High entropy materials as emerging electrocatalysts for hydrogen production through low-temperature water electrolysis

The production of hydrogen through water electrolysis (WE) from renewable electricity is set to revolutionise the energy sector that is at present heavily dependent on fossil fuels. However, there is still a pressing need to develop advanced electrocatalysts able to show high activity and withstand industrially-relevant operating conditions for a prolonged period of time. In this regard, high entropy materials (HEMs), including high entropy alloys and high entropy oxides, comprising five or more homogeneously distributed metal components, have emerged as a new class of electrocatalysts owing to their unique properties such as low atomic diffusion, structural stability, a wide variety of adsorption energies and multi-component synergy, making them promising catalysts for challenging electrochemical reactions, including those involved in WE. This review begins with a brief overview about WE technologies and a short introduction to HEMs including their synthesis and general physicochemical properties, followed by a nearly exhaustive summary of HEMs catalysts reported so far for the hydrogen evolution reaction, the oxygen evolution reaction and the overall water splitting in both alkaline and acidic conditions. The review concludes with a brief summary and an outlook about the future development of HEM-based catalysts and further research to be done to understand the catalytic mechanism and eventually deploy HEMs in practical water electrolysers.


Future perspectives
Hydrogen is set to play a pivotal role as an energy vector in the future energy system, enabling the storage of renewablyproduced electricity on the terawatt scale and the decarbonisation of a variety of industry sectors, which can help to achieve carbon-neutrality. Electrocatalysts are essential, indispensable components of low-temperature water electrolysers including both alkaline electrolysis (AEL) and proton exchange membrane water electrolysis (PEMWE), and play a significant role in governing the device's performance. In particular, considering the ambitious goals of deploying AEL and PEMWE on hundreds of gigawatt scale in the coming decade, the supply and criticality of raw materials constituting electrocatalysts are becoming serious concerns of the electrolyser industry. High entropy materials including alloys, oxides and other compounds, as emerging electrocatalysts, have shown substantial promise for use to catalyse the hydrogen evolution reaction (HER) and the oxygen evolution reaction (OER) in a wide pH range, thanks to their unique electronic structure allowing for continuous adsorption of reaction intermediates, slow atomic diffusion that can help improve catalytic stability, and a very high degree of flexibility in terms of catalyst design that permit to combine a large variety of elements, active or passive, into a single catalyst. Notwithstanding recent progress as reviewed in this article, the research on high entropy material (HEM)-based electrocatalysts is still in its infancy, and significant efforts need to be made to the following aspects: (1) In-depth understanding of the catalytic mechanisms of HEMs should be gained. In many publications, only apparent activity of HEM catalysts is considered and compared to the stateof-the-art benchmarks. We suggest that the intrinsic activity should be thoroughly evaluated and researchers should validate whether the developed HEMs are truly more active than PGM catalysts or not. According to table 1, it seems that HEMs containing PGM metals are more active towards the HER in both alkaline and acidic electrolytes, compared to PGM-free electrocatalysts. However, given the lack of detailed atomic structure characterisation and intrinsic activity comparison, it is not clear if the seemingly high HER activity of PGM-containing HEMs is originated truly from the high-entropy effect or from the surface aggregation of PGM (i.e. a PGM 'skin') like what happens in binary systems. For the OER, since HEM catalysts were mostly tested by far under alkaline conditions where transition metals such as Fe, Co and Ni are highly active, it is difficult to conclude to what extent PGMs contribute to the OER activity and stability. Therefore, comprehensive investigation on the reaction pathways for different HEM systems is critically needed, considering that the multiple catalytically active sites within the catalyst may play different roles. Such investigations will in turn aid in catalyst design. (2) HEMs represent a complex material system consisting of five or more elements, which will be difficult to study using the conventional trial-and-error approaches. In this respect, machine learning and artificial intelligence, in conjunction with experimental combinatorial chemistry, will play an important role in catalyst design. Remarkable advance has recently been made towards the mechanical applications of HEA materials, but for electrocatalysis a large database will be demanded to train models and eventually come up with optimised combinations of elements that can significantly boost HER and OER. It is also worth mentioning that HEMs are also promising candidates that may permit to break the 'scaling' relations and likely achieve unprecedented electrocatalytic performance. Besides the complexity of the bulk of HEMs, surface reconstruction and phase segregation typically happen under actual electrocatalytic conditions, which would pose substantial challenges for DFT simulations. In this case, experimental identification of surface compositions and elemental distribution becomes crucial to establish accurate models. To this end, atomic probe tomography and tomography based on aberration-corrected scanning transmission electron microscopy, as two important techniques enabling to resolve elemental distribution at atomic scale, may play an essential role. (3) The degradation mechanisms of HEM catalysts need to be elucidated. Although the lattice distortion and slow atomic diffusion of HEMs can help improve the stability of catalysts, corrosion and/or incidental dissolution in many cases are still inevitable, particularly under harsh operation conditions. To this end, more research should be carried out to scrutinize the degradation pathways and accordingly introduce some stabiliser element in HEMs to optimise the stability. (4) The research on HEM catalysts should take their practical applications into account. Many HEM catalysts containing PGMs have been targeted for use in AEL, showing catalytic performance even poorer than the current state-of-theart catalysts, without considering that PGMs are not critically required in industrial alkaline electrolysers at all and that the introduction of PGM would increase the production cost of hydrogen via AEL. We suggest researchers to test PGMcontaining HEM catalysts in acidic conditions with the aim to diminish the utilisation of PGM in PEMWE. (5) For practical applications, we suggest that, on one hand, the developed HEMs had better be assessed under industrial relevant conditions to validate their viability for use in AEL and PEMWE; on the other hand, simple, cost-effective and scalable methods should be further developed to enable massive production of HEM catalysts.

Introduction
A secure access to energy is basic to develop any quotidian activity, as well as being essential for the smooth operation of all industrial sectors. For example, without energy the manufacture and distribution of food, among others, would be disrupted. Since the industrial revolution, energy has been predominantly obtained through the consumption of fossil fuels (e.g. coal, oil, natural gas) [1]. The use of fossil fuels to produce energy is bound to unavoidable costs: environmental pollution, ecosystems degradation, intergovernmental friction due to the limited availability and specific geological localisation of natural resources, centralised energy production, and competition with other chemical industries for raw materials (e.g. the manufacture of goods such as polymers and drugs). Moreover, the current carbon-based energy production leads to unambiguous, and now inevitable, climate change consequences [2][3][4] due to the exponential rise in CO 2 emissions [5].
According to the Statistical Review of World Energy [1,6], in 2019 the global energy consumption reached 170 000 terawatts hours (TWh), with an associated CO 2 emissions of 34 040 million tonnes. Based on the economic growth perspectives (2%-3% annual) and the population growth trend (1%-2% annual), it is expected that energy consumption will reach 330 000-430 000 TWh by 2050 [7]. Thus, if the carbonneutral goals are to be met by the mid of this century (e.g. by 2050 within the EU [8]), energy must be produced from renewable and carbon-neutral resources (e.g. wind farms, hydropower plants and solar farms). Current technology allows to harvest renewable resources into electricity on the TW scale. However, due to the intermittent nature of renewable energy sources, the excess of electricity produced at the time of low demand must be stored, so that it is available for consumption when the energy demand surpasses its production. The use of hydrogen as an energy vector has the potential to fulfil this gap [9], where renewable energy in the form of electricity can be employed to split water into 'green' hydrogen (H 2 ) and oxygen (O 2 ). Thus-produced hydrogen can be stored or transported from the production site to the end use location, where the stored energy can be recovered in the form of electricity through fuel cells, or in the form of heat via controlled combustion [10]. In the energy sector, apart from the direct use of green hydrogen as an energy carrier, hydrogen can also be readily incorporated into the current fuel feedstock through its reaction with CO 2 , which will play a major role in the decarbonisation of the hard-to-abate chemical industry [11,12]. Moreover, green hydrogen can potentially substitute the 'grey' hydrogen used in industry (e.g. the Haber-Bosch process or in the metallurgic refinery), which is currently obtained from the steam reforming of hydrocarbons [13]. Figure 1 schematically illustrates the possible application scenarios of hydrogen in various sectors.
The process of splitting water into its elemental constituents, i.e. H 2 and O 2 , through an energy input, is known as water splitting. If electrical energy is applied, then the process is typically denoted as water electrolysis (WE) (equation (1)), The hydrogen evolution reaction (HER) and the oxygen evolution reaction (OER) occur simultaneously at the cathode and at the anode, respectively. Currently, there are two main low-temperature WE technologies: proton exchange membrane water electrolysis (PEMWE) and alkaline electrolysis (AEL) (figure 2). Both HER and OER are complex processes involving the transfer of 2e − and 4e − , respectively. While one intermediate is required for the HER to proceed, there are three intermediates generated during the OER [14][15][16]. Hence, the OER is kinetically slower than the HER and as a consequence a higher potential than that dictated by thermodynamics (denoted as overpotential) is required for the WE to occur, being thus the OER mainly responsible for both the efficiency loss on the conversion of electricity to hydrogen and the catalyst degradation associated with WE.
AEL is a mature technology which operates under alkaline conditions, normally in concentrated potassium hydroxide (KOH) with a concentration of 20-30 wt.%. In an AEL electrolyser a porous separator that allows the diffusion of hydroxide ions is placed between the cathode and the anode. The HER and the OER occurring in AEL and its potential against the standard hydrogen electrode are represented in equations (2) and (3), respectively. One of the main advantages of AEL with respect to PEMWE is that it employs earth abundant electrocatalysts at both electrodes, with a prolonged lifetime of 20-30 years [17][18][19]. However, the AEL technology suffers from three major disadvantages: (1) limited current density, which translates in bulky designs and low hydrogen yield; (2) low operating pressure, which requires a post pressurising step thus increasing the operating and distribution cost of hydrogen (and oxygen); (3) low partial load range, with gases diffusion between electrodes compromising efficiency and safety, making the combination of AEL with fluctuating renewable energy production challenging [20], Compared to AEL, PEMWE shows many advantages, for instance, it can operate at higher current densities, resulting in higher hydrogen yield in unit time. Moreover, the fast proton transport through the proton exchange membrane allows PEM electrolysers to work under a fluctuating power input, better adapting to the fluctuating nature of renewable electricity. The use of a thin and solid membrane with low hydrogen permeability enables a high pressure to be applied during electrolysis, which cannot only improve the energy efficiency but also allows the produced hydrogen to be stored and distributed without further compression, mitigating the safety hazards like those associated with AEL operation under similar conditions. The HER and the OER occurring in PEMWE are described in equations (4) and (5), respectively. The main drawback of PEM electrolysers is that under acidic and highly oxidising operating conditions, catalysts for OER are virtually limited only to formulations based on IrO 2 or RuO 2 , with precious metal scarcity and long-term catalytic stability still being a challenge [21][22][23]. Additionally, Pt-based electrocatalysts remain the state-of-the-art catalysts for HER in PEMWE, competing directly with other technological applications for Pt [24]. Hence, the amount of precious metals on the electrodes must be minimised as much as it can, or ultimately eliminated, in order to manufacture cost-competitive PEMWE on gigawatt (GW) scale for green hydrogen production [22], The electrochemical performance of water electrolysers is largely limited by the kinetics of both half reactions. Thus, many efforts have been made to develop new highperformance electrocatalysts enabling high activity and outstanding stability for HER and OER, such as, various platinum group metal (PGM) nanostructures [25][26][27], transition metal phosphides and phosphates [28][29][30][31], chalcogenides [32][33][34][35][36][37][38], carbides [39][40][41], and nitrides [41][42][43]. Interested readers can refer to some review articles published recently [28,33,34,39,40,42,44,45], For most metal-containing electrocatalysts, it is generally accepted that it is the metal sites that contribute mainly to catalysis and that the synergy among different metal species in the catalyst would enhance catalytic performance. In this context, high entropy materials (HEMs) comprising five or more metal elements have recently emerged as a new family of electrocatalysts and shown promise as an efficient and stable alternative to the current state-of-the-art catalysts for use in AEL and PEMWE [46][47][48][49][50][51]. Generally, HEMs possess high thermal, chemical, and structural stability due to the high entropy effect, comparatively large lattice distortion, and slow atomic diffusion; moreover, the presence of multi-components may provide a wide variety of absorption sites for reaction intermediates, all of which are favourable for electrocatalysis. For these reasons, the research on HEM electrocatalysts for WE recently experiences an exponential growth, as illustrated by the number of publications and citations shown in figure 3.
In this review, we first present a brief overview about HEMs and why they can serve as good electrocatalysts. This will be followed by a short introduction to the most commonly employed methods reported in literature for the synthesis of HEM-based catalysts. We will then comprehensively summarise the HEM catalysts that have been used for the HER and the OER processes under different conditions (both alkaline and acid), as well as those used as bifunctional catalysts for the overall WE. At last, we provide a perspective of future development of this new research direction, highlighting the efforts to be made to gain fundamental understanding of catalytic mechanisms of HEMs and realize practical applications in future AEL and PEM electrolysers.

HEMs
As mentioned before, HEMs are typically described as stable structures containing five or more metal components (individual concentration of 5-35 at. %) homogeneously distributed in a single phase. HEMs can be further categorised according to their chemical nature, and the most commonly investigated ones are high entropy alloys (HEAs) [52] and high entropy oxides (HEOs) [53]. Besides, other HEMs reported in the literature include high entropy fluorides [54], high entropy carbides [55] and nitrides [56], high entropy metal diborides [57] or high entropy phosphides [58]. The HEMs definition originates from the highly mixed configuration entropy (S conf ) of these materials, as described by the following equation (equation (6)): where R is the ideal gas constant, and x i represents the mole fraction of each element. S conf reaches its maximum value when all components are present in an equimolar amount, and in this situation the previous equation can be simplified as follows (equation (7)): where n represents the number of components in the material. For instance, the S conf of a HEM with five constituents in an equimolar concentration is 1.61R. Note that the previous equation is rightfully applied for HEAs, whilst for HEOs the anion contribution is commonly eluded for simplicity [53]. Nevertheless, when considered, the S conf value would become smaller because the concentration of oxygen is usually higher compared to that of other cations, and moreover, the anionic sub-lattice is generally ordered [59]. In many cases, when S conf ⩾ 1.5R the free energy of mixing (∆G mix ) is favoured by the entropy of mixing (∆S mix ), this being able to overcome the unfavourable enthalpy of mixing (∆H mix ) above a threshold temperature (equation (8)). Below this threshold temperature, phase segregation is thermodynamically favoured leading to a homogeneous distribution of components [60]. Thus, when reversible phase transformation between solid solution and mixed phases is observed with temperature HEMs can be described as entropy stabilised materials (ESMs). It is worth mentioning that (1) not all HEMs are entropy stabilised, (2) ESMs below the threshold temperature can retain its structure due to the slow kinetics, and (3) not all elemental configurations yield a single phase or HEMs, even if five or more elements are combined, What makes HEMs promising catalytic candidates for challenging reactions, such as the HER and the OER [47-49, 52, 61-64], are their unique physicochemical properties (figure 4). The high entropy contribution favours the distribution of elements within a stable solid solution, instead of phase segregation. The aleatory distribution of components with different electronegativity results in a wide variety of partially charged sites, and a near-continuum distribution of associated adsorption sites, leading to favourable synergistic effects. The different atomic size also gives rise to a high lattice distortion, with higher surface potential energy, which not only influences the thermal and electrical conductivity but also leads to slow atomic diffusion, improving the material's stability [61,[65][66][67][68][69][70][71].
The selection of elements to combine needs to be elucidated before deciding the synthetic methodology for HEMs. Otherwise, the material exploration would become enormous due to the high number of elements and quasi-infinite possible combinations. A wise strategy is to fix two to three components to the elements commonly used in the state-of-the-art catalyst for the desired reaction, and introduce other components that could help address identified drawbacks. For instance, iridium (Ir) is indispensable for use as the anode catalysts in PEM electrolysers, but there is a pressing need for lowering Ir concentration in the catalyst without compromising the catalytic activity and stability against corrosion. Thus, it is sensible to combine Ir with other electrochemically-active elements (e.g. Ni, Co, Fe, Se, Mo, Mn, Zn, V) [18], and/or potentially stable elements against dissolution during OER (e.g. Si, Ti, Mn, Fe, Co, Cu, Mo, Ag, Sn, Te, Ta, W, Tl, Pb and Bi) [72], to form HEMs that could fulfil requirements for minimisation of Ir utilisation. Another challenge that needs to be addressed is to design noble-metal-free cathodes with comparable activity and stability to Pt-based catalysts for the HER [73]. For successful engineering of HEMs, it is indispensable to integrate experimental research with computational approaches [74][75][76][77]. For example, Troparevsky et al through density-functional-theory (DFT) calculations developed a matrix to predict the stable formation of HEAs based on enthalpies of formation (∆H f ) for binary alloys (figure 5). A binary combination with a very stable ∆H f would likely result in the precipitation of this phase, whilst an unstable ∆H f indicates the immiscibility of those two elements. This matrix suggests, for example, that for Ru-based HEAs the addition of Cu would tend to form an immiscible phase, whilst the addition of V would likely result in a too stable binary phase, in both cases leading to phase segregation, rather than HEAs [74].

Synthesis of HEMs
Typically HEM-based electrocatalysts can be prepared using a variety of methods such as shock decomposition/reduction techniques [78,79], high energy mechano-chemical methods [80] and several conventional wet chemistry methods [81,82]. Details of these methods have been described recently in several publications on HEAs [46,47,50,52,62,67,83] and HEOs [83], to which readers are suggested to refer.
In the following, we will only present a short summary of the major synthetic methods used to prepare nanostructured HEM catalysts.

Flash decomposition techniques.
A common strategy for the synthesis of highly dispersed HEMs is through fast decomposition/reduction of mixed metal precursors, forcing homogeneously distributed metal ions to coalesce, while avoiding the generation of multiple heterostructures favoured by thermodynamics. Fast decomposition techniques usually result in small, homogeneous and highly dispersed nanoparticles (NPs) with a narrow size distribution.
In a typical pulse thermal decomposition or carbothermal shock (figure 6(a)), a solution containing the desired metal salt precursors is loaded onto a conductive support (e.g. carbon nanofibers), which is then heated above 1000 • C within the microsecond scale. Controlling the heating/cooling rate, the temperature and the hold time permits to fine-tune the HEAs' particle size and distribution [78]. In addition, Cui et al, reported the synthesis of (CrMnFeCoNi)S x high entropy metal sulphides (MSs) via a carbothermal shock in the presence of thiourea [84], suggesting that this method is not limited to the synthesis of HEAs. However, the carbothermal shock approach requires the support materials to be highly electrically conductive, which to a certain extent limits its application.
A process that enables the synthesis of HEAs but meanwhile is not restricted to conductive supports is the fast-moving bed pyrolysis method [85]. Recently, Gao et al, used this method to synthesise a variety of HEA NPs, from quinary CuPdSnPtAu to denary MnCoNiCuRhPdSnIrPtAu, on several different supports (carbon black, graphene oxide, Permutit zeolite and γ-Al 2 O 3 , see figure 6(b)) [85]. To do so, metal precursors were impregnated onto the support, dried, and under inert atmosphere placed into a pre-heated furnace for annealing [85].
A solution-based method that can be used to prepare HEA catalysts is the nanodroplet-mediated electrodeposition. For this method, a microemulsion generated for instance by ultrasonicating a mixed solution of dichloroethane and water containing the metal precursor salts, is directed towards an electrode under a cathodic potential. Metal precursors confined within the nanodroplets are reduced through an electric shock (c.a. 100 ms) when in contact with the electrode surface, resulting in the corresponding HEA formation [86]. In addition, ultrasonication has also been employed to prepare HEA electrocatalysts. The ultrasonication-assisted wet chemistry allows for fast reduction of the metal precursors in solution to the corresponding HEA through the acoustic cavitation phenomenon, which generates localised temperature as high as 5000 • C and pressure up to 2000 atm. In just less than a nanosecond [87,88], facilitating the flash formation of HEAs.
Besides, laser was also used to act as a high-energy source enabling the rapid formation of HEA nano-catalysts. For example, Waag et al first employed the kinetically-controlled laser ablation to synthesise colloidal CoCrFeMnNi HEAs with an average particle size of 5 nm and a narrow size distribution [79]. Before laser ablation, the individual metal nanopowders were mixed, pressed at 200 MPa into sheets, and annealed in inert atmosphere (20 h, 1000 • C). The multicomponent sheet was then placed under ethanol and irradiated with laser irradiation pulses of 10 ps [79]. As a consequence, a plume of surface atoms are ejected from the surface of the sheet at a high velocity, the deceleration of which causes the plume/solvent interface to heat at high temperature. This causes the liquid at the interface to evaporate forming a low density metal-solvent mixing region inside a cavitation bubble. As the solvent evaporates, the moltenmetal atoms at the surface cool down condensing into clusters, which can further grow into NPs [89]. Furthermore, by dropping an ethanol solution containing the mixed metal chloride precursors on a support (e.g. carbon, graphene, glass, copper foam), Wang et al adapted the kinetically-controlled laser ablation method for the synthesis of supported PtIrCuNiCr and PtAuPdFeNi HEAs NPs [90]. Additional treatment of the loaded mixed metal chloride materials with NaOH, borohydride, ammonium chloride or triphenyl phosphine, prior to laser ablation, also resulted in the synthesis of HEOs, borides, nitrides and phosphides, respectively, highlighting the versatility of this method for the synthesis of a wide variety of HEMs [90]. (a) Graphic illustration of the carbothermal shock method. From top to bottom, microscopy images of precursor salt particles before thermal shock and the formation of bimetallic PtNi nanoparticles after thermal shock. Particle distribution of PtNi nanoparticles as function of the thermal shock duration (5 ms-10 s), and representative temperature profile during a 55 ms thermal shock process. Elemental map of a PtPdNiCoFeAuCuSn HEA, scale bar: 10 nm. From [78]. Reprinted with permission from the AAAS. (b) Graphic illustration of the fast-moving bed pyrolysis method and the effect of fast or slow heating rate (h.t.) on the metal distribution to form HEAs or heterostructures. Reproduced from [85]. CC BY 3.0.

High-energy mechano-chemical techniques.
The complex process of fracturing, grinding, cold welding and high-speed plastic deformation of solid precursors in a closed container by hard inactive spheres (e.g. zirconia balls, stainless steel) when applying a combination of high-energy mechanical forces (e.g. rotation, translation, vibration) is commonly known as ball milling. The strong forces produced by collision and friction between balls and with the container wall generate high energy capable of not only modifying the morphology of materials but also to trigger chemical reactions. As such, ball milling has been widely employed for the synthesis of HEAs [80]. For example, HEAs such as AlCoCrTiZn, AlCoCrFeNi, CoCrFeMnNi, AlFeMnTiCr, AlFeMnTiCo and AlFeMnTiNi were prepared by ball milling (50-60 h at 300-400 rpm) the corresponding metal powder precursors [91,92]. Tailoring the ball milling parameters (e.g. temp. and time) enables to tune material's properties such as the particle size, the morphology or the phases present. Additionally, high-energy ball milling is not limited to the synthesis of HEAs. For instance, high entropy metal diborides [57], oxides [93,94] and fluorides [54] were synthesised through high-energy ball milling of their constituent metal diborides, metal oxides or metal fluorides (MFs), respectively, making ball milling a versatile, easilyscalable and cost-effective technique.

Conventional wet-chemistry techniques.
Common methods optimised for the synthesis of metallic or metal oxide NPs can also be adapted for the synthesis of HEMs. To this end, Wang et al, reported 5 nm (CoCuFeMnNi)O 2 HEOs with a spinel structure through a solvothermal synthesis (170 • C, 15 h) from the metal acetate precursors in a water/ethylene glycol/ethanol solution using Pluronic P12/hexamethylenetetramine as the stabilizing agents, followed by post-annealing in air at 400 • C for 2 h [95]. The solvothermal method is not limited to the synthesis of HEOs. By avoiding annealing in air, it enables the synthesis of HEAs, for instance, the noble-metal-based RuPtRhPdIr HEA, reported by Bondesgaard et al [96]. A widely employed method for the synthesis of metal oxides is through the precipitation of the corresponding metal precursors from an aqueous solution with base. In this line, Spiridigliozzi et al demonstrated the applicability of co-precipitation by preparing a library of 18 different rare earth HEOs through a simple and scalable method just by mixing an aqueous solution containing the metal nitrate precursors with an ammonia solution, followed by annealing in air at 1500 • C [81].
The polyol reduction is another common wet-chemistry method for the synthesis of metallic NPs that was adapted for the preparation of HEAs. As an example, Wu et al, synthesised noble metal-based IrPdPtRhRu HEA with a 5.5 nm particle size and a narrow size distribution by adding an aqueous solution containing the metal precursors to a pre-heated triethylene glycol solution with PVP at 230 • C [82]. A more classical approach for the preparation of metal NPs is through chemical reduction using a strong reducing agent such as sodium borohydride. This method was adapted by Feng et al for the synthesis of 2 nm NiCoFePtRh HEA supported on carbon [97]. It is worth mentioning that most wet-chemistry techniques can actually be adapted to synthesise HEMs. However, finetuning experimental conditions is essential and non-trivial to ensure that an HEM, rather than a mixture of separate phases, is attained.

Chemical/electrochemical dealloying.
Chemical and electrochemical dealloying has been extensively used to prepare nanoporous metals and alloys previously [98,99]. Recently, it was also shown to be an effective 'top-down' approach to preparing HEMs with a nanoporous structure [49]. This is typically realised by amalgamating different metal constituents of interest with a diluent phase, such as aluminium, using the conventional metallurgy method, followed by chemical or electrochemical etching, which leads to the final HEM with the desired composition. To this end, Qiu et al recently reported the synthesis of Al 96 Ni 1 Co 1 Ir 1 X 1 and Al 96 Co 1 Fe 1 Ni 1 X 1 (X = Mo, Cr, Cu, Nb or V) ribbons by melting the individual metals in an arc-melting furnace under Ar protection followed by melt spinning. The excess Al was then chemically etched, or dealloyed, with 0.5 M NaOH to obtain the corresponding nanoporous HEAs (figure 7) [100][101][102][103]. Thus-obtained HEAs are generally mechanically robust and possess hierarchical macro-and meso-pores, which is conductive to electrocatalysis.

HEMs for the HER
In acid and alkaline media, the HER proceeds through the reduction of protons (equation (2)) or water molecules (equation (4)), respectively. In both conditions the first step involves the electron transfer to a proton, or a water molecule, to generate an adsorbed hydrogen specie (H ad ) at the surface of the cathode catalyst (the Volmer reaction, equations (9) and (10)). A H 2 molecule can then evolve through either the recombination of two adsorbed hydrogen species (the Tafel reaction, equation (11)), or via a second electron transfer to a proton which recombines with a H ad species (the Heyrovsky reaction, equations (12) and (13)), or a combination of both, depending on the surface coverage of hydrogen on the catalyst [104,105], 2H ad ⇌ H 2 (acid and alkaline) (11) The rate of the overall reaction is strongly dependent on the hydrogen adsorption free energy (∆G H ). A weak interaction would make the first electron transfer and H ad formation difficult to occur, whilst a too strong interaction would result in slow hydrogen desorption [106][107][108]. Information about the rate-determining step (RDS) can be obtained experimentally from the Tafel slope (equation (14)), defined as the dependency of the ohmic drop-corrected overpotential (η) on the current density (j) [109], where R is the ideal gas constant, T the temperature, F the Faraday constant, α the transfer coefficient and z the number of transferred electrons during the RDS. For instance, a Tafel slope of 116 mV dec −1 , 39 mV dec −1 and 29 mV dec −1 indicate that the RDS is the Volmer, the Heyrovsky and the Tafel process, respectively [73,104]. A smaller Tafel slope usually indicates faster electron transfer kinetics, and hence, a lower η Reproduced from [103] with permission from the Royal Society of Chemistry.
value that is required to deliver the same current density increment. In most literature, the catalytic activity towards HER is broadly reported as the overpotential required to reach a current density of 10 mA cm −2 (η 10 ), since this corresponds to harvesting c.a. 12% of the sunlight in a photoelectrochemical (PEC) device [22], which would allow to fulfil the expected global energy demand by 2050 if 1% of the Earth surface were covered with 10% efficient PEC devices [110]. Pt is to date the best catalyst for HER due to its optimal absorption free energy of hydrogen species (∆G H ≈ 0) and outstanding electrochemical stability under harsh conditions [106][107][108]. Recently, non-Pt electrocatalysts such as transition metal nitrides [111], sulphides [112] and phosphides [45] were extensively investigated. Although they show reasonably good HER activity in a wide pH range, none of them exhibits an intrinsic HER activity outperforming that of Pt [113], let alone that the long-term catalytic stability of these transition metal compound catalysts in acid medium is still far from satisfactory [22,73], which is challenging to substitute Pt for PEM electrolysis. Therefore, it is believed that diluting Pt by developing Pt-based HEMs would be a viable approach allowing for substantial optimisation of Pt resources. Meanwhile, some researchers are optimistic towards the development of Pt-free HEMs with tailored hydrogen adsorption that can be potentially used as HER catalysts for industrial PEMWE [114,115]. It is worth mentioning that commercial AEL does not employ Pt-based HER electrocatalysts, and hence efforts to making Pt-based catalysts for AEL should be carefully assessed in a techno-economic perspective.

HEAs for hydrogen evolution
Besides Pt, there are a number of other elements active towards the HER. For the HEAs containing multiple active elements, it is important to assess whether the HER activity results from the formed HEA phase or is related to the collective contribution of all individual components. To Pt-based HEAs, Pt typically dominates in the HER activity because Pt is reported to be several orders of magnitude more active than other elements [73,108]. However, for Pt-free HEAs, elucidating the origin of HER activity becomes particularly important. In this respect, Arumugam et al compared the HER activity of mixed phase alloys and HEAs based on Cu, Co, Ni, Fe and Mn. Samples were prepared via high-energy planetary ball milling, with phase composition related to milling time. For instance, individual metals, multi-phase alloys, and HEA were obtained after milling for 1, 15 and 60 h, respectively. The required overpotential to reach a cathodic current density of 50 mA cm −2 in pH-neutral electrolyte decreased from 624 to 320 mV against the reversible hydrogen electrode (RHE) as the milling time increased, suggesting that the enhanced activity towards HER results from the synergistic effect of HEAs, instead of the collective contributions from individual components [116]. Li et al rationalised that the synergy originates from the multi-sites nature of HEAs for electrochemical reactions, including both HER and OER [68]. In the case of CuAlNiMoFe HEA, the synergistic effect observed for HER in alkaline solution was ascribed to Al and Cu providing structural stability and electric conductivity, respectively. In addition, Mo promoted the adsorption of water molecules and reduced the energy barrier of the Volmer process, Ni and Fe served as adsorption sites for H ad reaction intermediates, whilst Ni-Mo-Fe sites were responsible for H ad recombination to produce H 2 [68]. Apart from ensuring that metals are homogeneously distributed within a HEA to enable the synergy, the phase composition plays an important role in governing the HER activity. Zhang et al [117] prepared NiFeMoCoCr HEAs by arc melting, followed by subsequent post-heat treatment at 800 • C and 1150 • C, which led to the formation of multi-phase and single-phase HEAs, respectively (figure 8). The presence of multiple HEA phases increased the charge transfer resistance compared to the single-phase HEA, resulting in a higher overpotential in acid and alkaline media. To this end, single-phase HEAs are highly preferred for use as HER catalysts. Although both CuCoNiFeMn [116] and NiFeMoCoCr [117] were less active than commercial Pt/C catalysts towards the HER in acidic, neutral or alkaline media, they showed little metal dissolution and good stability, highlighting the potential of HEMs as stable catalysts for water splitting.
Self-supported HEAs also drew considerable attention recently and were directly used as the cathode to expedite the HER. For example, Liu et al mixed Ni, Co, Fe and Mo with a Mn matrix by arc melting and subsequent melt spinning. The pre-catalyst was cut into small pieces with an area of 10 × 2 mm 2 . The NiCoFeMoMn electrode was then electrochemically dealloyed to remove the excess of Mn and yield the corresponding NiCoFeMoMn nanoporous HEA. Thusobtained HEA was self-supported and therefore was employed directly as electrodes. When tested for the HER in 1.0 M KOH, the NiCoFeMoMn HEA electrode delivered a stable current density of 100 mA cm −2 for 180 h at a low overpotential of 104 mV [118]. Interestingly, the NiCoFeMoMn HEA electrode was also active (η 10 = 243 mV) and stable at a high current density of 100 mA cm −2 for 240 h for the OER in 1.0 M KOH, and therefore could act as bifunctional catalysts for overall water splitting [118]. Table 1 summarizes and compares the activity and stability of various HEMs catalysts reported recently in the literature. Due to the excellent performance of Pt, HEAs containing PGMs have been intensively investigated [82,[85][86][87]103]. For example, Wu et al [82] and Liu et al [87] prepared RuRhPdIrPt/C and RuRhPdAuPt/C HEAs, respectively. Both catalysts outperformed the benchmark Pt/C catalysts for the HER (0.1 M KOH) in terms of activity, whilst ternary and quaternary alloys (e.g. PtAuPd and PtAuPdRh) underperformed the HEAs and Pt/C catalysts. Particularly, the RuRhPdIrPt/C catalyst delivered 10 mA cm −2 at an overpotential of only 17.0 mV (η 10 ), substantially lower than that of Pt/C (η 10 = 77.4 mV). Hard x-ray photoelectron spectroscopy (XPS) was employed to reveal the valence band structure of the RuRhPdIrPt/C HEA. The higher activity of the RuRhPdIrPt/C HEA was attributed to the broad valence band observed, which suggests the hybridisation of orbitals among the HEA constituents [82]. Moreover, no apparent loss in activity was detected after 3000 cyclic voltammetry (CV) scans for RuRhPdIrPt/C, whilst an increase of 35 mV in η 10 was observed for commercially sourced Pt/C tested under the same conditions. Besides, RuRhPdIrPt/C also showed better stability than the Pt/C benchmark, which was ascribed to the slower metal diffusion of the former owing to the lattice distortion in HEMs [82].
It should be noted that in practice PGM-based catalysts are hardly employed in industrial alkaline water electrolysers, because transition metal based alternatives can offer reasonably good HER performance in alkaline solutions for a long term. Therefore, research on PGM-based HEAs should be directed towards HER electrocatalysis under acidic conditions for  [103].
Considering that the actual Pt loading in a PEM electrolyser is typically in the range of 0.5-1.0 mg cm −2 and assuming an operating current density of 10 mA cm −2 with an overpotential of 50 mV, approximately 100 tonnes of Pt (∼60% of the Pt annual production) would be required to store 1000 TWh of energy in the form of hydrogen through PEMWE [22]. Therefore, reducing the Pt loading in catalysts and operating PEM electrolysers at higher current densities can effectively decrease the amount of Pt required. Even so, Pt is a scarce and expensive metal and it has many other applications, e.g. as catalysts for the synthesis of bulk petrochemicals and nitric acid, decomposition of vehicle exhaust, and the fabrication of lightweight fibre glass reinforcement materials used in the wind power, telecommunications, construction and automotive industries [119]. Thus, it is of high importance to ultimately develop Pt-free electrocatalysts for the HER in acid media [105]. To this end, Zhang et al and Ma et al have explored NiFeMoCoCr [117] and CoCrFeNiAl [120] HEAs, respectively, for catalysing the HER in acidic electrolyte (e.g. 0.5 M H 2 SO 4 ). In this case, the NiFeMoCoCr and CoCrFeNiAl HEAs delivered low η 10 values of 107 and 73 mV, respectively, which favourably compare to that of Pt/C. Moreover, the NiFeMoCoCr alloy gave no signs of deactivation at a current density of 100 mA cm −2 for 8 h [117]. Notwithstanding some progress, the long-term catalytic stability of Pt-free HEA catalysts still needs to be remarkably improved before they can be used in practical PEMWE.

Other HEMs for hydrogen evolution
Recently, transition metal based compounds such as sulphides [149], carbides [55], nitrides [56], borides [150], selenides [35,151], and phosphides [58] have been extensively investigated for use as HER catalysts. Along this line, high entropy transition metal compounds were also explored. For example, Zhao et al [141] lately reported the synthesis of (CoCrMn-FeNi)P high entropy metal phosphide (HEMP) using a onepot process by heating the solution containing corresponding metal chloride precursors, tetrabutylphosphonium chloride and ethylene glycol in an inert atmosphere at 400 • C for 3 h. When assessed towards HER in 0.1 M KOH, the (CoCrMn-FeNi)P HEMP outperformed in terms of activity all monometallic metal phosphides, though the overpotential recorded (η 10 = 136 mV) was still slightly higher than that of the state-of-the-art Pt/C catalyst (η 10 = 55 mV). In addition, Wang et al synthesised a Co 0.6 (VMnNiZn) 0.4 PS 3 high entropy metal phosphorous trichalcogenide [137]. DFT calculations indicate that hydrogen preferably absorbs at S sites on single metal phosphorous trichalcogenides, whilst on the high entropy metal phosphorous trichalcogenides the P sites draw the electrons from S sites, resulting in the generation of P sites with a quasi-optimal ∆G H (figure 9), more favourable for the HER. Besides, the near-continuum absorption at metal sites associated with HEMs promoted the hydrogen dissociation and thereby a low η 10 value of 65.5 mV towards the HER in 0.1 M KOH was obtained [137]. It is worth noting that both (CoCrMnFeNi)P and Co 0.6 (VMnNiZn) 0.4 PS 3 HEM catalysts were not assessed in acidic conditions, and thus, whether these high entropy metal compounds are viable alternatives to Pt for PEMWE or not is unclear [137,141].
Although several high entropy transition metal based catalysts have been reported to be active for the HER, their HER performance still needs to be substantially improved, in order to compete with the existing HER catalysts and eventually the Pt benchmark [137,141]. To this end, the formulations of high entropy metal compounds should be better optimised, preferably with aid of machine learning methodology and/or artificial intelligence. In particular, attention should be paid to the HEMs that are electrochemically stable in acid media, to find good Pt-free electrocatalysts that can substitute Pt one day for the HER in PEMWE.

HEMs for the OER
The reaction mechanism for the OER is not as well established as that for the HER. To date, several mechanisms have been proposed, and which mechanism should apply to which category of catalysts remains an issue under debate [72,152]. Under acid conditions, the most widely accepted OER mechanism, i.e. adsorbate evolution mechanism (AEM), proceeds through the first formation of an absorbed OH (OH ad ) at the surface of the catalyst via an electron transfer from a water molecule physisorbed on the anode and simultaneous deprotonation (equation (15)). This is followed by a second electron transfer and consecutive deprotonation to generate an adsorbed oxygen (O ad ) species (equation (16)). The third step, is the nucleophilic attack of a second water molecule onto the pre-formed O ad species to form a hydroperoxo intermediate (OOH ad ) (equation (17)). Subsequently, a final electron transfer and deprotonation step leads to oxygen formation (equation (18)) [14], The AEM in alkaline media proceeds through the initial adsorption of a hydroxide ion (OH ad ) and electron transfer to the electrode (equation (19)). The following step involves another electron transfer of a second hydroxide ion and consecutive deprotonation of the previously formed OH ad to generate an O ad species (equation (20)). The last two electron transfer and deprotonation steps are shared with those described in the acidic OER mechanism (equations (17) and (18)) [14], According to equations (3) and (5), thermodynamically the OER is less energetically demanding in alkaline conditions than in acid conditions. This is because for the formation of an oxygen molecule two strong O-H bonds from a water molecule are broken in acid conditions, whilst under alkaline conditions oxygen evolves from hydroxide ions. Thus, materials employed as anode catalysts in acid media must endure strongly oxidising conditions, and the state-of-the-art acidic OER catalysts typically comprise IrO 2 and RuO 2 [153]. In particular, IrO 2 , despite being less active than RuO 2 , is preferred because of its higher stability against corrosion during OER. A low η 10 value of c.a. 270 mV and a small Tafel slope of c.a. 40 mV dec −1 are commonly reported for hydrous amorphous IrO x in the literature [153]. In addition, several nonpreciousmetal catalysts for acidic OER were developed as well, however, to date most of them show unsatisfactory performance compared to IrO 2 or RuO 2 , as illustrated by their typically higher overpotential (η 10 = 350-700 mV) and larger Tafel slope (c.a. 60 mV dec −1 ) [72]. The strong local acidic conditions generated at the electrode-electrolyte interface and the highly oxidising environment during the acidic OER generally result in the corrosion of almost all materials over a long term, even for Ir-based catalysts. Therefore, electrocatalytic stability is a main challenge to address for OER catalysts to be used in PEMWE, where HEMs would be good candidate helping to stabilise non-precious metal based catalysts due to the inherent slow atomic diffusion in HEMs [69]. Different from the acidic OER, under alkaline conditions IrO 2 and RuO 2 based catalysts usually show an insufficient activity and higher dissolution rates, making them unfeasible candidates [154]. It is worth stressing that in the literature many catalysts developed for alkaline OER are often compared to RuO 2 or IrO 2 , which are not good references under these conditions. Hence, such comparison is, to a large extent, of questionable significance. A more valuable assessment would be to compare developed catalysts under alkaline conditions to the stateof-the-art NiFeO x or CoFeO x catalysts [22,[155][156][157]. Nickel and cobalt oxy/hydroxides were reported to be conductive and stable under OER conditions in alkaline media. For instance, Ni(OH) 2 grown over commercial Ni foam through a simple hydrothermal treatment in just water (160 • C, 24 h) could reach 10 mA cm −2 at an ultralow overpotential of 170 mV in 0.1 M KOH, without signs of activity decay for 24 h at 50 mA cm −2 [158]. Besides, it was reported that the addition of Fe results in an enhanced activity towards OER, but there is ongoing discussion about whether Fe sites are the active sites for OER, promoting the oxidation of the host metal oxy/hydroxide towards an active phase (e.g. Ni 4+ ), or a combination of both works [159].

HEOs for oxygen evolution
Although OER involves the generation of multiple intermediates (OH ab , O ab , OOH ab ), the binding energy of these intermediates is linearly correlated. Man et al found that for different metal oxides (rutile, perovskite, spinel, rock salt and Bixbyite) the absorption energy difference between OH ad and OOH ad is always 3.2 eV [160]. In an ideal situation where all reaction steps have the same activation energy, the energetic difference between OH ad and OOH ad is 2.46 eV. Thus, the difference between actual catalysts and an ideal one, accounting for two electron transfers, results in a minimum overpotential of 370 mV. By plotting the absorption energy of OH ad against the absorption energy of OOH ad it was elucidated that both intermediates bind to the surface through a single oxygen bond at similar adsorption sites. Moreover, all computed structures fell within the same trend (figure 10), higher in energy than the ideal for OER to proceed at 1.23 eV [160]. This is the so-called 'scaling' relations. Through rational design of materials, e.g. developing new HEOs, the 'scaling' relations can be broken and accordingly an OER overpotential much lower than 370 mV be achieved.
In this regard, Svane and Rossmeisl computed by DFT the adsorption energies of OER reaction intermediates over Ru 0.2 Ti 0.2 Ir 0.2 Os 0.2 Rh 0.2 O 2 HEO with a rutile structure for coordinatively unsaturated and bridge sites [161]. A part of the reaction intermediates (OH ab , O ab , OOH ab ) involved in the OER mechanism described in equations (15)- (18), and a proton transfer from OH ab and OOH ab to a neighbouring bridging oxygen to form O ad and OO ad (+H ad ) was considered as an alternative reaction pathway ( figure 11(a)). The reaction pathway over each site was assumed to follow the minimum energy path, including a combination of both mechanisms. From the free energy of the four reaction steps the reaction overpotential was calculated ( figure 11(b)). The calculations showed that over Ru and Os sites the O ad + H ad step is favoured compared to OH ad , and the opposite is true for Ir and Rh sites. Moreover, OO ad + H is more stable than OOH ad on many of Ir and Rh sites. Interestingly, when plotting the overpotential histogram over different sites it was observed that for this specific composition (i.e. Ru 0.2 Ti 0.2 Ir 0.2 Os 0.2 Rh 0.2 O 2 ) OER can theoretically proceed at a lower overpotential than that dictated by the 'scaling' relations over Ir and Ru sites [161].
HEOs have been reported to outperform in activity and stability its single constituents. For instance, (CoNiMnZnFe)O with a rock salt structure was demonstrated to show an overpotential of 336 mV at 10 mA cm −2 , 27-119 mV lower than the corresponding monometallic metal oxides [94]. Nevertheless, further computational and experimental studies are required to aid in the design of HEOs in a wider composition landscape. Due to the faster kinetics of OER in alkaline media, as well as the feasibility of using earth-abundant elements as stable catalysts, most HEMs developed so far are targeted for use in AEL. Most of these HEMs contain principal components of Ni, Co and Fe [65,79,102,116,[162][163][164][165][166], with additional elements functioning as either active components or stabilisers.

HEOs with a spinel structure as OER catalysts.
Spinels have the general formula A 1 − δ B δ (A δ B 2 − δ )O 4 , with ions before and in the parenthesis occupying tetrahedral and octahedral sites, respectively. Spinels drew considerable interest as OER catalysts because of their compositional flexibility (nearly all main group metals and transition metals can be incorporated), which allows to easily tune their electronic structure. Moreover, spinels are generally thermally stable, electrically conductive and easy to prepare from cheap precursors. Zhao et al recently comprehensively reviewed the synthesis of spinels and their use in OER and other electrochemical processes [167], which readers interested to know more can refer to.
As far as the HEO spinel catalysts are concerned, Wang et al [95] lately reported unsupported (CoNiFeCuMn) 3 O 4 spinel NPs and those supported on multi-walled carbon nanotubes, prepared through a solvothermal synthesis, followed by post-annealing in air at 400 • C. The HEO constituents were selected given their high miscibility. When tested towards OER in 1.0 M KOH, the unsupported and supported (CoNiFeCuMn) 3 O 4 HEOs showed a η 10 value of 400 and 350 mV, respectively. No signs of catalyst degradation were observed during a 12 h chronopotentiometric (CP) test at a steady current density of 10 mA cm −2 . Interestingly, increasing the annealing temperature above 400 • C resulted in phase segregation with the concomitant decrease in activity [95], highlighting the importance of single-phase HEO in promoting the OER performance. Besides, Jin et al reported several nanoporous HEO spinels synthesised through an alloying-etching process, which allowed to elucidate the effect of elemental composition on the OER performance [101]. In this case, both ternary AlCoFe and quaternary AlCoF-eNi oxides were prepared following the same procedure to make a just comparison. The quaternary AlCoFeNi (c.a. η 10 = 330 mV) oxide was found to outperform in activity the ternary AlCoFe oxide (c.a. η 10 = 390 mV). Furthermore, the researchers found that adding a different fifth element to the HEO would lead to different trends in activity ( figure 12). For example, the addition of Mn (c.a. η 10 = 350 mV) resulted in a lower OER activity, whilst the addition of Cr or Mo gave rise to an improved activity (c.a. η 10 = 300 mV) compared to the quaternary AlCoFeNi oxide [101]. Intriguingly, when Pt was added at the beginning of the alloying-etching process, Pt clusters were segregated on the surface of (AlCoFeNiCr) 3 O 4 , instead of forming the corresponding senary HEO. The Pt modified HEO exhibited a small η 10 value of 250 mV and a Tafel slope of 50 mV dec −1 , when used to catalyse the OER in alkaline solution. The promoting effect of Pt was suggested to be related to the stabilisation of the OOH ad intermediate and favoured O 2 desorption [101]. In addition, Yang et al recently reported an inverse spinel HEO by combining OERactive Ni, Co and Fe elements with corrosion-resistant Cr and Al elements, which showed impressive performance, able to sustain continuous OER electrolysis at a high current density of 400 mA cm −2 in 0.1 M KOH for 150 h with no apparent sign of degradation [168].

HEOs with a perovskite structure as OER catalysts.
Single perovskites have a general formula of ABO 3 − δ , where alkaline or alkaline-earth cations occupy A sites, and transition metal cations occupy B sites. Perovskites are an important family of catalysts for OER, and readers are referred to the following reviews for a comprehensive picture [169][170][171][172]. Although some perovskites contain at least five different elements when counting A and B cations, high entropy perovskites encompass only those with five or more different cations on B sites. This is because B sites are generally the catalytically active sites for the OER. In perovskites, the transition metal d and oxygen 2p orbitals hybridise to form σ and π orbitals, σ and π antibonding orbitals being referred as e g and t 2g orbitals, respectively. The binding strength of  adsorbates correlates with the electron occupancy of e g orbitals, which in turn depends on the number of d electrons and the spin state of transition metal ion [169]. For OER, a volcano relationship was observed by plotting the e g electron filling against the activity, with an optimal e g filling near 1.2 [173]. For HEOs with a perovskite structure, tuning the composition can not only result in an optimal e g filling, but also lead to improved structural and electrochemical stability due to the high degree of lattice distortion which gives rise to slow atomic diffusion and the near-continuum distribution of adsorption sites.
Nguyen et al prepared through a facile precipitation process several single transition metal perovskites (A = La, B = Mn, Cr, Fe, Co or Ni) and compared them against the La(CrMnFeCoNi)O 3 HEO perovskite containing five elements at the B sites in an equimolar amount. Among the single-metal perovskites, LaCoO 3 and LaNiO 3 were the more active catalysts towards OER in 1.0 M KOH, with a η 10 value of 380 and 420 mV, respectively. In comparison, the La(CrMnFeCoNi)O 3 HEO perovskite outperformed both LaCoO 3 and LaNiO 3 , showing a η 10 of 359 mV [174]. It was further observed that the La(CrMnFeCoNi)O 3 HEO perovskite's OER activity was improved by doubling the concentration of one of the B-site elements, suggesting that the electronic structure of the HEO, and hence its activity, is influenced not only by the chemical composition but also by the concentration of B-site elements. With a higher concentration of Co, the best OER activity of La(Cr 0.17 Mn 0.17 Fe 0.17 Co 0.33 Ni 0.17 )O 3 was achieved with a η 10 of 325 mV [174]. To investigate the influence of the elemental concentration on OER, Sun et al developed Co-Ni-Fe ternary phase diagrams for double perovskites with a formula of A 2 BB ′ O 6 − δ . By varying the concentration of the OER-active elements (Fe, Ni and Co) in Sr 2 Fe 1.5 -x − y Co x Ni y Mo 0.5 O 6 − δ , they concluded that Fe was required as a scaffold to obtain a double perovskite structure, and the incorporation of Co led to improved stability. In addition, the incorporation of Co resulted in faster reaction kinetics as indicated by reduced Tafel slopes. As for Co, the incorporation of Ni improved the reaction kinetics. Whilst a synergistic effect in the activity was observed after co-doping with both Ni and Co (figure 13) [175].

HEAs for oxygen evolution
When HEAs are used as OER catalysts, oxo/hydroxides species will be established in situ on the catalyst surface as the active phase [155,156,159]. For instance, Chen et al showed that by doping a CrFeCoNiCu HEA with oxygen (1-3 at.%), the overpotential towards OER can be reduced by 100 mV [165]. Such oxo/hydroxide species can either be obtained by pre-activation of HEAs through CV cycles, or form spontaneously when HEAs are exposed to air [154]. Through an alloying-etching process, Fang et al synthesised nanoporous AlFeCoNiM (M = Mo, V, Cr, Nb, Ti, Cu, Mn) HEAs [102] and also the corresponding ternary AlFeNi and quaternary AlFeNiCo alloys as control catalysts for comparison. They observed that the OER activity in 1.0 M KOH increased from the ternary alloy to the quaternary alloy, and further improvement was observed with the addition of a fifth element and the formation of HEAs. They proposed that an increase in the electron occupancy at the Fermi level in HEAs resulted in an enhanced electrical conductivity, and hence, a higher activity. Although all HEAs outperformed in activity the quaternary alloys, the addition of some special elements like Cr, Mo and Nb led to a higher activity [101,102,163] compared to other elements such as Cu, Mn, V or Ti, as evidenced by the low η 10 value of 240 mV and small Tafel slope of 50 mV dec −1 obtained for AlFeCoNiCr, AlFeCoNiMo and AlFeCoNiNb in alkaline conditions. Additionally, HEAs were more stable than the quaternary alloy counterpart because of the high-entropy effect [102]. Through rational design of HEAs, Cui et al [66] reduced the overpotential of a AlCrFeCoNi HEA by substituting Al and Co by Cu and Mn. Al is a strong electron donor with high affinity towards oxygen, hampering O 2 release. In contrast, Cu has fully filled d orbitals which can weaken the oxygen binding energy when hybridised with the other elements. Moreover, Mn has a larger radius than Co which results in a lattice expansion that can better accommodate the hydroperoxo intermediate.
Following a similar alloying-etching process, nanoporous AlNiCoIrM (M = Mo, Cr, Cu, Nb, V) HEAs, and the corresponding ternary AlNiIr and quaternary AlNiCoIr alloys, were synthesised and assessed as catalysts for OER in acid conditions (0.5 M H 2 SO 4 ) [100]. Contrary to that observed in alkaline media, the addition of a fourth element into the ternary alloy resulted in a decrease in the activity, which the researchers assigned to the lower Ir mass on the electrode. Further adding a fifth element to form a HEA, a different effect on the activity was observed, depending on the metal introduced. The introduction of Cu had no effect on the activity compared to the quaternary AlNiCoIr alloy, even though the Ir loading was further decreased. The addition of Nb and Cr led to a decrease in the activity, whilst Mo and V gave rise to an enhanced activity, thus, highlighting the notable effect of HEA composition on the electronic structure and thereby the OER activity [176]. The higher activity of AlNiCoIrNb and AlNiCoIrCr was attributed to the increased covalency of Ir-O bonds and concomitant smaller free energy barriers for intermediates adsorption [100,177]. Besides, AlNiCoIrMo and AlNiCoIrV were stable under OER reaction conditions and showed only a slight increase in the overpotential after 7000 accelerated CV cycles (1.20-1.48 V RHE ), significantly lower than the deactivation observed for the ternary AlNiIr alloy. The best catalyst, AlNiCoIrMo, showed an impressive η 10 value of only 233 mV [100].

Other HEMs for oxygen evolution
As mentioned above, HEAs typically need a pre-activation treatment prior to OER to introduce a surface oxo/hydroxide layer on their surfaces, rendering a catalytically active phase. Comparably, other high entropy compounds such as phosphates [178], borides [179], selenides [180,181] and fluorides [182] would also undergo an activation process to yield a partially oxidised surface, when employed to catalyse OER [183]. Such formulations are promising because they often outperform their oxides/hydroxides counterparts, besides that such HEMs usually show better performance than their mono-, bi-, tri-and quaternary-metallic equivalents [94].
In the last few years, MFs have emerged as promising electrocatalysts for OER. This because the high electronegativity of fluorine draws the electron density from metal sites, leading to the formation of stabilised unsaturated sites. In this case, the surface metal oxy/hydroxides form spontaneously when exposed to the alkaline electrolyte, requiring no pre-activation treatment [182]. For instance, an iron-nickel fluoride catalyst was reported to exhibit a η 10 value as low as 225 mV [184]. Despite of outstanding activity, MFs only show reasonably good stability for OER under strong alkaline conditions. Thus, the development of high entropy fluorides with improved stability can make them more competitive against other formulations.
While fluoride compounds suffer from low electron conductivity, which can hinders their application as electrocatalysts, intercalation of oxygen in the anionic framework turns out to be able to improve their conductivity [186]. Recently, Wang et al prepared a high entropy perovskite oxide-halide catalyst by ball-milling a perovskite oxide (or a HEPO) with K(MgMnFeCoNi)F 3 [187]. This method allowed to successfully mix several oxide and fluoride perovskites within a single phase, where the mixtures showed OER performance better than individual oxide or fluoride perovskites ( figure 14). The improved activity was ascribed to the decrease in the resistance of the reaction intermediates absorbed at the electrode surface [187]. For perovskite fluorides, electrical conductivity can also be improved by varying the cation at A sites. For instance, the η 10 value for K(MgMnFeCoNi)F 3 could be decreased from 369 mV to 314 mV after Na is introduced forming K 0.8 Na 0.2 (MgMnFeCoNi)F 3 [185].
Transition MSs hold great potential for use in electrocatalysis, however, they suffer from strong corrosion under OER conditions [188]. The tendency towards dissolution under oxidising conditions can be overcome through designing high entropy sulphides (HES) because the lattice distortion and slow atomic diffusion of HES can help mitigate dissolution. Given the proven high alkaline OER performance of MS compounds containing Ni, Co and Fe, most of HESs reported so far are also based on the formulations consisting of Ni, Co and Fe as active components. However, the incorporation of Ni, Co and Fe to yield a high entropy MS composed of five or more elements without phase segregation is fairly challenging because of the immiscibility of NiS, CoS and FeS to form bimetallic phases [84]. To solve this problem, the pulse thermal decomposition (c.a. 1380 • C) method was developed, which allows for flexible synthesis of supported HES catalysts of tuneable size on the nanometre scale. For instance, (CrMnFeCoNi)S x with an average size of 12, 22, 34 and 41 nm were synthesised by varying the decomposition pulse time from 55 ms to 10 s. The constituent elements were selected according to DFT calculations because such a combination yielded a close-to-optimal adsorption energy for O ad and OOH ad intermediates. Using (CrMnFeCoNi)S x to catalyse the OER, a current density of 100 mA cm −2 was achieved at an overpotential of 295 mV in 1.0 M KOH, and meanwhile significantly improved stability was observed compared to the binary (NiFe)S counterpart [84].
Another HEM derivative with a great potential for use as OER catalysts is high entropy metal organic frameworks (HE-MOFs) since they can be made from abundant precursors and their synthesis is easily scalable. For example, Zhao et al synthesised 9 g of MnFeCoNiCu HE-MOF by mixing an aqueous solution containing the corresponding metal chloride precursors and 1,4-benzedicarboxylic acid with trimethylamine at room temperature [189]. When assessed towards OER in 1.0 M KOH, a small η 10 value of 254 mV was accomplished. Moreover, MnFeCoNiCu HE-MOF could retain 95% of its initial activity after operating at a constant current density of 10 mA cm −2 for 48 h. The HE-MOF outperformed in terms of activity and stability the mono-, bi-and tri-metallic MOFs prepared following the same procedure [189]. Table 2 summarizes the OER performance of the most prominent HEMs reported recently in literature.

Post-OER characterisation of HEM catalysts
It is well known that most electrocatalysts experience intricate morphological, compositional, and/or crystallographic transformation under harsh oxidative OER conditions [188], and the truly catalytically active species are typically generated in-situ on the catalyst's surface and may evolve dynamically during OER. Therefore, post-OER characterisation is important to understand the dynamic changes of HEM catalysts and correlate such changes with the catalyst's stability. Zhong et al reported that upon pre-activation during OER, soluble Cr +6 and V +5 species formed in FeCoNiCrVB HEA catalysts, which led to the leach of Cr and V in the subsequent OER process [210]. In addition, for FeMnCoCrNi HEA catalysts   [198,217,228]. Apart from dissolution, nonoxide HEMs (e.g. high entropy borides [217,219], or HESs [193]) would be converted into the corresponding oxides or hydroxides during OER, similar to monometallic and bimetallic compounds reported previously [188]. While currently the post-OER characterisation is mainly focused on the compositional changes of HEM catalysts including elements dissolution and oxidation, more comprehensive post-OER characterisation should be performed to elucidate microstructure and crystal structure changes as well; moreover, the stability number (i.e. S-number, characterised by the ratio of oxygen evolved over the catalyst dissolved), a newly proposed metric assessing the stability of electrocatalysts [248], had better be widely adopted in future, in order to obtain a full picture of HEM catalyst's stability [193].

HEMs for the overall WE
Electrocatalytic materials discussed above were engineered to catalyse one of the half reactions of water splitting, namely, HER or OER. But in fact, many HEMs potentially can be employed as bifunctional catalysts and are active simultaneously for both HER and OER [51,100,116]. In this respect, Glasscott et al reported the synthesis of CoFeLaNiPt NPs supported on carbon fibre via an electrodeposition process, and assessed their electrocatalytic performance towards overall water splitting in 0.1 M KOH [86]. They found that Pt acted as the main active site for HER, whilst Co, Ni and Fe were identified as active centres for OER. The role of La was to provide structural stability of the catalyst. With a metal loading of c.a. 10 ng cm −2 , an overpotential of 377 and 555 mV was required to attain 10 mA cm −2 for OER and HER, respectively. However, the researchers did not perform long-term stability test [86]. Other PGM-containing HEAs were also explored as bifunctional catalysts for overall water splitting. For example, PtIrCuNiCr HEA showed a combined HER and OER overpotential as low as 185 mV, and could continuously  [90]. Nevertheless, the aforementioned PGM-containing HEAs were assessed in alkaline media, where the usage of PGM is not absolutely necessary and may increase the cost of hydrogen production. Their performance towards overall water splitting should be critically evaluated in acidic conditions. As mentioned in previous sections, nanoporous HEAs can be obtained by metal melting and subsequent post chemical etching treatment. Using this method, nanoporous AlNiCoIrMo HEA catalysts were prepared and tested for overall water splitting in acid media (0.5 M H 2 SO 4 ). Prior to reaction, the catalyst was activated by CV scans in the potential range of 0.02-1.2 V RHE , which served to form an active oxy/hydroxide layer as well as to increase the concentration of Ir at the surface. Impressively, overall water splitting was accomplished at 10 mA cm −2 under a small overpotential of 250 mV, using AlNiCoIrMo as both HER and OER catalysts, which is lower than that of the state-of-the-art Pt/C-IrO 2 system (320 mV) [100]. Similarly, an AlNiCoRuMo HEA was also reported and was able to deliver a stable current density of 10 mA cm −2 at a 1.5 V cell voltage for 100 h, when employed as bifunctional catalysts for overall water splitting in 0.1 M KOH [135]. Besides, Arumugam et al also synthesised a PGM-free HEA consisting of CuCoNiFeMn and used it as bifunctional catalyst for overall water splitting in alkaline and neutral electrolytes [116]. For OER in alkaline electrolyte, CuCoNiFeMn exhibited an overpotential comparable to that of the RuO 2 benchmark catalyst. In comparison, it showed a relatively high overpotential for the HER, almost doubling that of the Pt/C benchmark. In terms of catalytic stability, it seems that the bifunctional CuCoNiFeMn did not show satisfactory performance: a 20% decay was observed after 20 h when tested in 0.1 M KOH, and more rapid decay occurred in cell voltage in neutral media, even working only at 10 mA cm −2 [116].
Apart from HEAs, other HEMs were also reported to catalyse the overall water splitting. For example, Zhao et al developed a one-pot wet chemical methodology for the synthesis of HEMPs NPs supported on carbon nanosheets. The prepared (CoCrFeMnNi)P/C was assessed for the overall water splitting in 1.0 M KOH electrolyte ( figure 15). For OER, the measured η 10 value of the HEMP (320 mV) was lower than that of the IrO 2 reference catalyst (440 mV), while for HER the (CoCrFeMnNi)P/C HEMP showed a η 10 of 136 mV, higher than 55 mV delivered by the Pt/C catalyst [141]. When the (CoCrFeMnNi)P/C HEMP was assembled in a cell and used as bifunctional HER and OER catalysts, a current density of 100 mA cm −2 could be delivered at a cell voltage of 1.78 V. In contrast, a higher cell voltage of 1.87 V was required to achieve the same current density when commercial Pt/C and IrO 2 were employed in the cathode and anode, respectively. The (CoCrFeMnNi)P/C HEMP cell showed a little decay of 1.2% after 24 h operation at 10 mA cm −2 [141]. Besides, Lei et al synthesised CoZnCdCuMn HES NPs on carbon nanofibres through ion exchange of a parent CoS/C sample. The CoZnCdCuMnS/C HES outperformed, in terms of activity and stability, the bimetallic (CoZnS/C, CoCuS/C, CoCdS/C, CoMnSS/C), trimetallic CoZnCdS/C and quaternary CoZnCdCuS/C catalysts towards the HER and the OER in 1.0 M KOH. Specifically, the CoZnCdCuMnS/C HES could reach 10 mA cm −2 at 173 and 220 mV for HER and OER, respectively. When tested as bifunctional catalysts, a constant current density of 10 mA cm −2 was attained at 1.63 V and could be maintained for 72 h [142]. Table 3 compiles a list of HEMs utilised as bifunctional HER and OER catalysts and the overall WE performance they delivered.