through-plane σ and η_NaCl of the pore-filling Nafion were two and four times higher, respectively, than for bulk recast Nafion. Later, the same group also reported on Nafion impregnation into PCTE (pore diameter, 600–1000 nm) membranes (Gloukhovski et al. 2017). The pore-filling Nafion demonstrated increased proton conductivity, presumably due to the lower tortuosity and better connectivity of water channels aligned by the pore walls. Methanol and sodium chloride permeability, on the other hand, decreased significantly, presumably due to the suppressed swelling, resulting in narrower water channels and stronger “salting-out” and Donnan exclusion effects. Presumably, confinement could also encourage formation of regions of closely packed sulfonic groups, which facilitate proton transfer but strongly exclude other solutes. The alignment of the pore-filling Nafion, however, could not be demonstrated by SAXS, presumably due to the insufficient amount of Nafion in the experimental membranes. Also, no evidence of enhanced structural stability at elevated temperatures could be found.

Examples of successful filling of PCTE and AAO pores with other polymers (not necessarily in FC context) also exist in the literature, e.g. pore-filling of PCTE (pore diameters, 30–400nm) with polyethylene oxide (Vorrey and Teeters 2003) and of AAO (pore diameters, 20–200nm) with poly(phenylene sulfonic acid) (White et al. 2008). Direct attachment of sulfonic groups to PCTE pore walls (pore diameter, 10–2000 nm) by graft polymerization was also reported (Chen et al. 2006). In all these examples, proton conductivity increased with decreasing pore diameter; however, even maximal conductivity was low and selectivity was not reported.

Oleksandrov et al. (2009) attached sulfonic groups to the pores of a commercial AAO membrane (Whatman Anodisc; pore size, 20nm) through silanization. The composite membrane demonstrated conductivity three times higher and methanol permeability of about 10 times lower than Nafion 117 membrane under the same conditions. However, no SEM visualization was provided. It must also be noted that the bulk section of a 20-nm Anodisc membrane is identical to a 200-nm membrane, with 20-nm pores present only in an additional extra thin layer on top of the bulk membrane. This is because the main application of Whatman AAO membranes is filtration, for which the thin top layer is sufficient for size separation (Zhang 2006).

Another example of silanization + sulfonation, with porous silica substrate (pore diameter, 5–7nm; mouth diameter, 2 nm) showed overall conductivity of about 10 S m⁻¹, almost independent of RH (Moghaddam et al. 2010). This weak dependence on RH, in contrast with Nafion, might arise from the stability of proton-conducting channels (Park et al. 2007), in which their structure is determined by support pore geometry and obviously cannot be affected by RH. It could also be related to capillary condensation of water within the nanopores, which were not filled with dense polymer, as in the above example, and evaporation of pore-filling water could be prevented by high curvature of the liquid-vapor interface at the pore mouth down to very low RH. It is hard to evaluate the intrinsic conductivity of the channels in the last two examples, though, as the pore fraction was not reported.