Alkaline fuel cell technology - A review

(cid:1) Alkaline-based discoveries mostly relate to anionic exchange membranes (AEMs). (cid:1) Improving the stability of abundant and low-cost catalyst materials is required. (cid:1) Notable progress has been made creating AEMs with good stability and conductivity. (cid:1) Ionomer-catalyst compatibility is essential during MEA creation. (cid:1) A transition from lab-scale to application-level is required to advance AEMs.


Introduction
This review article presents the state-of-the-art in the field of alkaline fuel cells.The main motivation is to provide a single source of information that can be accessed by researches and developers who wish to advance the know-how and create new alkaline fuel cell technologies.
Initially, a brief synopsis of the underlying electrochemistry is given before delving into electrocatalytic materials.The latter will be rigorously reviewed, considering both traditional liquid electrolyte alkaline fuel cells (AFCs) and the more recently polymer anionic exchange membrane fuel cells (AEMFCs) may utilise similar catalyst materials.The AFC description begins with a short historical account of its evolution and elaborates on the materials used, fabrication methods and stacking concepts.The recent trend towards polymer anionic exchange membranes (AEMs) is given thorough coverage as this is deemed the future of alkaline systems.The intrinsic chemistry of the AEM will be briefly detailed in this section, with additional coverage of the most common methods for creating AEMs.AEMFC performances are included and discussed, likewise with ionomer development.Finally, a summary is given where key remaining challenges are outlined, and the road ahead is mapped out.

Universal components to alkaline-based fuel cells
All fuel cells are based on the principle that fuel reacts at the anode, and air or oxidant at the cathode of an electrochemical system that produces electric power and some excess heat.In this regard, alkaline-based fuel cells are no different from acidic proton exchange membrane fuel cells (PEMFCs).The general workings of an alkaline-based fuel cell, utilising either a liquid or a polymer electrolyte is exhibited in Fig. 1 and the general characteristics in Table 1.
Humidified hydrogen gas is supplied to the anode, which reacts with the hydroxide ions in the electrolyte to produce water and electrons after penetrating the GDL and reaching the catalyst layer (CL), as displayed in Eq. ( 1) A humidified oxygen source, typically purified air/oxygen is supplied to the cathode together with water.Oxygen gas, solvated in water, is reduced at the cathode CL to form hydroxide ions, which diffuse through the electrolyte to participate in the hydrogen oxidation reaction (HOR) that takes place on the anode.The ideal oxygen reduction reaction (ORR) also labelled the direct 4-electron pathway is displayed in Eq. (2).
Additional highly relevant details to this reaction are further elaborated under the sub-section Oxygen Reduction Reaction.The red-ox reactions in Eqs.(1) and ( 2) are combined to form the overall mechanism given in Eq. (3).
Equivalently with every other reaction, there are various inefficiencies and losses which lower the achievable potentials in Eqs.(1)e (3).These losses are generally defined as overpotentials and are categorised as charge transfer, ohmic or mass-transport overpotential.These overpotentials variably affect a cell at specific current densities.Charge transfer overpotential is mostly linked with activation losses and is most visible at low current densities.Ohmic overpotential originates in the electrical resistance of the cell components, determines the slope of the current-voltage graph and is most visible at intermediate current densities.At high current densities the fuel cell will be limited by the speed at which reactants can reach the electrode, inducing mass transportation losses.
Table 1 below displays typical characteristics of both traditional AFCs and contemporary AEMFCs, where record performances and space-related values are included to give a complete view of the two technologies.Comparing the cost of the two technologies is hard, as there is a void in literature on the topic of total AEMFC system costs.One estimation of AEMFC power cost (V kW À1 ) projected it to lie in the range of ~600 V kW À1 (in 2015) [1].A conservative estimation is that this figure has most likely decreased by a third by 2021, given the increase in AEMFC performance since 2015.
Considering AEMFC technology competes with PEMFCs, complete AEMFC systems must achieve the targeted cost of ~25 V kW À1 as set by the US Department of Energy (US DOE) to excel past PEMFCs [2].Traditional AFC stack system costs were estimated to 100-155 V kW À1 in 2005e2007 by AFC Energy and experts in the field (~104e170 V kW À1 today) [3,4].

Alkaline catalyst materials
The following sub-section summarises the developments within the various classes of HOR and ORR catalyst materials in an alkaline environment.Focused reviews on this topic may be found in literature for the HOR [14e16] and the ORR [14,17e19].

Hydrogen oxidation reaction
The hydrogen oxidation reaction and its complement the hydrogen evolution reaction (HER) are two important reactions in several technologies such as fuel cells, chlorine manufacturing and water electrolysis.Comparative to the HOR, the HER has been thoroughly studied through the development of alkaline water electrolysers.There is a notable asymmetry between the anodic (HOR) and cathodic (ORR) fuel cell reactions in terms of applied R&D efforts due to the early focus on surmounting the low exchange current density (ECD) of the ORR (10 À4 mA cm À2 [20] for ORR vs 1 mA cm À2 [20,21] for HOR).Consequently, there are many remaining challenges related to developing and understanding electrocatalytic materials for the HOR.There is a large variety of possible fuels for AFC anodes, specifically borohydride, hydrazine, ammonia, methanol, ethanol, and ethylene glycol.The choice of fuel will in turn determine which catalyst materials are available as these depend on the source of hydrogen.The ensuing sub-section pertains only to catalysts developed for the HOR using pure hydrogen gas as fuel.
Studies on HER/HOR have revealed notably slower reaction kinetics in alkaline electrolytes compared to acidic ones, lagging with approximately two magnitudes using the exchange current density (ECD) as a measuring factor [2].A study on the pH dependence of the HER/HOR affiliated it to a change in the configurational entropy as the proton approaches the electrode surface [22].Other studies have suggested the hydrogen binding energy (HBE) as a possible sole descriptor of HER/HOR activity [23].These two theories are joined by a third, that the dissociation of water carries an addition energy barrier which determines the rate of the HER [24].The exact cause of the poor hydrogen reaction kinetics under alkaline conditions is Fig. 1 e The essentials of alkaline-based fuel cells, where layer 1 and 5 represent the anode/cathode gas diffusion layer (GDL), layer 2 and 4 the anode/cathode catalyst layer (CL) and layer 3 the liquid/polymer electrolyte.
Table 1 e Fundamental features of traditional alkaline fuel cells and the more contemporary anion exchange membrane fuel cells.

AFC [5e10]
AEMFC [5,11e13]  still unknown, and experimental/theoretical findings support all three theories, hence there is uncertainty as to which of the aforementioned theories is closest to describing the cause.The HOR mechanism in alkaline media involves the Tafel and/ or Heyrovsky reactions, followed by the Volmer reaction: The overall reaction is given in Eq. ( 1).Either the Tafel or Heyrovsky reaction may be the rate determining step (RDS) at low overpotential depending on the catalyst activity in alkaline media, whereas diffusion of dissolved H 2 in the electrolyte has been proposed as the RDS at high overpotential.Many publications specify the Volmer step as rate determining [23,25].
Platinum groups metals (PGM).Platinum (Pt) represents the pinnacle of electrocatalytic performance regardless of whether hydrogen is reduced or oxidised [21].This has been a favoured material in AFCs as a single, binary, ternary or bimetallic combination similarly with PGMs like palladium, iridium and ruthenium [7].The acidic ECD of the HOR on platinum lies two orders of magnitude above its alkaline counterpart, in the region of 1 mA cm À2 Pt for polycrystalline Pt and carbon-supported Pt [21,26].The experimental ECD was investigated on commercial Pt/C (46 wt.%) and polycrystalline Pt in alkaline conditions, resulting in values of 0.57 and 0.69 mA cm À2 respectively [21].
A breakthrough HOR performance rendering noble-metals obsolete has yet to be displayed, thus many studies have compromised in seeking to instead lower the loading of noblemetals.Hence the effects of Pt/C particle size were investigated [27] and mass activity peaked at 3 nm, though specific activity increased with particle size until levelling off to a plateau value.This increase was correlated with a decline in the ratio of less active edge-atoms [25,27].
Several PtM (M ¼ Fe, Co, Cu, Ru, Au) nanowire alloys performed admirably against the Pt/C (20 wt.%) benchmark e.g. the Pt 7 Ru 3 yielding an ECD of 0.493 mA cm À2 , insinuating increased HOR activity from single-dimensional materials [28].Moreover, a PteNi/C electrode displayed that 3.1 % Pt with 13.9 % Ni resulted in a greater ECD than pure Pt [16,29].Palladium (Pd) and Pt are similar elements as they both stem from the same group, however Pd is 50x more abundant and paradoxically twice as expensive at the moment [30].Particle size effects for Pd followed the same trend as Pt [27], the increasing activity levelled off at 19 nm producing an ECD of 0.122 mA cm À2 Pd which was attributed an increment of sites with low hydrogen binding energy (HBE).
Pd adlayers were deposited through electroplating Pd 2þ ions, and through galvanic displacement of underpotential deposited Cu-atoms on a polycrystalline Au substrate [31].The resulting surfaces comprised 3D Pd structures of unknown depth, where the measure q Pd/Au (Pd on Au coverage) varied inversely with HOR activity in 0.1 M NaOH.The ECD varied from 0.022 to 0.003 mA cm À2 Pd from low to high Pd coverage.
Encouraging results have been achieved with PdeCeO 2 /C anodes, where mass specific ECDs of 20 A g À1 Pd and average Tafel slopes of 60 mV dec À1 have been reported [32].The PdeCeO 2 interaction was found to be related to the HOR activity, thus it was optimised resulting in the aforementioned results.Density functional theory (DFT) calculations allude that the interaction between Pd and the oxygen atoms of CeO 2 weakens the HBE of the catalyst relative to Pd{111}, implying that the PdeCeO 2 framework weakens hydrogen adsorption, thereby enhancing HOR kinetics.
Despite the considerable price and scarcity of PGMs such as ruthenium and iridium, they are still used in the development of electrocatalytic materials however their employment for AFC applications is increasingly niche compared to the increasing R&D focus on non-PGM catalysts.One example includes work by Wang et al. [33] where alloys of iridium, palladium and ruthenium on carbon were found to display greater HOR activity measured against both Ir/C and Pt/C in their pure forms.Namely, the alloys Ir 9 Ru 1 /C and Ir 3 Pd 1 Ru 6 /C were found as the most active, exhibiting ECDs of 0.9 and 0.6 mA cm À2 cat: , compared to Ir/C (0.4 mA cm À2 cat: ) and Pt/C (0.5 mA cm À2 cat: ).The growth in catalytic activity was partly credited a decline in HBE.
Transitional metals.Pt is the best electrocatalyst for the HOR, where its performance remains the benchmark against which all non-PGM catalyst materials are compared.Nickel (Ni) based materials have been comprehensively studied as replacements for PGMs due to their high abundance, moderate cost and immunity towards corrosion at high pH.Raney nickel catalysts were an early recipient of attention as a non-PGM HOR catalyst due to its inherent high surface area.The group remains among the most active non-PGM catalysts toward the HOR and was an early non-PGM favourite in AFCs as they could endure the severity of 6.0 M KOH solutions [34,35].It should be emphasised that it is the high pH of AFC systems that allow metals such as nickel to be considered due to their immunity towards corrosion.Raney nickel is commonly used as support material due to poor HOR activity (ECD z 0.00045 mA cm À2 cat: ) in addition to notable stabilityrelated issues meaning alterations through e.g.doping is necessary [36].Doping with Ti 2 or Mo 21.3 increased the ECD by a factor of 3.9 and 2 respectively.However this material does not harmonise with contemporary AEMFCs, as the anode exhibits nickel oxide traits (passivation) causing deactivation [37].Furthermore, it is only functional at high KOH concentrations (z6.0 M) rendering the comparatively low alkalinities of AEMFCs (z0.1e1.0M KOH) insufficient.Additionally, Raney nickel's activation process is not realisable in AEMFCs, as it involves applying a cathodic current to reduce the nickel oxides (Ni(OH) 2 ) which elicits hydrogen evolution thus requiring a bifunctional ORR/OER catalyst [5,37,38].
The activity of nickel nanoparticles on carbon was found to increase significantly upon mildly oxidising the catalyst surface [39].The transformation from Ni/XC-72 to NieNiO/XC-72 results in a specific ECD increasing from 0.0062 to 0.056 mA cm À2 cat: .The partial oxidation of the catalyst surface results in NieNiO/XC-72, where hydrogen is adsorbed onto the nickel surface and the nickel oxide adsorbs hydroxide species.
Usually nickel will strongly adsorb hydrogen resulting in poor HOR activity, however the oxidation process causes an increase of the Gibbs energy of adsorption on the Nickel sites close to surface-NiO, effectively enhancing HOR activity through lowering the overpotential at which the HOR begins.
To increase the HOR activity and stability of nickel, a number of transitional metals (Mo, Co, Fe, Cu) were alloyed through various means creating catalysts such as NiMo [40] and NiCu [41].Using several in-situ spectroscopic analysis methods, it was determined that the molybdenum did not actively contribute to the HOR, though it affected the electronic structure of nickel by lowering the HBE to more thermoneutral values thereby increasing the alloy's HOR activity [40].The NiMo on Ketjenblack (KB) produced a ECD of 0.027 mA cm À2 cat: (4.5 A g À1 cat: ) which is highly competitive with other nickel-based electrodes.Furthermore, the resulting MEA remained stable for z 100 h, however significant degradation was displayed between 100 and 115 h.
An array of NiCu alloy films were created by employing a magnetron cosputtering method, where the films exhibited a volcano-like dependence on the increasing atomic fraction of copper [41].The ECD peaked at 40 at.%copper producing 0.034 mA cm À2 cat: and increased the material's resistance to oxidation.The role of transition metals (TMs) in binary Ni-TM alloys was investigated where the authors suggested that the binding energy of both hydrogen and hydroxide should be collectively considered as descriptors of HOR activity [42].
Moreover, an excellent Ni 3 Fe 1 /C catalyst material was created resulting in an ECD of 0.060 mA cm À2 Ni g À1 Ni , (0.96 A g À1 Ni , low due to the considerable concentration of catalytically inert TMs).Employing a chemical reduction synthesis method might be beneficial for creating efficient HOR electrocatalysts compared against e.g. a thermal reduction synthesis method resulting in the NiMo/KB electrode [40].
The stability of several binary Ni 3 TM (TM ¼ Mo, Fe, Co, Cu) alloys was investigated, where all except Ni 3 Mo retained initial activity following 1000 cycles (z720 h between 0 and 0.3 V RHE ), however extending the range to 0.7 V RHE caused the formation of Ni(OH) 2 through dissolution precipitation for all cases [43].
Ternary nickel alloys have also been explored as HOR electrocatalyst materials, where a CoNiMo electrode displayed an excellent HOR activity when electroplated onto a gold disk [44].The ternary material outperformed electroplated NiMo and pure Ni, and was responsive to the ratio of its components.The ideal ratio determined was 0.12:5.10:1.00,resulting in an activity 20x that of Ni and was expected to surpass that of Pt.While the catalyst displayed a noteworthy improvement over pure Ni, it still failed to match the Pt benchmark when measuring the ECD (0.015 mA cm À2 cat: vs 0.610 mA cm À2 cat: for Pt).The employment of heteroatoms such as nitrogen and boron as dopants in TM-based catalysts is an established practice in the development of alkaline ORR catalyst materials, and this trend has spread to HOR catalyst development.By utilising heteroatoms as dopants, the electronic structure of the TMs changes, thereby altering the HBE and increasing HOR activity.A Ni 3 N/C catalyst was developed using a hydrothermal method resulting in a specific ECD of 0.014 and 0.017 mA cm À2 cat: for carbon-supported and unsupported Ni 3 N respectively [45].Moreover, the supported catalyst yielded remarkable mass normalised activity of 12 A g À2 cat: .Investigations using spectroscopic techniques revealed a decrease in d-band energy when comparing Ni to Ni 3 N. Interfacial charge transfers between Ni 3 N and the carbon support and a decline in d-band energy resulted in a likewise lowering of the binding energies of oxygen species and hydrogen thus improving HOR activity.The ECD retained 59.8% of its original value after 5000 cyclic voltammetry (CV) scans, thereby displaying the stability of the Ni 3 N/C catalyst.
A similar nitrogen-doped TM catalyst was created through cathodic electrodeposition followed by thermal nitridation producing a Ni 3 N/Ni/NF (NF ¼ nickel foam) catalyst [46].The performance of the Ni 3 N/Ni/NF surpassed that of Pt/Ni and Ni/ NF by 1.4x and 17.6x respectively as shown in Fig. 2a, producing an ECD of 3.08 mA cm À2 geo: (calculated from the geometric area of the electrode) or 0.003 mA cm À2 cat: .The catalyst was efficient for both hydrogen evolution and oxidation, surpassing many established PGM and non-PGM HER catalyst materials as seen in Tab.S3 in Ref. [46].The geometric and electronic build of Ni 3 N/Ni resulted in an intersection between Ni 3 N and Ni which yielded excellent sites for hydrogen adsorption with free energy changes close to zero according to DFT calculations.The stability and COtolerance was tested, and the Ni 3 N/Ni/NF exhibited agreeable results compared to the Pt/NF benchmark, as displayed in Fig. 2b-d.
While the aforementioned efforts applied the nitrogen dopant on the catalyst itself, a different group elected to modify the carbon support [47].The electrocatalytic performance of Ni nanoparticles was increased by the N-doped carbon nanotube (CNT) support by 21x (ECD) and 33x (mass activity).The increase of ECD and a mass activity to yield 0.028 mA cm À2 cat: and 3.5 A g À1 cat: was explained by DFT calculations.The nitrogen-dopant (specifically the edge-placed N atoms) affected the d-band structure of Ni, resulting in lower HBEs and the support material stabilised the Ni nanoparticles against reconstruction.
Similar results were reported in a more recent study, for an array of Ni catalysts with carbon supports doped with either sulphur, boron or nitrogen created through a two-step pyrolysis and reduction method [15].
The three catalysts all showed an increase in electrochemical surface area (ECSA) compared to the pure Ni/C as seen in Fig. 3 .The augmented ECSA was associated with the dopants enhancing the production of small uniform particles due to an anchoring effect, resulting in additional nucleation sites and higher nucleation rates.The increased specific activity was affiliated with the interaction between the catalyst and the support lowering the HBE.
To provide perspective, the specific exchange current densities of all HOR catalyst materials hitherto mentioned in this review are summarised in Table 2. Further figures displaying HOR catalyst performances may be found in the following review papers [14e16,48].

Oxygen reduction reaction
The cathodic oxygen reduction reaction (ORR) has occupied the limelight in terms of fuel cell catalyst development ever since the inauguration of the fuel cell.This is due to the large possible overpotential associated with the laborious transfer of four electrons in addition to several possible intermediates.
The ideal ORR displayed in Eq. ( 2) is also named the direct 4-electron pathway, as it constitutes one of two possible ways this reaction may proceed by (as is commonly accepted).The ORR is a complicated process involving intermediates and the transfer of four coupled proton and electron steps, which leads to several viable reaction pathways.There is still uncertainty surrounding the exact progression of the mechanism, and its complete characterisation is viewed as one of the leading challenges of its field.Due to the high pH in alkaline ORR, superoxide and peroxide ions take part in the other reaction, the alternative "2 þ 2 electron" pathway displayed in Eqs. ( 7) and (8).
The peroxide ion reacts with the solvent and electrons to form hydroxide ions.
Another pathway for the produced peroxide includes catalytic decomposition, forming dioxygen and OH À as shown in Eq. ( 9).
Due to the two possible reaction pathways, emphasis is laid upon creating catalyst materials which promote catalysis through the most efficient way, namely the four-electron pathway.
Platinum groups metals.The exchange current density of the ORR on platinum is 10 À4 mA cm À2 , i.e. two magnitudes below that of the HOR which implies the possibility of a significant  i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 6 ( 2 0 2 1 ) 1 8 4 8 9 e1 8 5 1 0 overpotential [14,20].Platinum is the most efficient catalyst for the ORR with an onset potential of 0.94 V RHE , a kinetic current density of 31.5 mA cm À2 at 0.8 V RHE [50] and a fourelectron reduction of oxygen similarly to all other PGMs.However, at very low Pt/C ratios the overall number of electrons exchanged decreases to approximately two because of the carbon contribution.Keeping the Pt/C ratio equal to 60 wt.%Pt ensures a four-electron reaction [51].Various techniques have been utilised to reduce the loading due to the high cost of Pt and other PGMs, namely the usage of nanoparticles or alloying PGMs with TMs.Platinum alloys generally exhibit a higher activity and stability than pure Pt [52,53].The enhanced electrocatalytic activity of the alloys has been explained by a several theories, specifically a decline in bond distance of PtePt hence promoting the adsorption of oxygen.Additionally, the Pt 5d orbital electron density and the presence of surface oxide layers [51].
Palladium (Pd) is the principle PGM catalyst used to replace Pt used either as an alloy or in its pure form due to their similarity [30].Pt outperforms Pd in their pure forms, however a Pd-TM alloy reverses this trend [54].The increased catalytic activity is caused by a decline in the necessary Gibbs free energy affiliated with the electron steps caused by a coupling effect related to the d-orbital between the metals, as Pd and the TM have a low and high d-band occupancy respectively [55].Similarly with Pt, smaller Pd particle sizes don't always result in increased activity, as the specific activity of 27 nm Pd nanocubes was approximately 4x higher than 3 nm Pd spheres, thereby correlating with earlier reported size-related effects [56,57].This has been attributed to the active sites being blocked by strong OH À adsorption on small particles [58].While Pd might seem an appropriate replacement to Pt, the current associated cost is considerable as the raw cost of Pd greatly exceeds that of Pt.
Non-PGM noble metals.Contrary to its rarity, gold (Au) {100} is the most active electrocatalyst towards the alkaline ORR and the low overpotential of 0.34 V in 0.1 M NaOH even beats Pt {110} (0.48 V) in 0.1 M perchloric acid [59,60].While the four electron pathway was utilised by the {100} Au facet, it only occurred in a tight potential range.Three monofacet crystal Au nanocrystals were synthesised by Lu et al. [61] by regulating the form of the particles.It was displayed that clean single-crystalline surfaces could be utilised to produce ORR behaviour similar to bulk gold single-crystal surfaces by employing octahedra and cubes encased with {100} & {110} surfaces.Moreover, the narrow potential range was avoided by using a truncated ditetragonal prism (all sides with {310} facets) which displayed a four-electron charge transfer for the entire investigated range (0e0.9V RHE ).
To decrease catalytic expenses, gold alloys have also been explored.Through the chemical reduction of metal salt precursors 1-dodecyne, AgAu alloy nanoparticles were created with average core sizes in the range of 3e5 nm [62].The optimal performance was measured at 35.5 at.%Au as measured by kinetic current density, number of transferred electrons and onset potential, and was found comparable to the industrial standard Pt/C.The great performance was attributed to an alloying effect causing increased metal-oxygen affinity in addition to the metal-ligand interfacial bonding interactions.Through a simple one-pot hydrothermal method a self-supported AuCuCo catalyst was developed.The bifunctional catalyst displayed a similar electrocatalytic performance to the benchmarks 20 wt.% Pt/C and IrO 2 , in addition to enhanced stability [63].This performance was explained by the selfsupported 3D-structure, the alloying effect between the three components and rich twin crystals.
Silver (Ag) is another noble metal considered as a potential replacement for Pt because of its high activity towards the ORR and relatively low cost [64].The ORR can occur with either the "2 þ 2", the four-electron process or a mixture, contingent on the electrode potential and the oxidation state of the surface [51].The ideal weight % of Ag to achieve n ¼ 3.6e3.8(n ¼ number of electrons exchanged) was determined 20 wt.% [65].Studies on the shape of silver nanoparticles allured that the ORR on silver nanodecahedra contained by the {111}structure occurred through a direct four electron pathway, contrary to {100}-structure containment [66].Usage of the {111}-structure resulted in weaker OH-adsorption, yielding more active sites and higher ORR activity.

Carbon based catalysts. Non-metal doped carbon catalysts
represent a new possibility for replacing PGMs, as these materials have displayed curiously beneficial qualities against the alkaline ORR.Usage of heteroatoms as nitrogen and boron as dopants in support materials such as carbon black or CNTs has produced most agreeable results.This increase in activity has been bestowed onto the charge polarisation elicited by a difference in electronegativity between carbon and dopant, however a complete description has yet to arise [17].N-doped carbon nanocages displayed satisfactory results with the ORR mainly proceeding via the four electron pathway and had a comparable performance to Pt/C (20 wt.%) in regard to onset and half-wave potential and limiting current density [67].Moreover, the stability was shown by retaining 88.25% of its activity after 5000 CV cycles.A nitrogen on iron-doped carbon nanofibers and graphene (N/Fe-CG) catalyst was created through electrospinning, resulting in active FeeN 4 sites in the centers [68].The catalytic activity of the N/Fe-CG catalyst was equivalent to the Pt/C (20 wt.%) benchmark which was outperformed in terms of methanol resistance and stability (in both acidic and alkaline conditions).
N-doped CNTs were utilised as support for a NiCo catalyst where the great activity and stability was accredited a synergistic effect between the active sites of the bimetallic catalyst and the heteroatom-doped carbon support [69].The stability was determined via chronoamperometry (~28 h at 0.6 V RHE ) and cyclic voltammetry (5000 cycles between 0.5 and 1.1 V RHE ), where the NiCo/N-CNT catalyst outperformed the 46 wt.% Pt/ C benchmark in retaining 95% of its initial current density compared to the 69% the benchmark managed to withhold.
Porous carbons were decorated with atomically dispersed FeeN x species and B-centers (FeBNC) and pyrolysed at various temperatures.The FeBNC-800 catalyst displayed a high activity with a four electron ORR and a high onset potential of 0.968 V RHE , comparable to the 20 wt.% Pt/C benchmark (1.012 V RHE ).Furthermore, the current densities of the two materials at 0.40 V RHE were highly comparable at 5.51 and 5.41 mA cm À2 for the FeBNC-800 and 20 wt.% Pt/C respectively.DFT calculations implied that the porous carbon was embellished by the incorporation of the B dopant into FeeN x hence lowering the energy barrier for oxygen reduction and improving the overall performance [70].
Combining a solvothermal method with in-situ doping resulted in sulphur-doped carbon sphere electrodes with a varied atomic % of sulphur [71].The performance of the material peaked at 4 at.% with a four electron ORR and an onset potential and current density (at 0.5 V RHE ) of 0.882 V RHE and 5.06 mA cm À2 .Compared against the 20 wt.% Pt/C benchmark (onset ¼ 0.986 V RHE and 5.87 mA cm À2 at 0.5 V RHE ), the performance was most impressive.The stability of the sulphurdoped carbon sphere electrodes determined by CA (24 h at 0.7 V RHE ) was superior to the Pt/C benchmark, the former retaining 69% of its current vs. 41% for the Pt/C.Phosphorus was doped into hollow carbon spheres (P-HCP), employing a hydrothermal method where an optimised P-HCP electrode proceeded with a four electron reaction displaying an onset potential of 0.887 V RHE and a current density of 5.4 mA cm À2 (at 0.6 V RHE ).The performance was similar to the 20 wt.% Pt/C benchmark at same loading [72].
Transitional metal oxides.Transition metal oxides represent several sub-categories of catalyst materials, where this review includes spinels, perovskites and manganese oxides.Spinels are ternary oxides named after the mineral spinel (MgAl 2 O 4 ) with the general structure AB 2 O 4 which is displayed in Fig. 4. The B-cation is critical as it plays an important role in the activity of the catalyst.The anions of the oxide are organised in an enclosed cubic formation where the A-cation occupies an eighth of the tetrahedral sites, and the B-cation half of the octahedral sites [73e75].
A surprisingly high ORR activity was discovered using Co 3 O 4 on moderately reduced nitrogen-doped graphene oxide (Co 3 O 4 /N-mrGO) [76].The Co 3 O 4 /N-mrGO displayed a performance highly comparable to the 20 wt.% Pt/C benchmark, moreover with enhanced stability and durability.An onset potential of 0.930 V RHE and a four electron ORR illustrated the high electrochemical activity of this catalyst.
Perovskite oxides have an ABO 3 -type crystal structure (see Fig. 4) similar to calcium titanate (CaTiO 3 ) [77], where A is an alkali or more commonly a rare earth metal, and B a transitional metal.The perovskite structure consists of BO 6 octahedra (in the center) with cations from A, forming a cubic structure around the center as seen in Fig. 4 [78].The perovskite structure in Fig. 4 is adaptable and can be subject to considerable lattice mismatch, hence a variety of dopants are eligible on the A and B site [19].Doping La 0.8 Sr 0.2 MnO 3Àd with both scandium and phosphorus forming La 0.8 Sr 0.2 MnO 0.95 Sc 0.025 P 0.025 O 3Àd resulted in a significant enhancement in onset potential from 0.860 to 0.960 V RHE [79].Compared against the 20 wt.% Pt/C benchmark, the catalyst displayed a corresponding activity (Fig. 5) and improved stability during a current cycling test.
Manganese oxides have garnered attention due of their low cost, good catalytic activity toward the ORR and abundance, as Mn is the tenth most common element [80].Various manganese oxide crystallographic structures were investigated, where the activity of a-MnO 2 was superior to b, d and amorphous structures [81].The a-MnO 2 had an onset potential of 0.760 V RHE (compared to 0.860 V RHE for 20 wt.% Pt/C) and increased its stability from 3 to 8 h when doped with Ni.Utilising Koutechy-Levich plots, the electron transfer number of a-MnO 2 was calculated to 4.2, implying a four electron ORR.
Similarly to the HOR the onset potentials and current densities of all ORR catalyst materials previously mentioned in this review are summarised in Table 3 to provide perspective for ORR.Further figures displaying ORR catalyst performances may be found in the following review papers [14,17e19].

Summary
The focus on developing non-PGM HOR/ORR catalyst materials has increased momentously in the past decade, and accelerated in the period from 2016 to the present with the increase in recorded AEMFC performances.Increased use of in-situ spectroscopic techniques may hold the key to unmasking the particulars of both the HOR and the ORR in regards to intermediates and changes in structure [16].Insight gained from such advances can hold the key to elevating the performance and longevity of non-PGM catalysts.

Alkaline fuel cells with a liquid electrolyte
Historically the first real, "useful" fuel cells were based on liquid alkaline electrolytes and metal electrodes using a form Fig. 4 e The general structure of perovskite and spinel transition metal oxides respectively. of "zero gap technology" in which the metal electrodes were placed back to back against a thin electrically insulating diaphragm immersed in the liquid electrolyte.The idea of the diaphragm was to prevent the cross-over of hydrogen and air and allow the chemical reactions to take place.In order to supply fuel and air it was necessary to develop a gas permeable electrode that allowed the reactants to reach the active sites of the catalyst but at the same time prevent the flow of liquid electrolyte into the gas channels.This was the birth of the gas diffusion electrode and was a pivotal invention that has subsequently seen many changes over the years.The details of these basic ideas are explained in the following sections.

Gas diffusion electrodes
The Constitution of modern AFC gas diffusion electrodes (GDEs) typically include multiple PTFE-bonded carbon black (CB) layers, and the prevailing structure is the double-layer electrode structure shown in Fig. 6, embodying a backing material (BM), a gas diffusion layer (GDL), and a catalyst layer (CL).The location of the BM depends on the design, and can be placed in the GDL, in the CL or in between.A BM requires high gas permeability, structural strength, electronic conductivity and corrosive resistance.AFC current collectors may be composed of e.g.nickel due to its low cost and resistance to corrosion, where meshes or foams are frequently used.Similarly to PEMFC designs, porous carbon paper or carbon cloth may also be utilised [82].
Nickel foam with its substantial pore size and porosity increase liquid/gas flow and extend the electrochemical area, thus enhancing catalyst performance [83].Furthermore, coating it with various metals has resulted in improved performance.Compared to regular nickel foam, its silver- [84] and gold-plated [85] varieties exhibited lower ohmic and charge transfer resistances.
Reactant gas is supplied to the CL through the GDL, which stops the liquid electrolyte from passing through the electrode.Some leakage is still possible due to the formation of water, an occurrence termed "weeping".The GDL can be fabricated from pure porous PTFE (either PTFE suspension or powder) mixed with ammonium carbonate, where sintering  (<320 C) causes gas bubbles to form resulting in a porous PTFE film [10].Such a GDL has a high electronic resistance, hence should only be used for mono-polar stacks (Fig. 8) such as the Elenco/Zetek designs.
An ideal GDL should be hydrophobic and electronically conductive, where the latter may be achieved by adding CB to the mixture featuring 25e60 vol.%PTFE [86], enabling usage in a bipolar design (Fig. 8).The CL in Fig. 6 comprises the catalyst and its support material bonded with PTFE, where highsurface (500e1000 m 2 g À1 ) support materials like CB are frequently used to increase power density.Other carbon structures can substitute CB as support to great effect, such as carbon nanotubes (CNTs) [17,47,58], carbon spheres [71,72] and other carbon architectures [67,68].Stability-issues with carbon supports have birthed a new direction utilising TMbased support-materials for catalysts.
Carbon black is active due its high surface area, a facet which originates in the high temperatures involved in its creation and the pre-treatments specific to each CB type.The surface area of the CB type Ketjenblack was optimised in a N 2 atmosphere at 900 C, while Vulcan XC-72 required CO 2 [87].
Complete utilisation of carbon black's high surface area and the affiliated electrochemical activity is coupled to the catalyst's malleability toward uniform distribution over a substrate.Areas not covered are electrically conductive, however they do not participate in the electrochemical reaction.Furthermore the triple phase boundary (TPB, see Fig. 6) between gas, electrolyte and catalyst should cover the entire surface, as catalyst agglomeration causes uneven wear leading to decreased performance and endurance.
The PFTE creates a TPB on the substrate and works primarily as a binder, where the loading (5e25 vol.%) depends on the required hydrophobicity.A metal sheet/grid fixed to the GDL facilitates current collection [88,89].The traditional design in Fig. 6 has been challenged by integrated anodes, where the GDL and CL were combined through a PTFEadhered palladium-plated nickel foam [90].While this design increased the power density, the usage of palladium is questionable.Alternatives to PTFE include the amorphous fluoropolymer material Teflon-AF [91] and Gortex (expanded PTFE) [92] where both have displayed improved AFC performance when measured against PTFE GDEs.
Fabrication procedures are highly influential for electrode performance, thus requiring specific emphasis.Displayed in Fig. 7 is a general electrode creation procedure, where many variations exist.Electrode preparation methods are termed either dry or wet methods, pending on whether for instance the binder PFTE is used as a powder or a suspension liquid.The ink-paste method is a common method for creating electrodes, which involves mixing a dispersant and a binder, adding a catalyst powder to create a paste/ink which is subsequently distributed over a substrate either by rolling or screen-printing.Ultrasonication prior to distribution over the substrate ensures a uniform catalyst distribution in the paste/ ink, decreases diffusion path lengths and increases the number of gas dissolving sites in the CL.
Recently more advanced methods have been tested, namely electron beam physical vapour deposition which displayed comparable performance to the ink paste method despite a lower catalyst loading, though at an increased cost/ difficulty [83].
The process displayed in Fig. 7 is general and variations using filling materials such as ammonium carbonate and added washing/drying stages are hardly uncommon.Development of AEMFC electrodes is an emerging discipline with many approaches derived from traditional AFCs/PEMFCs [13,93].
Durability is naturally of paramount importance for an electrode.Anodic degradation rates for an AFC using a Pt/Pd on charcoal GDE were measured to 3.4 mV h À1 after 11,500 h at 100 mA cm À2 [7], while the rate for a Raney nickel-based GDE was calculated to 24 mV h À1 after 1500 h [8].Cathodic degradation rates for an AFC employing Pt-based cathodes have been quantified to 10e30 mV h À1 after 3500 h at 100 mA cm À2 [9], 5e10 and 20 mV h À1 at 100 and 200 mA cm À2 respectively after 2800 h [10] and 17 mV h À1 for Ag-based GDEs after 3500 h at 150 mA cm À2 [94].
There are three principle sources to AFC performance degradation, adsorption of CO 2 , corrosion and weeping.CO 2 is damaging to AFC performance as it reacts with free OH À ions to form CO 2À  3 in turn forming a carbonate species (Eq.( 10)), exemplifying one of the great downsides to AFCs.
Corrosion of carbon materials and deterioration of the hydrophobicity of PTFE due to the electrolyte has been observed, where HO À 2 radicals from the ORR oxidised the carbon [95].Carbon-supported catalysts suffer more in alkaline than in acidic electrolytes, where alkaline decline of the ECSA was 3x that of the acidic ECSA after potential cycling between 0.1 and 1.23 V RHE in 0.1 M NaOH [96].Time-dependant electrode performance reduction is generally caused by an electrode structure flooded with electrolyte, thereby reducing oxygen admittance to the active sites [10].This anomaly is evidenced by an increased cell capacitance due to a greater part of the electrode is now in contact with the electrolyte.Gortex membranes used in GDEs exhibited increased resistance to flooding compared to commercial GDEs, where 48 h of operation at ~0.858 V RHE (10 mA cm À2 ) in an 6.0 M KOH electrolyte pressurised up to 2.51 bar without leakage provides evidence of this [92].

Systems
AFC system designs are split between mono-and bipolar systems, where the former was created to simplify the manufacturing process of the GDL as pure PTFE (electrically insulating) could be used.Bipolar designs have the benefit of the interconnection between cells (bipolar plates), which requires an electrically conductive GDL (CB-PTFE composites).Both designs are displayed in Fig. 8.
Monopolar designs benefit from lower costs due to lack of bipolar plates, lower stack thickness due to the singular gas chamber between the electrodes and the ability to disconnect a cell with the remainder of the stack still operating.A significant disadvantage includes a 100 mA cm À2 current density limit due to ohmic losses from inefficient current collection [97].Bipolar designs feature a uniform current density and a higher terminal voltage, rendering it the preferred commercial design [5].
Complete AFC systems are branched into either spacial or terrestrial, the former being pressurised, without cost restraints and impelled by pure H 2 /O 2 , and the latter with lowcost components and powered by H 2 /air at ambient pressure and low temperature.

Achievements
The inauguration of all AFC systems came through Francis Thomas Bacon, who's cell was built with sintered nickel anodes and lithiated nickel oxide cathodes employing a circulating 30e45 wt.% KOH solution as electrolyte.The 5 kW monopolar system powered by pure H 2 /O 2 at 200 C at 45 bars displayed a notable performance [5].
The 1960s brought the Gemini space program where a PEMFC was utilised, though the system was found to be inefficient and unreliable due to membrane-related problems.An alkaline system based on Bacon's work was developed to replace the PEMFC for the Apollo missions.Specifications such as the extreme Pt loading of 40 mg cm À2 , an 85 wt.% KOH concentration and temperatures over 100 C at 3 bar pressure were applied in 93 cells divided between 3 parallel connected stacks resulting in a 4.5 kW AFC [98].
In the 1970s a 7 kW system was developed by Siemens comprising 70 cells driven by pure H 2 /O 2 operating at 0.4 A cm À2 at 0.8 V per cell in 7.0 M KOH at 2 bar, 80 C. The bipolar system employed a circulating electrolyte and asbestos diaphragms in each electrode to stop gas leakage from the electrolyte side.Development of systems for submarines were the main result of Siemens's R&D efforts [5].Another company involved in AFC development was the Belgian-Dutch company Elenco, who's R&D efforts resulted in a 200 kW AFC system for a hybrid bus.
Up to the 2000s, the UK company Zetek [97] (now Eident Energy) modified a London taxi to be powered by an AFC in 1999 (100 km range), in addition to the first AFC-powered boat in 2000.The majority of commercial interest in AFC development has dried up with some electing to switch to PEMFCs, AEMFCs or producing underlying components such as GDEs, membranes, ionomer, etc. [5].
Today, existing companies supplying AFC solutions include the UK-based company AFC Energy and GenCell Energy from Israel.While the former still supplies AFCs at the > 10 kW to > MW scale, the shift towards a solid-state electrolyte (AEMFCs) is pronounced.GenCell announced in 2018 a commercial system including a 4 kW AFC employing cracked ammonia (99.5%) as a hydrogen source (Project Alkammonia) as a stationary off-grid power supply [99].AFC Energy is a leading partner, as all their standard configurations have the cracked ammonia as a fuel option.

Summary
The future of AFCs appears to predominantly lie in transferring relevant experiences over to AEMFC technology to accelerate its progress.While there are some companies left that still supply complete AFC systems, the potential for development is deemed greater in PEMFCs and in the near future AEMFCs.

Anionic exchange membrane fuel cells
Anion exchange membranes (AEMs) are gaining popularity for use in alkaline electrolyser and fuel cell (FC) technologies to replace traditional AFCs.However, there are many challenges to be addressed before AEMs become mainstream, including performance, stability, durability, mechanical strength, and low-cost production methods.
Generally, AEMs offer several advantages over liquid-based alkaline systems, such as better CO 2 tolerance, decreased gas crossover, which in the case of electrolysis results in purer hydrogen and for FCs a higher open circuit voltage (OCV) and increased efficiency [12].Moreover, for FCs the problem of flooding and "weeping" is also largely avoided due to the solidstate electrolyte.One of the greatest AEM challenges is creating a thin membrane (~10 mm) which can reach a high enough conductivity (>100 mS cm À1 ) to allow it to deliver sufficient current density (>0.5 A cm À2 ) while simultaneously achieving the target of 5000 h of stable operation at 80 C [12].Average hydroxide conductivities in AEMs have reached proton conductivities in PEMs [100] (~100 mS cm À1 at 60 C [5]), and several exemplary papers report values in the range of 150e300 mS cm À1 [100e102].
Similarly to traditional PEMFCs, the conductivity of the membrane is adversely affected by low humidity, due to the mobility of the OH À -group in the polymer body of the membrane depends on the level of dissociation of the cationeanion groups which require a higher level of solvating water molecules than would be present in the dry membrane.Additionally, AEMs can suffer from embrittlement if stored in dry conditions after activation in 1.0 M KOH.For this reason AEMs are always kept humidified and may be refreshed using a dilute alkali e.g. as 0.1 M KOH, thereby removing any build-up of carbonate species.
Mechanical strength and durability is of great importance as many AEMs are easy to rupture through stresses caused by attaching it to electrode frames and especially in the process of creating MEAs, via the application of catalysts and current collectors.Thus emphasis must be laid upon creating AEMs with a high tensile strength when thin membranes in the range of ~10e30 mm are produced.Such low thicknesses enable AEM fuel cells (AEMFCs) to operate at high peak power densities (PPDs) due to lower mass transport resistance and improved water management [103,104], however this places extensive demands on the mechanical/chemical stability of the AEM [104].
Specifically, hot-pressing methods that are used to make PEM MEAs do not always work optimally with AEMs, hence other bonding methods are needed.Accordingly, some commercially available AEMs are supplied on a backing material providing additional mechanical strength, or supplied with fibre reinforcement for similar reasons.
Other companies producing AEM-related products include Tianwei (China), MEGA (Czech Republic) and Asahi Chemical Industry Co (Japan) [105e107].The Pention AEM from Xergy Inc [108,109].has been employed to produce one of the highest power density associated with commercial AEMs (3370 mW cm À2 ).Many of the commercially available membranes for FC and electrolyser applications have evolved from the chloralkali industry or desalination plants.Presently, the commercial exploitation of AEM technology lies mainly in water electrolysis, where companies such as Enapter already have small kW AEM units.For FCs, the UK company AFC Energy is developing FCs based on AEMs, however their performance data is yet to be published.
There are several possible pathways, where Fig. 9 exhibits two classic examples.First case, the main polymer chain containing a phenyl group such as a polysulphone is functionalised by a chloromethylation reaction.Secondly, a process known as "radiation grafting" is used to attach a vinyl group (vinyl benzyl chloride).The next step for both cases, the QA group is formed by treatment with an amine reagent (quaternisation).Finally, the hydroxide conductivity is induced by treatment with KOH to attach OH À to the cationic end groups as displayed in Fig. 9.
AEM ionic conductivity depends on the number of functionalised cationic groups, however there is a trade-off with mechanical stability because AEMs tend to swell up in volume as more groups are functionalised.This leads to the main challenge with AEMs e keeping a high anionic conductivity at the same time as high mechanical strength.
Another challenge lies in the paradox that the presence of OH À in the AEM causes long term degradation, because the OH À group also acts as a nucleophile that can attack the cation group, thus splitting it into methanol and an amine (Fig. 10a).The effect of nucleophilic degradation has found to be mitigated by placing the cation group at the end of the pendant alkyl chain [127].A second degradation process is where OH À attacks via an aeb Hoffman elimination reaction to form a carbonecarbon double bond and the elimination of water (Fig. 10b).
Fig. 10 displays only a fraction of the various degradation pathways to which AEMs in general are vulnerable, and more extensive accounts may be found in more focused literature [100,104,127e129].
Low-cost polymer backbones available from industry such as PEEK, PPO and polysulphone saw a fair amount of use in the early days of AEM development.However, progress has been made in the chemical stability of AEMs, where issues such as Fig. 9 e Generic chemical reaction steps to convert a polymer into an AEM with QA as cationic end-group.chain cleavage and resulting mechanical degradation common to these types of backbones have been circumvented with AEM aryl backbones without ether-bonds [104,111].Aryl ether-free AEM backbones can be approximated into two categories, aryl-aryl (poly(aryl piperidinium) (PAP) [111], polyphenylene [110], etc.) and alkyl-alkyl (polynorbornene [103,108,114], polyethylene [100,130], etc.) backbones.The latter generally displays higher peak power densities, a low solubility using most organic solvents and weak mechanical strength without PTFE-reinforcement (tensile strength ~10 MPa and > 200% elongation at break) compared to arylaryl backbone AEMs.AEMs based on aryl-aryl backbones do not require reinforcement having excellent mechanical strength (tensile strength > 60 MPa and 10e117% elongation at break).
A number of different strategies have been devised in recent years to overcome these issues, such as microphase separation, crosslinking and organic-inorganic composites [104].Microphase separation is utilised in order to increase the performance of anionic membranes through augmenting their ionic conductivity and adjusting their dimensional stability.Crosslinking and organic-inorganic composites have been utilised with great success to alter the mechanical stability of an AEM.
The concept of microphase separation implies the backbone of the polymer is hydrophobic and the cation-containing polymer is hydrophilic [128,131].In Nafion this is achieved by replacing the hydrogen atom with fluorine atoms in the main polymer chain, and non-fluorinated hydrocarbon-based polymers are used in AEMs.The microphase separation creates a region where water can be absorbed locally to improve the hydroxide ion conductivity while simultaneously controlling the amount of swelling that would otherwise degrade the mechanical strength.With this type of architecture AEMs can be fabricated with a higher anionic conductivity, while still possessing the same ion exchange capacity (IEC).There are 4 main types of AEM architectures that are based on microphase separation and these are shown in Fig. 11.These have contributed to great advancement in AEM performance in recent years.
In another approach a long alkyl chain was grafted onto a PPO backbone, forming a comb-like structure with microphase separation resulting in excellent stability maintaining 91% of its conductivity after a 700 h test in 2.0 M KOH at 60 C despite the PPO backbone [119].The comb-shaped morphology was considered responsible for the hydroxide conductivity (92.6/ 46.6 mS cm À1 at 80/30 C).Similarly, long side-chains are employed in an approach to induce microphase separation where using spacers which are chemically incorporated into polymer backbones to allow more room for the water of solvation to decrease swelling and mechanical degradation [132].An aromatic AEM with a long pendant spacer exhibited a good ionic conductivity of 91 mS cm À1 at 60 C and virtually no degradation after 960 h in 6.0 M NaOH at 60 C.
Several cationic groups were grafted onto a PES-backbone, where the 1-methylpyrrolidinium and 1-methylpiperidinium functionalised PES-AEMs displayed the greatest stability and AEMFC performance compared against two imidazoliums and a trimethylammonium at 60 C in 2.0 M NaOH [121].These results are in line with the degree of microphase separation.Limited R&D efforts are focused on PES, PEEK and polysulphone materials due to poor stability originating in the alkaline vulnerability of ether-bonds and poor FC performance (PPD ~109 mW cm À2 at 60 C).
Despite the several beneficial performances of AEMs with microphase separation there is still no evidence of causality between this method and in-situ AEMFC durability/peak power density [104].
With respect to functional cationic end-groups there are many approaches to making AEMs, including QA groups [103,108,119], quaternary phosphonium [117,126], imidazolium [115,122,124], guanidinium [123] and cobaltocenium [116].Historically QAs have received more attention, however initial instability and manufacturing difficulties caused attention to swing towards the other options.Today QA groups are still widely employed in AEMs where initial stability issues have been ameliorated with techniques such as crosslinking [103,108].
Very light crosslinking was utilised in creating a breakthrough AEM.Synthesising a block copolymerpoly(norbornene) (BCP-PNB) through vinyl addition polymerisation with QA cation groups attached by pendant chains, with PTFE added for mechanical strength resulted in an AEM with a ionic conductivity of 212 mS cm À1 at 80 C [103].The high performance of this AEM was partly attributed its very low thickness of 10 mm.
Creating an organic-inorganic composite by adding TiO 2 , SnO 2 , or layered double hydroxides as filler in AEMs represents another method for enhancing an AEMs mechanical properties [104,118].The high ionic conductivity of 122 mS cm À1 at 80 C was recorded after spray-coating QAfunctionalised layered double hydroxides (QA-LDH) onto the surface of triple-ammonium side chain PPO membranes [118].Due to inadequate fuel cell performance associated with this method, it has not garnered notable attention.
Recently, aryl ether-free PAP AEMs have amassed attention as stable, high-performance membranes [111].PAP functionalised with biphenyl and terphenyl yielded stable AEMs with high ionic conductivities of 169 and 193 mS cm À1 at 95 C respectively.Moreover, both AEMs maintained their conductivity after 2000 h in 1.0 M NaOH at 100 C in addition to displaying excellent mechanical qualities.Aryl ether-free PAP AEMs are currently being explored by many research groups.To decrease the adverse effects associated with phenyl groups, an aliphatic chain was integrated into a PAP AEM/ionomer resulting in aliphatic chaincontaining poly(diphenylterphenyl piperidinium) (PDTP) copolymers [113].Excellent mechanical and chemical stability was displayed with a tensile strength of 60e76 MPa and retaining 100% ionic conductivity after 1500 h in 1.0 M NaOH at 80 C, in addition to great hydroxide conductivity of 158 mS cm À1 at 60 C.
Another aryl-aryl type polymer backbone that has shown great alkaline stability is polyphenylene [104,110].Polyphenylene was functionalised with hexyltrimethylammonium (HTMA) and benzyltri methylammonium (BTMA) resulting in two AEMs [110].Following an initial loss of conductivity within the first 100 h, the HTMA membrane retained its conductivity over 5300 h in both 0.5 and 2.0 M NaOH at 80 C and surpassed the performance of the BTMA AEM.The stability test for the HTMA AEM was extended to 11,160 h, leaving the HTMA polyphenylene AEM with 28% of its initial hydroxide conductivity.
The realm of radiation grafted AEMs has been studied by Varcoe et al. for decades.Recently this group reported AEMs comprised of low density polyethylene (LDPE) yielding OH À conductivities in the region of 145 mS cm À2 at 80 C when fully hydrated [100].A variant of this was used to make a high temperature AEMFC operating at 110 C with a record high ionic conductivity of 300 mS cm À1 at 110 C [102].More advanced versions of the LDPE membrane were compared against a high density polyethylene (HDPE) AEM, where the latter displayed a greater ionic conductivity (214 mS cm À2 at 80 C), smaller thicknesses and greater PPDs [130].Similarly to the previously discussed aryl ether-free PAP AEMs, these AEMs possess poor microphase separation while still display great ex-and in-situ performance.This further underlines the lack of causality betweeen great AEMFC performance and microphase separation.
Membranes based on materials such as chitosan have attracted some attention in regard to sustainable AEM production.Chitosan is a non-toxic, bio-degradable cycloaliphatic polymer which contains hydroxide and active amino groups (-NH 2 ) on the polymer backbone which provide  hydrophilicity, high mechanical strength and chemical adaptability into the resulting membrane [124].A chitosan and 1-ethenyl-3-methyl-1H-imidazoliumchloride polymer with 1-ethyl-2-pyrrolidone composites (CS/EMImC-Co-EP) were developed using a thermochemical crosslinking technique [124].The membranes hydroxide conductivity remained stable following a 300 h test in a 8.0 M KOH solution at 85 C. Crosslinking is necessary to increase the chemical and mechanical stability of chitosan while simultaneously rendering it insoluble in aqueous media.Sustainably produced AEMs have yet to yield matching performances to state of the art AEMs.
Several FC systems incorporate both an AEM and liquid electrolytes in an attempt to maintain long term performance and at the same time reduce gas crossover [93].This latter idea has been embodied in a experimental AEM, using ion solvation polymers that absorb liquid alkalis such as KOH into their structure.A good example is polyvinyl alcohol that is crosslinked in the presence of strong KOH [133].The initial AEM performance was good, but KOH is leached out with time due to the formation of water, removing OH À from the body of the polymer.However, this approach caused decreased CO 2 tolerance and the adverse formation of carbonate (Eq.( 10)).

AEMFC performance and degradation
The peak power densities (PPDs) of AEMFCs have advanced from mW cm À2 to W cm À2 and durabilities surpassing > 1000 h in the last decade.The current PPD and stability records paint a promising picture of AEMFC development, however there is a wide variation in PPDs as displayed in Table 4 where the performance of most AEMs in this review are featured.The highest PPD registered to date employs the previously mentioned PTFE-reinforced polynorbornene AEM, which produced 3.50 W cm À2 at 80 C, in addition to the very high current density of 9.7 A cm À2 as shown in Fig. 12 [103].Additionally, its durability was tested in-situ at 600 mA cm À2 (~0.7 Fig. 12 e Current-voltage and power curves displaying the highest power and current density achieved to date [103].
V RHE ) at 80 C over 545 h resulting in no significant degradation.Such a high current density was possible due to the excellent water transport characteristics, in part afforded by the very low thickness of the AEM (10 mm).
Radiation grafted AEMFCs comprising low and highdensity polyethylene AEMs were compared yielding PPDs of 2.01 and 2.55 W cm À2 (fuelled with H 2 /O 2 at 80 C with anodic/ cathodic metal loadings 0.40 mg PtRu cm À2 /0.40 mg Pt cm À2 and no back-pressure) [130].This increase in performance was accredited increased water transport due to the change in polyethylene.More importantly, the stability was greatly enhanced as shown by a 440 h durability test (0.60 A cm À2 (~0.7 V RHE ), 70 C with H 2 /CO 2 -free air) where the LDPE AEMFC was stopped after 100 h due to a large increase in area specific resistance.The AEMFC with LDPE and HDPE AEM yielded degradation rates of 790 and 68 mV h À1 respectively.
The pre-discussed PdeCeO 2 /C HOR catalyst was employed as a substitute for PtRu and Pt in an AEMFC utilising a HPDE AEM with a variety of cathodes including Pd/C, AgeCo/C, Fe/C and a Pt/C benchmark [134].All cathodes with the PdeCeO 2 /C anode resulted in PPDs !1.0 W cm À2 , where the benchmark achieved a PPD of 2.0 W cm À2 (fuelled with H 2 /O 2 at 80 C with anodic/cathodic metal loadings 0.25 mg cm À2 Pd /0.4 mg Pt cm À2 with no back-pressure) and the Fe/C a PPD of ~1.0 W cm À2 (fuelled with H 2 /O 2 at 70 C with anodic/cathodic metal loadings 0.25 mg cm À2 Pd /30 mg Fe cm À2 with no back-pressure).The radiation-grafted LDPE AEM described in Ref. [100] was employed in the first high-temperature AEMFC, which performed admirably at 110 C, yielding a PPD of 2.1 W cm À2 (fuelled with H 2 /O 2 at 110 C with anodic/cathodic metal loadings 0.70 mg PtRu cm À2 /0.70 mg Pt cm À2 and 0.15/0.15MPa back-pressure) [102].High kinetic activity was exhibited through a high open circuit voltage (OCV) of 1.02 V RHE and a current density of 0.574 A cm À2 (at 0.8 V), comparing most favourably against high-temperature PEMFCs (OCVs z 0.95e0.96V RHE and 70e300 mA cm À2 at 0.8 V RHE ).Furthermore, the non-optimised high-temperature AEMFC was operated at 0.2 A cm À2 for 50 h (same conditions as previously stated) with a fairly stable voltage producing an approximate degradation rate of 2 mV h À1 .Few long-term (>1000 h) AEMFC degradation rates have been reported.A half-cell (ORR) AEMFC was utilised to measure the stability of commercial 40 wt.%Pt/C,where a degradation rate of 29 mV h À1 was reported after 1500 h at 0.31V SHE [137].Fig. 13a shows one of the most recently reported degradation rates from 2020 [13], where a 5 cm 2 AEMFC utilising a PNB-type AEM operated for 2000 h with a very small voltage loss.Fig. 13b displays two AEMFC rates from 2011 to 2019, compared to the degradation rate of a PEMFC [11].While the AEMFC degradation rate measured in 2011 (employing Tokuyama materials) is almost twice that of 2019, the increase in temperature and especially operating current density are most influential.Such a low rate of degradation despite these increases was made possible by employing a HDPE AEM and an EFTE ionomer.Furthermore, PO-CellTech reported a degradation rate of 7.5 mV h À1 after 1400 h at 0.6 A cm À2 at 75 C, where this AEMFC was still running at the time of publication [11].
Water management is a critical challenge on the system level, as optimising this aspect can make a great difference to the performance of the AEMFC [103].The current recordholder in AEMFC PPD labelled this facet as most critical, where rapid transport of water from the anode circumvents the issue of anodic catalyst flooding while simultaneously supplying water to the water-consuming cathode.Mechanical stability is of high importance in this regard, as producing very thin membranes enables fast water transport through the membrane and decreases the influence of the voltage drop across the membrane.Moreover, losses associated with poor water management are recoverable by increasing the dew points of the cell or by increasing the flow rate of the anode  [11] with permission from The Royal Society of Chemistry.
[11].The latter is impractical due to the expensive additional equipment required to sustain a high fuel efficiency.Further development of high-temperature AEMFCs will be deeply reliant on the experiences gained on this topic [102].
AEMFC degradation from a 100 h test under a variety of relative humidity (RH) conditions was recently investigated [138], where low anodic/cathodic RH (55%RH) induced a severe degradation of performance and failure after 20 h.The anode was most sensitive to low RHs and the most stable AEMFC was achieved with a 100%/50%RH anode/cathode.An anode with a low RH (50%/100%RH anode/cathode) caused a notable deterioration of performance, however adding 1.0 M NaOH into the FC resulted in reviving most of the FCs performance.

Ionomer development
Development of anionic ionomer is a blooming field of study, where similarly to the present status of AEMs, an alkaline analogue to the familiar and well-known Nafion® products has not yet been created.While ionomers are often created by dissolving the AEM using a mixture of solvent and water (thus yielding an ionomer with a non-competing chemistry to the membrane), ionomers should ideally be developed separate from the AEM [11].This is of great importance if the MEA preparation method requires a large difference in the solubility of the AEM and the ionomer.Due to the intrinsic characteristics of the anode/cathode, it is likely that optimised ionomers will be created separately from the AEM.
The ionomer-HOR catalyst interaction was studied, where the adsorption of ionomer components (phenyl groups) notably impacted on AEMFC anode performance [139].Specifically, the co-adsorption of cation-hydroxide-water hindered the diffusion of hydrogen gas to the catalyst surface and phenyl-group adsorption filled active sites on the catalyst, effectively reducing its activity.Increasing the anode flow rate decreased the influence of co-adsorption, however considering the ionomer-catalyst interaction in the catalyst design this is solution if impractical.Design of ionomers less reliant on phenyl groups is a topic requiring further R&D.
The effect of these two issues were investigated and compared by altering a quaternised poly(biphenyl alkylene) ionomer through two different approaches to increase hydrogen access to the anode [140].The first approach involved increasing the free volume in the quaternised poly(biphenyl alkylene) ionomer by inserting symmetric dimethyl groups into the backbone thereby increasing the hydrogen diffusion coefficient and also the PPD of the AEMFC by ~25%.Secondly, the functional group was changed from alkyl trimethylammonium to alkyl triethylammonium which resulted in a doubling of the hydrogen limiting current of the new ionomer in addition to increasing the PPD by 60% due to decreased co-adsorption and better hydrogen access to the surface of the catalyst.Co-adsorption of cation-hydroxidewater was therefore deemed as a superior source AEMFC performance degradation compared to slow diffusion of hydrogen gas due to phenyl adsorption.
A correlation between AEMFCs utilising Pt/PtRu catalysts and ionomers with various polyaromatic backbones was determined through DFT-determined adsorption energies of ionomer remnants onto metal [141].As displayed in Fig. 14, AEMFC performance increases as the adsorption of phenyl groups weakens.Adsorption of benzene rings parallel to the catalyst metal surface causes a strong adsorption of ionomer fragments, and also a strong interaction and hybridisation of the aromatic p-orbitals with metal electronic states.The adsorption energies of the ionomers were determined as follows: p-terphenyl !m-terphenyl > biphenyl > diphenyl ether > benzene !o-terphenyl > 9,9-dimethyl fluorene.This study displays that use of fluorene-based ionomers such as FLN-55 should be prioritised over intermediate performing ones containing p-biphenylenes (BPN45 and BPN100) and especially the poorly performing p-TPN ionomer containing pterphenyl remnants.
Generally speaking PtRu catalysts are less affected by phenyl groups, which is why this catalyst is generally preferred to Pt on the anode.
ORR catalysts can also be adversely affected by anionic ionomers as summarised in Ref. [107].Ex-situ studies on the interaction between fully dissolved cations (free from a polymer) and Pt/C as an ORR catalyst exhibited that imidazolium cations caused a terrible degradation of performance, causing an increase in the production of peroxide-species.High potentials resulted in the Pt/C catalyst oxidising the organic ionomer, where the degradation products were harmful to both the ORR and the HOR.Furthermore, the ORR was scarcely affected by anionic ionomers comprising benzene ring-free QA cations like tetramethylammonium. Use of ionomers containing benzene rings, such as benzyltrimethylammonium lead to poor ORR performances.
A recent study optimised the ionomer, thereby doubling the power density (1.6 / 3.2 W cm À2 ) [13].This paper compared the tetrablock poly(norbornene) copolymers (GT32, GT64, GT78), where the excellent aforementioned performance increment was brought about by using a combination of these ionomers.By utilising the hydrophilic GT78 in the anode and the hydrophobic GT32 in the cathode, the characteristics of the ionomers harmonise with the intrinsic water requirements in the anode and cathode.
Ionomers with different ammonia concentrations were synthesised to optimise anodic/cathodic water management, thus enabling the AEMFC to operate under low RH [138].Two quaternised poly(fluorene) (FLN) ionomers were Fig. 14 e AEMFC performance as a function of phenyl adsorption energy [141].Reprinted with permission from Ref. [141]).Copyright (2021) American Chemical Society.employed as binders in the creation of the MEAs, producing a comparable conclusion to Ref. [13], as using the more hydrophilic FLN-100 on anode side and FLN-55 on the cathode produced an improved AEMFC performance compared to most other configurations.Additionally, the two ionomers exhibited good alkaline stability with no structural changes after 720 h in 1.0 M NaOH at 80 C. Low RH-conditions are attractive for automotive applications, however this imposes challenges for the ionomer [11].The use of highly ionically conductive ionomers was suggested as an alleviating alternative.
The interaction between PGM/non-PGM catalysts (Pt/C, Fe 0.5 -950 (HOR)/PtRu/C and PdeCeO 2 /C (ORR)) and a multitude of ionomers (Nafion (DuPont), FAA3 (FuMA-Tech) and 4 PPObased anionic ionomers with different cationic groups) was investigated using a RDE [142].Both Pt catalysts produced high activities with all ionomers and Fe 0.5 -950 was highly efficient with the PPO-based ionomers compared to the commercial FAA3 ionomer.Generally, all catalyst-Nafion combinations resulted in greater diffusion limited current densities than any other ionomer.This was attributed to a lower H 2 /O 2permeability of the PPO-based ionomers and/or an incomplete ion-exchange (converting the PPO ionomers from bromide to hydroxide form).
Moreover, the 4 PPO-based anionic ionomers were designed free of fluorine, which similarly to using chitosan for creating AEMs [124,143] will be a meaningful aspect in regard to the environmental friendliness of AEMFC technology.

Summary
Great advancements in polymer science has resulted in highly conductive and mechanically stable AEM classes such as PNB, PAP and HDPE.Advancements such as these have engendered great AEMFC performances with notable durability.Both in-and ex-situ experiments have been effective in mapping out issues and determining solutions such as characterising the ionomer-catalyst relationship, though additional advancements on these topics is necessary.Comparing the featured degradation rates of AFCs and AEMFCs in regard to their maturity tells the tale of the progress AEMFC technology has undergone.However, it is important to note the average rate of degradation for AEMFCs is notably higher (~500 mV h À1 ), and accounting for these rates renders AEMFCs more typical of a technology at the beginning of its life.

Conclusion and outlook
Having gained a significant amount of attention, the progress of AEMFCs over the last five years has been momentous.Numerous challenges remain on all levels, from the chemical/ mechanical stability of the AEM and the interaction between the ionomer and the catalyst on the component level, to issues such as water management and carbonate build-up on the system level.The realisation of AEMFCs with PPDs > 1.0 W cm À2 operating at > 1000 h with low degradation rates (~5e10 mV h À1 ) paints a promising picture, however these targets will initially have to be met again with ultra-low PGM loadings (<100 mg cm À2 ) to capture and keep commercial interest and prove the viability of AEM technology compared to PEMFCs.Further on, these targets must be met a third time with a non-PGM AEMFC.The following points represent some of the main R&D areas.Correctly addressing the following comprehensive challenges will require cooperation between industry and academia.
Creating efficient and stable HOR/ORR non-PGM catalysts able to match or surpass the performance of the PGM benchmark.While encouraging performances with TMbased and heteroatom-doped materials have been reported, a significant body of work remains on this topic, specifically in regard to long-term stability (>1000 h).AEM performance has increased notably, though a performance which satisfies the rigorous demands has yet to be demonstrated.Similarly to catalyst development, stability is the principal barrier to overcome for AEMs also.Few large-scale (>100 cm 2 ) AEMFCs have been built, hence further efforts are required to face the optimisation challenges on the system level.Overall, streamlined design strategies are required for catalysts, ionomer, electrolyte and miscellaneous electrode components to avoid a mismatch of electrochemical properties.
Achieving the performance goals outlined by the US DOE [12] represents the first step only, as further progress yet is required.Each of these bullet points are composite challenges where every facet (e.g.AEM, ionomer, catalyst and its support, MEA components and preparation, FC fuel, dew points, etc.) requires optimisation.Realising AEMFCs as a cog in the future hydrogen economy is contingent on solving these challenges if they are to take their envisioned role in the global energy transition.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Fig. 3 e
Fig. 3 e HOR polarisation curves of Ni/SC, Ni/NC, Ni/BC, and Ni/C in hydrogen-saturated 0.1 M KOH (scan rate of 5 mV s ¡1 at 2500 rpm)[15].Reproduced from Ref.[15] with permission from The Royal Society of Chemistry.

Fig. 6 e
Fig. 6 e AFC electrode design displayed from the anode side, showing the triple phase boundary (TPB) between electrolyte, catalyst and voids.

Fig. 10 e
Fig. 10 e (a) The direct nucleophilic substitution reactions, and (b) the Hoffman elimination reaction.
i n t e r n a t i o n a l j o u r n a l o f h y d r o g e n e n e r g y 4 6 ( 2 0 2 1 ) 1 8 4 8 9 e1 8 5 1 0

Fig. 13 e
Fig. 13 e (a) Cell voltage and high-frequency resistance during a long-term durability test [13].Adapted with permission from John Wiley and Sons.(b) A durability comparison between two AEMFCs and a PEMFC, adapted from Ref. [11] with permission from The Royal Society of Chemistry.
. The specific activity (ECD normalised for catalyst loading) follows the sequence Ni/SeC > Ni/NeC > Ni/C > Ni/

Table 2 e
The specific exchange current densities of the various HOR electrocatalyst materials featured in this review.

Table 3 e
The onset potentials and current densities of the various ORR electrocatalyst materials featured in this review.The majority of the current densities were measured using a rotating disc electrode (RDE) at 1600 rpm in O 2 -saturated 0.1 M KOH.

Table 4 e
AEM fuel cell performances and associated performance-related aspects.*Only the Pt loading is specified, though a PtRu catalyst was employed.