Recent advances in catalyst materials for proton exchange membrane fuel cells

Research on fuel cell technology is constantly gaining importance, while global emission requirements are becoming more and more restrictive. For environmentally neutral proton exchange membrane fuel cells (PEMFCs) to become a competitive technology, sustainable infrastructures need to be established. One of the main showstoppers is the utilization of the rare and therefore costly precious metal Pt as the key element in the electrocatalysis of hydrogen and oxygen. A huge amount of research is done on immensely reducing or even replacing Pt for future PEMFC technology. In this research update, the progress on oxygen reduction reaction catalysts in acidic media over the past two years is reviewed


I. INTRODUCTION
Proton exchange membrane fuel cell (PEMFC) vehicles are emerging into the commercial market as the drive for zero-emission vehicles increases. 1However, the cost of the PEMFC is still high, with one of the major cost components being the noble metal catalyst.The rather sluggish oxygen reduction reaction (ORR) at the cathode requires catalysts with a higher surface area and optimized structure in order to minimize the use of platinum group metals (PGMs).
Today, Pt and its alloys are the most commonly used catalysts for PEMFCs at both the anode and the cathode.In 2017, the U.S. Department of Energy (DOE) set the technical target of reaching a total loading of platinum group metals below 0.125 mg/cm 2 by 2020.Their target activities at 0.9 V vs reversible hydrogen electrode (RHE) are a PGM mass activity (MA) of 0.44 A/mg PGM for PGM catalysts and a current density of 44 mA/cm 2 for PGM-free catalysts. 2hile the Pt loading can easily be reduced below 0.05 mg/cm 2 at the anode, 3 the stability and performance of the electrocatalyst at the cathode is critical.In the acidic and oxidizing conditions at the cathode, Pt nanoparticles (NPs) tend to agglomerate and grow, thus losing the surface area.Although some research papers have reached the 2020 goals, commercial fuel cells still operate with around 0.35 mg Pt /cm 2 when used in vehicles. 1here are several strategies for improving the cathode catalyst in terms of efficiency and durability (Fig. 1).The size of commercial Pt NPs has already been minimized to 3-6 nm.To further reduce the cost and/or improve the catalytic activity of Pt, research groups are aiming at the shape control of Pt NPs, alloying of Pt, core-shell NP structures with Pt-rich surfaces, and the synthesis of hollow NP structures.The durability can be improved by confining the PGM NPs in a carbon based structure, as well as by making use of different support materials.Recycling of membrane electrode assemblies (MEA) containing PtCo has also shown that the global impact of the catalyst can be reduced by extracting the metals from the aged MEA and synthesizing new catalysts. 4n the other hand, Pt-free catalysts such as single-atom catalysts (SAC) are the focus of investigations (Fig. 1).These are most commonly of M-N-C type, where M is a metal, most commonly Fe or Co. 3 Optimization of these catalysts is currently followed up by creating edge-rich and porous structures, making use of favorable structures such as graphene or carbon nanotubes (CNTs), and by increasing the catalyst's hydrophobicity.Additionally, transition metal dithiolenes (M-S-C structures) comprising Fe or Co have been recently proposed by density functional theory (DFT). 5A number of different compounds have likewise been determined by DFT as sufficiently stable two-dimensional materials for ORR, 6 although experimental demonstration is yet to be performed.
In the following, the most recently investigated electrocatalysts for ORR will be discussed for each strategy, being split into PGM-based catalysts on the one hand and PGM-free catalysts on the other hand.The focus is on their synthesis, performance, and durability.Due to the vast number of publications in the past two years, the focus lies on those works providing membrane electrode assembly (MEA) tests in fuel cells.Additionally, works on catalyst materials that have not been tested at this level but in the view of the authors present promising performance and applicability in the PEMFC have also been considered.The goal of this work is to give a brief overview of the latest developments and to point out the most promising approaches for future research.
While ORR studies in alkaline media are currently very popular, the focus of this work is on the research dedicated to ORR in acidic media in order to identify suitable materials and synthesis processes for the successful development of PEMFCs.The primary objective does not necessarily lie in outperforming commercial Pt/C, but in achieving high reliability and durability of the catalyst in a cost-efficient synthesis process.

II. PGM-BASED CATALYSTS
Four different approaches are mainly pursued on PGM-based catalysts: the alloying of Pt or other PGM, the optimization of the pure Pt catalyst, an optimization of the supporting substrate, and the protection of the catalyst in an organic structure.Naturally, these approaches are often combined to acquire superior results.

A. PGM alloys
PGM alloys allow for reducing the use of the PGM, and electronic properties improving the catalytic activity of the PGM can be FIG. 1.Current strategies for improving the efficiency and durability of PGMbased (left) and PGM-free (right) ORR catalysts for PEMFCs.scitation.org/journal/apmobtained.The alloying elements may also contribute with their own catalytic activity.
One of the most popular approaches is the alloying of Pt with transition metals such as Fe, Co, Ni, and Cu, which are usually employed in the form of NPs with a Pt-rich surface.The Pt-rich and electrochemically stable surface is usually formed by either a specific etching process or the dissolution of transition metal atoms during operation; the result is often referred to as a core-shell structure.][18][19][20] Gong et al. successfully synthesized Pt-Ni as dumbbell-shaped particles, which did not show increased electrochemically active surface area (ECSA) but were able to achieve both higher ORR MA and higher retention of activity after accelerated degradation testing (ADT) when compared to globular Pt-Ni NPs 21 (Fig. 2).
Apart from alloying, a few works concentrated on the phosphorization 52,53 or nitrogenation 54  Wang et al. investigated ternary alloys of Pt with the three transition metals Fe, Ni, and Co, finding that the Pt-Co-Fe alloy was the most favorable among the possible combinations in terms of both activity and durability. 43he non-platinum containing PGM alloys recently reported for acidic ORR are Pd-based 55,56 or Ir-based. 57Pd-Mo nanosheets investigated by Luo et al. showed extremely high ORR activity in alkaline media and also have superior performance in acidic media; however, the authors considered that the stability in acidic media was insufficient for practical applications. 55][66][67][68][69] Nan et al. investigated different Pd-M/Pt core-shell catalysts with Ni, Co, and Fe as alloying elements.The Pd-Fe core provided the best ORR activity, while all alloy cores outperformed the catalyst containing a pure Pd core. 68A ternary alloy of exclusively noble metals (Pt-Pd-Ir) was investigated by Deng et al. 70 For most reported alloy catalysts, an improvement due to alloying with additional elements is reported.Deng et al. found that binary Pt-Pd on carbon nanowires (CNWs) outperformed ternary Pt-Pd-Au, however, the Au phase was mostly segregated so that the effect of alloying could not be studied. 59

B. Optimization of pure Pt catalysts
The optimization of pure Pt focuses on shape-tuning, to optimize ECSA and to expose the most active crystal planes, and on the tuning and control of the particle size.In terms of shape, rhombic dodecahedral NPs show the highest activity in perchloric acid, while the cubic shape is the least favorable 71 (Fig. 4).
Dong et al. studied the mechanism of ORR on Pt(100), Pt (110), and Pt (111) facets and were able to monitor the intermediates in situ by Raman spectroscopy.On Pt (111), where ORR is improved, the reaction takes place via the formation of HO 2 radicals, whereas OH radicals are formed on Pt(100) as well as Pt (110). 72Chen et al. synthesized NPs in a multipod structure showing good performance; however, degradation was still an issue. 73heng et al. used an Se film on the carbon support in order to seed Pt NPs smaller than 2 nm, resulting not only in superior performance but also in much improved durability. 74Another approach was pursued by Chen et al., who synthesized supramolecular Ptcontaining structures with promising results for ORR. 75

C. Optimization of the supporting layer
The interface between the PGM catalyst and its supporting layer is crucial since poor adhesion can lead to catalyst detachment.In addition, the support must provide an electrical connection to the catalyst and may influence the density of states in the d-band of the catalyst.Carbon black (CB) is the most commonly used support for the PEMFC catalyst.9][90][91] Those can be used to either replace CB or act as an intermediate support between the catalyst and CB.
Doping of CB with N, as well as the use of Mn as a precursor results in more graphitic and corrosion-resistant carbon. 92,93s reported by Yang et al., Pt particles deposited on N-CB were smaller and better dispersed, resulting in an increase in ECSA and The most promising approach in order to limit degradation and increase the durability of PGM catalysts is their protection in a carbon-based matrix, where the PGM catalyst is not in direct contact with the polymer membrane. 96,97Zhao et al. successfully "confined" Pt 3 Co NPs in mesoporous carbon derived from the zeolitic imidazolate framework (ZIF-8).With the NPs trapped in the mesoporous structure, both detachment and agglomeration of NPs could be largely avoided, while the catalytic activity was not compromised. 98i et al. used an ionic liquid (IL) film to cover Pt/C NPs.In addition to the protective effect of those films, their electronic properties also improved ORR activity with respect to the bare Pt/C. 99improving the PEMFC performance at a low Pt loading of 0.07 mg/cm 2 ; however, the durability was not studied. 96hoi et al. synthesized Pt-Fe NPs in a block copolymer matrix and achieved the formation of a carbon shell around the NPs.This catalyst achieved extremely high Pt mass activity in a half cell of 9 A/mg, and in PEMFC, an extremely low Pt loading of 0.01 mg/cm 2 was used to achieve a performance comparable to commercial Pt/C. 30iao et al. investigated the ORR of the noble metals Ir and Ru in a PGM-N-C configuration as SACs, reporting extraordinarily high ORR MAs including promising results at the PEMFC level. 100,101iu et al. determined that Ir and Rh showed favorable performance when applied as SAC, in contrast to Pt and Pd. 102

E. Synthetic strategies
The most common approach for the synthesis of PGMbased catalysts is the solvothermal reduction method in autoclave.Meanwhile, alternative synthesis routes include pyrolysis, galvanic replacement, selective etching, and electrodeposition (Fig. 5).Moreover, the synthesis may comprise those multiple synthesis steps.In the solvothermal reduction method, metal precursors (usually acetyl acetates) are dissolved and mixed with a reducing agent and other additives.Carbon is often mixed in before the reduction process, which takes place in an autoclave at elevated temperatures.The product is then dried.For application in an MEA, the catalyst is usually dispersed in a solvent and sprayed onto the electrode.
Zhao et al. studied the size dependence of icosahedral Pt NPs synthesized using tetraethylene glycol as both the solvent and reductant and were able to synthesize the catalyst within 20 min.Interestingly, the highest kinetic current was reached for the largest NPs (14 nm), which also exhibited the best durability. 103ei et al. demonstrated the synthesis of Pt-Co, Pt-Fe, and Pt-Ni NPs through the same route, allowing a direct comparison between their ORR activities, finding that Pt-Co was the most active at the ORR between them. 104A similar observation was made by Wang et al. 105 Yin et al. reported a room temperature synthesis for mesoporous Pt-Co and Pt-Ni nanotubes (NTs). 106Here, the reduction of Pt-M (M = Co, Ni) takes place using ascorbic acid as a reducing agent and by galvanic replacement of the previously synthesized Te nanowires (NWs).In addition, a polymeric surfactant introduces mesoporosity in the material (Fig. 6).The same approach is used by the group to prepare mesoporous Pt-Te NTs. 37ang et al. proposed a silica-assisted pyrolysis method for the reduction of Fe-Pt NPs. 27After pyrolysis, the silica is removed and the Pt-Fe NPs are etched to obtain a hollow structure (Fig. 7).Alternative synthesis routes for Pt-Fe NPs are the  impregnation-reduction method 28 and pyrolysis from a single precursor. 29Higher process temperatures lead to the formation of an ordered structure, more favorable for ORR activity. 28,29hen et al. achieved the synthesis of ternary Pt-Pd-Cu alloy NPs in a hexapod shape (Fig. 8).The formation of this unique shape was triggered by a combination of etching and selective growth. 64esoporous structures are reported in many cases, and their catalytic activity is certainly promising due to their generally higher ECSA and MA. 17,37,65,67,69,70,106However, their applicability in a PEMFC is questionable, as its performance depends on the number of triple points, which do not necessarily increase due to mesoporosity of the catalyst.In these cases, high ORR activities at the rotating disk electrode (RDE) level need to be backed up by PEMFC testing to prove the effect of a mesoporous structure on the FC.Otherwise, porous structures can still lead to improved MAs due to a decrease in the amount of non-surface material and can be ultimately optimized by the synthesis of hollow structures, as long as the stability is not compromised. 12,27,69 more promising approach is the combination of the catalyst with a mesoporous support layer, where the catalyst is confined within the mesopores. 94,98,107n alternative for the synthesis of PGM NPs is through pulse electrodeposition onto the carbon support. 11,108Wang et al. successfully deposited Pt NPs within a desirable particle size of 3-10 nm after determining the optimum deposition parameters.The electrodeposited Pt/C showed higher half-wave potential E 1/2 , kinetic current density j k , and stability with substantially lower Pt loading in comparison to commercial Pt/C. 108ussain et al. produced an Nb-Ti oxide substrate by atomic layer deposition (ALD), followed by the deposition of Pt by magnetron sputtering, and high ORR activity at a very low Pt loading of 9 μg/cm 2 was reached. 86A sputter deposition approach was also followed by Ergul-Yilmaz et al. for the deposition of Pt. 109

F. Durability issues
Many PGM-based catalysts suffer from constant performance degradation. 12For pure Pt, this generally refers to the dissolution and redeposition of Pt, leading to particle growth and thus loss of the total catalyst area. 110For alloys, leaching of non-noble metals can cause compositional and structural changes.A detailed in situ examination of the degradation of octahedral Pt-Ni was performed by Beermann et al., identifying coalescence and particle motion as the main degradation mechanisms. 111Another considerable problem of pure Pt is its methanol-dependent performance.Mom et al. investigated the oxidation behavior of Pt under PEMFC conditions and found that oxidation of Pt was favored under wet conditions at oxidizing potentials, especially on Pt NPs.Furthermore, the authors showed that the use of a Br-containing membrane instead of a regular Nafion membrane could suppress oxide formation. 112utstanding durability has been reported by Tian et al. on bunched Pt-Ni nanocages, with only marginal decrease in MA after 50 000 cycles of ADT (Fig. 9).This performance has been related to the protective Pt skin structure of the catalyst. 12Wu et al. improved the durability of Pt 3 Ni NWs by Au doping, while the ORR activity was unaffected by the addition of Au. 113 Corroborated by DFT studies, the authors suggest that Au occupies surface defects of the Pt 3 Ni matrix and immobilizes Ni atoms.A similar observation was made by Shen et al. using In in order to limit the diffusion of Ni. 41 Although the initial performance was inferior to that of a Pt-Ni alloy, an improved durability of Pt-Ni-In was found, and the optimum In content is the result of a trade-off between activity and durability.
Feng et al. produced Pt-Te nanosheets, which were then submitted to electrochemical dissolution of Te.The so-obtained porous Pt nanosheets showed very high MA and durability even after 30 000 potential cycles. 114Similar observations were made by Kong et al. after dealloying Pt-Ni NWs 115 and referred to the higher defect density of dealloyed Pt, leading to improved ORR performance. 116n a different approach, a significant improvement was made by Guo et al. to Pt/C by applying a phosphorization treatment, leading to the formation of Pt 2 P. 52 Tu et al. found that the addition of W to Pt-Cu led to the outstanding stability of the catalyst after 30 000 potential cycles, even though the performance has been notably decreased after 5000 cycles. 42This example shows that sufficiently long durability tests are indispensable for the evaluation of an ORR catalyst and 10 000 potential cycles, as suggested by the DOE, should be the absolute minimum. 2 The fact that MA of the PGM-based catalysts is often reported with respect to Pt mass leads to the questionable assumption that ORR solely takes place on Pt atoms.Especially, if other PGMmetals such as Pd or Ir are involved, but also in the case of alloying of Pt with non-noble elements, their contribution to ORR activity is neglected and unrealistically high MA are reported.Although the Pt MA is therefore little meaningful from a scientific viewpoint, it can give a good estimation on the required amount of Pt and therefore the cost efficiency.From a purely electrochemical view, the ECSA-normalized specific activity (SA) should provide better comparability (Table I).A sound guidance for measuring and reporting the catalytic activity was produced by Wei et al. 117 Zalitis et al. targeted the issue that although Pt alloys often show higher activity around the half-wave potential, they often do not exhibit as high current densities at higher ORR overpotential, close to the working potential of the PEMFC. 118It is therefore important to document the MA at a higher potential such as 0.9 V vs RHE.
Ir-based catalysts 57 and Pt-Pd alloys 35 show very good CO tolerance in contrast to Pt. Bak et al. further found that carbon corrosion was suppressed when Ir was present. 34scitation.org/journal/apm

III. PGM-FREE CATALYSTS
Research on PGM-free catalysts for the ORR in acidic media includes ceramic catalysts such as CoS 2 , 122 NiN 123 and Fe 3 C; 124 however, the majority of the PGM-free catalysts are carbon-based.Carbon-based electrocatalysts profit from the huge abundance of carbon as well as from the highly advanced existing technologies able to tune carbon to the desired electronic properties.Carbon doped with elements such as nitrogen, 125,126 sulfur, phosphorous, and their combinations [127][128][129] has proven to work as ORR catalysts in both half-cell and PEMFC experiments.However, the current densities of these catalysts are not comparable to their PGM counterparts yet.Therefore, co-doping carbon with nitrogen and a transition metal to form SAC M-N-C structures where M=(Fe, Co, Cr, Ni, Mn, Ti, Zn) [130][131][132][133][134][135] is the focus of current PGM-free research.These SACs utilize abundant materials and thus have the potential to significantly reduce the cost of PEMFCs.8][139][140][141][142][143][144] To facilitate the optimization of SACs, studies to understand which are the active sites have been performed.Zhang et al. found pyrrole-type FeN 4 to be most active through both DFT calculations and from high open circuit voltage (OCV) and peak power density from tests in PEMFC.Chen et al. reached the same conclusion 145 (Fig. 10).Mun et al. investigated the effect of the carbon support on the activity of FeN 4 and found that the activity can be tuned by adding electron-withdrawing/donating groups. 146

A. Synthetic strategies
The synthesis of M-N-C catalysts is most commonly done through pyrolysis, where nitrogen containing the organic precursor and metal precursor are mixed and subsequently heated in an inert atmosphere. 138,147,1489][160][161][162][163] In principle, several different precursors may be combined (Fig. 11).Zhu et al. combined the use of a sacrificial polystyrene template and ZIF-8 to obtain a hierarchically porous structure. 164The first pyrolysis step is followed by etching to remove the support, surface oxides, and unstable metal clusters 138,147,148 (Fig. 12).An additional pyrolysis step ensures carbonization and yields a more uniform morphology. 147,148Several groups have added excess Fe to form NPs of iron carbides or nitrides within the Fe-N-C. 141,165,166ao et al. combined the pyrolysis and etching step by adding H 2 to atmosphere during pyrolysis of a MOF-based catalyst.Tunable hierarchical porosity was obtained. 131Li et al. added additional metal ions to the pyrolysis.The metal ions were oxidized during heat treatment and ensured increased porosity while being easily removed during subsequent acid leaching. 154A hierarchically open-porous structure has been obtained by the addition of graphitic carbon nitride to ZIF-8 and iron precursor prior to pyrolysis. 139,167,168imilarly, ammonium chloride or sodium chloride has been added to the pyrolysis to achieve an edge rich structure with more accessible active sites. 149,169Wang et al. also achieved an edge rich structure by using iron chloride as a precursor, and both experiments and DFT calculations showed higher catalytic activity of the Fe-Nx sites at edges compared to those at the center of the carbon matrix. 138Another strategy is to increase the density of active sites as done by Wan et al. by making concave structures. 142ang et al. produced iron oxide NPs in a mesoporous N-doped carbon, which after pyrolysis formed a core-shell structure with iron oxide encapsulated in graphitic carbon, and x-ray photoelectron spectroscopy analysis showed Fe-N active sites. 170FIG.12. Formation of Fe-N-C by pyrolysis using L-aspartic acid, dicyandiamide (DCD), and FeCl 3 as precursors. 138 As an alternative to the pyrolysis of MOF and sacrificial support, Wang and Berthon-Fabry synthesized Fe-N-C using the sol-gel method to produce an aerogel. 171ighest performance is, however, found for Fe-N-C catalysts using CNTs and/or graphene as the carbon precursor, with the advantage of not needing a sacrificial support. 172Li et al.
fabricated carbon nanospheres connected with CNTs to improve the electrical conductivity of the catalyst. 173Liu et al. also anchored Fe single atoms on CNTs 174 but further improved the performance by exchanging CNTs for graphene 175 (Fig. 13).The highest performance and durability of a PGM-free catalyst were achieved by Chen   Fe and N on RGO with minimal loss after 15 000 potential cycles. 146][178] Although Co-N-C so far has shown lower catalytic activity than Fe-N-C (Table II), Co ions are less critical when released into both the membrane and ionomer since they are not Fenton active, and hence, a better durability is expected. 161However, the Co-N-C needs advanced engineering to reach potentials comparable to Fe-N-C and Pt.Although half-wave potentials of up to 0.84 V vs RHE have been reached by a surfactant-covered Co-N-C structure, 176 they need to be further stabilized to achieve the superior durability expected.
Promising results have also been obtained by adding more than one metal to the M-N-C such as Fe-Ni, 164,179 Co-In, 180 Co-Ni, 181 and Co-Zn. 150,182,183Zang et al. reported increased activity, both experimentally and from DFT calculations, due to Co-Ni dual active sites, 182 similar to the results produced by Ye and coworkers using Fe-Fe dual sites. 184Fe-Fe was also studied by Zhang et al. by thermal migration of adjacent FeN 4 sites to form Fe 2 N 6 dual sites.These dual sites show high selectivity toward the four electron ORR due to the ability to adsorb two O atoms simultaneously. 185On the other hand, Zhu et al. found that Fe and Ni formed single atom sites. 164This shows the immense number of variations that can be found for SACs, while only a handful has yet been explored, and that further research is needed to find a durable catalyst with a high catalytic efficiency and low production cost.

B. Durability issues
He et al. observed an initial high rate of degradation before an apparent stabilization of the Co-N-C.This is similar to what is found in PGM based catalysts and is due to an initial loss of unstable catalytic sites and oxidation of carbon. 176umar et al. found that carbon corrosion is occurring also in Fe-N-C catalysts.The deactivation of the catalyst is increased by the H 2 O 2 byproduct from the ORR on the Fe-N-C, as there is significant increase in corrosion when cycled in O 2 compared to Ar. 144 Bae et al. studied the effect of H 2 O 2 with varying pH.With the decrease in pH in the presence of H 2 O 2 , the amount of oxygen in the sample after the test increased.This is in accordance with the fact that Fenton's reaction is pH dependent and results in reducing the turnover frequency of nearby active sites. 186oping graphene, CNTs, or CNWs with a transition metal and nitrogen has been proven by several groups to increase the durability, 146,154,178 with a further advantage of ensuring good conductivity in the catalyst layer 182 A major advantage of the SACs is that they are superior to Pt/C when it comes to methanol tolerance. 148,151This would lower demands on gas purity for the PEMFC, which further reduces the operating costs.

IV. PEM FUEL CELL TESTING
While the electrochemical activity of a catalyst can be tested in a three electrode setup, PEMFC testing is required to evaluate the  performance of the catalyst when the electrolyte is changed from liquid to solid.However, comparison of the catalyst performance in a PEMFC is more complex.In Tables III and IV, test conditions and reported performance are shown for PEMFC tests using H 2 /O 2 and H 2 /air, respectively.Many parameters of PEMFC testing affect the results, although they are reported in varying degrees.The ionomer content in the catalyst layer and coating technique on membrane affect the proton conduction from the catalyst layer.The type of support, choice of gas diffusion layer (GDL), and compression of the cell affect the electrical conductivity.Finally, the porosity in GDL and catalyst layer and the flow field design and size of the cell affect how homogeneously the reactant gases are distributed across the catalyst layer.
Half of the reported tests use air at the cathode, while the rest use O 2 .Several groups, testing both PGM and PGM-free catalysts, have tested their catalyst in both and found that the peak power density is almost twice as high when using O 2 compared to air. 22,130,169,172Using pure oxygen mitigates some of the mass transport limitations and is more comparable to using air.Contaminants in air are rarely specified, but it is known that CO and SO 2 both poison the catalyst. 187,188or PGM containing catalysts, the catalyst layer has a thickness of 10 μm and the catalyst loading varies between 0.06 and 0.4 mg/cm 2 .3][94][95] Daş et al. tested Pt-M alloys as the anode, cathode, and both and found that the highest performance gain was obtained when the bimetallic catalysts were applied as the anode. 189 highly desirable durability in PEMFC application was achieved by Gao et al., where the performance did not decrease even after 50 000 potential cycles of ADT for a Pt-Co/CNW catalyst (Fig. 14).9 For PGM-free catalysts, low volumetric activity leads to ten times thicker catalyst layers.Consequently, mass transport and resistance problems that are minimized or already overcome for noble metal catalyst needs to be solved.It should also be noted that although 4 mg/cm 2 is a common loading for SACs, some of the best performances come from catalysts with half the loading.142 In a recent review by Satjaritanun and Zenyuk, pooling of water at interfaces between the catalyst layer and the microporous layer or the membrane as a result of non-homogeneous fabrication was found as an obstacle for these catalysts.190 Liu et al. concluded that the hydrophobicity of the catalyst layer was the reason for water accumulation in a study where both the catalyst coated membrane and gas diffusion electrode assemblies were analyzed. Furthermore, it was concluded that although flooding of micropores is often cited as a major degradation mechanism and it can block O 2 transport to the active site, it is not the most important.In the review by Satjaritanun and Zenyuk, Fe-N 4 demetalization was instead pointed out as the major degradation mechanism.190 Electrochemical impedance spectroscopy of aged PEMFCs with Fe-N-C catalysts has confirmed that there is a loss of active sites, which further leads to Fe-ion poisoning of the ionomer and membrane.192,193 It should be noted that in alkaline environment, many of the PGM-free catalysts have higher performance than the Pt-based catalyst.Therefore, these catalysts would have an advantage if the PEM was changed to an anion exchange membrane (AEM), ensuring a less corrosive environment in the cell.Seeberger et al. demonstrated the feasibility of a bipolar membrane assembly for PGMfree FCs, combining the alkaline ORR with acidic hydrogen oxidation reaction (HOR) by joining a PEM with an AEM as a double membrane.194 Fe-N-C has also been pointed out to be an alternative to Pt/C in high temperature PEMFCs.In these cases, where the membrane is polybenzimidazol doped with phosphoric acid, Pt is poisoned by the phosphate ions, while Fe-N-C is immune to the poisoning effect.

V. RECYCLING OF PEMFC
Recycling of PEMFC components and particularly of PGM electrocatalysts is of utmost importance considering the current striving toward sustainability.Yet, this is an aspect that deserves larger attention from the scientific community in the field.
Lotrič et al. performed a life cycle assessment of the PEMFC stack and found that Pt is the component in the PEMFC with the largest environmental impact, and hence, recycling of the Pt can reduce the environmental footprint of the PEMFC. 195In general, the global recycling of PGM has the potential to over 95% recovery. 196For PGM used in PEMFC specifically, Sharma et al. have   shown that Pt can be dissolved from a gas diffusion electrode by electrochemical dissolution in HCl, with a recovery efficiency of more than 90% when producing a new catalyst from the recycled material. 197,198However, the gas diffusion electrode was never assembled or cycled before dissolution, and the recovery efficiency may therefore be lower considering Pt dissolution and migration into the membrane and the fact that Pt particles adhered to the membrane upon disassembly.gave Pt yield of 77% and 83%, respectively, while the ion exchange resin is advantageous with respect to synthesizing a new catalyst material from the product. 4

VI. SUMMARY
Pt remains as the most important component of reliable and efficient PEMFC ORR electrocatalysts-and will continue to do so for the time being.The cost of the fuel cell stacks can be efficiently reduced by lowering the amount of Pt used and by increasing its efficiency through alloying, shape-tuning, and creating NPs with Ptrich surfaces.
Among the most recent studies, the most promising way for maximizing the efficiency and durability lies in the combination of different approaches: the alloying of Pt with a transition metal, the synthesis of hollow NPs, and the fixation of NPs in an enclosing carbon-based structure, which can prevent the most prominent durability issues, namely, the growth of Pt NPs, and the leaching of Pt alloys.With further reduction of the amount of Pt used and recycling of the MEA after utilization, the impact of the Pt in the PEMFC can be reduced to levels comparable to that from PGMs used in the catalysts of internal combustion engines.
Another important durability issue occurring at high potential cycling, the corrosion of the carbon support, can be tackled by replacing or modifying the carbon support material.Many different approaches and materials have been studied to modify the support of PGM-based catalysts, showing desirable improvements.Considering the high amount of studies on the field in general, optimized support structures can soon be combined with optimized PGM catalysts in order to eliminate all durability issues and to provide reliable performance at low Pt loading.As there is a well-established industry for the production of carbon supported Pt NPs, these improvements can be implemented without needing major changes, and so, the way to commercialization is relatively short.
Regarding PGM-free catalysts, sufficient durability is the key issue.As a result of this being a new technology, current work focuses on performance optimization, while works focusing on their durability are rather scarce.For the SACs, solutions have to be found to effectively prevent metal dissolution in acidic environments and increase durability during potential cycling.Furthermore, optimization of the catalyst layer in the PEMFC, achieving a more homogeneous deposition, and upscaling of the fabrication are key to bringing these catalysts closer to broad commercialization.Otherwise, SACs are readily applicable in AEMFCs.

AUTHORS' CONTRIBUTIONS
L.M. and K.E.contributed equally to this work.

FIG. 4 .
FIG. 4. Shape dependency of ORR activity for pure Pt NPs in perchloric acid.

FIG. 6 .
FIG. 6. SEM micrographs of Pt-Co mesoporous NTs synthesized using Te NWs and F127 block copolymer as templates.The inset in (b) shows the cross section of a NT.Interestingly, the Pt-Co NTs outperformed Pt-Ni NTs synthesized through the same route. 106Reprinted with permission from Yin et al., ACS Sustainable Chem.Eng. 7, 7960 (2019).Copyright 2019 American Chemical Society.

FIG. 7 .
FIG. 7. Scheme for the formation of hollow Pt-Fe NPs from Pt and Fe 3 O 4 precursors by silica-assisted pyrolysis.SiO 2 prevents the agglomeration of NPs during the heat treatment at 900 ○ C at which formation of Pt-Fe NPs takes place.Subsequent etching in HF removes SiO 2 and dealloys the NPs to obtain an electrochemically stable composition of Pt-Fe.The Pt-Fe/C catalyst reached a MA of over 1 A/mg Pt and 91% retention of MA after ADT. 27Reprinted with permission from Yang et al., Chem.Eur.J. 26, 4090 (2020).Copyright 2019 Wiley-VCH Verlag GmbH & Co. KGaA, Weinheim.

FIG. 10 .FIG. 11 .
FIG. 10.X-ray photoelectron spectra of single-atom Fe anchored to nanographene (NG) and reduced graphene oxide [(RGO, (a) and (b)] and schematic of the Fe-N 4 , Fe-N 5 , Fe-N 3 , and Fe-C 4 active sites where Fe is indicated in gold, N is indicated in gray, and C is indicated in dark orange (c).The free-energy diagram of these sites also at the Pt(111) facet in acidic media from DFT calculations (d) and a volcano plot of the ORR activity by plotting the overpotential as a function of the binding energy of OH of these sites, Pt(111), nitrogen-doped graphene, and graphene (e). 146Reprinted with permission from Chen et al., Nano Energy 66, 104164 (2019).Copyright 2019 Elsevier.
Duclos et al. studied the recovery of PtCo from an MEA by two different hydrometallurgical routes.Both methods include incineration of fluorinated polymers, causing emissions, and the authors point out the need for recycling of the fluorinated polymers as well.The ion exchange resin and solvent separation methods

TABLE I .
Characteristics of ORR for different PGM-based electrocatalysts in 0.1 M HClO 4 .j k , MA, and SA are given at 0.9 V vs RHE.ADT refers to 10 000 potential cycles between 0.6 and 1.0 V unless stated otherwise.MA and ECSA usually refer to Pt mass only.All potentials refer to RHE.

TABLE I .
(Continued.) a Normalised by total PGM mass.bAfter5000 potential cycles of ADT.c After 10 000 seconds of chronoamperometry.d After 1000 potential cycles of ADT. e After 50000 potential cycles of ADT.f After 3000 potential cycles of ADT.g After 20000 potential cycles of ADT.h After 30000 potential cycles of ADT.i After 25000 potential cycles of ADT.j In H2SO4.k After 2000 potential cycles of ADT.l At 60 ○ C. m After 15000 potential cycles of ADT.n Normalized by mass of Pt and Ir.o After 6000 potential cycles of ADT.p After 8000 potential cycles of ADT.q After 4000 potential cycles of ADT.r Normalized by total catalyst mass.s After 5 h of chronoamperometry.t Recalculated from Ag/AgCl assuming a pH of 1.

TABLE II .
Characteristics of the ORR for different PGM-free electrocatalysts.ORR was performed in 0.1 M HClO 4 unless stated otherwise.MA and SA are given at 0.8 V.All ECSA were determined by the H upd method.ADT refers to 10 000 cycles between 0.6 and 1.0 V unless stated otherwise.All potentials refer to RHE.

TABLE II .
(Continued.) c After 10000 s of chronoamperometry.d After 40000 s of chronoamperometry.e After 5000 potential cycles of ADT.f After 30000 potential cycles of ADT.g After 50000 s of chronoamperometry.h After 20000 potential cycles of ADT.i After 20000 s of chronoamperometry.j After 24 h of chronoamperometry.k With respect to BET surface area.l After 3000 potential cycles of ADT.m After 15000 potential cycles of ADT.

TABLE III .
Characteristics of PEMFC tests in H 2 /O 2 for different electrocatalysts at 80 ○ C and 100% relative humidity.ADT refers to cycling between 0.6 V and 1.0 V unless stated otherwise.

TABLE III .
(Continued.) j With respect to Pt mass.k At 60 ○ C. l After 100 h of constant operation. 156

TABLE IV .
Characteristics of PEMFC tests in H 2 /air for different electrocatalysts at 80 ○ C and 100% relative humidity.ADT refers to cycling between 0.6 V and 1.0 V unless stated otherwise.
a Voltage retention after 10 h at constant-current operation.bTestsperformed at 60 ○ C. c After 30000 potential cycles of ADT.dPt loading e After 50000 potential cycles of ADT.f Tests performed at 80% relative humidity g After 100 h at constant voltage.h After 180 h at constant voltage.i After 1000 potential cycles of ADT.j Pd-Au NPs are coated onto the Nafion membrane, while a conventional Pt/C catalyst is used in addition.k After 15000 potential cycles of ADT.l After 5000 potential cycles from 1.0 to 1.5 V. m After 20000 potential cycles of ADT.n After 10000 potential cycles of ADT.