Earth-Abundant Electrocatalysts in Proton Exchange Membrane Electrolyzers

In order to adopt water electrolyzers as a main hydrogen production system, it is critical to develop inexpensive and earth-abundant catalysts. Currently, both half-reactions in water splitting depend heavily on noble metal catalysts. This review discusses the proton exchange membrane (PEM) water electrolysis (WE) and the progress in replacing the noble-metal catalysts with earth-abundant ones. The efforts within this field for the discovery of efficient and stable earth-abundant catalysts (EACs) have increased exponentially the last few years. The development of EACs for the oxygen evolution reaction (OER) in acidic media is particularly important, as the only stable and efficient catalysts until now are noble-metal oxides, such as IrOx and RuOx. On the hydrogen evolution reaction (HER) side, there is significant progress on EACs under acidic conditions, but there are very few reports of these EACs employed in full PEM WE cells. These two main issues are reviewed, and we conclude with prospects for innovation in EACs for the OER in acidic environments, as well as with a critical assessment of the few full PEM WE cells assembled with EACs.


Introduction
Currently, 81% of the global energy demand is met by fossil fuels and it is estimated that more than 540 EJ was supplied for the total global energy demand in 2014.This figure is expected to increase by 40% towards 2050 [1].The CO 2 emissions from combustion of fossil fuels are large enough to severely alter the Earth's climate and global ecosystem, forcing mankind to accelerate the return to renewable energy.This is amplified by the reserves of fossil fuels estimated to last only 50-60 years [2][3][4].
Hydrogen (H 2 ) can meet our future energy demands as a clean and sustainable fuel, but costeffective ways need to be developed for a successful turn towards the hydrogen economy [5][6][7][8][9].Water electrolysis is an environment friendly scheme for conversion of renewable electricity (e.g., solar, wind) into high purity hydrogen, but at present electrolysis accounts for only 4% of the total hydrogen production [10].The rest is covered by transformation of fossil fuels, such as natural gas steam reforming, coal gasification and partial oxidation of hydrocarbons [11][12][13][14], however, all these routes involve the release of CO 2 .Polymer electrolyte membrane water electrolysis (PEM WE) has the advantages of simplicity, compact design, fast response, high current densities, production of ultrapure hydrogen that can be electrochemically pressurized, and small footprint.The PEM WE concept was first investigated and demonstrated in the 1960s [15][16][17].Since then, substantial research has been dedicated to improve the different PEM WE components, and as a result, this technology is approaching commercial markets [18].What hinders the implementation of PEM WE on a large scale is its acidity, which necessitates the use of noble metals, such as Ir, Pt, or Ru as electrocatalysts.Additionally, acidic conditions are more preferable as the concentration of reactant protons is higher [19,20].The high cost of the polymeric membrane is another obstacle.Currently, the capital investment cost (CAPEX) for a PEM WE system is around $1500 per kWe (kW electricity input) and the cost per kg of H 2 is $7.1, taking into account that the electricity is provided by renewables [21][22][23].
Catalysts 2018, 8, x FOR PEER REVIEW 2 of 43 the advantages of simplicity, compact design, fast response, high current densities, production of ultrapure hydrogen that can be electrochemically pressurized, and small footprint.The PEM WE concept was first investigated and demonstrated in the 1960s [15][16][17].Since then, substantial research has been dedicated to improve the different PEM WE components, and as a result, this technology is approaching commercial markets [18].What hinders the implementation of PEM WE on a large scale is its acidity, which necessitates the use of noble metals, such as Ir, Pt, or Ru as electrocatalysts.
Additionally, acidic conditions are more preferable as the concentration of reactant protons is higher [19,20].The high cost of the polymeric membrane is another obstacle.Currently, the capital investment cost (CAPEX) for a PEM WE system is around $1500 per kWe (kW electricity input) and the cost per kg of H2 is $7.1, taking into account that the electricity is provided by renewables [21][22][23].In comparison, the H2 cost through steam methane reforming (SMR) is only $1.40 [24] and the optimistic break-even year for renewable PEM H2 production based on learning curves is around 2033 (Figure 1) [25].The same study underlines that the major cost of PEM lies in the electricity consumption [25].This is directly connected to the overpotential required for efficient water electrolysis, i.e., the overpotential of the electrocatalysts to reach certain current densities.It is established that in terms of efficiency and stability, the platinum group metals (PGMs) are the best choices for electrodes in a PEM electrolyzers, however, the question is at what cost.For example, the annual global production of Pt in 2017 was 1.7 × 10 5 kg, while the total demand for Pt in the same year was over 2.2 × 10 5 kg.If the recycled Pt is also considered as part of the production, the annual production of Pt just met the total demand (USGS 2016 Mineral Years Report).Therefore, widespread installation of Pt-Ir based PEM electrolyzers will dramatically increase the total demand of PGMs.As an example, a Terra Watt (TW) hydrogen production system requires 0.5 and 10 years of annual global production of Pt and Ir, respectively [26].One has also to take into consideration that Ir is typically produced as a minor by-product of Pt [27].In other words, the annual production of Ir is also determined by the production rate of Pt.As a result, the increasing demands of Ir will increase their cost due to its dependence on Pt mining.
We performed our own calculations, using the state-of-the-art PEM electrolyzer that we will return to in Chapter 3. In this system, the cathode has 0.4 mgPt/cm 2 of Pt, and the anode 1.54 mgIr/cm 2 and 0.54 mgRu/cm 2 of Ir and Ru, respectively.Our calculations (see SI for more information) suggest that such a PEM system with a power density of 1.18 W/cm 2 requires 1.5, 180 and 12 years of annual production of Pt, Ir and Ru, respectively, to cover 1 TW of hydrogen production.It is evident, that the replacement of the noble metal electrocatalysts for both the hydrogen evolution reaction (HER) and oxygen evolution reaction (OER) will have a tremendous impact on the future scale-up activities for PEM WE.Furthermore, competition will be avoided with other industrial activities, such as the automobile and electronics sectors, where the demand for PGMs is big.In comparison, the H 2 cost through steam methane reforming (SMR) is only $1.40 [25] and the optimistic break-even year for renewable PEM H 2 production based on learning curves is around 2033 (Figure 1) [24].The same study underlines that the major cost of PEM lies in the electricity consumption [24].This is directly connected to the overpotential required for efficient water electrolysis, i.e., the overpotential of the electrocatalysts to reach certain current densities.It is established that in terms of efficiency and stability, the platinum group metals (PGMs) are the best choices for electrodes in a PEM electrolyzers, however, the question is at what cost.For example, the annual global production of Pt in 2017 was 1.7 × 10 5 kg, while the total demand for Pt in the same year was over 2.2 × 10 5 kg.If the recycled Pt is also considered as part of the production, the annual production of Pt just met the total demand (USGS 2016 Mineral Years Report).Therefore, widespread installation of Pt-Ir based PEM electrolyzers will dramatically increase the total demand of PGMs.As an example, a Terra Watt (TW) hydrogen production system requires 0.5 and 10 years of annual global production of Pt and Ir, respectively [26].One has also to take into consideration that Ir is typically produced as a minor by-product of Pt [27].In other words, the annual production of Ir is also determined by the production rate of Pt.As a result, the increasing demands of Ir will increase their cost due to its dependence on Pt mining.

REN H2 (PEM)
We performed our own calculations, using the state-of-the-art PEM electrolyzer that we will return to in Chapter 3. In this system, the cathode has 0.4 mg Pt /cm 2 of Pt, and the anode 1.54 mg Ir /cm 2 and 0.54 mg Ru /cm 2 of Ir and Ru, respectively.Our calculations (see Supplementary Materials for more information) suggest that such a PEM system with a power density of 1.18 W/cm 2 requires 1.5, 180 and 12 years of annual production of Pt, Ir and Ru, respectively, to cover 1 TW of hydrogen production.It is evident, that the replacement of the noble metal electrocatalysts for both the hydrogen evolution reaction (HER) and oxygen evolution reaction (OER) will have a tremendous impact on the future scale-up activities for PEM WE.Furthermore, competition will be avoided with other industrial activities, such as the automobile and electronics sectors, where the demand for PGMs is big.
This field of research is very active, and according to Web of Science, 2043 reports have been published during 2017 on both OER and HER catalysts (Figure 2).Motivated by these figures, as well as the challenging electrochemistry under the intense conditions required by the PEM WE, we wanted to see how many of these reports referring to EACs were actually applied in PEM WE devices, replacing in fact the noble-metal catalysts.Therefore, the main purpose of this article is not primarily an exhaustive report on EACs developed for the HER and OER in acidic conditions, which were tested and studied in half-cells, typically involving measurements in three electrodes with rotating disc electrodes (RDE), but to focus on those applied and tested in full PEM WE cells.Do the catalysts perform as expected from the half-cell measurements, or are there deviations related to differences in configuration, supply of reactants, deposition on porous substrates, leaching of electroactive elements (i.e., stability), and surface area exposed?Moreover, what are the recent advances on EACs for the OER under strongly acidic conditions?In the current article we document the very first reports on EACs for the OER in acidic environment, as well as one applied EACs-based PEM WE system.Reprinted with permission from [55].Copyright 2018, The Royal Society of Chemistry.

Principles of PEM Water Electrolysis
The electrochemical conversion of water to hydrogen and oxygen is known as water electrolysis, and was discovered already in 1800 [56].Since then, the use of two electrodes immersed in an aqueous caustic solution of KOH electrolyte, known as alkaline water electrolysis, was developed and utilized for industrial applications [57].Although some improvements as current density and operating pressure are foreseeable [54], this well-established technology is still the most cost-effective choice for electrochemical hydrogen production on an industrial scale at present.
Another promising water electrolysis cell that operates at low temperatures (normally below 80 °C) is the proton exchange membrane (also known as polymer electrolyte membrane) (PEM) electrolyzers.The concept of PEM water electrolysis was idealized by Grubb in the early fifties [15,16] and first manufactured by the General Electric Co. in 1966 [17], where they take the advantage of a solid polymer perfluorinated sulfonic membrane as electrolyte for hydrogen production.Some

Principles of PEM Water Electrolysis
The electrochemical conversion of water to hydrogen and oxygen is known as water electrolysis, and was discovered already in 1800 [56].Since then, the use of two electrodes immersed in an aqueous caustic solution of KOH electrolyte, known as alkaline water electrolysis, was developed and utilized for industrial applications [57].Although some improvements as current density and operating pressure are foreseeable [54], this well-established technology is still the most cost-effective choice for electrochemical hydrogen production on an industrial scale at present.
Another promising water electrolysis cell that operates at low temperatures (normally below 80 • C) is the proton exchange membrane (also known as polymer electrolyte membrane) (PEM) electrolyzers.The concept of PEM water electrolysis was idealized by Grubb in the early fifties [15,16] and first manufactured by the General Electric Co. in 1966 [17], where they take the advantage of a solid polymer perfluorinated sulfonic membrane as electrolyte for hydrogen production.Some typical pros and cons for PEM water electrolyzers compared with the classic alkaline water electrolyzers are summarized in Table 1.
We highlight again that a cost reduction by developing earth-abundant electrocatalysts with comparable performance and a further improvement in the energy efficiency of the PEM water electrolyzers are essential factors before PEM WE becomes a competitive solution for large-scale hydrogen production.

Operating Principles
When a PEM electrolysis cell is in operation, an excess of water is supplied to the anode, where water decomposes into protons, electrons and oxygen gas by the electrical energy (Equation ( 1)).The protons are transported to the cathode by passing through the polymer electrolyte, while the generated electrons travel along an external circuit and combine with the protons into hydrogen gas, as described in Equation (2).The amount of hydrogen gas generated is twice that of oxygen, as defined by the overall reaction, Equation (3), whereas ∆G 0 is the standard Gibbs free energy of the net water splitting reaction.

Thermodynamics
The standard theoretical open circuit voltage (OCV), also referred as standard reversible cell voltage, U 0 rev , required by water electrolysis under standard conditions, can be derived from; the standard Gibbs free energy (∆G 0 R ) of + 237.2 kJ/mol H 2 , Faraday's constant (F), and the number of electrons (n = 2) exchanged during water splitting under standard conditions; p = 1 bar, T = 298.15K (Equation ( 4)) [61].
The positive Gibbs free energy change reflects that the water electrolysis reaction is thermodynamically unfavorable.For the reaction to proceed at finite rate, overpotentials for the OER and HER, as well as the electrolyte resistance, must be added to U 0 rev [58].These represent losses generating heat.At an overpotential of 0.25 V, i.e., an applied cell voltage of 1.48, V this heat balances the heat consumed by the reaction under standard conditions, and the cell operates in thermoneutral mode; 1.48 V is termed the thermoneutral voltage [62] and is reasonable to use when calculating the voltage efficiency of the cell.Thus, the actual operating cell voltage is the sum of all the different overpotentials (Equation ( 5)) [54,63].
U op is the operational voltage, U 0 rev is the standard reversible potential, η a , η c , η el and η sys are the overpotentials related to the anode, cathode, ionic conductivity of the electrolyte membrane, and system losses (resistance in contacts, interconnects, current collectors, wires, etc.), respectively.It should be emphasized that the half-reactions described in Equations ( 1) and ( 2) are simplifications of more complex multistep and parallel electrochemical reaction pathways [64].

Main Cell Components and Requirements
The core component of a PEM electrolysis cell is the membrane electrode assembly (MEA), which is composed of a solid polymer electrolyte (SPE) sandwiched between two electronically conductive electrodes, as shown in Figure 3.
The standard theoretical open circuit voltage (OCV), also referred as standard reversible cell voltage,  , required by water electrolysis under standard conditions, can be derived from; the standard Gibbs free energy (∆ ) of + 237.2 kJ/mol H2, Faraday's constant (F), and the number of electrons (n = 2) exchanged during water splitting under standard conditions; p = 1 bar, T = 298.15K (Equation ( 4)) [61].
The positive Gibbs free energy change reflects that the water electrolysis reaction is thermodynamically unfavorable.For the reaction to proceed at finite rate, overpotentials for the OER and HER, as well as the electrolyte resistance, must be added to  [58].These represent losses generating heat.At an overpotential of 0.25 V, i.e., an applied cell voltage of 1.48, V this heat balances the heat consumed by the reaction under standard conditions, and the cell operates in thermoneutral mode; 1.48 V is termed the thermoneutral voltage [62] and is reasonable to use when calculating the voltage efficiency of the cell.Thus, the actual operating cell voltage is the sum of all the different overpotentials (Equation ( 5)) [54,63].
is the operational voltage,  is the standard reversible potential,  ,  ,  and  are the overpotentials related to the anode, cathode, ionic conductivity of the electrolyte membrane, and system losses (resistance in contacts, interconnects, current collectors, wires, etc.), respectively.It should be emphasized that the half-reactions described in Equations ( 1) and ( 2) are simplifications of more complex multistep and parallel electrochemical reaction pathways [64].

Main Cell Components and Requirements
The core component of a PEM electrolysis cell is the membrane electrode assembly (MEA), which is composed of a solid polymer electrolyte (SPE) sandwiched between two electronically conductive electrodes, as shown in Figure 3.The SPE must fulfil particular requirements, such as high chemical and mechanical stability, low gas permeability, and high proton conductivity.In this regard, Nafion ® is the most commonly used polymer membrane due to high proton conductivity, good mechanical stability and acceptable gas crossover.The electrodes are composed of a porous catalyst layer (CL) with electrocatalysts dispersed The SPE must fulfil particular requirements, such as high chemical and mechanical stability, low gas permeability, and high proton conductivity.In this regard, Nafion ® is the most commonly used polymer membrane due to high proton conductivity, good mechanical stability and acceptable gas crossover.The electrodes are composed of a porous catalyst layer (CL) with electrocatalysts dispersed on a nanoporous support to lower the activation energy and promote charge transfer kinetics.Next comes a more openly porous gas diffusion layer (GDL), also acting as a current collector alone or by the help of additional metallic meshes or sinters.These are finally encased by bipolar plates (BPPs) which direct and distribute gases in flow channels, separate the anode and cathode side environments, and connect a cathode electrically to the next anode [58].
Due to the acidic environment of the cell, the catalysts for the hydrogen evolution reactions (HER) on the cathode and the oxygen evolution reactions (OER) on the anode are essentially dependent on noble metals and their alloys.Pt nanoparticles on carbon support is by far the best catalyst material for the HER because of their good catalytic activity and high corrosion resistance.Besides, Pd and Ir nanoparticles supported on carbon materials are also commonly utilized as HER electrocatalysts [65].Less expensive earth-abundant materials such as sulfides, phosphides, carbides and nitrides [18], cobalt clathrochelate [66], polyoxometallates [62] have been proposed as alternative HER catalysts.
The slowest reaction is normally the OER of the anode, determining the reaction rate of the overall process.Non-noble catalysts such as Ni and Co in contact with the acidic electrolyte would start to corrode, meanwhile the Pt surface will be covered by a low conducting oxide film, which reduces the catalytic activity for the OER.In this respect, Ir and Ru oxide-based catalysts are typically used for the OER because of their high structural stability.As reported by Ahn and Holze [67], Ru oxide appears to be the most catalytically active electrode with the smallest activation overpotential at 353 K, followed by Ir/Ru-oxide, Ir-oxide, Ir metal, Rh-oxide, Rh metal and Pt.However, Ir is scarce, its average mass fraction in crustal rock is only 0.001 ppm [54].

State-of-the-Art Devices
After General Electric developed the PEM WE technology, its application was mostly limited to oxygen production [68], e.g., for submarine and spacecraft applications.In the late 1980s, the first pressurized PEM electrolyzers for H 2 production up to 100 bar with efficient MEAs, were constructed and tested [69,70].Since then, MEAs with Ir, Ru-and Pt-based electrocatalysts and Nafion ® proton conductor polymer electrolyte have dominated frontier PEM electrolyzer cell designs [18,71].
The state-of-the-art OER catalyst for PEM electrolyzers is an oxide mixture composed of Ru 2 O and IrO 2 [72], e.g., Ir 0.7 Ru 0.3 O 2 [73] and Ir 0.4 Ru 0.6 O 2 [74], with slight differences in overpotential and stability when varying the composition of each oxide.Although RuO 2 has shown the best OER performance among all the other materials [54,74], its poor stability due to the corrosion [75] from the strong local acidity at the perfluorosulfonic membrane, and high anodic potential, it requires the addition of the more stable IrO 2 [76,77].However, Ir is one of the rarest elements on earth, and this sets the requirement to reduce/replace the Ir content in order to lower the price, such as by adding other elements that are more earth abundant, e.g., Co [78], Ta [79], and Sn [80].A recent study reported the state-of-the-art OER performance of fluoride doped MnO 2 , IrO 2 solid solution ((Mn 1−x Ir x )O 2 :F), with even lower onset potential than IrO 2 [81], may further reduce the Ir loading of the OER catalysts.
For the cathode, it is established that Pt, especially highly dispersed C-based Pt, is the benchmark HER catalyst for PEM electrolyzer [18].In fact, less research efforts have been made on the cathode material for PEM electrolyzers [54].The reason is partially that the exchange current density of H + /H 2 on Pt is almost 1000 times larger than that of H 2 O/O 2 on Ir [82], and Ir is also more precious than Pt, therefore research has been mainly focused on how to reduce the cost and increase the efficiency of OER catalyst.However, as the cathode side also contributes to a large extent in the cost of a PEM electrolyzer, it is necessary and important to reduce the loading of Pt [83], or replace it with efficient earth abundant electrocatalysts, such as MoS 2 [84] or CoP [85].This effort is briefly summarized below and as we set out earlier, our main target is to document actual application of EACs in full PEM WE cells.
Reported PEM electrolyzers with state-of-the-art electrocatalysts are summarized in Table 2.One can notice that the performance of a PEM electrolyzer is not only determined by the electrocatalysts, but also by other elements, e.g., operation temperature, cell area and membrane type.However, those elements are out of the scope of this review, hence they are not to be discussed here.

Earth-Abundant Cathode Materials
Thus far, we have explored the theory and principles of PEM WE and summarized the state-of-the-art devices demonstrated in the literature.In the following sections, we will explore the most promising earth-abundant electrocatalyst materials that have been used in full PEM WE cells, replacing noble metal-based anodes and cathodes, especially under acidic conditions.

Molybdenum Sulfide, MoS 2
Molybdenum sulfide (MoS 2 )-based materials are among the most extensively studied materials as catalyst for HER over the past decade due to their excellent stability, high activity, earth abundancy and low price.MoS 2 exists in nature with an atomic structure resembling that of graphite, a layered structure where each layer consists of a molybdenum layer sandwiched between two sulfur layers.Alternatively, the monolayers can be described as consisting of either edge sharing trigonal prisms (2H) or octahedrons (1T).Packing of these layers gives the basis for the three polytypes of bulk MoS 2 (Figure 4).Despite the early indications of low HER activity for bulk MoS2 [90], molybdenum sulfides turned out to be promising for replacing Pt.Theoretical work by Hinnemann et al. in 2005 showed that the edges are in fact catalytically active [91].Using Density Functional Theory (DFT) they calculated the hydrogen binding energy of the Mo(1 010) edge, where sulphur is unsaturated, and found it to be close to ideal value of 0 eV [89].In addition, they fabricated an MEA using Nafion ® , Despite the early indications of low HER activity for bulk MoS 2 [90], molybdenum sulfides turned out to be promising for replacing Pt.Theoretical work by Hinnemann et al. in 2005 showed that the edges are in fact catalytically active [91].Using Density Functional Theory (DFT) they calculated the hydrogen binding energy of the Mo(1010) edge, where sulphur is unsaturated, and found it to be close to ideal value of 0 eV [89].In addition, they fabricated an MEA using Nafion ® , nanoparticle MoS 2 on graphite as cathode, and Pt as anode, which achieved a current density of 10 mA/cm 2 at only 175 mV of overpotential.This was the best activity shown for an acid-stable and earth abundant catalyst at that time.Two years later, their theoretical prediction of the edges being the activity centers was confirmed experimentally by Jaramillo et al. [42].They deposited monolayer MoS 2 on Au (111) with physical vapor deposition in an H 2 S environment.After finding total edge lengths with STM and comparing with catalytic activity for various samples, they found that the reaction rate scaled with particle perimeter and not area.These findings sparked an interest in improving the catalytic activity in MoS 2 that is still growing today.
Since the main objective of the present review is to review the literature on device-tested electrodes, we will not go deep into the vast literature on MoS 2 -based electrocatalysts.We will rather briefly mention some of the methods that have been identified for increasing the HER activity of MoS 2 .One of the first and obvious approaches was to maximize the edge sites by making small particles.This led to investigations of the activity of [Mo 3 S 4 ] 4+ -clusters that showed HER activity but were less stable [92].Some years later, [Mo 3 S 13 ] 2− -clusters became a hot topic after results showing one of the highest per site activities [33].Another approach that has produced promising results is to deposit molybdenum sulfide onto something highly conducting and/or with high surface area, like nanotubes, nanowires, reduced graphene oxide etc. [93][94][95][96].Depending on the methods used, one often ends up with amorphous MoS x .Efforts to improve the activity of the semiconductor phase comprise doping, introducing vacancies, and strain engineering, which can activate the basal plane and edges that are not intrinsically active [97][98][99][100].The 1T phase is metastable, however, the metallic nature makes it highly conductive compared to the 2H phase, and, in addition, the basal plane is active as well, resulting in promising HER activity [101,102].For more in-depth reviews the reader is referred to a number of reviews [84,89,103,104].Despite all these efforts to improve the catalytic properties over the past decade, there are, to the best of our knowledge, only the following few reports on molybdenum sulfide-based cathodes implemented in a PEM cell.
In 2014, Corrales-Sánchez et al. were the first to report the performance of a PEM cell using MoS 2 -based cathodes [84].They reported the performance of three different types of MoS 2 -based electrodes, bare pristine MoS 2 , MoS 2 mixed with commercial conductive carbon, Vulcan ® XC72, and MoS 2 nanoparticles on reduced graphene oxide.The MEA used in the PEM cell consisted of IrO 2 particles and anode material that was spray deposited on each side of a Nafion membrane.Porous titanium diffusion layer and titanium current collectors on both sides of the MEA were sandwiched by the cell housing.The pristine MoS 2 was the worst performing cathode investigated achieving a current density of approximately 0.02 A/cm 2 at 1.9 V. Their best performing MoS 2 /rGO electrode achieved a current density of 0.1 A/cm 2 , while the best mixture of MoS 2 and Vulcan ® (47 wt.% MoS 2 ) reached almost 0.3 A/cm 2 at 1.9 V in the initial test.The latter electrode went through a stability test for 18 h at 2.0 V.The current density actually increased steadily for 15 h and reached 0.35 A/cm 2 .The authors speculated that the increase might be due to hydration effects.Furthermore, they also tested the effect of hot pressing of the MEA, which is recommended to ensure good contact between electrode and membrane.For three different MoS 2 /Vulcan mixes, the unpressed MEAs performed better than the hot-pressed ones.
Ng et al. identified three types of Mo-based cathode materials with excellent HER activity from three electrode measurements in 2015 [105].They later loaded the materials onto carbon black and tested them as cathodes in a PEM electrolyzer with Nafion as membrane and Ir on Ti-mesh as anode.One of their electrodes was based on molybdenum sulfide with an excess of sulfur according to the XPS measurement.The electrode exhibited a good performance and required 1.86 V to reach 0.5 A/cm 2 in addition to good stability.Furthermore, the current density reached over 0.9 A/cm 2 at 2 V. Another cathode, based on Mo 3 S 13 clusters, required only 1.81 V to reach 0.5 A/cm 2 , while at 2 V the current density reached almost 1.1 A/cm 2 .In the stability test, however, the current density dropped by approximately 120 mA/cm 2 over a period of 14 h at 1.85 V most likely due to detachment from the support or degradation of the clusters.The third and last material they tested was based on sulfur doped molybdenum phosphide and performed slightly better than the Mo 3 S 13 electrode.These are the best performances reported for molybdenum sulfide cathode in PEM electrolyzers to this day.
In early 2016 Kumar et al. reported that a cell with a MoS 2 nanocapsule cathode maintained a current density of approximately 60 mA/cm 2 for 200 h at 2.0 V [106].The cell consisted of a Nafion membrane and IrO 2 anode.The low performance is likely due to low conductivity and is comparable to that reported for bare MoS 2 [84].A study of this system mixed with carbon black should follow to allow comparison with other systems reviewed here.
The same year, Lu et al. reported the performance of an electrolyzer using amorphous molybdenum sulfide coated on a carbon cloth as cathode [107].The cathode was synthesized by using thermolysis to form amorphous MoS x on the carbon cloth.A post treatment with remote H 2 plasma introduced sulfur vacancies.The cell consisted of a Nafion membrane and RuO 2 nanoparticles on carbon paper as the anode.The cell required 2.76 V to reach 1 A/cm 2 and the current density at 2.0 V was slightly above 0.3 A/cm 2 .Earlier this year, Kim et al. published work on a similar cathode.They deposited amorphous molybdenum sulfide on carbon paper using electrodeposition.The PEM cell used a Nafion membrane and electrodeposited IrO 2 on carbon paper as anode.They investigated the effect of deposition potential and time on the performance.The best performing electrode reached a current 0.37 A/cm 2 at 1.9 V [108].

Nickel Phosphide, Ni 2 P
Nickel phosphide (Ni 2 P) has been demonstrated as one of the best earth-abundant electrocatalysts for HER [34,109].Extensive investigations on Ni 2 P have been performed in a three-electrode electrochemical cell and Ni 2 P exhibits the superior activity to split water with low overpotentials, while sustaining high current densities [110][111][112][113][114][115].However, after a thorough literature review, there are no reports, to our best of knowledge, that have implemented Ni 2 P in a PEM device.Nevertheless, we compare Ni 2 P with other earth-abundant electrocatalysts, and the recent developments on Ni 2 P as electrocatalysts for HER are briefly reviewed.
Ni 2 P can be synthesized by a variety of methods including solution-phase synthesis and gas-solid synthesis.The solution-phase synthesis is performed by using tri-n-octylphosphine (TOP) as a phosphorus source to react with Ni precursor [116].At elevated temperatures (above 300 • C), the TOP vaporizes rapidly and then phosphorizes different precursors, such as bulk Ni or Ni thin films, by forming Ni 2 P. For instance, Read et al. successfully synthesized Ni 2 P thin film on Ni substrate by the solution-phase synthesis method [113].Figure 5a shows SEM images of representative Ni 2 P film formed on the surface of Ni foil and the resulting Ni 2 P is highly porous.The corresponding powder XRD pattern in Figure 5c, clearly shows that both Ni 2 P and Ni are present without other impurities.The EDS element maps in Figures 5d and 2e further confirm the presence of Ni and P at the surface and the existence of a sharp interface between the Ni 2 P coating and the underlying Ni substrate.Figure 5f shows polarization data for the HER in 0.5 M H 2 SO 4 for a few transition metal phosphides (Ni 2 P, Fe 2 P, Co 2 P, Ni 2 P, Cu 3 P, and NiFeP) as cathodes.Ni 2 P showed the best HER performance in acidic solutions among those and required overpotentials of only −128 mV and −153 mV to reach a current density of −10 mA/cm 2 and −20 mA/cm 2 , respectively.However, in alkaline media, all tested metal phosphide electrodes exhibit lower electrocatalytic HER activity compared to those in acidic conditions.Ni 2 P films require overpotentials of around −200 mV to reach current densities of −10 mA/cm 2 in 1.0 M KOH.
substrate.Figure 5f shows polarization data for the HER in 0.5 M H2SO4 for a few transition metal phosphides (Ni2P, Fe2P, Co2P, Ni2P, Cu3P, and NiFeP) as cathodes.Ni2P showed the best HER performance in acidic solutions among those and required overpotentials of only −128 mV and −153 mV to reach a current density of −10 mA/cm 2 and −20 mA/cm 2 , respectively.However, in alkaline media, all tested metal phosphide electrodes exhibit lower electrocatalytic HER activity compared to those in acidic conditions.Ni2P films require overpotentials of around −200 mV to reach current densities of −10 mA/cm 2 in 1.0 M KOH.Gas-solid synthesis has also been implemented to synthesize Ni2P, where hypophosphites, for instance NH4H2PO2 and NaH2PO2, can decompose and release PH3 at elevated temperatures; The PH3 can further react directly with Ni precursors, such as metal oxides and metal hydroxides, to form Ni2P [117][118][119][120][121]. For instance, Sun et al., reported one porous multishelled Ni2P, which was successfully synthesized by a gas-solid method [120].The porous multishelled NiO precursor was reacted into Ni2P by using NaH2PO2 as the phosphorus source, as shown in Figure 6a.Electrochemical measurements were performed in a 1 M KOH solution.Figure 6b shows the linear sweep curves for carbon, nanostructured Ni2P, hierarchical Ni2P, multishelled Ni2P, and Pt/C.The multishelled Ni2P exhibits a small overpotential of 10 mV (at current density of 1.0 mA/cm 2 ) and a (g) Gas-solid synthesis has also been implemented to synthesize Ni 2 P, where hypophosphites, for instance NH 4 H 2 PO 2 and NaH 2 PO 2 , can decompose and release PH 3 at elevated temperatures; The PH 3 can further react directly with Ni precursors, such as metal oxides and metal hydroxides, to form Ni 2 P [117][118][119][120][121]. For instance, Sun et al., reported one porous multishelled Ni 2 P, which was successfully synthesized by a gas-solid method [120].The porous multishelled NiO precursor was reacted into Ni 2 P by using NaH 2 PO 2 as the phosphorus source, as shown in Figure 6a.Electrochemical measurements were performed in a 1 M KOH solution.Figure 6b shows the linear sweep curves for carbon, nanostructured Ni 2 P, hierarchical Ni 2 P, multishelled Ni 2 P, and Pt/C.The multishelled Ni 2 P exhibits a small overpotential of 10 mV (at current density of 1.0 mA/cm 2 ) and a rapid cathodic current increase as more negative potentials were applied.The overpotential driving a cathodic current density of 10 mA/cm 2 was 98 mV, which is much lower than that observed on hierarchical Ni 2 P (298 mV) and nanostructured Ni 2 P (214 mV). Figure 6c shows the Tafel plots of the tested samples.At lower overpotentials, Tafel analysis on the multishelled Ni 2 P exhibits a slope of 86.4 mV/decade, which is much smaller than those of hierarchical Ni 2 P (108.4 mV/decade) and nanostructured Ni 2 P (125.4 mV/decade), suggesting faster HER kinetics of the multishelled Ni 2 P. At the high-overpotential regime, a slightly upward deviation is observed in Tafel plots of Pt/C and hierarchical Ni 2 P, which could stem from the rate-limiting step gradually changing from the Heyrovsky to the Volmer mechanism at high current densities [122].This porous multishelled structure endows Ni 2 P with short charge transport distances and abundant active sites, resulting in superior catalytic activity than those of Ni 2 P with other morphologies [120].
nanostructured Ni2P (125.4 mV/decade), suggesting faster HER kinetics of the multishelled Ni2P.At the high-overpotential regime, a slightly upward deviation is observed in Tafel plots of Pt/C and hierarchical Ni2P, which could stem from the rate-limiting step gradually changing from the Heyrovsky to the Volmer mechanism at high current densities [122].This porous multishelled structure endows Ni2P with short charge transport distances and abundant active sites, resulting in superior catalytic activity than those of Ni2P with other morphologies [120].A catalytic reaction is highly sensitive to the surface of the catalyst.One of the most common strategies to enhance the catalyst performance is by increasing the active facet of the catalyst.Several computational studies have suggested that Ni2P(001) surface is an active facet for HER due to an ensemble effect, whereby the presence of P decreases the number of metal-hollow sites, providing a relatively weak binding between protons and Ni−P bridges, the sites to facilitate catalysis of the HER [123,124].Later on, Popczun et al. successfully synthesized Ni2P nanoparticles which possessed a high density of exposed (001) facets (as shown in Figure 7) and then these Ni2P were tested as cathodes for the HER in 0.50 M H2SO4 [125].The overpotentials required for the Ni2P nanoparticle to produce cathodic current densities of 20 and 100 mA/cm 2 were 130 and 180 mV, respectively.These overpotentials are lower than those of none-preferred facet Ni2P [113] and other non-Pt HER electrocatalysts, including bulk MoS2 [94] and MoC [126].Figure 7c  A catalytic reaction is highly sensitive to the surface of the catalyst.One of the most common strategies to enhance the catalyst performance is by increasing the active facet of the catalyst.Several computational studies have suggested that Ni 2 P(001) surface is an active facet for HER due to an ensemble effect, whereby the presence of P decreases the number of metal-hollow sites, providing a relatively weak binding between protons and Ni−P bridges, the sites to facilitate catalysis of the HER [123,124].Later on, Popczun et al. successfully synthesized Ni 2 P nanoparticles which possessed a high density of exposed (001) facets (as shown in Figure 7) and then these Ni 2 P were tested as cathodes for the HER in 0.50 M H 2 SO 4 [125].The overpotentials required for the Ni 2 P nanoparticle to produce cathodic current densities of 20 and 100 mA/cm 2 were 130 and 180 mV, respectively.These overpotentials are lower than those of none-preferred facet Ni 2 P [113] and other non-Pt HER electrocatalysts, including bulk MoS 2 [94] and MoC [126].Figure 7c displays corresponding Tafel plots for Ni 2 P electrodes.Tafel analyses of the Ni 2 P nanoparticles show an exchange current density of 3.3 × 10 −5 A/cm 2 and a Tafel slope of ~46 mV/decade in the overpotential region of 25-125 mV.At higher overpotentials (150-200 mV), the Tafel slope and exchange current density increased to ~81 mV/decade and 4.9 × 10 −4 A/cm 2 , respectively.Again, this Tafel slope behavior reflects the change in the rate-limiting step of the HER [122].
Cation doping is an effective strategy to improve the HER activity of electrocatalysts.A few cations, such as Mn, Fe and Mo, have been reported to dope Ni 2 P [110,111,[127][128][129].For instance, Li et al. synthesized a series of (Ni x Fe 1−x ) 2 P by varying the amount of Fe doping ratio [128].They found out that HER activities for (Ni x Fe 1−x ) 2 P electrodes show a volcano shape as a function of Fe doping ratio (see Figure 8); HER activities first increased as Fe content increased until the composition reaches (Ni 0.33 Fe 0.67 ) 2 P.Then, by further increasing the Fe content, HER performance decreased gradually.(Ni 0.33 Fe 0.67 ) 2 P shows the best performance among the tested (Ni x Fe 1−x ) 2 P samples, with a small overpotential of 214 mV to reach cathodic current densities of 50 mA/cm 2 .Such an interesting behavior could stem from an increase in the electrocatalytically active surface areas, as well as a change in the electronic structure with increasing Fe content [128,130].Cation doping is an effective strategy to improve the HER activity of electrocatalysts.A few cations, such as Mn, Fe and Mo, have been reported to dope Ni2P [110,111,[127][128][129].For instance, Li et al. synthesized a series of (NixFe1−x)2P by varying the amount of Fe doping ratio [128].They found out that HER activities for (NixFe1−x)2P electrodes show a volcano shape as a function of Fe doping ratio (see Figure 8); HER activities first increased as Fe content increased until the composition reaches (Ni0.33Fe0.67)2P.Then, by further increasing the Fe content, HER performance decreased gradually.(Ni0.33Fe0.67)2Pshows the best performance among the tested (NixFe1−x)2P samples, with a small overpotential of 214 mV to reach cathodic current densities of 50 mA/cm 2 .Such an interesting behavior could stem from an increase in the electrocatalytically active surface areas, as well as a change in the electronic structure with increasing Fe content [128,130].

Iron Sulfides, FexSy
Metal chalcogenides have received interest as HER electrocatalysts over the past decades such as molybdenum sulfide MoS2 [42], tungsten sulfide WS2 [143], iron phosphide FeP [47] or nickel phosphide Ni2P [125].Among them, iron sulfides (generally noted as FexSy) show great interest, especially being the most abundant mineral on the Earth's surface, and pyrrhotite Fe9S10 being the most abundant iron sulfide in the Earth and solar system [144,145].
To our knowledge, the only study of iron sulfide electrocatalysts in a PEM WE device has been published by Di Giovanni et al. [145].In this paper the authors describe the synthesis and characterization of different stoichiometries of iron sulfide FexSy nanomaterials and their activity toward the HER.Pyrite FeS2, greigite Fe3S4, and pyrrhotite Fe9S10 crystalline phases were first prepared using a polyol synthetic route.Morphological and electronic properties of the prepared nanoparticles were characterized, as well as their electrochemical properties.Greigite is formed of micrometer-sized gypsum flowerlike particles consisting of thin platelets with a very high aspect ratio.Pyrite particles have a hierarchical morphology consisting of large micrometer-sized spheres of aggregated smaller particles.Their performances were investigated in situ in a PEM electrolyzer single cell.MEA were prepared using pyrite, pyrrhotite, or greigite as the anode catalyst and tested

Iron Sulfides, Fe x S y
Metal chalcogenides have received interest as HER electrocatalysts over the past decades such as molybdenum sulfide MoS 2 [42], tungsten sulfide WS 2 [143], iron phosphide FeP [47] or nickel phosphide Ni 2 P [125].Among them, iron sulfides (generally noted as Fe x S y ) show great interest, especially being the most abundant mineral on the Earth's surface, and pyrrhotite Fe 9 S 10 being the most abundant iron sulfide in the Earth and solar system [144,145].
To our knowledge, the only study of iron sulfide electrocatalysts in a PEM WE device has been published by Di Giovanni et al. [145].In this paper the authors describe the synthesis and characterization of different stoichiometries of iron sulfide Fe x S y nanomaterials and their activity toward the HER.Pyrite FeS 2 , greigite Fe 3 S 4 , and pyrrhotite Fe 9 S 10 crystalline phases were first prepared using a polyol synthetic route.Morphological and electronic properties of the prepared nanoparticles were characterized, as well as their electrochemical properties.Greigite is formed of micrometer-sized gypsum flowerlike particles consisting of thin platelets with a very high aspect ratio.Pyrite particles have a hierarchical morphology consisting of large micrometer-sized spheres of aggregated smaller particles.Their performances were investigated in situ in a PEM electrolyzer single cell.MEA were prepared using pyrite, pyrrhotite, or greigite as the anode catalyst and tested in a PEM electrolysis single cell.The catalysts were not supported, but were mixed with 20% of carbon black.Nafion 115 (125 µm) was used as the membrane and IrO 2 as the anode catalyst.A cross section SEM image is presented in Figure 10 left.For the same catalyst loading, both ex situ and in situ (Figure 10 right) electrochemical experiments showed that pyrite (FeS 2 ) is the most active compared to greigite Fe 3 S 4 and pyrrhotite Fe 9 S 10 , with the electrocatalysis starting at an overpotential of ca.180 mV.These three materials exhibited a very stable behavior during measurement, with no activity degradation for at least 5 days.All catalysts have been tested in a PEM electrolysis single cell, and pyrite FeS 2 allows a current density of 2 A/cm 2 at a voltage of 2.3 V.It is noteworthy that FexSy based materials have been studied as electrocatalysts for the HER and showed promising results.
FeS pyrrhotite has been prepared by a solvothermal route and showed hexagonal shaped nanoparticles with size ranging from 50 to 500 nm, achieving electrocatalysis for molecular hydrogen evolution with no structural decomposition or activity decrease for at least six days at an overpotential of 350 mV in neutral water [146].
Fe2S pyrite has been prepared by Faber et al. by electron-beam evaporation on borosilicate substrates following by a thermal sulfidation [147].The cathodic overpotential to drive the HER at 1 mA/cm 2 for Fe2S pyrite was 217 mV.
Miao et al. prepared mesoporous Fe2S materials with high surface area by a sol-gel method followed by a sulfurization treatment in an H2S atmosphere [148].An interesting HER catalytic performance was achieved with a rather low overpotential of 96 mV at a current density of 10 mA/cm 2 and a Tafel slope of 78 mV/decade under alkaline conditions (pH 13).
Jasion et al. proposed the synthesis of nanostructured Fe2S [149].By changing the Fe:S ratio in the precursor solution, they were able to preferentially synthesize either 1D wire or 2D disc nanostructures.The HER electrocatalytic activity of the nanostructured FeS2 (drop-casted on a glassy carbon electrode) was measured via linear sweep voltammetry (LSV) and showed the best results for the 2D disc structures with an overpotential of just 50 mV larger than that of Pt.
Chua and Pumera investigated the electrochemical hydrogen evolution of natural FeS2 [150].Interestingly, they focused on the susceptibility of natural FeS2 hydrogen evolution performances towards sulfide poisoning, a major issue for cathodic hydrogen evolution.The results showed a better response of the FeS2 electrodes than platinum.
A hybrid catalyst of cobalt-doped FeS2 nanosheets-carbon nanotubes for the HER was proposed by Wang  It is noteworthy that Fe x S y based materials have been studied as electrocatalysts for the HER and showed promising results.
FeS pyrrhotite has been prepared by a solvothermal route and showed hexagonal shaped nanoparticles with size ranging from 50 to 500 nm, achieving electrocatalysis for molecular hydrogen evolution with no structural decomposition or activity decrease for at least six days at an overpotential of 350 mV in neutral water [146].
Fe 2 S pyrite has been prepared by Faber et al. by electron-beam evaporation on borosilicate substrates following by a thermal sulfidation [147].The cathodic overpotential to drive the HER at 1 mA/cm 2 for Fe 2 S pyrite was 217 mV.
Miao et al. prepared mesoporous Fe 2 S materials with high surface area by a sol-gel method followed by a sulfurization treatment in an H 2 S atmosphere [148].An interesting HER catalytic performance was achieved with a rather low overpotential of 96 mV at a current density of 10 mA/cm 2 and a Tafel slope of 78 mV/decade under alkaline conditions (pH 13).
Jasion et al. proposed the synthesis of nanostructured Fe 2 S [149].By changing the Fe:S ratio in the precursor solution, they were able to preferentially synthesize either 1D wire or 2D disc nanostructures.The HER electrocatalytic activity of the nanostructured FeS 2 (drop-casted on a glassy carbon electrode) was measured via linear sweep voltammetry (LSV) and showed the best results for the 2D disc structures with an overpotential of just 50 mV larger than that of Pt.
Chua and Pumera investigated the electrochemical hydrogen evolution of natural FeS 2 [150].Interestingly, they focused on the susceptibility of natural FeS 2 hydrogen evolution performances towards sulfide poisoning, a major issue for cathodic hydrogen evolution.The results showed a better response of the FeS 2 electrodes than platinum.
A hybrid catalyst of cobalt-doped FeS 2 nanosheets-carbon nanotubes for the HER was proposed by Wang et al. [151].The pyrite phase of Fe 1−x Co x S 2 /CNT showed a low overpotential of ~120 mV at 20 mA/cm 2 , a low Tafel slope of ~46 mV/decade, and long-term durability over 40 h of HER operation.Huang et al. employed carbon black as a support to prepare a cobalt-doped iron sulfide electrocatalyst with high-electrical conductivity and maximal active sites [152].Electrochemical results showed an enhancement in the HER activity of Co-doped FeS 2 in comparison to undoped FeS 2 in acidic electrolyte (pH = 0).The overpotential necessary to drive a current density of 10 mA/cm 2 is 150 mV and only decreases by 1 mV after 500 cycles during a durability test.
Bi-functional iron-only electrocatalysts for both water splitting half reactions are proposed by Martindale et al. [153].Full water splitting at a current density of 10 mA/cm 2 is achieved at a bias of ca. 2 V, which is stable for at least 3 days.
Iron sulfide alloys have also shown potential catalytic activity.Yu et al. report the 3D ternary nickel iron sulfide (Ni 0.7 Fe 0.3 S 2 ) microflowers with a hierarchically porous structure delivering an overpotential of 198 mV at a current density of 10 mA/cm 2 [154].Zhu et al. proposed bimetallic iron-nickel sulfide (Fe 11.1% -Ni 3 S 2 ) nanoarrays supported on nickel foam having a η 10 of 126 mV [155].
A patent has also been filed for the use of iron sulfide in an electrolytic cell [156].

Carbon-Based Materials
Due to the earth abundancy and high electronic conductivity, carbon based materials, such as carbon nanoparticles (CNPs), carbon nanotubes (CNTs), graphene, etc., are mostly used as the supporting material for the electron transfer between the substrates and the electrocatalysts [157].One of the most successful carbon material used as electrocatalyst support is carbon black, which is a commercially available product with high surface area (ca.200-1000 m 2 /g) [158].By uniformly dispersing electrocatalyst NPs on carbon black, the electrochemically active surface area (EASA) of the electrocatalyst can be maximized, and the amount of the catalyst, such as Pt, can be minimized.Pt/C is actually the benchmark HER catalyst for PEM electrolysis [159].
In order to further reduce the cost of H 2 produced by the PEM electrolyzer, other carbon-supported electrocatalysts, especially those only consist of earth abundant elements, such as Mo 2 C/CNTs [126], A-Ni-C (atomically isolated Ni anchored on graphitic carbon) [160], Co-doped FeS 2 /CNTs [151], CoFe nanoalloys encapsulated in N-doped graphene [161], Ni 2 P/CNTs [136], WO 2 /C nanowires [162], etc., have been studied as potential HER catalysts alternative to Pt.However, carbon-supported and Pt-free HER catalysts that have actually been tested in a real PEM device are rarely reported, and only a few can be found in the literature, and they are summarized in Table 3.
Nevertheless, the usage of C-based materials is not only limited to the anode.A recent study shows that carbon nitride (C 3 N 4 ) resist the harsh conditions at the anode side and therefore can be used as the supporting material for OER catalysts, such as IrO 2 , hence to reduce the Ir content at the anode [163].

Co-Clathrochelates
The interest in Co-clathrochelates as electrocatalysts is prompted by their ability to maintain the same ligand environment for Co in different oxidation states [164].However, only a few studies can be found implementing Co-clathrochelates in PEM electrolyzers.As can be seen from Table 3, the cell performance when cathodes are impregnated with such stable Co-containing electrocatalyst complexes is comparable to other earth-abundant catalyst systems, achieving current densities of 0.65 and 1 A/cm 2 at 1.7 and 2.15 V, respectively (Dinh Nguyen et al. [165] and Grigoriev et al. [166]).In both these works, the Co-clathrochelates were implemented in 7 cm 2 cells, but with different loadings.Figure 11  complexes is comparable to other earth-abundant catalyst systems, achieving current densities of 0.65 and 1 A/cm 2 at 1.7 and 2.15 V, respectively (Dinh Nguyen et al. [165] and Grigoriev et al. [166]).In both these works, the Co-clathrochelates were implemented in 7 cm 2 cells, but with different loadings.In Figure 11 left, the Co(dmgBF 2 ) 2 shows better electrochemical performance than [Co(dmg) 3 (BF) 2 ]BF 4 in the three-electrode configuration.However, when the two electrode modifications above were implemented in single cells for i-V characterization and stability testing under operational conditions, the [Co(dmg) 3 (BF) 2 ]BF 4 catalyst shows the best performance.The results are given in Figure 12 for current-potential and stability, respectively.The results reveal an increased cell voltage of 0.2-0.25 V when substituting the HER catalyst from Pt to Co-clathrochelates.The catalysts show no sign of degradation after 60 h of operation at 0.2 A/cm 2 .In Figure 11 left, the Co(dmgBF2)2 shows better electrochemical performance than [Co(dmg)3(BF)2]BF4 in the three-electrode configuration.However, when the two electrode modifications above were implemented in single cells for i-V characterization and stability testing under operational conditions, the [Co(dmg)3(BF)2]BF4 catalyst shows the best performance.The results are given in Figure 12 for current-potential and stability, respectively.The results reveal an increased cell voltage of 0.2-0.25 V when substituting the HER catalyst from Pt to Co-clathrochelates.The catalysts show no sign of degradation after 60 h of operation at 0.2 A/cm 2 .In an earlier study by Millet et al. [62], a 23 cm 2 cell was prepared with a Co(dmgBF2)2 catalyst with a loading of 1 mg/cm 2 in carbon black as compared to the 2.5 mg/cm 2 of the same catalyst in in the study of Dinh Nguyen et al., resulting in an increased cell voltage of 0.1 V to obtain a current density of 0.5 A/cm 2 .It must here be noted, however, that Miller's cell ran at 90 °C, while Dinh Nguyen's was operated at 60 °C, so the real difference under equal condition can be expected to be somewhat larger, suggesting that a 150% increase in the catalyst loading makes a significant impact on cell efficiency.The discrepancy between the results in half-cell and full-cell testing clearly underlines the need for testing in operation conditions before concluding on electrochemical performance.Co-and Fe-hexachloroclathrochelates have also been applied by Grigoriev et al. in a full cell, impregnated on Vulcan XC72 Gas Diffusion Electrodes (GDEs) with a surface area of 7 cm 2 [166].The main outcome is that substituting Co with Fe improves the electrocatalytic performance of In an earlier study by Millet et al. [62], a 23 cm 2 cell was prepared with a Co(dmgBF2)2 catalyst with a loading of 1 mg/cm 2 in carbon black as compared to the 2.5 mg/cm 2 of the same catalyst in in the study of Dinh Nguyen et al., resulting in an increased cell voltage of 0.1 V to obtain a current density of 0.5 A/cm 2 .It must here be noted, however, that Miller's cell ran at 90 • C, while Dinh Nguyen's was operated at 60 • C, so the real difference under equal condition can be expected to be somewhat larger, suggesting that a 150% increase in the catalyst loading makes a significant impact on cell efficiency.The discrepancy between the results in half-cell and full-cell testing clearly underlines the need for testing in operation conditions before concluding on electrochemical performance.Coand Fe-hexachloroclathrochelates have also been applied by Grigoriev et al. in a full cell, impregnated on Vulcan XC72 Gas Diffusion Electrodes (GDEs) with a surface area of 7 cm 2 [166].The main outcome is that substituting Co with Fe improves the electrocatalytic performance of the same macromolecule (Figure 13).One can also see that the overvoltage is around 0.25 V higher for the hexachloroclathrochelates than for the carbon supported Pt cathode used as reference.Comparing the results of Grigoriev et al. to the results reported by Dinh Nguyen et al. is difficult, since no information is given with respect to ohmic contributions to cell resistance for the former, while ohmic contributions are subtracted for the latter.However, the same difference in overvoltage can be seen with respect to carbon supported Pt.
Catalysts 2018, 8, x FOR PEER REVIEW 18 of 43 from strongly electron-withdrawing fluorine, via moderately electron withdrawing phenyl to electron donating methyl groups.The electron affinity of the ligands did, however, affect the reduction potential of Co to surface nanoparticles, which in turn improved the HER [167].Xile Hu et al. reported a more ambiguous effect of manipulating electron affinities by substituting phenyl-for methyl ligands.In this study, a more positive potential for H2 evolution correlated with a decreased activity for electrocatalysis.Complex red-ox behavior was also reported in this study, such as Co(III) hydride intermediates formed upon reduction in acidic media [168].Zelinskii et al. utilized perfluorophenyl-ribbed substituents to stabilize Co(I) in an effort to enhance the HER, but although the reduced Co(I) was successfully stabilized, the resulting Co-clathrochelate complex was not electrochemically active in the HER [169].
(a) (b) One of the main challenges for non-noble metal catalysts in aqueous electrolyzer cathodes is their stability in harsh acidic conditions.The Co-clathrochelates show good stability in the reported works, exemplified by a stable overvoltage of 240 mV and a faradaic efficiency of 80%, remaining stable for more than 7 h in pH = 2 and at 1 mA/cm 2 and 0.9 V [170].Grigoriev et al. reported that the HER performance of Co-encapsulating macromolecules is improved by adding electron-withdrawing ligands, but otherwise changing ligands makes little difference as long as the electronic structure is similar.This can be seen for different aryl and alkyl apical substituents in [166].El Ghachtouli et al. reported that the exchange of ligands between fluorine and phenyl or methyl groups has negligible effect on i-V behavior, although the ligands go from strongly electron-withdrawing fluorine, via moderately electron withdrawing phenyl to electron donating methyl groups.The electron affinity of the ligands did, however, affect the reduction potential of Co to surface nanoparticles, which in turn improved the HER [167].Xile Hu et al. reported a more ambiguous effect of manipulating electron affinities by substituting phenyl-for methyl ligands.In this study, a more positive potential for H 2 evolution correlated with a decreased activity for electrocatalysis.Complex red-ox behavior was also reported in this study, such as Co(III) hydride intermediates formed upon reduction in acidic media [168].Zelinskii et al. utilized perfluorophenyl-ribbed substituents to stabilize Co(I) in an effort to enhance the HER, but although the reduced Co(I) was successfully stabilized, the resulting Co-clathrochelate complex was not electrochemically active in the HER [169].
One of the main challenges for non-noble metal catalysts in aqueous electrolyzer cathodes is their stability in harsh acidic conditions.The Co-clathrochelates show good stability in the reported works, exemplified by a stable overvoltage of 240 mV and a faradaic efficiency of 80%, remaining stable for more than 7 h in pH = 2 and at 1 mA/cm 2 and 0.9 V [170].

Density Functional Theory (DFT) for HER Catalysts
Density functional theory (DFT) is an essential tool for understanding the mechanisms and active sites of novel catalysts as it enables evaluation of the thermodynamics of the individual steps in HER.Modelling reaction barriers is however computationally demanding, and most studies as such, rather adopt a "∆G approach".As HER involves both proton transfer and charge transfer, the activity of a catalyst is intrinsically linked to its crystal and electronic structures.In that respect, the hydrogen bonding strength/adsorption energy (∆G H ) has been widely used as descriptor of catalyst activity.Following the Sabatier principle, too strong or weak interactions with the catalyst surface tends to lower the overall catalyst activity yielding the typical volcano type behavior (Figure 14).

Density Functional Theory (DFT) for HER Catalysts
Density functional theory (DFT) is an essential tool for understanding the mechanisms and active sites of novel catalysts as it enables evaluation of the thermodynamics of the individual steps in HER.Modelling reaction barriers is however computationally demanding, and most studies as such, rather adopt a "ΔG approach".As HER involves both proton transfer and charge transfer, the activity of a catalyst is intrinsically linked to its crystal and electronic structures.In that respect, the hydrogen bonding strength/adsorption energy (ΔGH) has been widely used as descriptor of catalyst activity.Following the Sabatier principle, too strong or weak interactions with the catalyst surface tends to lower the overall catalyst activity yielding the typical volcano type behavior (Figure 14).MoS2 and similar layered transition metal dichalcogenides (TMD) crystallize in two structures, the 2H and 1T polymorphs (Figure 15), with trigonal prismatic and octahedral coordination, respectively.The thermodynamically stable 2H polymorph of single layer MoS2 is semiconducting with a band gap of 1.74 eV [177], and its (0001) basal plane exhibits negligible catalytic activity towards HER due to a ΔGH of ~2 eV [176].On the other hand, Hinnemann et al. [91] showed that the (-1010) Mo edge sites of single trilayer MoS2 can be highly active towards HER, and that they resemble the active sites of the hydrogen-evolving enzymes nitrogenase and hydrogenase [42,91].The Mo edge exhibits a calculated ΔGH of merely 0.08 eV (Figure 16), compared to 0.18 eV of the (10-10) S edge [178], and is as such, close to thermoneutral (for low H coverages).The increased activity of the edge sites is MoS 2 and similar layered transition metal dichalcogenides (TMD) crystallize in two structures, the 2H and 1T polymorphs (Figure 15), with trigonal prismatic and octahedral coordination, respectively.

Density Functional Theory (DFT) for HER Catalysts
Density functional theory (DFT) is an essential tool for understanding the mechanisms and active sites of novel catalysts as it enables evaluation of the thermodynamics of the individual steps in HER.Modelling reaction barriers is however computationally demanding, and most studies as such, rather adopt a "ΔG approach".As HER involves both proton transfer and charge transfer, the activity of a catalyst is intrinsically linked to its crystal and electronic structures.In that respect, the hydrogen bonding strength/adsorption energy (ΔGH) has been widely used as descriptor of catalyst activity.Following the Sabatier principle, too strong or weak interactions with the catalyst surface tends to lower the overall catalyst activity yielding the typical volcano type behavior (Figure 14).MoS2 and similar layered transition metal dichalcogenides (TMD) crystallize in two structures, the 2H and 1T polymorphs (Figure 15), with trigonal prismatic and octahedral coordination, respectively.The thermodynamically stable 2H polymorph of single layer MoS2 is semiconducting with a band gap of 1.74 eV [177], and its (0001) basal plane exhibits negligible catalytic activity towards HER due to a ΔGH of ~2 eV [176].On the other hand, Hinnemann et al. [91] showed that the (-1010) Mo edge sites of single trilayer MoS2 can be highly active towards HER, and that they resemble the active sites of the hydrogen-evolving enzymes nitrogenase and hydrogenase [42,91].The Mo edge exhibits a calculated ΔGH of merely 0.08 eV (Figure 16), compared to 0.18 eV of the (10-10) S edge [178], and is as such, close to thermoneutral (for low H coverages).The increased activity of the edge sites is The thermodynamically stable 2H polymorph of single layer MoS 2 is semiconducting with a band gap of 1.74 eV [177], and its (0001) basal plane exhibits negligible catalytic activity towards HER due to a ∆G H of ~2 eV [176].On the other hand, Hinnemann et al. [91] showed that the (1010) Mo edge sites of single trilayer MoS 2 can be highly active towards HER, and that they resemble the active sites of the hydrogen-evolving enzymes nitrogenase and hydrogenase [42,91].The Mo edge exhibits a calculated ∆G H of merely 0.08 eV (Figure 16), compared to 0.18 eV of the (10-10) S edge [178], and is as such, close to thermoneutral (for low H coverages).The increased activity of the edge sites is attributed to in-gap surface states near the Fermi level, implying that 2D MoS 2 with a high edge site concentration can be activated towards HER [179].Significant computational studies have been devoted to exploring strategies to increase the density of activity sites in MoS 2 , and to optimize ∆G H through electronic structure manipulation.Bonde et al. [178] for instance, showed that Co-promotion decreases the ∆G H of the S edge to 0.07 eV, but not of the Mo edge, and as such leads to increased number of active sites.Tsai et al. [180] showed that various supports can also be used to tailor the hydrogen bonding to MoS 2 ; for Mo edges.Increasing the catalyst adhesion to the support was found to weaken the hydrogen bonding, and is attributed to downward shifts of the S p-states, which in turn lead to filling of H 1s antibonding states.Efforts have also been made to understand how the basal plane of MoS 2 can be activated towards HER through defect chemical, structural and strain engineering [181][182][183][184]. Li et al. [181] showed that ∆G H of basal plane MoS 2 decreases with increasing S vacancy concentration (Figure 16), and that vacancy formation induces in-gap defect states stemming from undercoordinated Mo (Figure 16b), which allows for favourable hydrogen binding.Straining the vacancies was furthermore shown to decrease the ∆G H (Figure 16c) even further.Ouyang et al. [184] showed that other native point defects such as V MoS3 and MoS 2 and extended defects, such as grain boundaries affect hydrogen bonding and as such the HER performance.In addition, Deng et al. [185] showed that single atom transition metal substitution creates in-gap states that lower the ∆G H , with Pt-MoS 2 yielding a close to thermoneutral binding energy.
While the basal plane of 2H-MoS 2 is semiconducting [177], its metastable 1T phase is metallic [186] and even its basal plane is highly active towards HER.The metallicity and high HER activity stems from the partially filled Mo 4 d and S states at the Fermi level, leading to favorable ∆G H [187]. DFT calculations reveal that ∆G H is highly coverage-dependent due to H induced surface reconstructions, reaching values between −0.28 and 0.13 eV for 12.5 to 25 % coverage [187].The phase stability of the 1T phase, and its band gap and as such HER activity, can be tuned by surface functionalization, by e.g., −CH 3 , CH 3 , OCH 3 , and NH 2 , which all were shown to bind more strongly to the 1T surface compared to the 2H basal plane [188].
Catalysts 2018, 8, x FOR PEER REVIEW 21 of 43 attributed to in-gap surface states near the Fermi level, implying that 2D MoS2 with a high edge site concentration can be activated towards HER [179].Significant computational studies have been devoted to exploring strategies to increase the density of activity sites in MoS2, and to optimize ΔGH through electronic structure manipulation.Bonde et al. [178] for instance, showed that Co-promotion decreases the ΔGH of the S edge to 0.07 eV, but not of the Mo edge, and as such leads to increased number of active sites.Tsai et al. [180] showed that various supports can also be used to tailor the hydrogen bonding to MoS2; for Mo edges.Increasing the catalyst adhesion to the support was found to weaken the hydrogen bonding, and is attributed to downward shifts of the S p-states, which in turn lead to filling of H 1s antibonding states.Efforts have also been made to understand how the basal plane of MoS2 can be activated towards HER through defect chemical, structural and strain engineering [181][182][183][184]. Li et al. [181] showed that ΔGH of basal plane MoS2 decreases with increasing S vacancy concentration (Figure 16), and that vacancy formation induces in-gap defect states stemming from undercoordinated Mo (Figure 16b), which allows for favourable hydrogen binding.
Straining the vacancies was furthermore shown to decrease the ΔGH (Figure 16c) even further.
Ouyang et al. [188] showed that other native point defects such as VMoS3 and MoS2 and extended defects, such as grain boundaries affect hydrogen bonding and as such the HER performance.In addition, Deng et al. [185] showed that single atom transition metal substitution creates in-gap states that lower the ΔGH, with Pt-MoS2 yielding a close to thermoneutral binding energy.
While the basal plane of 2H-MoS2 is semiconducting [177], its metastable 1T phase is metallic [186] and even its basal plane is highly active towards HER.The metallicity and high HER activity stems from the partially filled Mo 4 d and S states at the Fermi level, leading to favorable ΔGH [187].DFT calculations reveal that ΔGH is highly coverage-dependent due to H induced surface reconstructions, reaching values between −0.28 and 0.13 eV for 12.5 to 25 % coverage [187].The phase stability of the 1T phase, and its band gap and as such HER activity, can be tuned by surface functionalization, by e.g., −CH3, CH3, OCH3, and NH2, which all were shown to bind more strongly to the 1T surface compared to the 2H basal plane [188].Realizing the importance of crystal and electronic structure with respect to HER activity of MoS2, a range of other layered TMD have attracted attention both experimentally and computationally.Tsai et al. [189] showed also that for MoSe2 and WeSe2, the Mo and Se edge sites are more active than the Realizing the importance of crystal and electronic structure with respect to HER activity of MoS 2 , a range of other layered TMD have attracted attention both experimentally and computationally.Tsai et al. [189] showed also that for MoSe 2 and WeSe 2 , the Mo and Se edge sites are more active than the basal planes, and that the selenides generally exhibit weaker H binding than their sulphur counterparts.Tsai et al. [176] furthermore explored the electronic structure, ∆G H and the energy of HX adsorption, ∆G HX (i.e., descriptor for stability) for the basal planes of a range of 2D MX 2 (M = Ti, V, Nb, Ta, Mo, W, Pd, and X = S or Se) TMDs.The 2D TMDs vary from semiconducting to metallic (Figure 17), with group 7 TMDs (Mo and W) changing from semiconducting to metallic from the 2H to the 1T phase.The metallic TMD were in general found to exhibit stronger H bonding (lower ∆G H ) than the semiconducting phases (Figure 17).The semiconducting TMDs span a wider range of ∆G H than the metallic phases, reflecting the importance of the electronic structure with respect to the HER activity (Figure 17).They found an inverse correlation between ∆G H and ∆G HX for both semiconducting and metallic phases, reflecting the general understanding of the relationship between HER activity and.Furthermore, the metallic TMDs were in general found to exhibit stronger H bonding (lower ∆G H ) than the semiconducting phases.
Catalysts 2018, 8, x FOR PEER REVIEW 22 of 43 basal planes, and that the selenides generally exhibit weaker H binding than their sulphur counterparts.Tsai et al.
[176] furthermore explored the electronic structure, ΔGH and the energy of HX adsorption, ΔGHX (i.e., descriptor for stability) for the basal planes of a range of 2D MX2 (M = Ti, V, Nb, Ta, Mo, W, Pd, and X = S or Se) TMDs.The 2D TMDs vary from semiconducting to metallic (Figure 17), with group 7 TMDs (Mo and W) changing from semiconducting to metallic from the 2H to the 1T phase.The metallic TMD were in general found to exhibit stronger H bonding (lower ΔGH) than the semiconducting phases (Figure 17).The semiconducting TMDs span a wider range of ΔGH than the metallic phases, reflecting the importance of the electronic structure with respect to the HER activity (Figure 17).They found an inverse correlation between ΔGH and ΔGHX for both semiconducting and metallic phases, reflecting the general understanding of the relationship between HER activity and.Furthermore, the metallic TMDs were in general found to exhibit stronger H bonding (lower ΔGH) than the semiconducting phases.Of the HER active transition metal phosphides, those of especially Ni and Mo have been the subject of extensive computational investigations.Bulk Ni2P is metallic with a crystal structure consisting of alternating Ni3P and Ni3P2 planes along the (0001) axis.The HER activity of Ni2P (0001) surfaces were originally predicted computationally by Liu and Rodriguez [124] showing that the P Of the HER active transition metal phosphides, those of especially Ni and Mo have been the subject of extensive computational investigations.Bulk Ni 2 P is metallic with a crystal structure consisting of alternating Ni 3 P and Ni 3 P 2 planes along the (0001) axis.The HER activity of Ni 2 P (0001) surfaces were originally predicted computationally by Liu and Rodriguez [124] showing that the P sites on the phosphide surface play an important role in producing a weak-ligand effect involving Ni → P charge transfer, resulting in suitable ∆G H and high activity for the dissociation of H 2 .The bare Ni 3 P terminated surface exhibits a strongly binding Ni 3 hollow site with ∆G H of ~−0.5 eV for the first H, and several sites of lower H binding strength [190,191].DFT calculations show that the surfaces prefer a P-covered reconstruction of the Ni 3 P termination in which a P ad-atom binds on-top of the strongly H binding Ni 3 hollow site [190] and that this P ad-atom reduces the bindings strength of the site, and can bind up to 3 H atoms [192,193].Hakala and Laasonen [191] showed that the H adsorption properties can be modified through Al substitutions, leading to ∆G H close to 0 eV.
In a joint experimental-computational effort, Xiao et al. [194] studied hydrogen binding at the Mo, Mo 3 P and MoP surfaces showing that 001-Mo surface binds H strongly with a ∆G H ranging from −0.54 to −0.46 eV for 1  4 to 3 4 monolayers.The Mo and P terminated (001) MoP surfaces was found to exhibit values of −0.63 to −0.59 and −0.36 to 0.34, respectively, indicating that the P terminated surface can adsorb H at low coverages and desorb at high coverages, reflecting the importance of P also in these catalysts.

Earth-Abundant Anode Materials
As mentioned previously, the only stable and well-established catalysts for the OER in acidic media are noble metal oxides such as IrO x and RuO x [195].A recent study (2016) on benchmarking of water oxidation catalysts (WOC) revealed that there are no EACs that can reach the target metric of short-term acid stability, which is defined as operation at 10 mA/cm 2 for 2 h [196].We also expected that there are no PEM WE reports based on EACs anodes for the OER side, but this is also not the case.Herein, we report on recent advances and current trends on EACs for the OER that show promising results in terms of performance and stability in acidic media, which exceeded the short-term target of 2 h in just two years.The presented materials and their performance are summarized in Table 4.
Manganese oxide (MnO x ) was reported to be functional under acidic conditions and before activation exhibited a Tafel slope of approx.650 mV/decade, but after potential cycling and activation of the MnO x film the slope was improved to approx.90 mV/decade [197].The authors reported a galvanostatic stability of 8 h in 0.5 M H 2 SO 4 at a current density of 0.1 mA/cm 2 and overpotential of 540 mV.The same group introduced Mn in CoO x with the former acting as a stabilizing structural element and the CoMnO x showed a Tafel slope of 70-80 mV/decade and a stability of more than 12 h without any dissolution [198].The overpotential for a galvanostatic operation at 0.1 mA/cm 2 , which is 2 orders of magnitude lower than the target values though, was approx.450 mV.In another work, MnO 2 was stabilized by introduction of TiO 2 in the undercoordinated surface sites of MnO 2 .Frydendal et al. applied a 5 nm layer of Ti-modified MnO 2 on a 35 nm think layer of pure MnO 2 [199].The composite material exhibited a Tafel slope of 170 mV/decade and a moderate overpotential of approx.490 mV at 1 mA/cm 2 .The Mn dissolution in 0.05 M H 2 SO 4 was suppressed by roughly 50% after the TiO 2 modification.The authors came up with this strategy after an initial DFT study, which indicated that guest oxides such as GeO 2 and TiO 2 should improve the stability of MnO 2 .The reason is that both GeO 2 and TiO 2 have lower surface formation energies than MnO 2 and are more favorable for termination at the undercoordinated sites on MnO 2 .Another Mn-containing system is reported by Patel et al., and is based on nanostructured Cu 1.5 Mn 1.5 O 4 :x wt.% F (x = 0, 5, 10, 15) [200].The Cu 1.5 Mn 1.5 O 4 :10F electrocatalyst in 0.5 M H 2 SO 4 at 40 • C exhibited an onset potential at 1.43 V vs. RHE for the OER and reached 9.15 mA/cm 2 at 1.55 V vs. RHE.Interestingly, the in-house made IrO 2 showed the same onset overpotential and 7.74 mA/cm 2 at 1.55 V.The reported Tafel slope for the EAC is 60 mV/decade and it should be noted that the current-voltage curves were iR corrected.In a report by Anantharaj et al. it is suggested that the method used to calculate the iR drop compensation should be reported, along with the uncompensated i-V curves [55].The material showed also very good stability for almost 24 h of operation at constant current density of 16 mA/cm 2 .This material is also suitable for the oxygen reduction reaction (ORR), where it showed again similar activity to IrO 2 .The authors did not apply the Cu 1.5 Mn 1.5 O 4 :10F as a anode in full PEM WE cell, but they did so for the cathode in a PEM fuel cell (PEM FC) mode.The results are promising and the performance is the same as with IrO 2 and quite close to the operation in a 3-electrode mode.It should be noted though that the loading of the EAC was 6.7 times higher than for IrO 2 .
Another Mn-based electrocatalyst was a Mo-and Co-modified electrolytic manganese dioxide (MEMD) developed by Delgado et al. [201].According to the authors, Mo was incorporated in EMD for improvement of its mechanical stability, while the role of Co was to reduce the water content in EMD and increase the electrical conductivity.The catalyst was synthesized by a simple electrodeposition method and the Mo and Co elements were electrodeposited simultaneously with the EMD.The optimized catalyst had Mo and Co contents of 12.77 wt.% and 0.33 wt.%, respectively, and exhibited an overpotential of 305 mV at 100 mA/cm 2 in 2 M H 2 SO 4 .The authors found that this catalyst outperformed a commercial IrO 2 -based dimensionally stable anode (DSA ® ) electrode, which under the same conditions had an overpotential of 341 mV.The improved stable performance is explained by the lower charge transfer resistance induced by the Co incorporation, as well as to an increase of the roughness factor.The stability of the catalyst was assessed at a lower overpotential, i.e., at 280 mV, and the degradation of the material was small after approx.3500 s of operation.In principle, the authors should have reported the stability at 100 mA/cm 2 in order to provide more convincing results.
Moreno-Hernandez et al. developed a quaternary oxide, Ni 0.5 Mn 0.5 Sb 1.7 O y , which exhibited an initial OER overpotential of approx.675 mV vs. RHE in order to reach 10 mA/cm 2 in 1.0 M H 2 SO 4 [202].The overpotential stabilized at approx.735 mV and the electrocatalyst performed for 168 h of continuous operation (Figure 18).The authors reported a full cell application in a two-compartment electrolysis cell with Nafion as the separating membrane, but they did not use the catalyst in a full PEM WE cell.The stability of the Ni 0.5 Mn 0.5 Sb 1.7 O y is comparable to the noble metal oxides and is related to the fact that Ni, Mn and Sb oxides are stable in acidic conditions at OER potentials according to Pourbaix diagrams [203,204].
Catalysts 2018, 8, x FOR PEER REVIEW 24 of 43 same as with IrO2 and quite close to the operation in a 3-electrode mode.It should be noted though that the loading of the EAC was 6.7 times higher than for IrO2.Another Mn-based electrocatalyst was a Mo-and Co-modified electrolytic manganese dioxide (MEMD) developed by Delgado et al. [201].According to the authors, Mo was incorporated in EMD for improvement of its mechanical stability, while the role of Co was to reduce the water content in EMD and increase the electrical conductivity.The catalyst was synthesized by a simple electrodeposition method and the Mo and Co elements were electrodeposited simultaneously with the EMD.The optimized catalyst had Mo and Co contents of 12.77 wt.% and 0.33 wt.%, respectively, and exhibited an overpotential of 305 mV at 100 mA/cm 2 in 2 M H2SO4.The authors found that this catalyst outperformed a commercial IrO2-based dimensionally stable anode (DSA ® ) electrode, which under the same conditions had an overpotential of 341 mV.The improved stable performance is explained by the lower charge transfer resistance induced by the Co incorporation, as well as to an increase of the roughness factor.The stability of the catalyst was assessed at a lower overpotential, i.e., at 280 mV, and the degradation of the material was small after approx.3500 s of operation.In principle, the authors should have reported the stability at 100 mA/cm 2 in order to provide more convincing results.
Moreno-Hernandez et al. developed a quaternary oxide, Ni0.5Mn0.5Sb1.7Oy,which exhibited an initial OER overpotential of approx.675 mV vs. RHE in order to reach 10 mA/cm 2 in 1.0 M H2SO4 [202].The overpotential stabilized at approx.735 mV and the electrocatalyst performed for 168 h of continuous operation (Figure 18).The authors reported a full cell application in a two-compartment electrolysis cell with Nafion as the separating membrane, but they did not use the catalyst in a full PEM WE cell.The stability of the Ni0.5Mn0.5Sb1.7Oy is comparable to the noble metal oxides and is related to the fact that Ni, Mn and Sb oxides are stable in acidic conditions at OER potentials according to Pourbaix diagrams [203,204].Another important element in the aqueous electrochemistry is cobalt (Co).Co oxide-based catalysts have shown excellent performance in alkaline and near neutral pH solution [205,206].Under strong acidic conditions, however, they show fast dissolution, sluggish kinetics and high overpotentials [196,207,208].Mondaschein et al. developed a highly crystalline Co3O4 nanostructured film, which was deposited on FTO by electron-beam evaporation followed by annealing at 400 °C [209].The overpotential for 10 mA/cm 2 was 570 mV vs. RHE in 0.5 M H2SO4, and the catalyst maintained an OER with near-quantitative Faradaic yield for over 12 h.Unfortunately, the dissolution rate of Co at this high current density was considerable and further studies are needed for corrosion protection of such structures.To this end, Yan et al. have recently reported the synthesis of mesoporous Ag-doped Co3O4 nanowires, which showed improved stability over 10 h operation at 1.6 V vs. RHE in 0.5 M H2SO4, as Ag is known to be stable in acidic media [210].The Ag-doped Co3O4 nanowires were synthesized by electrodeposition-hydrothermal process, which was followed by Another important element in the aqueous electrochemistry is cobalt (Co).Co oxide-based catalysts have shown excellent performance in alkaline and near neutral pH solution [205,206].Under strong acidic conditions, however, they show fast dissolution, sluggish kinetics and high overpotentials [196,207,208].Mondaschein et al. developed a highly crystalline Co 3 O 4 nanostructured film, which was deposited on FTO by electron-beam evaporation followed by annealing at 400 • C [209].The overpotential for 10 mA/cm 2 was 570 mV vs. RHE in 0.5 M H 2 SO 4 , and the catalyst maintained an OER with near-quantitative Faradaic yield for over 12 h.Unfortunately, the dissolution rate of Co at this high current density was considerable and further studies are needed for corrosion protection of such structures.To this end, Yan et al. have recently reported the synthesis of mesoporous Ag-doped Co 3 O 4 nanowires, which showed improved stability over 10 h operation at 1.6 V vs. RHE in 0.5 M H 2 SO 4 , as Ag is known to be stable in acidic media [210].The Ag-doped Co 3 O 4 nanowires were synthesized by electrodeposition-hydrothermal process, which was followed by calcination at 400 • C. The nanostructured catalysts showed a Tafel slope of 219 mV/decade and an overpotential of approx.680 mV at current density of 10 mA/cm 2 .The authors do not provide any dissolution products analysis or any post-operation analysis of the material, as well as no comparison with IrO x .
Another Co-based catalyst in the form of cobalt-iron Prussian blue-type analogues (PBAs) thin film was developed by Han et al. [211].In this work, the PB-type water oxidation catalyst promoted the OER over a wide pH range spanning from pH = 1 to pH = 13.According to the authors, an especially interesting aspect of this universal catalyst was the very good behavior at pH = 2.The overpotential required at 10 mA/cm 2 was approx.0.83 V with a Tafel slope of 108 mV/dec.The oxygen evolution was quite close the theoretical one and the catalyst was stable for at least 2 h of operation at 10 mA/cm 2 .Co-containing polyoxometallates (Co-POMs) have shown promising catalytic properties for water splitting at near-neutral pH [212].To this end, Blasco-Ahicart et al. developed the Ba salt of Co-phosphotungstate polyanion (Ba[Co-POM], see Figure 19a) that outperformed IrO 2 at pH < 1, showing an overpotential of 189 mV vs. RHE at 1 mA/cm 2 (Figure 19b) with a faradaic efficiency of 99%.The Tafel slope was 66 mV/decade at the long-term stability was assessed at an overpotential of 250 mV vs. RHE.The initial current was more than 2 mA/cm 2 but decreased down to 0.35 mA/cm 2 after 24 h of operation.This degradation is assigned to charge localization that reduces the overall performance, which can be retrieved after charge delocalization at open-circuit potential.The authors could not assess the performance of the material at 10 mA/cm 2 as the carbon paste, which acted as a binder was not stable.
Catalysts 2018, 8, x FOR PEER REVIEW 25 of 43 calcination at 400 °C.The nanostructured catalysts showed a Tafel slope of 219 mV/decade and an overpotential of approx.680 mV at current density of 10 mA/cm 2 .The authors do not provide any dissolution products analysis or any post-operation analysis of the material, as well as no comparison with IrOx.Another Co-based catalyst in the form of cobalt-iron Prussian blue-type analogues (PBAs) thin film was developed by Han et al. [211].In this work, the PB-type water oxidation catalyst promoted the OER over a wide pH range spanning from pH = 1 to pH = 13.According to the authors, an especially interesting aspect of this universal catalyst was the very good behavior at pH = 2.The overpotential required at 10 mA/cm 2 was approx.0.83 V with a Tafel slope of 108 mV/dec.The oxygen evolution was quite close the theoretical one and the catalyst was stable for at least 2 h of operation at 10 mA/cm 2 .Co-containing polyoxometallates (Co-POMs) have shown promising catalytic properties for water splitting at near-neutral pH [212].To this end, Blasco-Ahicart et al. developed the Ba salt of Co-phosphotungstate polyanion (Ba[Co-POM], see Figure 19a) that outperformed IrO2 at pH < 1, showing an overpotential of 189 mV vs. RHE at 1 mA/cm 2 (Figure 19b) with a faradaic efficiency of 99%.The Tafel slope was 66 mV/decade at the long-term stability was assessed at an overpotential of 250 mV vs. RHE.The initial current was more than 2 mA/cm 2 but decreased down to 0.35 mA/cm 2 after 24 h of operation.This degradation is assigned to charge localization that reduces the overall performance, which can be retrieved after charge delocalization at open-circuit potential.The authors could not assess the performance of the material at 10 mA/cm 2 as the carbon paste, which acted as a binder was not stable.An interesting work conducted by Rodriguez-Garcia et al. combines the Co and Sb elements in an anode made of cobalt hexacyanoferrate supported on Sb-doped SnO2 [213].In this work the synergistic effect of the OER catalysts (CoHFe) and the support, antimonite tin oxide (ATO) is highlighted and the "winning" configuration is when 17 wt.% of CoHFe is deposited on ATO.The onset of the OER was approx.at 1.75 V vs. RHE as determined by RDE experiments.Interestingly, the authors assembled a full PEM WE cell and they have found the onset potential as from the RDE experiments.A current density of the order of 50-100 mA/cm 2 was reached at 2 V cell voltage (Figure 20a).The PEM WE cell with this earth-abundant anode showed a rather stable performance at 2 V during a stability testing of 22 h (Figure 20b) The authors studied the Sn and Sb leaching rates during PEM operation and they observed increases leaching rates for cell voltages above 2 V. To our knowledge this the first report on full PEM WE cells using EACs for the anode, showing very promising stability under WE operation.[213].In this work the synergistic effect of the OER catalysts (CoHFe) and the support, antimonite tin oxide (ATO) is highlighted and the "winning" configuration is when 17 wt.% of CoHFe is deposited on ATO.The onset of the OER was approx.at 1.75 V vs. RHE as determined by RDE experiments.Interestingly, the authors assembled a full PEM WE cell and they have found the onset potential as from the RDE experiments.A current density of the order of 50-100 mA/cm 2 was reached at 2 V cell voltage (Figure 20a).The PEM WE cell with this earth-abundant anode showed a rather stable performance at 2 V during a stability testing of In their report, Mondschein et al. found that arc-melted Ni 2 Ta rods combine the OER activity of Ni and the corrosion resistance of Ta and the intermetallic compound needed 980 mV to reach 10 mA/cm 2 (Figure 23b), a behavior assigned to the low electrochemically active surface area (EASA).The authors prepared a polycrystalline Ni-Ta electrode in order to increase the EASA and indeed, the overpotential at 10 mA/cm 2 was improved to 570 mV.The polycrystalline electrode showed improved corrosion resistance compared to a Ni pellet electrode prepared in a similar way, as the Ni content in the electrolyte after 36 h operation was below the detection limit of ICP-MS, while for the Ni pellet was 350.5 ppm.A superaerophobic bifunctional N-doped tungsten carbide nanoarrays catalyst was synthesized on carbon paper with a combination of hydrothermal and CVD methods by Han et al. [217].The OER onset is at an overpotential of approx.120 mV vs. RHE, while a high current density of 60 mA/cm 2 was reached at approx.470 mV overpotential.This catalyst outperformed IrO2 in 0.5 M H2SO4 under 3-electrode configuration as well as in a two-electrode water splitting cell, where both the anode and the cathode were the N-WC nanoarrays (Figure 22c).Unfortunately, the stability of the material is limited and after 1 h of operation at 10 mA/cm 2 the overpotential increased from 120 mV to 320 mV vs. RHE, but the faradaic OER efficiency is not reported.Mondschein et al. reported the intermetallic Ni2Ta for the OER in 0.5 M H2SO4 [218].Intermetallic alloys are metallic conductors and Ni2Ta has been used as a corrosion resistance coating [219,220].In their report, Mondschein et al. found that arc-melted Ni2Ta rods combine the OER activity of Ni and the corrosion resistance of Ta and the intermetallic compound needed 980 mV to reach 10 mA/cm 2 (Figure 23b), a behavior assigned to the low electrochemically active surface area (EASA).The authors prepared a polycrystalline Ni-Ta electrode in order to increase the EASA and indeed, the overpotential at 10 mA/cm 2 was improved to 570 mV.The polycrystalline electrode showed improved corrosion resistance compared to a Ni pellet electrode prepared in a similar way, as the Ni content in the electrolyte after 36 h operation was below the detection limit of ICP-MS, while for the Ni pellet was 350.5 ppm.A superaerophobic bifunctional N-doped tungsten carbide nanoarrays catalyst was synthesized on carbon paper with a combination of hydrothermal and CVD methods by Han et al. [217].The OER onset is at an overpotential of approx.120 mV vs. RHE, while a high current density of 60 mA/cm 2 was reached at approx.470 mV overpotential.This catalyst outperformed IrO2 in 0.5 M H2SO4 under 3-electrode configuration as well as in a two-electrode water splitting cell, where both the anode and the cathode were the N-WC nanoarrays (Figure 22c).Unfortunately, the stability of the material is limited and after 1 h of operation at 10 mA/cm 2 the overpotential increased from 120 mV to 320 mV vs. RHE, but the faradaic OER efficiency is not reported.Mondschein et al. reported the intermetallic Ni2Ta for the OER in 0.5 M H2SO4 [218].Intermetallic alloys are metallic conductors and Ni2Ta has been used as a corrosion resistance coating [219,220].In their report, Mondschein et al. found that arc-melted Ni2Ta rods combine the OER activity of Ni and the corrosion resistance of Ta and the intermetallic compound needed 980 mV to reach 10 mA/cm 2 (Figure 23b), a behavior assigned to the low electrochemically active surface area (EASA).The authors prepared a polycrystalline Ni-Ta electrode in order to increase the EASA and indeed, the overpotential at 10 mA/cm 2 was improved to 570 mV.The polycrystalline electrode showed improved corrosion resistance compared to a Ni pellet electrode prepared in a similar way, as the Ni content in the electrolyte after 36 h operation was below the detection limit of ICP-MS, while for the Ni pellet was 350.5 ppm.Lastly, Schäfer et al. developed a steel-based catalyst by anodization of Ni42 steel in LiOH [221].The unmodified Ni42 steel proved to be unstable in acidic media and OER potentials, whereas the LiOH-anodized Ni42 steel was quite stable.The main reason is the oxide formation in the latter, which protects the material from dissolution.This finding was explained by the higher Tafel slope of the treated sample (150 mV/decade) compared to the unmodified steel electrode (88 mV/decade).This supports the concept of the "oxide route" regarding the origin of oxygen formation during electrolysis.A higher activity, i.e., lower Tafel, leads to higher dissolution rates.The optimized sample (Ni42Li205) had an overpotential of 445 mV at 10 mA/cm 2 at pH = 0.Moreover, the faradaic efficiency for the OER at 10 mA/cm 2 was 79%, and a weight loss of 20 µg/mm 2 after 14 h of operation at pH = 1 was observed.

Summary, Challenges, Perspectives and Future Directions
In this review article, we briefly introduced the energy problem humanity is facing due to the depletion of fossil fuels and the climate changes resulting from their excessive usage.The "hydrogen economy" will become part of our future energy solutions and hydrogen fuel produced by water electrolysis represents a viable, renewable and environmentally friendly option that can replace fossil fuels.We presented a brief technoeconomic analysis and from the learning curves it is estimated that PEM water electrolysis will break even with the cost of hydrogen from fossil fuels around 2030, under an optimistic scenario.Currently, the high cost of hydrogen from PEM WE is related to the polymer exchange membrane, the noble metal electrocatalysts and the high overpotentials for water splitting.With this in mind, we have documented the progress done so far in the discovery and development of EACs both for the OER and HER sides of a PEM electrolyzer.We have not attempted an extensive literature review of EACs, because only for 2017, there were 2043 reports on the development of electrocatalysts.In addition, there are several other reviews, which the reader can refer to in this article, on EACs available in the literature covering either the whole range of new EACs or more specific classes, such as sulfides, phosphides etc.Instead, we reported the state-of-the-art full PEM WE cells based on noble metal catalysts and more importantly, we aimed in documenting how many of the newly developed EACs are actually used in full PEM WE cells, replacing the noble metal-based catalysts.This is equally important during the development stages of any catalyst, in order to observe and record efficiencies, stability and limitations under operating conditions, facts that may differ from the idealized measurements in half cells and rotating disc electrodes (RDE).This point is also emphasized by Spöri et al. in their review for benchmarking OER catalysts [222].An adequate examination of OER catalysts should be performed both at low and high current densities and moreover, several other parameters should be carefully considered when reporting water oxidation catalysts [55].For example, iR compensated curves, especially in the high current density region can dramatically differ from the uncompensated ones, reporting overestimated and inaccurate overpotential values.To our surprise, we found only 16 reports on HER EACs employed in full PEM WE cells and only one report for the OER.Of course, the great challenge is to find stable EACs for the OER in acidic environment, as currently the only stable and efficient catalyst is IrO 2 .
On the other hand, we are among the first to compile the very first EACs with promising efficiencies and stability for the OER under acidic environment.The reader can find the very first 17 breakthrough papers, which we hope that will motivate more research in order to develop and improve the stability of transition metal elements, such as Ni, Co, Fe and Mn for operation under anodic current flow at strongly acidic conditions.Transition metal antimonates of rutile type, as the Ni 0.5 Mn 0.5 Sb 1.7 O y reported by Moreno-Hernandez et al. show very good stability, which is related to the fact that Mn, Ni and Sb oxides are stable in acid, according to their Pourbaix diagrams [202].The strategy to integrate unstable catalysts with inactive counterparts, i.e., mixed polymorphs, may lead to stable electrocatalysts.Kwong et al. presented a fine example [215].The authors combined maghemite and hematite and they achieved a stable operation for more than 24 h at 10 mA/cm 2 in 0.5 M H 2 SO 4 at an overpotential of 650 mV vs. RHE, while maghemite and hematite alone are unstable and not active, respectively.The faradaic efficiency for the OER was also close to 100%.Another strategy is to combine a stable oxide with an unstable one, as the Ti-stabilized MnO 2 shown by Frydendal et al. [199].In this work, a DFT work predicted that TiO 2 can be inserted for termination at the undercoordinated sites on MnO 2 and in fact, the stability of MnO 2 increased by more than 50%.Apart from TiO 2 , the authors suggested GeO 2 as well, as it also has a lower surface formation energy than MnO 2 .
Intermetallic alloys, such as Ni-Ta, have been used as corrosion protective coatings already from the 90's.Mondschein et al. reported the polycrystalline Ni 2 Ta alloy, which was stable for more than 66 h at a current density of 10 mA/cm 2 in 0.5 M H 2 SO 4 [218].The challenge with such alloys is to increase their surface area by nanostructuring.The same applies for steel-based anodes as presented by Schäfer et al. [221].The reported Ni42 steel anode modified by a surface oxide layer appears to be both efficient and stable under OER potentials in strong acids.Other steel-based materials should also be tested, especially those containing Fe and Ni, as both elements are active towards OER electrocatalysis.Although surface passivation will increase the charge transfer resistance of the material, on the other hand is a promising trade-off for stable operation.
On the other side, the HER, one can find an enormous amount of EACs both for acidic and basic conditions.We very selectively touched upon the current state-of-the-art and the most promising HER EACs, and our main conclusion is that many more applied systems must be reported.Sixteen works out of thousands make a small sample to draw concrete conclusions from.It is encouraging to see that the HER and OER EACs tested in PEM WE showed similar performances to that expected by measurements in half-cells.There are cases though where the results do not correlate well, as we observed for some Co-clathrochelates.Moreover, in the development of both OER and HER catalysts, it is of utmost importance to report on faradaic efficiencies, as well as post-screening analysis of the electrolyte and catalyst.As we have seen from Tables 3 and 4, it is common to neglect such an analysis.Faradaic efficiency will provide a more accurate and trustable picture of the stability and performance of the catalyst and in addition, will help to identify and distinguish between corrosive and parasitic currents and actual HER and OER currents.There is also a big mismatch on the stability testing times reported on both HER and OER EACs.We generally see that researchers are more confident for the HER catalysts, as the testing times are by far longer than that of OER catalysts.Currently, it is reasonable, as EACs for the OER are in a primitive state with their stability way more challenging.On the other hand, a consensus should be established regarding performance protocols.We encourage the reader to study two benchmarking reviews, which cover a wide range of applied systems for more in-depth information [196,222].Moreover, a recent perspective article revisits and discusses the validity of ten important parameters that are commonly employed in HER and OER [55].
We take some of the best PEM electrolyzers based on noble metals and EACs, and a valuable comparison is given in Table 5.It is very encouraging to see that EACs, especially for the HER, have already reached efficiencies very similar to those with noble metal catalysts.Apart from the importance of the transition metals, the non-metallic elements, such as P and S, are also key elements in the development of earth abundant catalysts.DFT works also highlight the noble metal-like activity of the TMD and transition metal phosphides, and in some instances, it is also comparable to the activity and turnover frequencies of enzymes, such as hydrogenases.Furthermore, computational works indicate that P and S, as well as their vacancies, create such an electronic environment that induces favorable binding energies for the adsorption desorption of the H atom.
There is a long way to go for the OER ones, especially concerning their stability, but nevertheless, these results highlight even more the need to employ and operate EACs in full cells.It is also interesting to notice that a PEM WE based on purely EACs can already be realized.It is difficult to say whether the cost of hydrogen from PEM WE breaks even with fossil fuels around 2033, but this review endorses this optimistic scenario.It also provides ways for materials' optimization and development, in order to move forward PEM electrolyzers made purely by EACs, bringing/implying a significant cost reduction to the produced hydrogen.

Figure 1 .
Figure 1.Learning curve for renewable PEM H2 production showing the projected levelized costs until 2050 per kg H2 in USD.Reprinted with permission from [24].Copyright 2018, The Royal Society of Chemistry.

Figure 1 .
Figure 1.Learning curve for renewable PEM H 2 production showing the projected levelized costs until 2050 per kg H 2 in USD.Reprinted with permission from [24].Copyright 2018, The Royal Society of Chemistry.

Figure 2 .
Figure 2. Histogram showing the number of scientific reports on OER and HER from 2005 to 2017.Reprinted with permission from [55].Copyright 2018, The Royal Society of Chemistry.

Figure 2 .
Figure 2. Histogram showing the number of scientific reports on OER and HER from 2005 to 2017.Reprinted with permission from [55].Copyright 2018, The Royal Society of Chemistry.

Figure 3 .
Figure 3. Basic, key components of a PEM WE.

Figure 3 .
Figure 3. Basic, key components of a PEM WE.

Figure 4 .
Figure 4. Figure showing the structures of MoS2.(a) Illustration of the layer packing in the three polytypes: 2H, 3R and 1T; (b) Top view of MoS2.Reprinted with permission from [89].Copyright 2014 American Chemical Society.

Figure 4 .
Figure 4. Figure showing the structures of MoS 2 .(a) Illustration of the layer packing in the three polytypes: 2H, 3R and 1T.(b) Top view of MoS 2 .Reprinted with permission from [89].Copyright 2014 American Chemical Society.

Figure 5 .
Figure 5. (a,b) SEM images of a representative Ni2P film on Ni; (c) Experimental powder XRD pattern of a Ni2P sample (black), with the simulated (sim.)patterns of Ni (green) and Ni2P (red) shown for comparison.The y-axis was truncated to highlight the Ni2P as the Ni signal would otherwise dominate; (d,e) EDS elemental maps of a cross-section of the sample showing the presence of both Ni (green) and P (red) in a 2:1 ratio; (f) Polarization data for the HER in 0.5 M H2SO4 and (g) 1 M KOH for a series of metal phosphide films, along with a Pt mesh electrode for comparison.Reprinted with permission from [113].Copyright 2017 The Royal Society of Chemistry.

Figure 5 .
Figure 5. (a,b) SEM images of a representative Ni 2 P film on Ni.(c) Experimental powder XRD pattern of a Ni2P sample (black), with the simulated (sim.)patterns of Ni (green) and Ni 2 P (red) shown for comparison.The y-axis was truncated to highlight the Ni 2 P as the Ni signal would otherwise dominate.(d,e) EDS elemental maps of a cross-section of the sample showing the presence of both Ni (green) and P (red) in a 2:1 ratio.(f) Polarization data for the HER in 0.5 M H 2 SO 4 and (g) 1 M KOH for a series of metal phosphide films, along with a Pt mesh electrode for comparison.Reprinted with permission from [113].Copyright 2017 The Royal Society of Chemistry.

Figure 6 .
Figure 6.(a) Synthetic schematic illustration and material characterization of the multishelled Ni 2 P. (b) Linear sweep voltammetry (LSV) polarization curves of bare carbon, nanostructured Ni 2 P, hierarchical Ni 2 P, multishelled Ni 2 P, and benchmark Pt/C in 1 M KOH at a scan rate of 5 mV s −1 .(c) Corresponding Tafel plots with linear fittings.Reprinted with permission from [120].Copyright 2017 American Chemical Society.
Catalysts 2018, 8, x FOR PEER REVIEW 12 of 43 mV/decade and 4.9 × 10 −4 A/cm 2 , respectively.Again, this Tafel slope behavior reflects the change in the rate-limiting step of the HER[122].

Figure 7 .
Figure 7. (a) High resolution TEM image of a representative Ni2P nanoparticle, highlighting the exposed Ni2P (001) facet and the 5.2 Å lattice fringes that correspond to the (010) planes; (b) Proposed structural model of the Ni2P nanoparticles; (c) Polarization data for three individual Ni2P electrodes in 0.5 M H2SO4, along with glassy carbon, Ti foil, and Pt in 0.5 M H2SO4, for comparison; (d) Corresponding Tafel plots for the Ni2P and Pt electrodes.Reprinted with permission from [125].Copyright 2013 American Chemical Society.

Figure 7 . 43 Figure 8 . 43 Figure 9 .
Figure 7. (a) High resolution TEM image of a representative Ni 2 P nanoparticle, highlighting the exposed Ni 2 P (001) facet and the 5.2 Å lattice fringes that correspond to the (010) planes.(b) Proposed structural model of the Ni 2 P nanoparticles.(c) Polarization data for three individual Ni 2 P electrodes in 0.5 M H 2 SO 4 , along with glassy carbon, Ti foil, and Pt in 0.5 M H 2 SO 4 , for comparison.(d) Corresponding Tafel plots for the Ni 2 P and Pt electrodes.Reprinted with permission from [125].Copyright 2013 American Chemical Society.Catalysts 2018, 8, x FOR PEER REVIEW 13 of 43

Figure 9 .
Figure 9. (a) TEM image of Ni 2 P/CNT.(b) Tafel plots of the Ni 2 P/CNT, Ni 12 P 5 /CNT, Ni/CNT and Pt/C.(c) LSV curves of the Ni 2 P/CNT, Ni 12 P 5 /CNT, Ni/CNT, Pt/C, CNT and bare GCE in 0.5 M H 2 SO 4 with a scan rate of 5 mV/s.(d) Calculated TOFs for the Ni 2 P/CNT, Ni 12 P 5 /CNT and Ni/CNT in 0.5 M H 2 SO 4 .Reprinted with permission from [136].Copyright 2017 The Royal Society of Chemistry.
Catalysts 2018, 8, x FOR PEER REVIEW 15 of 43 degradation for at least 5 days.All catalysts have been tested in a PEM electrolysis single cell, and pyrite FeS2 allows a current density of 2 A/cm 2 at a voltage of 2.3 V.
left shows how a clean glassy carbon electrode (GCE) (a) is improved by addition of [Co(dmg) 3 (BF) 2 ]BF 4 (a) and Co(dmgBF 2 ) 2 (b) in a 0.5 M H 2 SO 4 aqueous solution.The two Co clathrochelate molecules are shown in Figure 11 right.

Figure 11
left shows how a clean glassy carbon electrode (GCE) (a) is improved by addition of [Co(dmg)3(BF)2]BF4 (a) and Co(dmgBF2)2 (b) in a 0.5 M H2SO4 aqueous solution.The two Co clathrochelate molecules are shown in Figure 11 right.

Figure 14 .
Figure 14.Volcano plot of the exchange current density as a function of the DFT-calculated Gibbs free energy of adsorbed atomic hydrogen for nanoparticulate MoS2 and the pure metals.Reprinted with permission from [42].Copyright 2007 Science.

2 )Figure 14 .
Figure 14.Volcano plot of the exchange current density as a function of the DFT-calculated Gibbs free energy of adsorbed atomic hydrogen for nanoparticulate MoS 2 and the pure metals.Reprinted with permission from [42].Copyright 2007 Science.

Figure 14 .
Figure 14.Volcano plot of the exchange current density as a function of the DFT-calculated Gibbs free energy of adsorbed atomic hydrogen for nanoparticulate MoS2 and the pure metals.Reprinted with permission from [42].Copyright 2007 Science.

Figure 16 .
Figure 16.(a) Free energy vs. reaction coordinate for HER on basal plane MoS2 for various vacancy concentrations; (b,c) Corresponding band structure; and (d) effect of strain and vacancies on ΔGH.Reprinted with permission from [181].Copyright 2015 Nature Publishing Group.

Figure 16 .
Figure 16.(a) Free energy vs. reaction coordinate for HER on basal plane MoS 2 for various vacancy concentrations.(b,c) Corresponding band structure, and (d) effect of strain and vacancies on ∆G H . Reprinted with permission from [181].Copyright 2015 Nature Publishing Group.

Figure 17 .
Figure 17.(a) p-projected density of states on the S or Se atom for 2D TMDs in the 2H and 1T structures relative to the Fermi level, with blue indicating metallic basal planes, while grey ones are semiconducting; (b) Semiconducting single-layered TMD basal planes; (c) Metallic single-layered TMD basal planes.Reprinted with permission from [176].Copyright 2015, Elsevier.

Figure 17 .
Figure 17.(a) p-projected density of states on the S or Se atom for 2D TMDs in the 2H and 1T structures relative to the Fermi level, with blue indicating metallic basal planes, while grey ones are semiconducting, (b) Semiconducting single-layered TMD basal planes, (c) Metallic single-layered TMD basal planes.Reprinted with permission from [176].Copyright 2015, Elsevier.

Figure 18 .
Figure 18.(a) Stability of the Ni 0.5 Mn 0.5 Sb 1.7 O y electrodes at 10 mA/cm 2 in 1 M H 2 SO 4 .(b) Cyclic voltammetry at 10 mV/s in between the stability test.Reprinted with permission from [202].Copyright 2017 The Royal Society of Chemistry.

Figure 19 .
Figure 19.(a) Molecular structure of the Co-POM cluster; (b) Linear sweep voltammetry of different Co-POM electrocatalysts compared with different carbon paste/IrO2 blends in 1 M H2SO4.Reprinted with permission from [212].Copyright 2017 Nature Publishing Group, Macmillan Publishers Limited.

Figure 19 .
Figure 19.(a) Molecular structure of the Co-POM cluster.(b) Linear sweep voltammetry of different Co-POM electrocatalysts compared with different carbon paste/IrO 2 blends in 1 M H 2 SO 4 .Reprinted with permission from [212].Copyright 2017 Nature Publishing Group, Macmillan Publishers Limited.An interesting work conducted by Rodriguez-Garcia et al. combines the Co and Sb elements in an anode made of cobalt hexacyanoferrate supported on Sb-doped SnO 2[213].In this work the synergistic effect of the OER catalysts (CoHFe) and the support, antimonite tin oxide (ATO) is highlighted and the "winning" configuration is when 17 wt.% of CoHFe is deposited on ATO.The onset of the OER was approx.at 1.75 V vs. RHE as determined by RDE experiments.Interestingly, the authors assembled a full PEM WE cell and they have found the onset potential as from the RDE experiments.A current density of the order of 50-100 mA/cm 2 was reached at 2 V cell voltage (Figure20a).The PEM WE cell with this earth-abundant anode showed a rather stable performance at 2 V during a stability testing of

43 Figure 21 .
Figure 21.(a) Photos of the flexible porous membranes of MoSe2 NS/MoO2 NB/CNT-M and the individual components; (b) Stability in acidic media using as anode and cathode the MoSe2 NS/MoO2 NB/CNT-M electrode.Reprinted with permission from [216].Copyright 2018, The Royal Society of Chemistry.

Figure 22 .
Figure 22.(a) Synthesis route of the N-doped WC nanoarrays; (b) i-V curves of water splitting with the N-WC as anode and cathode electrodes compared with N-WC as the cathode and Ir/C as the anode; (c) Video snapshot of the water electrolysis with a 1.5 V commercial battery.Reprinted with permission from [217].Copyright 2018 Nature Publishing Group.

Figure 21 . 43 Figure 21 .
Figure 21.(a) Photos of the flexible porous membranes of MoSe 2 NS/MoO 2 NB/CNT-M and the individual components.(b) Stability in acidic media using as anode and cathode the MoSe 2 NS/MoO 2 NB/CNT-M electrode.Reprinted with permission from [216].Copyright 2018, The Royal Society of Chemistry.

Figure 22 .
Figure 22.(a) Synthesis route of the N-doped WC nanoarrays; (b) i-V curves of water splitting with the N-WC as anode and cathode electrodes compared with N-WC as the cathode and Ir/C as the anode; (c) Video snapshot of the water electrolysis with a 1.5 V commercial battery.Reprinted with permission from [217].Copyright 2018 Nature Publishing Group.

Figure 22 .Figure 23 .
Figure 22.(a) Synthesis route of the N-doped WC nanoarrays.(b) i-V curves of water splitting with the N-WC as anode and cathode electrodes compared with N-WC as the cathode and Ir/C as the anode.(c) Video snapshot of the water electrolysis with a 1.5 V commercial battery.Reprinted with permission from [217].Copyright 2018 Nature Publishing Group.Catalysts 2018, 8, x FOR PEER REVIEW 28 of 43

Table 1 .
Advantages and disadvantages of PEM WE over alkaline water electrolysis.

Table 3 .
Summary of full PEM WE cells with EACs as cathodes.
* Weight of whole complex.** Weight of catalyst.

Table 5 .
Comparison between full PEM WE cells based on purely noble metal catalysts and those withEACs in the cathode or anode.