The oldest approach for alkaline water electrolysis, using a diaphragm, dates from the end of the 19th century. The purpose of the diaphragm was dual: to avoid the mixing of the generated gases H₂ and O₂ and to allow the passage of the ionic current from the anode to the cathode. By definition, a diaphragm has no ion-selective property and is just a physical barrier for the gases while being permeable to the ions and to water. Historically, inorganic diaphragms such as asbestos [chrysotile, Mg₃Si₂O₅(OH)₄] were largely used. More modern diaphragms are polymer based such as polyphenylene sulfide (PPS), polysulfones (PSU), etc., and can include an inorganic filler such as zirconia or titania particles. Alternatively, some thinner so-called ion-solvating membranes (ISMs) based on polymers such as polybenzimidazole (PBI) have recently been proposed. The field of membranes for alkali-water electrolysis and anion-exchange membrane (AEM) electrolysis has recently been thoroughly reviewed [1], and the focus of this article is dedicated to the ions transporting the current from the cathode to the anode.

Historically, the electrolyte of choice for diaphragm-based systems was KOH. The main reason was solubility, meaning a high concentration of OH⁻ in solution, as the KOH solubility in water can reach 8 M. However, the conductivity of KOH solutions reaches a maximum at about 7 M. Below this value, conductivity increases with concentrations as the number of ionic charge carriers increases, but at higher concentrations, ion–ion interactions hinder the conductivity, which decreases above 7 M also due to ion pairing effects (see fig. S2) [2]. The concentration of 7 M corresponds to a eutectic in the phase diagram water–KOH, meaning that the electrolyte would remain liquid at −60 °C, which is beneficial for electrolyzers operating in cold countries such as Norway or Switzerland. Moreover, as the viscosity of alkaline aqueous solutions typically decreases toward higher temperatures, thereby fostering ionic conductivity, alkaline electrolyzers operate in the range 80 to 90 °C.

As for the electrodes, early electrolyzers used nickel electrodes for both anode and cathode, while some modern electrolyzers used electrocatalyst-coated electrodes.

In a porous diaphragm, the pores are large enough to accommodate the surrounding KOH solutions. Ion transport is considered to occur via Brownian motion within the pores of more than 150 nm diameter. In a dilute KOH solution, the transport number of K⁺ and OH⁻ is respectively 0.27 and 0.73 (λ°_K+ = 73.5 cm²·Ω⁻¹·mol⁻¹, λ°_OH− = 198.3cm²·Ω⁻¹·mol⁻¹), meaning that hydroxide ions carry about 3 quarters of the current being more mobile than hydrated potassium ions. Indeed, the transport of OH⁻ partially occurs through a Grotthuss mechanism, involving the formation and disruption of hydrogen bonds in a cluster of water molecules to effectively transport either the protons or the hydroxide ions. As the concentration increases, the evolution of the transport number could not be found in the literature, but it is likely that the hydroxide ion remains the main current carrier. As a matter of fact, it is interesting to notice that the transport number of K⁺ is 0.03 in molten KOH [3].

In electrolysis conditions, the question of the transport number of the 2 ions through the diaphragm becomes more intricated.

The classical methodology to analyze transport phenomena through membranes is based on irreversible thermodynamics [12]. The gist of this approach is to relate a driving force acting on a species, being charged or not, to a linear sum of frictional interactions. It was first proposed by Maxwell in the 19th century, and the resulting equations are often referred to as the Stefan–Maxwell equations. For transport processes in ionic membranes, the early work of Spiegler [13] establishes the basis of the theory. For binary electrolytes such as NaCl crossing a porous, but nonselective cellulose acetate membrane, the seminal work of Bennion and Rhee [14] provides equations not only for the flux of the different species, Na⁺, Cl⁻, and water but also for the current. It is interesting to notice that the conductivity in the membrane was less than a third that of the bulk solutions, and that the transference number of the sodium cation decreases from 0.85 at 3.5 mM NaCl concentration to about 0.3 at 75 mM. Sinha and Bennion [15] further studied the transfer of KOH through a Permion 2291 cation-exchange membrane and observed that the transport number of K⁺ decreases from 0.846 to 0.453 when the concentration of KOH was varied from 1 to 10 M, showing the loss of selectivity with increasing lye concentration. These early works clearly show that the transport numbers vary with the experimental conditions.

Some membranes often referred to as ISMs are nonporous. These are polymer films that contain polar or ionizable groups (such as benzimidazoles or imidazoles) and that swell when immersed in KOH to achieve ionic conductivity. As diaphragms, they have no selectivity to OH⁻ transport. A key advantage of the ISMs is to alleviate some drawbacks of porous membranes, namely, pore blocking either physically when pressing the electrodes in the case of the zero-gap cell or chemically by particles floating in the lye such as iron oxide particles resulting from some corrosion processes. ISMs are often used based on low-cost arguments and good gas barrier efficiency.

When plotting the cell voltage versus the applied current, and by comparing the linear region associated with the resistive nature of the cell, and hence of the diaphragm, it has been reported that ionic solvating membranes are less resistive than commercial membranes such as Zirfon as schematized in Fig. 4 [21].

The most obvious reason can be the membrane thickness, which can be reduced to tens of micrometers. Another more speculative reason is due to different ionic conductivity mechanisms, where the presence of perhaps more direct ion pathways may replace the tortuous classical diaphragm. Also, it could be argued that in case of hydrophobic polymeric ISMs ions are less hydrated and hence less subjected to friction forces with the water molecules. Moreover, Babcock et al. [22] showed by neutron scattering experiments on PBI-based ISMs that the polymer immediately after doping exhibits a polymer chain separation range of approximately 10 to 20 nm. The important steric hindrance in those membranes must play a key role on the resulting hydration structure.

Recently, a wide range of new approaches to membrane fabrication has been proposed. One can cite an ultrathin PPS film on reinforced PPS mesh [23] (thickness = 200 μm, resistance = 0.21 Ω·cm², cell voltage = 2.33 V at 500 mA·cm²) or a polyvinyl alcohol (PVA)-skin layer of a Zirfon membrane [24].

A recent publication has reported a sol–gel process to produce high-performance (1,4-naphthalene)-5,50-bibenzimidazole (NPBI)-based ISM with high conductivity in low concentrations of KOH electrolyte and excellent alkaline stability (2 M KOH at 80 °C for 285 d) and high ion conductivity even in 2 M KOH solutions (190 mS·cm⁻¹ at 80 °C) [25]. This NPBI membrane can further engineered to include porogen additive to enhance further the conductivity and the durability [26].

Also, novel polymer compositions have been proposed, such as poly(oxindole biphenylene)-based ISMs with highly stable oxindole/KOH complex ion pairs, able to operate more than 2,500 h at high operating electrolysis conditions such as high temperatures (≥80 °C) and voltages (2.3V) with ultralow gas permeation [27]. Progress in ISM has been rapid recently, and Dayan et al. [28] have even suggested that ISM, such as a thermally cured sulfonated para-polybenzimidazole (MS-PBI) membrane, can substitute AEMs in water electrolyzers.

Finally, a biosourced ISM based on watermelon skin is worth mentioning. Not only for water electrolysis but also for electrochemical CO₂ reduction, where a micropore and continuous hydrogen-bonding network of cellulose fiber and pectin enable a high ion conductivity of 282.3 mS·cm⁻¹ (room temperature, saturated with 1 M KOH) [29].

If a membrane is so-called ion solvating, it can be argued that the polymer is a solvent for the salt KOH. In this case, there must be a distribution potential difference at the aqueous electrolyte | organic polymer interface. This distribution potential is given as the half-sum of the standard ion transfer potential of K⁺ and OH⁻ [30].where {\mathrm{\Delta }}_{\mathrm{P}}^{\mathrm{W}}{\varphi }_{{\mathrm{K}}^{+}}^{\mathrm{o}} is the standard potassium ion transfer potential defined from the Gibbs energy of transfer from the aqueous lye to the polymer

The Gibbs energy of transfer of K⁺ and OH⁻ cannot be found in the literature, but these values have been listed for a series of water | organic solvents, and to give an order of magnitude, the standard transfer potentials of K⁺ and OH⁻ are about 0.5 and −0.5 V, respectively, for the water | 1,2-dichloroethane interface [31], meaning that even if the Gibbs energy of transfer of the salt is about 100 kJ·mol⁻¹, the distribution potential is negligeable, and we can consider that the lye | ISM is not polarized. The de-hydration of KOH in the organic polymer may, of course, not be complete. Indeed, it has been shown computationally that mono- and di-hydrate complexes of the potassium cation in nonpolar media are rather stable species [32].

Alternatively, the ISM can be considered as a cation-exchange membrane. Indeed, in the case of PBI, strong bases like KOH can deprotonate the benzimidazole amine groups (pK_a = 12.8), leading to the formation of the corresponding benzimidazolides anions [33]. As for zirconia particles illustrated in Fig. 2, the negative charges on the polymer backbone will be compensated by a Helmholtz layer of potassium ions, thereby transforming the polymer as a weak anion exchanger. This is schematized in Fig. 5.

The role of cations is indeed interesting and a strategy to make the PBI membrane a K⁺ conductor is to incorporate metal ions such as Zn²⁺, as proposed for assembling zinc–iron flow batteries [34].

Figure 6 summarizes the different fluxes during electrolysis using a water-solvating membrane. The major differences with Fig. 3 stem from a very reduced flux of salt and water across the membrane, and a reduced gas permeation.

Ion-exchange membranes being either cationic or anionic play an important role for the design of compact electrolyzers. Proton conducting membranes were developed initially for hydrogen fuel cells where the protons produced by hydrogen oxidation carry the current to the cathode where oxygen is being reduced. With time, they were used to assemble PEM (polymer exchange membrane) electrolyzers. The advantages of PEM electrolyzers include the compactness, i.e., a low footprint, the only reactant is pure water, and they can be operated at relatively high pressure, e.g., 30 bars. The disadvantages stem mainly from the materials used such as precious metal electrocatalysts, e.g., iridium oxide for the anode and fluorinated proton conducting membranes, which can potentially generate polyfluoroalkyl substances (PFAS). The electrode connectors are mainly composed of metallic or carbon foam electrodes [also porous transport layers (PTLs) or gas diffusion layers (GDLs)]. The membrane is sandwiched between 2 electrodes, forming a membrane electrode assembly (MEA) onto which electrocatalysts are deposited. Microporous layers are sometimes used between the PTL and the MEA.

Another class of ion-exchange membrane-based systems is the AEM electrolyzers. These systems are gaining a lot of interest as they share some of the advantages of the PEM electrolyzers in terms of compactness and ease of use. They also rely on transition metal electrocatalysts and on more readily available anion-exchange polymers. The major limitation of AEM electrolyzers is the polymer membrane chemical stability with a lack of data on their lifetime, and therefore, the calculations of the operating costs remain elusive.

The major difference between an alkaline electrolyzer and an AEM electrolyzer is the concentration of KOH used: 7 M in the former case and generally 1 M in the latter case. The KOH solutions are therefore much less corrosive and more user-friendly.

Nonetheless, as recently reviewed by Mamlouk [35], AEM can degrade due to both mechanical and chemical stress. The former stems from the membrane swelling and the applied pressure on the membrane such that a tensile strength of >15 MPa is usually recommended to avoid the formation of cracks and possible intrusion of fibers from the porous transport layer (PTL). The latter stems mainly from the nucleophilic attack of OH⁻, not so much on the polymer backbone but on the cationic group by substitution (SN₂) or elimination (E₂) reactions. The presence of radicals during electrolysis is also a cause of membrane degradation, which leads to the loss of cationic headgroups. Recently, membranes featuring γ-amine-piperidinium-functionalized polystyrene, featuring a carbazole segment to tune ion-exchange capacities, have shown robustness features to envisage industrial applications [36].

When using AEM, the transport number of hydroxide ideally tends toward unity. The conductivity target expressed by the US Department of Energy is more than 100 mS·cm⁻¹.

As discussed above, studying ion motion in confined solutions such as AEM membranes remains a challenge as water dynamics and ion motion occur over time and length scales ranging from femtosecond to picosecond for the former and microseconds to milliseconds for the latter. Recently, 2D IR spectroscopy and semiclassical simulations were used to examine how water molecules are arranged into successive solvation shells and the interplay with bromide ion transport [37].

Chen et al. [38] have proposed a model for hydroxide solvation and transport in AEMs. They distinguished the so-called vehicular mechanism associated to classical ionic motion in solution (Brownian motion) and the Grotthuss mechanism (structural transport) involving bond breaking and bond making, establishing that the former contributes considerably more than the latter for hydroxide transport in the AEM. This is different to PEM systems, which show the opposite for proton transport. They also noted that the activation energy barrier for hydroxide diffusion in the AEM is greater than that for proton diffusion in PEMs, implying a more significant enhancement of ion transport in the AEM at elevated temperatures.

In 2018, using computer simulations, Dong et al. [39] have shown that the Grotthuss mechanism is more important in very confined environment and bottlenecks in the polymer structure. More recently, Dubey and Daschakraborty [40] have further observed a significant role of hydration level and temperature in the mechanism of vehicular diffusion. In the presence of water and at a higher temperature, both the hydroxide ion and water diffuse via mainly small steps, while at a lower hydration level and temperature, the diffusion becomes more governed by the jump-diffusion mechanism involving large steps.

A key parameter in the design of an AEM is the degree of amination and water uptake that promotes the transport of water and hydroxide anions in interconnected water channels, but at the detriment of the mechanical stability of the membrane. In the case of AEMs, the concentration of fixed cationic groups is given in meq·g⁻¹ usually in the range 1 to 3 meq⁻¹ [41]. This corresponds to a concentration of less than 3 M (mol·L_membrane⁻¹) and a water fraction of less than 0.5. Although the water content is higher than for an ISM, the presence of fixed cationic sites favors hydroxide transport by jump diffusion. In AEM, it is very likely that all types of hydroxide transport occur concomitantly: viscous Brownian motion with ion–ion interactions, Grotthuss-type mechanisms involving both free water molecules or electrolyte chains, and jump diffusion between fixed sites.

Recently, Tomasino et al. [42] have computed that a degree of 30% amination was optimum. Numerical simulations have also been used to characterize commercial membranes such as Sustanion (Dioxide materials), which is an imidazolium-based membrane, for which the authors concluded that vehicular diffusion is the predominant mechanism for chloride transport by diffusion, but contributes only 15% for hydroxide diffusivity [43]. Poly(arylene indole piperidinium) AEMs have also been studied theoretically, in particular the influence of the side chains. The results clearly show that flexible side chains favor OH⁻ transport due to the overlapping of the hydration shells, and how the molecular structure of the polymer influences the formation of water clusters [44]. Very recently, Menga et al. [45] have shown by simulations that the addition of crown-ether moieties in poly(arylene piperidinium) improves the OH⁻ mobility by increasing the solvation shell of the cationic groups facilitating the formation of continuous and uniform ion transport channels.

If numerical simulations provide insightful information at the molecular level, the gap with industrial applications of AEM water electrolysis remains important, as in the “real world” the large current densities used generate large electric fields and high fluxes of ions and water molecules. The presence of nanobubbles further complicates the membrane transport processes.

Park et al. [46] recently studied a series of alkali metal hydroxides. The transport number of the alkali cations was measured, and the results obtained showed that naturally the transport numbers of the cations increase with the electrolyte concentration, the membrane being not ideally selective, but also varied with the nature of the cations: t_LI+ = 0.131, t_Na+ = 0.145, t_K+ = 0.160, t_Cs+ = 0.174 for a 0.01 M solution, and t_LI+ = 0.201, t_Na+ = 0.214, t_K+ = 0.221, t_Cs+ = 0.230 for a 0.1 M solution. It can be observed that the less hydrated the cation, the higher the transport number, in line with the predictions of Eqs. 11 and 12 for a simple binary electrolyte. They also showed in a previous publication that the addition of non-electroactive nitrate salts to an AEM electrolyzer operating with a 0.01 M NaOH solution altered the water electrolysis performance [47]. They emphasized the role of cations in delivering water to a dry cathode and concluded that alkaline water electrolysis with low pH (e.g., pH 12) but with an additional high ionic strength electrolyte may be an interesting strategy for long-term electrolysis durability at low electrolysis overpotential (1.73 V for 800 h), since excessive hydroxide concentrations can be avoided. Figure 7 schematically illustrates an AEM and the presence of mobile hydroxide ions.

When assembling PEM or AEM electrolyzers, ionomers are used at the membrane | electrode interface. The role of ionomers in ion-exchange membrane electrolyzer designs is often discussed in relation to the improvement of the catalyst layer structure and the increased adhesion between the MEA and the porous electrode support, hence a reduced contact resistance and finally to improve the stability of the catalysts by providing a stable environment. Ionomers can play a role both in the anodic compartment and on the cathodic side. For the anodic side, the key role of the ionomer is to provide a right balance between ionic and electronic conductivity between the catalysts and the electrode feeder. For example, it was recently shown that a variation of 25% in electrolyzer performance can be obtained just by varying the ionomer content [48]. On the cathodic side, it is important to distinguish electrolyzers operating a dry cathode mode and those where the cathodic compartment is fed with the electrolyte solution. In the former case, the ionomers play a key role as a “sponge” to keep a certain water content and maintain ionic conductivity. Counterintuitively, Titheridge et al. [49] have shown that the cation exchanger Nafion can play a useful role as a cathodic ionomer, consistently performing better than other ionomers tested in systems circulating KOH. This was attributed to Nafion providing more free volume for the electrolyte. This topic raises still many questions, as a recent study [50] has clearly shown that for supporting electrolyte-fed AEM devices, the ionomer is not really useful for ion conduction through the catalyst layer as the ionic strength is large enough to ensure the current. However, the ionomer was shown to play a key role for the catalyst layer structure and its stability. Intermediate ionomer contents lead to the lowest overpotentials, highest effective surface areas, and lowest catalyst layer resistances. The stability was linked to catalyst adhesion and ionomer loss. One open question is to know if the ionomer solution should be similar to the precursors used to synthetize the ion-exchange membrane. This study suggests that different chemistries can be used, but it remains difficult to say if it is a general conclusion. It is interesting to notice that ionomers on both side of the membrane allow the electrolysis of pure water [51].

Water management in AEM was studied by in situ neutron imaging, showing that a higher ion-exchange capacity of the ionomer provided a better water retention, reducing the drying out of the cathode at high current densities [52]. Indeed, dry cathode operation leads to a water distribution characterized with a high-water content at the anode and therefore a gradient inside the MEA toward the dry cathode. Further, this gradient increases with current density, as water is consumed in the hydrogen evolution reaction at the cathode and is further transported to the anode by both electro-osmotic flow and dragging by the migrating hydroxide anions.

Figure 8 shows a schematic view of a porous cathode AEM cell comprising porous electrodes, metallic or carbon based, sandwiching the membrane. The major difference here is the importance of the water fluxes to the cathode where it is reduced to produce H₂ and OH⁻. In AEM electrolysis, K⁺ plays no active role, but its solvation helps to increase the water content.