Review of characterization and modeling of polymer electrolyte fuel cell catalyst layer: The blessing and curse of ionomer

Abstract

Ionomer impregnation represents a milestone in the evolution of polymer electrolyte fuel cell (PEFC) catalyst layers. Ionomer acts as the binder, facilitates proton transport, and thereby drastically improves catalyst utilization and effectiveness. However, advanced morphological and functional characterizations have revealed that up to 60% of Pt nanoparticles can be trapped in the micropores of carbon support particles. Ionomer clusters and oxygen molecules can hardly enter into micropores, leading to low Pt utilization and effectiveness. Moreover, the ionomer thin-films covering Pt nanoparticles can cause significant mass transport loss especially at high current densities. Ionomer-free ultra-thin catalyst layers (UTCLs) emerge as a promising alternative to reduce Pt loading by improving catalyst utilization and effectiveness, while theoretical issues such as the proton conduction mechanism remain puzzling and practical issues such as the rather narrow operation window remain unsettled. At present, the development of PEFC catalyst layer has come to a crossroads: staying ionomer-impregnated or going ionomer-free. It is always beneficial to look back into the past when coming to a crossroads. This paper addresses the characterization and modeling of both the conventional ionomer-impregnated catalyst layer and the emerging ionomer-free UTCLs, featuring advances in characterizing microscale distributions of Pt particles, ionomer, support particles and unraveling their interactions; advances in fundamental understandings of proton conduction and flooding behaviors in ionomer-free UTCLs; advances in modeling of conventional catalyst layers and especially UTCLs; and discussions on high-impact research topics in characterizing and modeling of catalyst layers.

Graphical abstract

Keywords

polymer electrolyte fuel cell / ultra-thin catalyst layer / electrostatic interactions / characterization and modeling / structure-property-performance relation / water management

1 Introduction

Polymer electrolyte fuel cells (PEFC) are critical enabling technique in today’s portfolio of clean energy solutions and envisioned as the ultimate power source of future vehicles owing to its unrivaled efficiency and environmental friendliness. Over the past three decades, great progress has been made in lowering the Pt loading from 4 mg/cm² to 0.4 mg/cm², crushing the perception that PEFC is just a laboratory toy, and making us now stand on the brink of large-scale commercialization [1–10]. Notwithstanding this huge success, the PEFC technology may need to reduce the Pt loading by another factor of ten to warrant economic viability (detailed estimates can be found in recent reviews [4,9]). To this end, material innovation and structural design of the catalyst layer are recognized to be of vital importance.

At a crossroads concerning ionomer in the catalyst layer, impregnating the catalyst layer with ionomer has become a decisive step in improving catalyst utilization and thus reducing catalyst loading in the history of PEFC development. However, recent advances reveal that the ionomer surrounding Pt nanoparticles is detrimental to the oxygen reduction activity. More importantly, it causes significant mass-transport loss at high-current densities (see a recent review [9]). Therefore, many works have been conducted to eliminate ionomer from the catalyst layer and largely reduce the thickness of the catalyst layer, leading to the birth of ionomer-free ultra-thin catalyst layers (UTCLs) that emerge as game-changing approaches to improve catalyst utilization, effectiveness and durability [4,5]. On the other hand, researchers still hope that the performance can be improved, the cost reduced, and durability prolonged by delicately optimizing the conventional ionomer-impregnated catalyst layer. For example, Uchida et al. have significantly improved Pt utilization by uniformly dispersing Pt nanoparticles of a sharp particle-size distribution on the exterior surface of a carbon support with few interior pores. They assume that the utilization and effectiveness of Pt nanoparticles in ionomer-free micropores is rather low due to impeded proton conduction and oxygen transport [11,12].

A plethora of open questions emerge at the crossroads: staying ionomer-impregnated or going ionomer-free. In terms of the conventional ionomer-impregnated catalyst layers, macro- and microscopic distributions of constituents (ionomer, Pt particles, and support particles) remain elusive; interlinking structure, properties and performance are formidable and long-standing issues. Ionomer-free UTCLs are posing great challenge to the old belief that ionomer is essential to proton conduction in the CL. How do protons transport in ionomer-free UTCLs? What is the proton conductivity thereof? How efficiently are the Pt catalysts utilized therein? Can the UTCLs survive extreme conditions such as low RH, flooding and cold start? Just to name only a few. The present paper aims at reviewing recent progresses on characterization and modeling of both conventional ionomer-impregnated catalyst layers and the emerging UTCLs, and identifying common experimental and theoretical challenges for these two competing paths.

A wealth of reviews in the realm of PEFC have been published [1–10]. The present review features advances in characterizing microscale distributions of Pt particles, ionomer, support particles and unraveling their interactions in catalyst layers; advances in fundamental understandings of proton conduction and flooding in ionomer-free UTCLs; advances in modeling of conventional catalyst layers and especially UTCLs; and identifying the high-impact research topics in characterizing and modeling of catalyst layers. The progress summary is inevitably filtered through the authors’ biases. Consequently, readers are strongly advised to also refer to other relevant reviews, so as to obtain a balanced understanding, for example, a comprehensive review of catalyst layer modeling by Eikerling et al. [3], a perspective focusing on ionomer in catalyst layer by Holdcroft [8], an overview of effects of constituents in catalyst ink on the durability of catalyst layer by Zamel [10].

2 Basics of PEFC

2.1 Principles, components and advantages

Fuel cell is an electrochemical device that converts chemical energy of the fuel directly into electric energy [13]. A PEFC, also termed as polymer electrolyte membrane fuel cell (PEMFC) or proton exchange membrane fuel cell (PEMFC) in the literature, uses a thin (~20 mm) proton-conducting membrane, such as a perfluorosulfonic acid polymer (the famous DuPont’s Nafion^®), to separate catalytic porous electrodes (usually Pt nanoparticles supported on carbon) [14].

In a PEFC, hydrogen from the fuel (pure hydrogen or hydrogen-rich gas mixture) at the anode is oxidized to protons as charge carriers,

(1){{\displaystyle \mathrm{H}}}_{\mathrm{2}}\leftrightharpoons \mathrm{2}{\mathrm{H}}^{+}+\mathrm{2}{\mathrm{e}}^{-}.

Protons transport through the electrolyte and react with oxygen at the cathode to form water and some waste heat,

(2)\mathrm{4}{\mathrm{H}}^{+}+\mathrm{4}{\mathrm{e}}^{-}+{{\displaystyle \mathrm{O}}}_{\mathrm{2}}\leftrightharpoons \mathrm{2}{{\displaystyle \mathrm{H}}}_{\mathrm{2}}\mathrm{O}.

Figure 1 shows the layout of a PEFC. A PEFC is usually composed of two bipolar plates that introduce gaseous reactants to the catalyst layer, harvest the generated electric current, and transport the by-produced water; two pieces of gas diffusion layer (GDL) that transport reactants, conduct electrons, and aid in water management; two pieces of catalyst layer (CL) where the electrochemical reactions take place; and a proton exchange membrane (PEM) that transports proton and prohibits electron conduction. The assembly of the PEM, CLs, and GDLs is termed as the membrane electrode assembly (MEA) [1].

The most outstanding advantages of a PEFC are high efficiency: 83% in terms of theoretical thermodynamic efficiency at standard state, 50%–60% in terms of practical stack efficiency, well above the efficiency of heat engines (20%–40%); environmental friendliness: water as the only by-product; low-temperature operation: below 100°C; and long lifetime: 3000–4000 h for transportation applications and over 10000 h for stationary applications [17].

2.2 Evaluation tools, transport and electrochemical processes

How is the performance of a PEFC evaluated? Most often, a polarization curve, viz. the cell voltage versus applied current (or current density) curve, is drawn as a starting point, as illustrated in Fig.2(a). The polarization curve embodies the convoluted effects at all scales and in all components in a PEFC. The impacts of 50–100 parameters related to materials, mass transport, reactions, and structure are condensed into this single curve [3,6]. Usually, the polarization curve can be divided into an activation regime in the low range of current density, an ohmic regime in the mediate range of current density, and a mass-transfer regime in the high range of current density. This somewhat crude separation makes the polarization curve a practical tool.

Electrochemical impedance spectroscopy (EIS) is a powerful method to distinguish multiple processes over a wide range of time constants in electrochemical systems. Figure 2(b) depicts a typical EIS of a PEFC. Upon decreasing the frequency, the ohmic, activation and mass-transfer regimes unfold successively. The high frequency resistance (HFR, usually at 1 kHz), defined as the impedance at the point where the EIS curve intersects with the real axis, is widely used to infer the water content in the PEM. The following 45° line corresponds to proton transport in the porous catalyst layer. The intermediate-frequency semicircle is usually ascribed to electrochemical reactions at Pt/solution interface. The low-frequency loop mirrors mass-transfer processes in porous media. Interested readers may refer to a recent review [20].

Behind the curtain of the polarization curve and EIS plot are complex transport and electrochemical processes in the PEFC. Figure 2(c) delineates transport paths of so-called triple phases (electrons, protons, and oxygen) in a PEFC, as well as the equivalent electric circuit. Electrons move through the GDL and then via the carbonous network in the CL. Protons, oxidized from hydrogen at the anode, mitigate through the PEM and then via the ionic network in the CL. Oxygen, introduced from flow channels grooved in the bipolar plates, transports via molecular diffusion and convection in the GDL, and then mainly via Knudsen diffusion in the CL. It is usually believed that Pt particles on the carbon support are covered by a thin ionomer film or water film. In this scenario, oxygen needs to dissolve into ionomer/water phase and then diffuse across the thin film before participating in the oxygen reduction reaction (ORR), as indicated in Fig. 2(d). Resolving the microscopic distributions of Pt nanoparticles, carbon and ionomer lays the basis for understanding reactions and transport in catalyst layers, which will be discussed in Subsection 3.2.

The ORR in Eq. (2) is a multi-electron electrocatalytic reaction (see an excellent review in Ref. [21].). The sluggish kinetics of the ORR necessities the use of Pt (or its alloy) particles as the catalyst. Even facilitated by today’s best catalyst, the ORR still possesses an overpotential of ~0.4 V. This staggeringly large overpotential makes continued efforts on elucidating the mechanism of the ORR and finding more active catalysts, constituting vigorous research topics in electrocatalysis. Phenomenologically, the Butler-Volmer equation or the Tafel equation by neglecting the anodic branch can be used to describe the relation between overpotential and current of the ORR.

2.3 A brief history and current status

In a letter to the editors of the Philosophical Magazine published in December 1838, Christian Friedrich Schoenbein, the Swiss professor of chemistry, reported his voltaic polarization research of metals (Pt, Au, and Ag) [22]. First, oxygen and hydrogen were produced by water electrolysis and collected inside two glass vessels. Then, two Pt wires, connected with a galvanometer, partially immersed into an acid solution, and acting as electrodes, were placed in the glass vessels. Schoenbein observed a “secondary current”, which was ascribed to “chemical combination” of oxygen and hydrogen in presence of Pt wires. This is considered as the first observation of the “fuel-cell effect” [6] and this experimental apparatus, as shown in Fig. 3, is regarded as the first fuel cell. In January 1839, William Robert Grove reproduced the experiments of Schoenbein [23]. In December 1842, Grove further connected several fuel cells in series and he was recognized as the inventor of the fuel cell power source [24].

In 1889, Mond and Langer built the first practical H₂/O₂ fuel cell, as shown in Fig. 3 [18]. For the first time, powered electrocatalyst in the form of Pt black was used. Mond and Langer found that it was necessary to keep Pt black “comparatively dry”. In doing so, they used “an electrolyte in a quasi-solid form”, i.e., a porous ceramic media impregnated by sulphuric acid. Special attention should be paid to the peculiar structure of the catalyst layer which was made up of thin Pt leafs perforated with small holes that were filled with Pt black particles (~0.1 mm). Moreover, Pt black particles formed a thin layer sandwiching between the Pt leafs and the diaphragm. At 40°C, Mond and Langer observed an electromotive force of 0.97 V at open circuit; the cell was able to produce a power density of ~2 mW/cm² (@0.73V) with a Pt loading around 2 mg/cm². However, the performance degraded very quickly, which was ascribed to transport of sulphuric acid to anode, resulting in a more and more concentrated solution [18].

Since 1950s, General Electric (GE) have initiated a program aimed at developing solid polymer electrolyte (SPE) fuel cells, leading to the birth of PEFCs [25–27]. Appleby and Yeager excellently reviewed the development of fuel cell technology at GE at that time [28]. It was W. T. Grubb at GE who first suggested the use of cation-exchange membranes as the acid electrolyte in fuel cells [25]. After a long list of sulfonated phenolformaldehyde, polyvinyl compounds, and Kel-F film, P(S-DVB) SA membrane became the favorite SPE membrane and was used in the fuel cell for the Gemini space mission in 1962 [28]. The Gemini fuel cell used Pt black particles bonded by Teflon with a Pt loading of 35 mg/cm², and produced 37 mg/cm² at 0.78 V under conditions of H₂/O₂, 50°C and 2 atm [27]. At that time, the major problem was the hot spots in the cells, caused by chemical degradation of the P(S-DVB)SA membrane [28]. To solve the hot spot problem, Nafion^® membrane developed by DuPont was used in the project of Biostatellite 2 for the first time in 1967. The use of Nafion^® membrane also made high temperature operations of up to 100°C, possible, partially mitigating the cathode flooding.

Around 1970s, development of commercial applications of PEFC was started, promoting the research of highly dispersed Pt/C catalysts [29–32]. In 1980s, another important development was the synthesis of solubilized ionomers, resulting in a commercial product of 5% Nafion^® in ethanol/water liquid mixture [33]. Breakthroughs were made at Los Alamos National Laboratory in late 1980s. Raistrick innovated the gas diffusion electrode composed of PTFE-bonded dispersed Pt/C by painting Nafion^® solution onto the electrode, resulting in a remarkable reduction in Pt loading from 4 to 0.4 mg/cm² [34]. Later, Wilson and Gottesfeld radically eliminated PTFE from the catalyst layer with further improved catalyst utilization [35,36]. Since that time, PEFC development has undergone continued adjustments, from material innovation to structural design, but no drastic modifications [6].

Figure 3 is a glimpse of the long and winding history of fuel cell technology in 19th and 20th centuries. The primary aim of giving a brief account of the PEFC history is to find out how and why today’s PEFC has evolved from the simple and crude experimental setups of Schoenbein and Grove. Interested readers may refer to more detailed reviews by Kocha [37], Appleby and Yeager [28], and Perry and Fuller [15].

Figure 4 displays the cell voltage, E_cell, as a function of the specific current density, j_cell/m_Pt with j_cell being the current density, and m_Pt being the Pt loading. The polarization curves of fuel cells at different times are compared, including the first practical H₂/O₂ fuel cell built by Mond and Langer in 1889 [18]; the first PEFC invented by Grubb and Niedrach at General Electric in 1960 [26]; the innovative design of impregnating the catalyst layer with ionomer by Raistrick at the Los Alamos National Laboratory in 1989 [34]; Gore-select^® MEA after systematic model-based optimization of the structure of porous electrodes in 2003 [38]; the radical design of nanostructure thin-film (NSTF) catalyst layer developed by Debe et al. at Minnesota Mining and Manufacturing Company (3M) in 2006 [39]; and the conventional catalyst layer improved by locating all Pt particles on the exterior surface of carbon support particles from Masahiro Watanabe group in 2016 [12]. Note that the data in high potential range are omitted for clarity for the curve of Watanabe’s group. The stretch target of 20 kW/g_Pt is adopted from Kongkanand and Mathias [9]. There may be different opinions on selecting the milestones of PEFC development in 21st century. In this paper, the work of Debe at 3M and that of Watanabe’s group at Yamanashi University are selected because they represent two radically different trends: ionomer-free and ionomer-impregnated electrodes. These two approaches are sharply different in whether ionomer is essential to proton conduction and high utilization and effectiveness of Pt catalysts or not.

Notwithstanding the tremendous progress made in the half century since the birth of PEFC, challenging gaps, in terms of cost and durability, between current status and 2020 targets exist, as demonstrated in Fig. 5. More information can be found in recent DOE reports [17,40]. Pt (or its alloy) based catalyst constitutes 45% of the fuel cell stack cost. As a result, reducing Pt loading attracts special attentions, involving concerted efforts from increasing intrinsic activity of the catalyst via materials innovation, improving utilization and effectiveness of the catalyst via structural optimization, and enabling high-current-density operation via system control. Improving durability involves synergistic efforts from many sectors including material innovation (catalyst, support, membrane, bipolar plate, and so on), structural design (MEA, bipolar plates with flow field, and end plates, etc.), operation optimization (start/stop, load cycling, idling, cold start, etc.), and system control. MEA is the key component determining the durability of fuel cell stack. Continued efforts have been made to the unravel aging phenomenon (attenuated electrochemical surface area (ECSA), growth of Pt particles, Pt band in membrane, increased protonic resistance, release of fluorine and carbon dioxide, membrane creep and microstructure fracture, etc.), elucidate degradation mechanisms (carbon oxidation, Pt dissolution, Ostwald ripening, radical degradation, etc.) and conceive mitigation strategies (robust catalyst support, Pt particles with a narrow size distribution, radical-resistive membrane etc.) [42,43].

3 Conventional ionomer-impregnated catalyst layer

A conventional catalyst layer is a heterogeneous composite made up of carbon supported Pt (or its alloy) nanoparticles held together by a perfluorosulfonic acid (PFSA) ionomer binder, usually Nafion^®. This subsection briefly introduces the preparation methods, microstructure, functional properties and modeling of the conventional CL. Experimental evidences and modeling of ionomer-free zones in conventional CLs are highlighted.

3.1 Preparation methods

Conventional CLs are usually formed from a catalyst ink consisting of Pt/C catalysts, ionomer (Nafion^®), and solvent [44]. The catalyst inks are either mechanically/magnetically stirred and/or ultrasonically dispersed to ensure spatial uniformity. Pt/C nanoparticles form heterogeneous aggregates in the catalyst ink. Moreover, there are evidences indicating aggregation of ionomer in the solution [45]. The catalyst-ionomer-solvent interactions, yet not adequately studied, have important effects on the structure and performance of the CL [10]. By means of brushing, or screen printing, or electrospray, or electrospinning, the catalyst ink is deposited onto either the GDL or the PEM, with the former termed as gas diffusion electrode (GDE) and the latter as catalyst coated membrane (CCM). The assembly of two pieces of GDE with a PEM or a CCM with two pieces of GDLs results in a MEA. Figure 6 shows a common preparation method of MEA and schematic illustrations of the electrospray and electrospinning techniques. Comprehensive reviews provide more detailed and systematic information on this aspect [1,10,44,46].

3.2 Microstructure

Resolving of the complicated structure of conventional CLs attracts tremendous attention. The key questions are: what is the pore size distribution? where do Pt particles locate? are all Pt particles covered by an ionomer film? how does the ionomer network distribute? can ionomer enter into primary carbon particles? what are interactions between distributions of Pt, ionomer, carbon and pores? This subsection does not intend to give a comprehensive review but merely tries to construct an up-to-date picture of the microstructure of conventional CLs based on most recent modeling and characterization results.

3.2.1 Support

Carbon, especifically carbon black and graphitized carbon, is the most common support employed in PEFCs due to its low cost, high surface area (10¹–10³ m²/g), chemical stability and affinity for Pt (or its alloy) nanoparticles [8,10]. As displayed in Fig. 7, primary carbon black particles (~20 nm) are formed by pseudo-spherical assembly of micro-crystallites with micropores (<2 nm) in between. Primary carbon black particles are inclined to aggregate, forming agglomeration particles (~200 nm) with mesopores (2–20nm) between primary particles. Inter-agglomerate pores are macropores (>20 nm) [8]. Typically, N₂ absorption experiments exhibit a bimodal distribution of the pore size. The observed two regimes are referred to as primary pores (<20 nm), and secondary pores (>20 nm), respectively. In this context, the primary pores are intra-agglomerate pores including micropores and mesopores [8,10].

Ketjen Black and Vulcan XC-72 are two commonly used types of carbon support in PEFCs. Soboleva et al. found that Ketjen Black possesses a larger surface area (~890 m²/g_carbon) among which (~480 m²/g_carbon) corresponds to micropores (<2 nm) [47]. In contrast, Vulcan XC-72 has a much smaller surface area (~220 m²/g_carbon) as a result of a much smaller portion of micropores [47]. The porous structure of carbon support is revealed to have important effects on the distributions of ionomer and Pt particles, water sorption, and ORR activity, as discussed below [47,48].

3.2.2 Pt particles

There are three key questions concerning the distribution of Pt particles in conventional CLs. First, what are the frequencies of Pt particles located on the exterior and interior surfaces of the agglomerate, respectively? Second, where are these intra-agglomerate Pt particles located? Third, how many Pt particles are electrochemically active?

Park et al. have answered the first question by employing a scanning transmission electron microscopy (STEM) [12]. When the STEM instrument is operated under the SEM mode, the Pt particles located on the exterior surface of the agglomerate can be counted. By fixing the location of the sample and turning the STEM instrument to the TEM mode, all the Pt particles located on both exterior and interior surfaces of the agglomerate can be counted. As shown in Fig. 8, 62% of Pt particles are located on the interior surface of the agglomerate in commercial carbon black supported Pt catalysts (Pt/CB) [12]. For the case of acetylene black supported Pt catalysts (Pt/AB, surface area of ~115 m²/g_carbon), all Pt particles are on the exterior surface of the agglomerate, which is ascribed to the greatly decreased number of interior micro- and meso-pores [12].

To infer the location of Pt particles inside the agglomerate, Soboleva et al. compared the pore size distribution of carbon support before and after loading Pt particles [47]. For the case of Ketjen Black, Soboleva et al. observed a significant decrease in the detectable surface area in N₂ adsorption measurement, resulting from the blocking of micropores by Pt particles [47]. Soboleva et al. speculated that Pt particles (2–4 nm)) are prone to being located at the mouth of micropores. However, they are unlikely to enter into the micropores due to size limitation, as described in Fig. 7 [47]. Note that Pt distribution is dependent on the porous structure of carbon support. For example, negligible loss in surface area was found for Vulcan XC-72 after deposition of Pt particles [47].

3.2.3 Ionomer

Ionomer distribution plays a key role in determining proton conductivity, gas transport, water transport, and Pt utilization in conventional CLs. Therefore, researchers are keen on knowing the ionomer distribution in CLs. However, imaging ionomer in CLs poses great challenge due to the fact that ionomer usually forms an ultra-thin film (<10 nm) covering the Pt/C agglomerates and that ionomer and carbon have similar density, therefore, presenting little difference in contrast. The key questions are: what is the ionomer coverage in conventional CLs? how do ionomer films distribute? can ionomer enter into the primary pores?

Computational approaches provide valuable insights to above questions. Malek et al. have studied the microstructure of conventional CLs by employing the coarse-grained molecular dynamics approach [49,50]. The computations show that ionomer forms a thin adhesive film, 3–4 nm in thickness and ~10 nm in length, partially covering Pt/carbon agglomerates. The conclusions are reached that the ionomer cluster consisting of aggregates of several backbone chains are unlikely to penetrate into primary pores of Pt/C agglomerates due to size limitation, and that the ionomer film is not directly adhered to Pt/C agglomerates, instead, a water film forms in between. This water film wets Pt particles, and acts as a proton-conducting medium. Experimental studies corroborate the above results as given in detail below.

The ionomer distribution can be inferred from pore size distribution via N₂ adsorption. By adding 10%(wt) ionomer to Ketjen Black supported Pt catalysts, Soboleva et al. have found that the detectable volumes of micropores and mesopores decrease significantly [47]. It is postulated that ionomer is primarily distributed on the exterior surface of Pt/C agglomerate at low-to-middle ionomer content (<30%(wt)). As N₂ is nearly insoluble in the ionomer, micropores and mesopores inside the Pt/C agglomerates that are covered by ionomer films are undetectable.

Only until recently can ionomer distribution in conventional CLs be imaged [51,52]. Lopez-Haro et al. stained the ionomer phase with Cs⁺ ions to increase the electronic contrast as compared with the carbon black [51]. In addition, Pt nanoparticles, possessing high-contrast, are eliminated in order to avoid perturbing the images Pt. As shown in Fig. 9, ionomer forms a thin-film with a thickness of ~7 nm on the carbon black, which is consistent with previous computation results. In addition, increasing the ionomer content by a factor of two does not change the thickness of the ionomer thin-film but increases the ionomer coverage.

3.2.4 Interactions between constituents

The adsorption of Pt particles onto carbon significantly changes the microstructure of the CL [50]. Carbon, Pt, and ionomer have different wettabilities. Carbon surface is usually highly hydrophobic; Pt particles are hydrophilic; Nafion^® ionomer consists of hydrophobic backbones and hydrophilic side chains. The computational work of Malek et al. have demonstrated that in the absence of Pt particles, backbones of the ionomer are strongly attached to the hydrophobic carbon surface in a well-ordered form with side chains being oriented toward the macropores, and that in the presence of Pt particles, the ionomer phase becomes disordered and the side chains are prone to being oriented toward the carbon surface [50]. Figure 10 illustrates the effect of Pt particles on the microstructure of the CL [50]. The distribution of water molecules is important to the proton conductivity of the CL. As delineated in Fig. 10, water is more likely to be dispersed in the micro- and meso-pores in presence of hydrophilic Pt particles. For comparison, water adsorbs primarily on the ionomer film in absence of Pt particles [50].

In summary, there are sufficient evidences suggesting that Pt particles can penetrate into ionomer-free zones inside the agglomerate. Therefore, there are two types of Pt particles, those at the Pt/ionomer interfaces and those at Pt/water interfaces. The content of the ionomer and the porous structure of the carbon support are key properties determining the ratio between these two types of Pt particles. The proton conduction in ionomer-free zones and the activity of Pt particles at the Pt/water interfaces are key, yet controversial, issues.

3.3 Functional properties

Key functional properties of the CL include proton conductivity, gas transport resistance, wettability, water uptake capability, and ORR activity etc. This subsection is focused on the first two terms. Discussion on the ORR activity of interior Pt nanoparticles is deferred to Section 5.

3.3.1 Proton conductivity

In the CL, proton conductivities at two scales, the CL itself and the ultra-thin ionomer film covering Pt/C agglomerates, are often studied. Regarding the former, Boyer et al. reported their early results in 1998 [53]. Since then, various methods have been developed, and the proton conductivity, k_CL, is used to calculate tortuosity of the ionic network in the CL. Not until recently have researchers explored the latter, which is attracting more and more attention.

There are broadly three methods to measure k_CL, the EIS method [54–56], the H₂ pump method [53,57], and the in-situ potential method [58], of which the EIS method is the most commonly used one. Figure 11(a) shows a representative EIS of a CL in H₂/N₂, which consists of a 45° line in the high frequency range and a nearly vertical line in the low frequency range [55]. The transmission line model is usually employed to fit the EIS and estimate k_CL. Liu et al. have found that k_CL is independent of the thickness of and the Pt loading in the CL, provided that the I/C weight ratio and the carbon type are fixed [56]. In addition, k_CL decreases with increasing the relative humidity (RH), as shown in Fig. 11(b) [56].

Based on {{\displaystyle \kappa }}_{\mathrm{C}\mathrm{L}}={{\displaystyle \kappa }}_{\mathrm{P}\mathrm{E}\mathrm{M}}{{\displaystyle ϵ}}_i/{{\displaystyle \tau }}_i with k_PEM being the proton conductivity of bulk membranes, ε_i the ionomer volume fraction in the CL, and τ_i the tortuosity of the ionic network in the CL, Iden et al. found that [57]. As is known, the ionomer phase forms an ultra-thin film with a thickness of less than 10 nm in the CL, which is comparable to the ionic domain size (~4 nm) in bulk ionomer membranes but is approximately three-orders smaller than the membrane thickness (~20 mm). The relation gives rise to the question: can the intrinsic proton conductivity and the conduction mechanism of an ultra-thin ionomer film be assumed to be the same as that of a bulk ionomer membrane [57,59]?

In 2011, Paul et al. built a model system of a 50 nm Nafion^® thin-film adsorbed on a SiO₂ substrate [60]. The EIS measurements have revealed that the proton conductivity of the thin-film is 31 mS/cm (60°C, 96%RH)), which is significantly smaller than that of bulk membranes, ~100 mS/cm, under the same conditions. In a subsequent study, Paul et al. have measured the proton conductivity of ultra-thin Nafion^® films with the thickness varying between 4 and 300 nm [61]. Figure 11(c) exhibits an exponentially decreasing trend with the reduction of the film thickness. Moreover, the activation energy, a parameter reflecting the mechanism of proton conduction, of the thin-film is much larger than that of bulk Nafion^® membranes, as shown in Figure 11(d) [60]. Paul et al. argued that the mechanisms of proton conduction in ultra-thin ionomer films are dominated by the surface diffusion mechanism instead of by the Grotthuss mechanism occurring in bulk membranes [61]. Recent attention has been paid to the underlying causes. Ono et al. have observed highly oriented sulfonic acid groups in the substrate/film interface which may be related to the low proton conductivity [62]. The notion that the proton conductivity of an ionomer thin-film is co-determined by the microstructure of the Pt/ionomer interface, the proton conduction mechanism, and the water content is the basis for a comprehensive understanding. Kusoglu et al. have found that the water up-take capability of an ultra-thin ionomer film is much lower than that of a bulk membrane [63].

3.3.2 Oxygen transport resistance

Subsection 3.2 indicates that there are secondary pores (>20 nm) between agglomerates and primary pores (<20 nm) inside agglomerates in the CL. Moreover, an ultra-thin ionomer film covers Pt/C agglomerates. Therefore, oxygen molecules transport through secondary and primary pores, dissolve into the ionomer phase, diffuse through the ionomer film, adsorb onto Pt particles, and finally participate in the ORR. The O₂ transport in primary pores is dominated by Knudsen diffusion, and its resistance, {{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{K}\mathrm{n}\mathrm{u}\mathrm{d}\mathrm{s}\mathrm{e}\mathrm{n}}, is non-Fickian and independent of pressure. The O₂ transport resistance related to the ionomer thin-film, {{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}}, is also non-Fickian and independent of pressure. Therefore, the total non-Fickian O₂ transport resistance in CL, R_NF , is the sum of two parts, {{\displaystyle R}}_{\mathrm{N}\mathrm{F}}={{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{K}\mathrm{n}\mathrm{u}\mathrm{d}\mathrm{s}\mathrm{e}\mathrm{n}}+{{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}}.

The total O₂ transport resistance of a PEFC, R_total, is usually measured by using the current limiting method. R_NF can be differentiated from the pressure-dependent O₂ transport resistance by extrapolating the R_total versus gas pressure curve to zero gas pressure [64]. Remarkably, R_NF significantly increases at low Pt loadings. Kongkanand and Mathias gathered literature data in Fig. 12 [9]. As {{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{K}\mathrm{n}\mathrm{u}\mathrm{d}\mathrm{s}\mathrm{e}\mathrm{n}} is approximately invariant with the Pt loading (or the roughness factor), this peculiar behavior is ascribed to {{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}}. This postulation is strongly supported by the NSTF data. The NSTF electrode has a relatively small roughness factor (~15) and is intrinsically ionomer-free. By adding ionomer to NSTF electrode, its R_NF dramatically increases from ~0.1 S/cm to 0.5–1 S/cm [9].

Why is {{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}} so large? {{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}}can be further divided into three parts: resistance at the gas/ionomer interface, resistance through the ionomer-film, and resistance at the ionomer/Pt interface. Suzuki et al. have revealed that the resistance at the gas/ionomer interface is the dominating term [65]. Owejan et al. have further emphasized the important of resistance at the ionomer/Pt interface [66]. Liu et al. have discovered that O₂ permeability at the gas/ionomer interface alone cannot explain the large {{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}} because {{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}} is relatively small when the ionomer is not in direct contact with the Pt surface [67]. Recent first-principle calculations have indicated that the ionomer film becomes denser near the Pt surface due to enlarged polymer-metal attractive interactions [68].

Why does {{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}} depend on Pt loading? Similar trend has been observed either by decreasing the thickness of the CL or diluting Pt/C catalysts with carbon black particles [66,69]. A common explanation is that the oxygen flux per Pt particle is increased because the current is shared by a smaller number of Pt particles at a low Pt loading. This leads to an elongated transport path, and thereby, a larger {{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}}. From polarization curves, it is observed that the overpotential due to O₂ transport is greater with decreasing the Pt loading at a given current density (defined with respect to the electrode surface area rather than the Pt surface area). However, this phenomenon cannot directly suggest the fact that {{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}}is greater at a lower Pt loading, because {{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}} is measured by using the current limiting method under conditions that are distinct from the polarization curve measurement. Some researchers argue that {{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}} should be an intrinsic property that is independent of the oxygen/current flux, or let’s say, the Pt loading. Regarding the poison effect of Pt, contradictions may be encountered as follows. If the presence of Pt particle “poisons” the ionomer-film, it is expected that {{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}}shall increase when there are more Pt particles on the carbon surface beneath the ionomer film. That is, a larger {{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}} is expected at a higher roughness factor, which is inconsistent with Fig. 12. This contradiction means the current understanding needs further modifications.

A very important role that is neglected in previous studies is the water film between the ionomer film and Pt particles, as shown in Fig. 10. As Pt/C agglomerates are partially covered by the ionomer film, O₂ is preferable to transport through the water phase instead of the impeded ionomer phase. By considering the role of water film, the above contradiction might be resolved. As Pt particles are hydrophilic and the carbon surface is hydrophobic, more Pt particles mean more water between the ionomer thin-film and the Pt surface. Therefore, more O₂ can transport through the water film with a much larger diffusivity, resulting in a reduced O₂ transport resistance at a higher Pt loading. To prove or disprove this explanation, the RH dependence of {{\displaystyle R}}_{\mathrm{C}\mathrm{L}}^{\mathrm{i}\mathrm{o}\mathrm{n}\mathrm{o}\mathrm{m}\mathrm{e}\mathrm{r}}is informative.

3.4 Agglomerate model: is it outdated?

Modeling is an integral part in the development of PEFC technology, especially in the design of the CL, the thinnest but probably the most complicated component, where electrochemical reactions take place and all different types of species (oxygen, water vapor, liquid water, protons, electrons etc.) exist. The modeling of PEFC was comprehensively reviewed by Weber and Newman in their widely cited paper in 2004 [2], which was supplemented by a recent review from Weber et al. in 2014 [7].With these two high-quality reviews, it is unnecessary to give another overview on this topic. Instead, the agglomerate model, the currently prevailing model, is focused on by introducing its basic ideas, presenting recent progresses, and discussing the remaining challenges.

3.4.1 Basic ideas

Continuum approaches are the mainstream to model the CL. These macrohomogenous approaches do not account for the heterogeneous and microscopic details of the CL. Instead, they assume that all phases exist at all points in the volume with properties averaged over a representative volume element. Macrohomogenous models improve over the interface model that treats the CL as an infinitely thin layer by accounting for distributions of species and potential across the CL, as shown in Fig. 13. The simple macrohomogenous model considers only one length scale, viz., the thickness of the CL. It employs the porous electrode theory to describe multi-component transport and electrochemical reactions. The so-called agglomerate model is essentially a macrohomogenous model; it improves over the simple macrohomogenous model by introducing a microscopic scale, the agglomerate scale. Such a refinement aims at accounting for the effect of mass transport limitation at the agglomerate scale on the ORR. It can be considered as a compromise between structural fidelity and computational cost.

At the scale of the CL thickness, the agglomerate model considers the CL as an effective and homogenous medium. At steady state, the controlling equations of oxygen diffusion, proton conduction, and potential are

(3){\displaystyle \frac{\mathrm{\partial }j}{\mathrm{\partial }x}}={{\displaystyle -i}}_{\mathrm{a}\mathrm{g}\mathrm{g}},

(4)j={{\displaystyle \kappa }}_{\mathrm{C}\mathrm{L}}{\displaystyle \frac{\mathrm{\partial }\eta }{\mathrm{\partial }x}},

(5){{\displaystyle D}}_{{\mathrm{O}}_{\mathrm{2}}}{\displaystyle \frac{{\mathrm{\partial }}^{\mathrm{2}}c}{\mathrm{\partial }{x}^{\mathrm{2}}}}={\displaystyle \frac{{{\displaystyle i}}_{\mathrm{a}\mathrm{g}\mathrm{g}}}{\mathrm{4}F}},

where x is the distance between the PEM/CL interface (m); j and h the local proton current density (A/cm²) and the ORR overpotential (V), respectively; c, the local O₂ concentration in the CL (mol/cm³); k_CL, the effective protonic conductivity of the CL (S/cm); {{\displaystyle D}}_{{\mathrm{O}}_{\mathrm{2}}}, the diffusion coefficient of O₂ in the CL (cm²/s), i_agg the volumetric ORR current density generated in agglomerates (A/cm³); and F the Faraday constant.

Equation (3) says that the change in j is caused by the ORR. Equation (4) describes the relation between j and h using the Ohmic equation. Equation (5) describes the O₂ consumption in the ORR.

In a classical picture of the agglomerate, O₂ transports across the ionomer thin-film, and then diffuses inside the agglomerate where the ORR takes place. At the agglomerate scale, the primary task is to express i_agg as a function of c, h and properties of the agglomerate, including the thickness of the ionomer thin-film, d_ion, the diffusion coefficient of O₂ in the ionomer thin-film, {{\displaystyle D}}_{{\mathrm{O}}_{\mathrm{2}},\mathrm{i}\mathrm{o}\mathrm{n}}, the diffusion coefficient of O₂ inside the agglomerate, {{\displaystyle D}}_{{\mathrm{O}}_{\mathrm{2}}\mathrm{a}\mathrm{g}\mathrm{g}}, and the radius of the agglomerate, R_agg.

By solving the diffusion equation of O₂ inside the agglomerate coupled with the ORR described by the Tafel equation, Eq. (6) can be obtained.

(6){{\displaystyle i}}_{\mathrm{a}\mathrm{g}\mathrm{g}}=\mathrm{4}Fc\left({\displaystyle \frac{\mathrm{1}}{{\displaystyle \frac{\mathrm{1}}{{{\displaystyle R}}_{{\mathrm{O}}_{\mathrm{2}}\mathrm{i}\mathrm{o}\mathrm{n}}}}+{\displaystyle \frac{\mathrm{1}}{{{\displaystyle k}}_{\mathrm{O}\mathrm{R}\mathrm{R}}{{\displaystyle E}}_{\mathrm{a}\mathrm{g}\mathrm{g}}}}}}\right),

where {{\displaystyle R}}_{{\mathrm{O}}_{\mathrm{2}}\mathrm{i}\mathrm{o}\mathrm{n}} is the O₂ transport resistance through the ionomer thin-film, {{\displaystyle R}}_{{\mathrm{O}}_{\mathrm{2}}\mathrm{i}\mathrm{o}\mathrm{n}}={{\displaystyle \delta }}_{\mathrm{i}\mathrm{o}\mathrm{n}}/{{\displaystyle D}}_{{\mathrm{O}}_{\mathrm{2}}\mathrm{i}\mathrm{o}\mathrm{n}} and k_ORR is the ORR rate (s⁻¹),

(7){{\displaystyle k}}_{\mathrm{O}\mathrm{R}\mathrm{R}}={\displaystyle \frac{{{\displaystyle a}}_{\mathrm{a}\mathrm{g}\mathrm{g}}{{\displaystyle j}}_{\mathrm{0}}}{\mathrm{4}F{{\displaystyle c}}_{\mathrm{r}\mathrm{e}\mathrm{f}}}}\exp \left(-{\displaystyle \frac{\eta }{b}}\right),

with c_ref being the reference O₂ concentration in the CL (mol·cm⁻³), b the Tafel slope (mV/(° ), a_agg the specific ECA (m⁻¹), a_agg= m_PtA_Pt/l_CL, m_Pt the Pt loading (mg·cm⁻²), A_Pt the experimentally measured ECSA ({\mathrm{m}}^{\mathrm{2}}{{\displaystyle \mathrm{g}}}_{\mathrm{P}\mathrm{t}}^{-\mathrm{1}}), and l_CL the thickness of the CL (m).

The effectiveness factor is given by

(8){{\displaystyle E}}_{\mathrm{a}\mathrm{g}\mathrm{g}}=\mathrm{3}{\displaystyle \frac{\varphi \coth \varphi -\mathrm{1}}{{\varphi }^{\mathrm{2}}}},

with \varphi ={{\displaystyle R}}_{\mathrm{a}\mathrm{g}\mathrm{g}}\sqrt{{{\displaystyle k}}_{\mathrm{O}\mathrm{R}\mathrm{R}}/{{\displaystyle D}}_{{\mathrm{O}}_{\mathrm{2}}\mathrm{a}\mathrm{g}\mathrm{g}}} being the Thiele’s modulus.

When R_agg→0, and {{\displaystyle R}}_{{\mathrm{O}}_{\mathrm{2}}\mathrm{i}\mathrm{o}\mathrm{n}}=\mathrm{0}, Eq.(6) is reduced to i_agg=4Fck_ORR. Therefore, Eqs.(3) and(5) are transformed to

(9){\displaystyle \frac{\partial j}{\partial x}}=-{{\displaystyle a}}_{\mathrm{a}\mathrm{g}\mathrm{g}}{{\displaystyle j}}_{\mathrm{0}}{\displaystyle \frac{c}{{{\displaystyle c}}_{\mathrm{r}\mathrm{e}\mathrm{f}}}}\exp \left(-{\displaystyle \frac{\eta }{b}}\right),

(10){{\displaystyle D}}_{{\mathrm{O}}_{\mathrm{2}}}{\displaystyle \frac{{\partial }^{\mathrm{2}}c}{\partial x^{\mathrm{2}}}}={\displaystyle \frac{{{\displaystyle a}}_{\mathrm{a}\mathrm{g}\mathrm{g}}{{\displaystyle j}}_{\mathrm{0}}}{\mathrm{4}F}}{\displaystyle \frac{c}{{{\displaystyle c}}_{\mathrm{r}\mathrm{e}\mathrm{f}}}}\exp \left(-{\displaystyle \frac{\eta }{b}}\right).

Therefore, the agglomerate model is reduced to the simple macrohomogenous model when the radius of the agglomerate is sufficiently small, , and the O₂ transport resistance through the ionomer thin-film is negligible, {{\displaystyle R}}_{{\mathrm{O}}_{\mathrm{2}}\mathrm{i}\mathrm{o}\mathrm{n}}=\mathrm{0}. In comparison with the simple macrohomogenous model, the agglomerate model takes into account two additional mass-transport loss: O₂ transport resistance through the ionomer thin-film, and O₂ transport resistance inside the agglomerates.

3.4.2 Recent progresses

The concept of agglomerate can be traced back to a study on a Teflon-bonded gas diffusion electrode by Giner and Hunter in 1969 [71]. Based on experimental characterizations, they proposed a cylindrical agglomerate model in which catalyst particles form porous agglomerates that were fully filled with electrolyte. Outside the cylindrical agglomerate were hydrophobic gas channels created by Teflon binder. In their model, gas transport in the gas channel was not considered. In 1980, Iczkowski and Cutlip applied the cylindrical agglomerate model to a PTFE-bonded, platinum-on-carbon cathodic CL in a phosphoric acid fuel cell [72]. In comparison with the work of Giner and Hunter, two modifications were made. First, the cylindrical agglomerate was assumed to be covered by a thin electrolyte film. Second, gas diffusion in the gas channel was modeled considering both molecular diffusion and Knudsen diffusion. In 1989, Ridge et al. employed the cylindrical agglomerate model to a Teflon-bonded, Pt black CL [73]. By accounting for proton transport in their model, Ridge et al. highlighted the importance of proton transport to catalytic sites in a Teflon-bonded CL.

In 1991, Celiker et al., probably for the first time, developed a spherical agglomerate model, when studying the effect of intragrain diffusion on operation of Teflon-bonded Raney metal gas diffusion electrodes [74]. In 1997, the spherical agglomerate model was applied to a cathodic CL of a PEFC with Nafion^®ionomer as the electrolyte by Broka and Ekdunge [75]. SEM images indicate that the CL is made up of spherical agglomerates with a radius of ~1 mm. A comparison show that the agglomerate model succeeds in capturing significant voltage loss at high current densities while the simple macrohomogenous model fails and significantly overestimates the cell voltage. Since then, the agglomerate model has been widely applied to understand mass-transport loss in the CL, optimize composition and structure of the CL, and study water management of the CL.

Understanding of mass-transport loss. Jaouen et al. have revealed that limitations by proton conduction on the CL scale or by oxygen diffusion on the agglomerate scale would lead to a doubling of the Tafel slope at higher current densities [76]. Sun et al. have found that the oxygen transport resistance across the Nafion^® film covering the agglomerate is the main factor controlling the limiting current density [77]. Kamarajugadda and Mazumder have found that at intermediate current densities, the cell performance is limited by agglomerate-scale (or local) loss, while at high current densities, the cell performance is governed by CL-thickness-scale (or global) loss [78]. Suzuki et al. have emphasized the importance of resistance of oxygen dissolution from the gas phase into the ionomer, and added this dissolution resistance to the conventional agglomerate model [65].

Optimization of CL structure. Siegel et al. have found that the void fraction, ε_void, in CL significantly influences the cell performance, and a value of 0.04 is optimal for their case [79]. Moreover, model results shows that a smaller R_agg results in better cell performance primarily due to more efficient mass-transport at the agglomerate scale. Sun et al. have examined the effect of Nafion loadings on the cathode performance [77]. The Nafion^® volume fraction in the CL is optimal at ~50% which is consistent with experimental finding. Yin have investigated the effect of Pt loading in the CL, and pointed out that an excessive Pt loading under a fixed ionomer loading is also a waste due to the limitation of proton transport [80]. Kamarajugadda and Mazumder have also studied the effect of Nafion^® content and Pt loading, and concluded that the Pt loading should be at a moderate level in order to optimize the fuel cell performance [78]. Secanell et al. have conducted a multi-objective optimization study, and indicated that the optimal CL structure actually depends on the operating condition [81,82]. Different from analyzing one effect at a time in previous studies, Khajeh-Hosseini-Dalasm et al. have applied the analysis of means and interaction plots to compare the relative importance and explore the potential interactions among parameters [83]. Strong interactions among the Pt loading, the Nafion^® content, the CL thickness are revealed. The water saturation and the CL thickness are found to be the most influential parameters affecting the cell performance.

Water management. Rao et al. have considered the liquid water effect in the agglomerate model by assuming a liquid water film covering agglomerates [84]. The thickness of the water film is determined by the water saturation in the CL, which is solved using a multiphase flow model on the CL scale. Eikerling and co-workers have stressed the importance of the bimodal pore size distribution in the CL on water management [85,86] and introduced the statistical theory of random composite media to CL modeling.

In addition to extensive applications, the agglomerate model itself has undergone significant modifications and refinements, as graphically illustrated in Fig. 14.

The effect of the agglomerate shape (cylindrical, spherical or plate-like) on the CL performance has been studied by Jain et al. [87] and Kamarajugadda and Mazumder [78]. It is concluded that the shape effect is insignificant when the agglomerate radius is smaller than 200 nm [78].

It is usually assumed that an ionomer thin-film covers the agglomerate. Rao et al. have added a water thin-film on the periphery of the ionomer thin-film to account for the liquid water effect at the agglomerate-scale [84].

Different arrangements of agglomerates, electrolyte and Pt particles inside the agglomerate have been examined. Jain et al. have proposed an agglomerate model with multi-zones, in which different electrochemical reactions and mass-transport properties are used for each zone [87]. Kamarajugadda and Mazumder have considered the overlapping effect of agglomerates with unequal radii [78]. Epting and Litster further have taken into account the size distribution of agglomerates that is measured using the nano-CT tomography, and revealed that the usual simplification of a single agglomerate size would cause an error as large as 70% [88]. Cetinbas et al. have unraveled the inconsistencies among previous agglomerate models on evaluating the effect of Pt loading [89]. Cetinbas et al. have introduced a stochastic distribution of Pt particles inside the agglomerate and obtained improved agreement with experimental data.

Eikerling and co-workers have brought paradigmatic changes to the agglomerate model [3,6,85,86,90–93]. Agglomerates are filled with electrolyte in conventional models. Eikerling and co-workers have developed water-filled agglomerate models [90–93]. As discussed in Subsection 3.2, there are sufficient evidences suggesting that ionomer clusters are unlikely to penetrate into the agglomerate, thus creating ionomer-free nanopores inside the agglomerates that are prone to being water-filled as a result of the capillary condensation effect. Therefore, water-filled agglomerate models developed by Eikerling et al. are better representations of the CL microstructure. Such a change from electrolyte-filled to water-filled agglomerates is non-trivial, because it brings into play proton transport inside the agglomerate as a key factor determining the performance of the agglomerate and the CL. In water-filled nanopores, proton transport is described using the Poisson-Nernst-Planck (PNP) equations

(11)\nabla ·\left({{\displaystyle ϵ}}_{\mathrm{S}}\nabla {{\displaystyle \varphi }}_{\mathrm{S}}\right)=-{{\displaystyle F}}_{C_{{\mathrm{H}}^{+}}},

(12){\displaystyle \frac{\partial {{\displaystyle C}}_{{\mathrm{H}}^{+}}}{\partial t}}=-\nabla ·{{\displaystyle N}}_{{\mathrm{H}}^{+}}=\mathrm{0},

(13){{\displaystyle N}}_{{\mathrm{H}}^{+}}=-{{\displaystyle D}}_{{\mathrm{H}}^{+}}\left(\nabla {{\displaystyle c}}_{{\mathrm{H}}^{+}}+{\displaystyle \frac{{{\displaystyle Fc}}_{{\mathrm{H}}^{+}}}{RT}}\nabla {{\displaystyle \varphi }}_{\mathrm{S}}\right),

where R is the gas constant, T the temperature, {{\displaystyle ϵ}}_{\mathrm{S}} the dielectric constant of water (solution), and φ_S the solution phase potential. {{\displaystyle c}}_{{\mathrm{H}}^{+}},{{\displaystyle N}}_{{\mathrm{H}}^{+}},{{\displaystyle D}}_{{\mathrm{H}}^{+}} refer to the concentration, flux and diffusion coefficient of proton, respectively.

The surface charge density on the pore wall made of Pt, σ_M, is a key property determining {{\displaystyle c}}_{{\mathrm{H}}^{+}} and φ_S. The key boundary condition at the Pt/water interface is [92,93]

(14)\overrightarrow{n}{{\displaystyle ϵ}}_{\mathrm{S}}\nabla {{\displaystyle \varphi }}_{\mathrm{S}}={{\displaystyle C}}_{\mathrm{d}\mathrm{l}}\left({{\displaystyle \varphi }}_{\mathrm{M}}-{{\displaystyle \varphi }}_{\mathrm{p}\mathrm{z}\mathrm{c}}-{{\displaystyle \varphi }}_{\mathrm{S}}+{\varphi }^{\mathrm{0}}\right),

where \overrightarrow{n} is normal to the Pt/water interface, C_dl the double-layer capacitance of the Pt/water interface, φ⁰ the reference potential, usually set as zero, and φ_pzc the potential zero charge of Pt.

In this case, the ORR rate is modified as

(15){{\displaystyle k}}_{\mathrm{O}\mathrm{R}\mathrm{R}}={\displaystyle \frac{{{\displaystyle a}}_{\mathrm{a}\mathrm{g}\mathrm{g}}{{\displaystyle j}}_{\mathrm{0}}}{\mathrm{4}F{{\displaystyle c}}_{\mathrm{r}\mathrm{e}\mathrm{f}}}}{{\displaystyle \left({\displaystyle \frac{{{\displaystyle c}}_{{\mathrm{H}}^{+}}}{{{\displaystyle c}}_{{\mathrm{H}}^{+}}^{\mathrm{r}\mathrm{e}\mathrm{f}}}}\right)}}^{{{\displaystyle \gamma }}_{{\mathrm{H}}^{+}}}\exp \left(-{\displaystyle \frac{\eta }{b}}\right),

where {{\displaystyle c}}_{{\mathrm{H}}^{+}}^{\mathrm{r}\mathrm{e}\mathrm{f}} is the reference proton concentration, and {{\displaystyle \gamma }}_{{\mathrm{H}}^{+}} the reaction order of protons in the ORR.

Using the above framework, the effects of structural parameters of the agglomerate, including R_agg, the radius of nanopores in the agglomerate, r_np (designated as a in Ref. [93]), ionomer coverage, q_ion, and electrochemical parameters, especially φ_pzc, and operating conditions on the effectiveness factor of the agglomerate, Г_agg, have been studied. Sadeghi et al. have built a water-filled agglomerate model, as shown in the rightmost image in the fourth row in Fig.14 [92,93]. Figure 15(a) shows the dimensionless proton concentration relative to the proton concentration in ionomer membranes, {{\displaystyle {{\displaystyle c}}_{{\mathrm{H}}^{+}/}c}}_{{\mathrm{H}}^{+}}^{\mathrm{r}\mathrm{e}\mathrm{f}}, on the surface of the agglomerate at q_ion = 0.5 and {{\displaystyle \varphi }}_{\mathrm{M}}-{{\displaystyle \varphi }}_{\mathrm{p}\mathrm{z}\mathrm{c}}=\mathrm{0.03}\mathrm{V}. {{\displaystyle {{\displaystyle c}}_{{\mathrm{H}}^{+}}=c}}_{{\mathrm{H}}^{+}}^{\mathrm{r}\mathrm{e}\mathrm{f}} on the surfaces is covered by an ionomer thin-film, and {{\displaystyle {{\displaystyle c}}_{{\mathrm{H}}^{+}}\mathrm{0.1}c}}_{{\mathrm{H}}^{+}}^{\mathrm{r}\mathrm{e}\mathrm{f}}on the surface is covered by a water thin-film. Figure 15(b) shows {{\displaystyle {{\displaystyle c}}_{{\mathrm{H}}^{+}/}c}}_{{\mathrm{H}}^{+}}^{\mathrm{r}\mathrm{e}\mathrm{f}} in xy-planes. Parametric studies have revealed that Г_agg strongly depends on φ_pzc, as exhibited in Fig. 15(c). In a more general sense, Г_agg strongly depends on the surface charging relation of Pt. The framework developed by Eikerling and co-workers has been adopted by other groups [94,95], constituting a new direction in CL modeling.

3.4.3 Pitfalls and challenges

Advanced characterization techniques improve the understanding of microscopic structure of conventional ionomer-impregnated CLs, as discussed in Subsection 3.2, attacking basic assumptions and the applicability of the conventional agglomerate model.

First, parameters of the agglomerate are very diverse in the literature. Table 1 lists the values of four parameters, namely, R_agg, d_ion, the volume fraction of ionomer inside the agglomerate, {{\displaystyle ϵ}}_{\mathrm{i}\mathrm{o}\mathrm{n},\mathrm{a}\mathrm{g}\mathrm{g}}, and the porosity inside the agglomerate, {{\displaystyle ϵ}}_{\mathrm{v}\mathrm{o}\mathrm{i}\mathrm{d},\mathrm{a}\mathrm{g}\mathrm{g}}, in the literature. A wide range of 20–5000 nm has been used for R_agg. Recent morphology results, for example Fig. 16, indicate that R_agg is approximately 100 nm. The value of d_ion is usually much larger than the experimentally observed value of ~6 nm. In addition, {{\displaystyle ϵ}}_{\mathrm{i}\mathrm{o}\mathrm{n},\mathrm{a}\mathrm{g}\mathrm{g}} should be very small as ionomer is unlikely to impregnate into the agglomerates. Larger values of R_agg and d_ion would result in a greater mass-transport loss. This means that there are significant transport loss that is not considered in conventional agglomerate models, necessitating the unreasonably large R_agg and d_ion, among which the transport resistance of O₂ through the ionomer thin-film can be a major source.

Second, the conventional agglomerate model deviates from the key features of the CL microstructure. The conventional agglomerate model assumes that an agglomerate is made up of a homogeneous mixture of Pt, carbon, ionomer, and void, upon which a thin ionomer film forms. However, as discussed in Subsection 3.2, the inside of an agglomerate is likely to be ionomer-free, and the outside of an agglomerate is merely partially covered by the ionomer film, as delineated in Figure 17. In this case, the proton transport inside the agglomerate, which is neglected in conventional agglomerate models, becomes an essential process, which is mediated by the surface charge on Pt and carbon as discussed in Subsection 4.4. The framework developed by Eikerling and co-workers tackles the problem of proton transport in water-filled nanopores. In doing so, distinct surface charging behaviors of Pt catalyst and carbon support should be considered. Moreover, regarding a partially ionomer-covered agglomerate, O₂ transport induces new complexities, because O₂ can diffuse through both the ionomer film and the water film.

Third, the validity of introducing an agglomerate scale is questionable. An increasing number of researches indicate that R_agg ≈ 100 nm, implying that an agglomerate consists of approximately 3–4 carbon particles (20–30 nm in diameter) in the radius direction. This means that continuum approaches cannot be used to model transport processes and electrochemical reactions inside the agglomerate. Therefore, the fundamental assumption of conventional agglomerate models is invalid. Indeed, an agglomerate model can be built consisting of several discrete carbon particles, as is the case in Refs. [92,93]. Such a practice does not allow an analytical solution at the agglomerate scale. The high computational cost and limited generality devalue this approach.

Finally, the agglomerate model is reduced to the simple macrohomogenous model when . A recent study by Hao et al. shows that differences between the simple macrohomogenous model and the agglomerate are negligible when R_agg <150 nm [107]. By accounting for the local O₂ transport resistance, Hao et al. have employed the simple macrohomogenous model to simulate the polarization curves of a low Pt loading cathode CL under a wide range of operating conditions [107].

4 Ionomer-free ultra-thin catalyst layer

4.1 Preparation methods

The ionomer-free CL, referring to PEFC electrodes that contain no ionomer, is usually prepared by sputter deposition [44,108–118]. This technique involves ejecting material from a target that is a source onto a substrate in an inert gas, such as argon, at a very low pressure of ~1 Pa, and at room temperature with a sputtering power of 10–100 W [114]. The sputtering target is usually metallic Pt. It is also flexible to prepare Pt alloy catalysts, such as Pt-Co, by using a multi-target sputtering system [110]. Both GDL [115,116] and PEM [109] can be used as the sputtering substrate. Sputter deposition technique has many variants, including direct current sputtering and radio frequency sputtering [44]. The common magnetron sputtering method refers to the process using a magnetic field to control the plasma, generated by either direct current or radio frequency excitation, to focus it on the sputtering substrate [44].

The sputtering of PEFC electrodes can be dated back to Weber et al. who sputtered a thin-layer of Pt onto a GDL in 1987 [108]. Since then, this method has gained continued efforts but is not outspread yet. The advantages are three-fold [114]. First, it is a dry process in which the Pt loading can be easily controlled by tuning the sputtering parameters such as the time and the power. Second, it allows preparation of very thin ionomer-free CLs (<1 mm), where the mass transport loss can be largely reduced. Third, it allows sputtering Pt (or its alloy) without a catalyst support or onto carbon-free support materials, which avoids carbon corrosion and benefits durability.

The largest disadvantage of sputtered PEFC electrodes is the low ECSA (), and the low Pt utilization (<10%) [114], which can be improved by sputtering the catalyst onto a nano-sized support structure, such as the PR149 whisker in NSTF electrodes of 3M [5,39]. Another possible approach is to create a porous Pt structure by adding a pore-forming material target, like alumina, and then removing the pore-forming material in a leaching process. The alternating catalyst layer structure (ACLS) of Toshiba is an example of the latter method [115,116]. Both NSTF and ACLS electrodes have a ECSA of \sim {{\displaystyle \mathrm{20}\mathrm{m}}}_{\mathrm{P}\mathrm{t}}^{\mathrm{2}}/{{\displaystyle \mathrm{g}}}_{\mathrm{P}\mathrm{t}}.

4.2 Microstructure

The foremost structural feature of ionomer-free CLs is the significantly reduced thickness, which is in the range between 10 nm and 1 mm, one to three orders smaller than that of conventional CLs. Ionomer-free CLs can also be made carbon-free. The microstructure of ionomer-free CLs is further divided into agglomerates [111], thin-film [111], porous leafs [115,116], and supported thin-film [5,110], as shown in Fig. 18. At an ultra-low Pt loading (<10 mg_Pt/cm²), the sputtered Pt (or its alloy) CLs are in the form of agglomerates (~10 nm). A Pt thin-film forms (with cracks) when increasing the Pt loading above a critical Pt loading which increases with the roughness of the sputtering substrate. The ACLS electrode of Toshiba consists of Pt leafs with pore layers in between [115,116]. In the NSTF electrode of 3M, a polycrystalline thin-film is deposited onto the lath-shaped whiskers [5].

4.3 Functional properties

In comparison with MEAs using conventional ionomer-impregnated CLs, the MEAs based on the NSTF technology have demonstrated several advantages including higher specific activity, high specific rated power, and improved durability under start-stop and voltage cycling conditions [5]. Due to its ultra-thin structure and absence of a resistive ionomer thin-film (see Sub-subsection 3.3.2), the O₂ transport resistance in ionomer-free CLs is significantly reduced. However, the proton conductivity becomes a key concern due to the absence of the ionomer. Another formidable challenge is water management. The water production rate per unit volume in UTCLs is much higher than that in conventional CLs. As a result, UTCLs are more susceptible to flooding behaviors, leading to reduced performance especially under wet and low temperature conditions. This subsubsection discusses recent progress made on these two issues.

4.3.1 Proton conduction

In 1978, Stucki and Menth observed for the first time that the hydrogen adsorption charge, {{\displaystyle Q}}_{{\mathrm{H}}_{\mathrm{a}\mathrm{d}}}, is identical for a Pt screen on a Nafion^® membrane and that immersed in a 1.0 mol/L H₂SO₄ [119]. It is worth noting that only a fraction of the screen was in contact with the Nafion^® membrane. This gave rise to the question how current generation takes place on and protons move to the rest part of the screen that is remote from the membrane.

In 1985, McBreen confirmed the observation of Stucki and Menth and designed ingenious experiments to further scrutinize the mechanisms [120]. First, {{\displaystyle Q}}_{{\mathrm{H}}_{\mathrm{a}\mathrm{d}}} and the charge corresponding to oxide reduction, Q_Ox, were doubled when two Pt screens were pressed onto a membrane, as shown in Fig. 19(a). Second, there was significant distortion of the cyclic voltammogram in both hydrogen adsorption/desorption and oxide formation/reduction regions when a graphite felt was interposed between two Pt electrodes and a membrane, as shown in Fig. 19(a), indicating a highly resistive electrolytic path. These results profoundly demonstrate that Pt electrodes that are not in direct contact with a Nafion^® membrane, viz. ionomer-free Pt electrodes, can be electrochemically active. Two plausible mechanisms have been speculated. In the first mechanism, H_ad is formed at the Pt/electrolyte interface and then diffuses along the Pt wall, as shown in Fig. 19(c). This mechanism is supported by a considerable reduction in {{\displaystyle Q}}_{{\mathrm{H}}_{\mathrm{a}\mathrm{d}}} and a noticeable distortion of the cyclic voltammogram when a graphite felt is interposed between two Pt electrodes and a membrane, considering the fact that H_ad is believed to be nearly immobile on carbon surface. The second mechanism involves transport of protons in the water film adsorbed on the hydrophilic Pt wall. This mechanism is endorsed by the finding that the nitrogen gas humidity should be sufficiently high in order to obtain reproducible results. Liquid water is a very good transport media for protons, mainly via the Grotthuss mechanism. Moreover, the second mechanism can also explain why the interposed graphite felt distorts the cyclic voltammogram and decreases {{\displaystyle Q}}_{{\mathrm{H}}_{\mathrm{a}\mathrm{d}}}, considering the fact that water can hardly adsorbed onto the hydrophobic carbon surface.

In 1998, Cha and co-workers quantitatively calculated the diffusion coefficient of H_ad on the Pt wall [121]. It was found that with v being the voltage scan rate when v was sufficiently slow (<5 mV/s in their study), and {{\displaystyle Q}}_{{\mathrm{H}}_{\mathrm{a}\mathrm{d}}}\propto v^{\mathrm{0.5}} with increasing v above 5 mV/s. The relation {{\displaystyle Q}}_{{\mathrm{H}}_{\mathrm{a}\mathrm{d}}}\propto v^{\mathrm{0.5}} could be described by a semi-infinite diffusion process. {{\displaystyle D}}_{{\mathrm{H}}_{\mathrm{a}\mathrm{d}}} was estimated to be \sim \mathrm{5}\times {\mathrm{10}}^{-\mathrm{8}}{\mathrm{c}\mathrm{m}}^{\mathrm{2}}/\mathrm{s} by fitting the {{\displaystyle Q}}_{{\mathrm{H}}_{\mathrm{a}\mathrm{d}}}\propto v^{\mathrm{0.5}} curve. The work of Cha and co-workers totally excluded the transport of protons in the adsorbed water film. The effect of relative humidity was not studied.

In 2003, Scherer and co-workers designed an ingenious model system to study the catalyst utilization in PEFC electrodes [122]. Two types of PEFC electrode, namely the Pt black and Pt/C electrode, were compared. A continuous Pt thin film coated onto a comb-shaped glassy carbon substrate (with a gap width of 30 mm and a gap depth of 100–150 mm) emulates the Pt black electrode, as shown in Fig. 19(b). By removing the Pt from channel walls, an interrupted Pt film is obtained, which is corresponding to the Pt/C catalyst, as shown in Fig. 19(b). It is observed that all Pt surfaces are electrochemically active in the continuous type electrode, while only the Pt surface in contact with the membrane is electrochemically active in the interrupted type electrode. This observation is consistent with the results of McBreen that the carbon surface is highly resistive. Scherer and co-workers highlighted the importance of impregnating the Pt/C catalyst with ionomer. A subsequent study from Jiang and Yi termed the diffusion of H_ad on the Pt wall as “H_ad spillover” [123].

In 2010, Osaka and co-workers reported on the proton conductivity of a water-filled Pt thin-film (with a thickness of 210±20 nm) with mesopores (with a radius of 5 nm) placed onto a Nafion membrane [124]. Compared to a non-porous flat Pt electrode, the water-filled porous Pt electrode exhibited a transmission line behavior, i.e., a ~45° line arose in the high frequency range of the EIS. The model fit gave an ionic resistivity of 0.18 MW·cm at 0.5 V (RHE). Note that this ionic resistivity is two orders smaller than that of pure water (18 MW·cm) and two to three orders larger than that of conventional PEFC electrodes (0.1– 1.0 kW·cm). Therefore, it is argued that Pt particles in ionomer-free zones in conventional CLs are ineffective for ORR. It is important to notice that EIS measurement was conducted at only one potential (0.5 V). The proton conductivity of water-filled Pt nanopore shall be a function of the electrode potential. Given that the coverage of H_ad approaches zero at 0.5 V, the mechanism of H_ad diffusion is ruled out and the mechanism of proton transport is singled out for this case. A two-order increase of proton conductivity for water-filled Pt nanopore electrodes implies that proton concentration therein is much higher than that in pure water. Underlying causes were unexplained.

In 2011, Thompson and Baker measured the proton conductivity of ionomer-free Pt electrodes at different RHs [125]. Three kinds of Pt black electrode were compared: unbound, PTFE-bound, and Nafion ionomer bounded. First, all Pt electrodes displayed transmission line behaviors under the condition of 0.16 V, H₂/N₂, and 35°C. Second, impedance curves at different RHs collapsed into a single curve after normalization with respect to the proton conduction resistance, R_proton, pointing to a single controlling mechanism of proton transport in each electrode. Third, the activation energy of PTFE-bound Pt black electrodes was relatively constant, in the range between 10 and 15 kJ/mol, at various RHs, further supporting the judgement that there was a single controlling mechanism of proton transport in ionomer-free Pt electrodes. Figure 20(a) shows the proton conductivity as a function of RH for PTFE-bound Pt black electrodes at three temperatures, PFSA membranes, conventional PEFC electrodes, and distilled water. Proton conductivities of PTFE-bounded electrodes decreases at lower RHs, and are two to three orders higher than that of distilled water, suggesting that interactions between the Pt surface and the adsorbed water film enhance proton conduction. Thompson and Baker concluded that adsorbed water on the Pt wall contributed to the proton conduction of ionomer-free Pt electrodes. However, how Pt interacts with adsorbed water and how these interactions dictate proton conduction remain unknown. Note that the potential dependence of proton conductivity was not investigated.

In 2013, An and Litster conducted in-situ measurements of proton conductivity for ionomer-free Pt black electrodes using a direct current method based on the microstructured electrode scaffold technique [126]. Potential dependence of proton conductivity was attempted. However, as shown in Fig. 20(b), large measurement errors mask the potential dependence. The activation energy of proton conduction in ionomer-free Pt black electrodes was estimated and claimed to be similar to that of liquid-saturated Nafion membranes. Therefore, it was concluded that the proton conduction mechanism in ionomer-free Pt black electrodes was similar to that in bulk Nafion membranes.

In summary, there are sufficient evidences for the argument that ionomer-free Pt surfaces are electrochemically active if they are sufficiently humidified, and the proton conductivity of water-filled Pt nanopore is several orders higher than that of pure water. However, it is still unclear which mechanism, H_ad diffusion on Pt wall or H⁺ transport through water, dominates. Strong RH dependence of the proton conductivity indicates that H⁺ transport through water shall play a significant role. Precise measurement of proton conductivity as a function of electrode potential is important for understanding the proton conduction mechanism. Recent progress on the key question why Pt enhances the proton conductivity is reviewed in Subsection 4.4.

4.3.2 Water management

In a report from 3M in 2007, Debe and Steinbach reported that flooding behavior of ultra-thin NSTF electrodes was noticed at the early stage of development in 1990s [127]. Initial attempts to mitigate this issue involved modifications of the GDL, because experimental data pointed out that the limiting current density at the onset of flooding was a strong function of GDL characteristics [127]. An effective water vapor diffusion coefficient of the GDL was defined and found to be directly correlated with the flooding caused limiting current density. In addition, their analysis implied that water transport in the third generation GDL, which significantly improved cell performance at 65°C as compared with the first generation GDL, involved both vapor and liquid paths [127].

In a subsequent study, cell performance was evaluated at 30°C. At such a highly liquid-water-prone temperature, MEAs with NSTF electrodes produced much lower steady state limiting current densities as compared with conventional CLs [128]. Anode water removal was proposed as a new paradigm of water management for UTCLs. The basic idea is to reduce the anode pressure from the typical 1 atm or higher to sub-atmospheric pressures as low as 20 kPa, resulting in several-fold increase in the flooding-caused limiting current density [128]. Further gains can be achieved through optimizing the anode GDL, as demonstrated in a following report in 2011 [129]. Debe et al. summarized strategies of water management at low temperatures for NSTF electrodes in a status report published in 2011 [130], and later in a tutorial published in 2013 [5].

A team at GM revealed two conflicting regimes of NSTF electrode [131–134]. Under dry conditions, the performance of NSTF electrodes is limited by low proton conductivity, termed as proton conduction regime [132]. Under wet conditions, the performance of NSTF electrodes is impaired by flooding, termed as flooding regime [131]. This team conceived several approaches to enhance the robustness of NSTF electrodes under wide-ranging operating conditions [131]. To mitigate the low proton conductivity under dry conditions, an ionomer thin-film was coated onto the NSTF surface. However, this approach will induce significant mass transport loss at high current densities due to the oxygen transport resistance through the ionomer thin-film as discussed in Sub-subsection 3.3.2. An alternative approach was to coat silica nanoparticles onto the NSTF surface, so as to maintain thin water films and improve electrode proton conductivity under dry conditions. The third approach was to add an interlay consisting of ionomer-impregnated carbon or ionomer-impregnated Pt/C sandwiched between the NSTF electrode and the PEM or the GDL, leading to increased water storage capacity and improved water removal capability [131,133].

Much less understood are transient responses of NSTF-based MEAs. Kongkanand and Sinha reported that the 0.3 mm ultra-thin NSTF electrode could exhibit negative cell voltages during wet up-transients, namely instantaneously increasing the current from 0.02 to 1.0 A/cm² at 80°C and RH>80% [134]. A non-isothermal transient model elucidates that the ultra-thin NSTF electrode has a limited water storage capacity and thereby easily gets completely flooded with liquid water, resulting in less-than-zero cell voltages during wet up-transients. Failed wet up-transients can be prevented by increasing the electrode thickness, viz., increasing the water storage capacity, or reducing the membrane thickness, viz., increasing water removal capability [134].

4.4 Modeling

The modeling of ionomer-free CLs features the description of proton conduction. In models for conventional CLs (see Subsection 3.4), Ohmic law is used to describe potential distribution in the electrolyte phase, φ_S. This treatment implicitly assume that the proton concentration, {{\displaystyle c}}_{{\mathrm{H}}^{+}}, is uniform across the CL. Therefore proton transport is dominated by the migration mechanism. This assumption is valid since there is ionomer in the conventional CLs and {{\displaystyle c}}_{{\mathrm{H}}^{+}} therein is approximately equal to {{\displaystyle c}}_{{\mathrm{H}}^{+}} in the Nafion bulk. In contrast, {{\displaystyle c}}_{{\mathrm{H}}^{+}} in ionomer-free CLs is unknown.

In 2004, Wang et al. modeled a water-filled ultra-thin CL using a simple macrohomogenous approach, in which {{\displaystyle c}}_{{\mathrm{H}}^{+}} and φ_S were solved from the PNP equations, Eqs.(11)–(13) [90]. The simple macrohomogenous approach assumed a well-intermixed phase composed of water-filled pores and catalyst. Given this assumption, interactions between protons in the water-filled pore and surface charges on the pore wall were neglected.

In 2011, Chan and Eikerling developed a pore-scale model for ionomer-free CLs in PEFCs [135]. In this model, electrostatic interactions between the charged Pt surface and protons in the water-filled pore were taken into account. Simply put, global electro-neutrality warrants that the magnitude of total charge carried by protons in the water-filled pore should be commensurate with that of the surface charge on the Pt wall. Therefore, the description of surface charge on the Pt wall, σ_M, is crucial. Chan and Eikerling employed a linear relation

(16){{\displaystyle \sigma }}_{\mathrm{M}}={{\displaystyle C}}_{\mathrm{d}\mathrm{l}}\left({{\displaystyle \varphi }}_{\mathrm{M}}-{{\displaystyle \varphi }}_{\mathrm{p}\mathrm{z}\mathrm{c}}\right),

by assuming that C_dl is independent of φ_M.

Model parameterization gave φ_pzc in the range of [0.45,0.9] V [135]. Identified values of φ_pzc are much higher than that measured by the CO displacement experiments which is around 0.3 V (see a recent review [136]). This discrepancy encouraged Zenyuk and Litster to incorporate an alkaline ORR mechanism into models for ionomer-free UTCLs [137]. In Zenyuk and Litster’s model, the surface charge on the Pt wall was positive at potentials higher than φ_pzc which was equal to 0.26 V (RHE). Positive surface charges on Pt wall were balanced by excess hydronium ions in the nanopore. In this scenario, the alkaline ORR was believed to be the dominating current generation mechanism.

The discrepancies between Refs. [135,137] originate from different values of φ_pzc in Eq. (16). Is it sufficient to use the linear charging relation parameterized with φ_pzc to describe the surface charging behavior of Pt?

Frumkin and Petrii measured surface excesses of sodium cations and sulfonate anions at a Pt electrode as a function of the electrode potential [138]. Regarding a Pt electrode in 0.002 mol/L Na₂SO₄ at pH= 6, surface excesses of Na²⁺ were detected at potentials of 0–0.5 V and above 0.8 V, meaning that σ_M was negative in these two potential ranges. In the medium potential range of 0.5–0.8 V, surface excesses of {{\displaystyle \mathrm{S}\mathrm{O}}}_{\mathrm{4}}^{\mathrm{2}-}existed at the Pt electrode, suggesting that σ_M was positive. Garcia-Araez et al. employed the laser pulsed method to study the potential-dependent water dipole orientation [139]. Surface charging, that is changes of σ_M as a function of the potential, is much faster than charge transfer processes, namely hydrogen adsorption and Pt oxidation reactions. By virtue of this time difference, the transient potential response on the scale of ~1 ms induced by a temperature jump can be attributed to the temperature coefficient of the potential difference across the electrochemical double layer. This temperature coefficient is related to the change in the orientation of interfacial water dipoles, which is further dictated by the change in the sign of σ_M. In a word, the sign of the transient potential response reflects the sign of σ_M. In terms of a Pt(111) electrode in 0.1mol/L KClO₄ +1 mmol/L HClO₄ electrolyte solution, negative potential responses were measured at potentials of 0.1–0.45V and above 1.05 V. On the contrary, transient potential responses were positive in the potential range of 0.48–0.95V. This negative-positive-negative evolution of transient potential responses implies a negative-positive-negative transition of σ_M.

The above two experiments defy the linear relation in Eq.(16) as a sufficient description of surface charging behavior, calling for an overhaul of description of the metal-electrolyte interface. Recently, a refined structural model of electrified interfaces has been developed accounting for the formation of surface oxide and the orientational ordering of interfacial water molecules, as shown in Fig. 21 [140]. Based on this refined structure, the metal surface charge density can be solved in a self-consistent manner. The analytical solution of the model reveals a non-monotonic charging behavior. The planar extended Pt surface may (depending on the solution pH) exhibits a negative charge in a low potential of 0–0.3V, a positive charge in an intermediate potential of 0.3–0.7 V and a negative charge in a high potential range due to surface oxide dipoles (0.7–1.0 V), as shown in Fig. 21. This non-monotonic behavior is consistent with experimental work of Frumkin and Petrii [138] and Garcia-Araez et al. [139].

In a subsequent paper, a water-filled nanopore with walls made of Pt was modeled, sandwiched between a GDL and a PEM, as shown in Fig. 22(a) [141]. The refined structural model of the electrified interface was incorporated into the nanopore model. The surface charging relation of the curved Pt surface was solved self-consistently. Figure 22(b) shows that Pt wall in the nanopore is always negatively charged. Moreover, the surface charging pattern of a Pt nanopore is distinct with that of an extended Pt plane, implying that the surface charging behavior depends on the geometry of the electrode. Consequently, a “universal” φ_pzc cannot be indiscriminately used to describe the surface charging behavior of Pt in various electrochemical systems. The surface charging relation must be solved self-consistently.

Due to the negative surface charge on the Pt wall, protons accumulate in the water-filled nanopore, resulting in a remarkably high proton concentration and thereby an ionic conductivity that is three-six orders higher than that of the pure water, as shown in Fig.22(c). By this point, the puzzle how protons transport in an ionomer-free Pt surface, as discussed in Sub-subsection 4.3.1, is resolved. The key point is that the negative surface charge on the Pt wall attracts and draws protons from the PEM through electrostatic interactions. Figure 22(d) shows the model fitting of the polarization curve of a NSTF MEA with model parameters found in Ref. [141].

One important message conveyed in Refs. [140,141] is that electrostatic effects at the scale of the Debye length (1–100 nm) are essential in the understanding of Pt electrocatalysis and the engineering of next-generation catalyst layers, in addition to geometric and electronic factors at the atomic scale (0.1–1 nm) and mass transport at the electrode scale (>0.1 mm). Figure 23 illustrates a hierarchical modeling framework to rationalize the electrocatalytic phenomena in complex electrodes for polymer electrolyte fuel cells and electrolyzers, consisting of DFT studies of reaction mechanisms and pathways at the lower end and porous electrode theory rationalizing the interplay of mass transport and electrochemical reaction at the upper end of the spectrum of scales. The electrostatic module is an important part in this framework.

The substrate effects on the ORR activity have usually been neglected in previous studies. A recent theoretical study have employed the modeling framework in Fig. 23 to explain the particle proximity effect in nanoparticle electrocatalysis [142]. A key point is that surface charging relations of Pt and the carbon support are distinct. To be specific, the Pt surface with negative charge attracts protons, while the carbon surface with positive charge repels protons at the ORR-relevant potentials (such as 0.9 V). By describing electrostatic reaction conditions around Pt nanoparticles, the model reveals that the surface specific activity of ORR decreases when the interparticle spacing of Pt particles increases due to the intensified protonic depletion effect of the carbon support. This study indicates that the substrate effect may be a significant factor to be considered in characterization, modeling, and optimization of PEFC catalyst layers.

It is safe to say that modeling of ionomer-free UTCLs is still in its infancy. The above models treat a single fully water-filled nanopore, and posit that protons transport via structural diffusion like they do in bulk water. It should be pointed out that not all of the nanopores in a UTCL are fully water-filled. In addition, a nanopore may not be fully water-filled. Instead, the nanopore can be wetted by an adsorbed water film, namely a partially water-filled state. Scaling from one nanopore to a CL requires an exquisite treatment of water management of the UTCL. The pore size distribution and the wettability of electrode surface are two key structural properties determining the number of water-filled nanopores. Several models on water management of UTCLs have been published [143–146]. In previous studies on ionomer-free UTCLs, low proton conductivity at the lower end of the spectrum of water content and flooding behavior at the upper end of the spectrum of water content have been investigated separately. Future efforts are needed to compile these two pressing concerns into a self-consistent model framework.

5 Concluding remarks

Changing roles of ionomer represent an interesting phenomenon in the evolution of PEFC catalyst layers. In the late 1980s, ionomer was introduced into CLs and drastically improved the catalyst utilization. Almost 30 years later, ionomer was revealed to cause significant mass transport loss at high current densities; and there are vigorous activities to improve gas transport properties of ionomer or even to eliminate ionomer from the CL, such as ionomer-free UTCLs. Conventional ionomer-impregnated CLs possess the merit of high operational robustness in wide-ranging and dynamic conditions. However, they suffer from the impeded mass transport that is related to the ionomer. In contrast, ionomer-free UTCLs usually have ordered paths for mass transport while its extreme thinness causes pressing concerns about the operational robustness. Delicate balance of water storage and evaporation capability (related to robustness, and dictated by the CL thickness and the pore size), mass transport (related to effectiveness and robustness, and dictated by the CL thickness, the tortuosity and the pore size), and catalyst dispersion (related to utilization and effectiveness, and dictated by the catalyst nanoparticle size distribution and microstructure of the support) is needed to achieve an optimal CL.

The ionomer-free zones inside the Pt/C agglomerates may contain a significant fraction of total Pt particles in conventional CLs, depending on the microstructure of carbon particles. Emerging UTCLs are altogether ionomer-free. Utilization and effectiveness of Pt catalysts in these ionomer-free regions are key and common concerns in the development of low Pt-loading PEFC CLs. Therefore, characterization and modeling of ionomer-free regions are one of the high-impact research directions in the realm of PEFC.

One way to characterize the activity of ionomer-free Pt nanoparticles in conventional ionomer-impregnated CLs is to examine the relation between the ionomer content/coverage and the ORR activity. Furuya et al. have found that the ORR activity decreases with increasing the ionomer coverage [147]. Modestov et al. have studied the dependence of ECSA and ionic resistance on the ionomer content [148]. However, Pt/C catalysts and conventional ionomer-impregnated CLs are essentially random composites, causing tremendous difficulties in precise characterization and modelling. Model catalyst and electrode with well-controlled distribution of Pt particles, ionomer and carbon particles are desirable.

One of the key issues in the modeling of ionomer-free regions is to understand the proton conduction mechanism. Progresses over the past three decades have revealed that protons can transport through the adsorbed water film on Pt and the proton concentration is dictated by electrostatic interactions between the charged Pt electrode and protons in water. Regarding a water-filled Pt nanopore, a mean-field model has indicated that the Pt wall is always negatively charged, underpinning a remarkably high proton conductivity and superior electrocatalysis performance of ionomer-free UTCLs [141]. In this direction, it is interesting to go beyond mean-field approximation and turn to first-principles calculations (see a recent multi-scale modeling work by Nouri-Khorasani et al. [150]). A recent comprehensive perspective on what first-principles theory and modeling can do for the understanding of fuel cell electrocatalysis has been published [151]. In addition, refined models that account for electrostatic interactions can be applied for ionomer-impregnated CLs, such as in a recent study by Muzaffar et al. [152].

EIS is not only a powerful tool to characterize transport and reactions in CLs, but also a useful tool in model verification. The above theoretical insights regarding proton conduction in ionomer-free regions are lacking of strong experimental evidences. There have been several attempts to measure the EIS of Pt black electrodes or NSTF electrodes [125,126]. However, the experimental error was are too large to obtain meaningful results. New methods are needed to accurately measure the EIS of ionomer-free Pt electrodes. In addition, previous EIS measurements have usually been conducted at a certain potential in the double-layer charging region [125]. It is necessary to extend the EIS to the potential range of the oxide formation in order to clarify the proton conduction mechanism. In doing so, it is important to eliminate the influence of pseudo-capacitances.

Staying ionomer-impregnated or going to ionomer-free, the following questions are critical:

1) Is ionomer essential to high catalyst utilization? The answer is “No”. Pt catalysts can be electrochemically assessable if they are wetted by water.

2) Is ionomer essential to high catalyst effectiveness? The answer is also “No”. In contrast, ionomer, a condensed matter, impedes oxygen transport and poisons Pt catalysts.

3) Is ionomer essential to high operational robustness? It is true that ionomer can act as a water reservoir in the CL and hence beneficial for a wide-ranging operation window. Nevertheless, UTCLs is still in its early stage of development. Extending operation window should be a major research issue. The absence of ionomer is not the root cause of unsatisfactory operational robustness of ionomer-free UTCLs. The problem is caused by the extreme thinness of UTCLs, that is, a limited water storage capacity, rendering that UTCLs easily flood under wet and cold conditions. Remedy approaches have been reported. For example, flooding can be mitigated by increasing the water storage capacity via increasing the electrode thickness, or by increasing water removal capability via using a thinner membrane thickness, reducing the anode pressure, and modifying GDL properties.

4) Is it possible to find an ionomer that has negligible impediment on mass transfer and no poisoning effect? Breakthrough in material, though occurring sporadically, can change the game.

References

[1] Litster S, Mclean G F. PEM fuel cell electrodes. Journal of Power Sources, 2004, 130(1–2): 61–76

[2] Weber A Z, Newman J. Modelling transport in polymer-electrolyte fuel cells. Chemical Reviews, 2004, 104(10): 4679–4726

[3] Eikerling M H, Malek K, Wang Q. Catalyst Layer Modelling: Structure, Properties and Performance, PEM Fuel Cell Electrocatalysts and Catalyst Layers. London: Springer, 2008: 381–446

[4] Debe M K. Electrocatalyst approaches and challenges for automotive fuel cells. Nature, 2012, 486(7401): 43–51

[5] Debe M K. Tutorial on the fundamental characteristics and practical properties of nanostructured thin film (NSTF) catalysts. Journal of the Electrochemical Society, 2013, 160(6): F522–F534

[6] Eikerling M, Kulikovsky A. Polymer Electrolyte Fuel Cells: Physical Principles of Materials and Operation. Boca Raton: CRC Press, 2014

[7] Weber A Z, Borup R L, Darling R M, Das P K, Dursch T J, Gu W, Harvey D, Kusoglu A, Litster S, Mench M M, Mukundan R, Owejan J P, Pharoah J G, Secanell M, Zenyuk I V. A critical review of modelling transport phenomena in polymer-electrolyte fuel cells. Journal of the Electrochemical Society, 2014, 161(12): F1254–F1299

[8] Holdcroft S. Fuel cell catalyst layer: a polymer science perspective. Chemistry of Materials, 2014, 26(1): 381–393

[9] Kongkanand A, Mathias M F. The priority and challenge of high-power performance of low-platinum proton-exchange membrane fuel cells. Journal of Physical Chemistry Letters, 2016, 7(7): 1127–1137

[10] Zamel N. The catalyst layer and its dimensionality — a look into its ingredients and how to characterize their effects. Journal of Power Sources, 2016, 309: 141–159

[11] Uchida M, Park Y C, Kakinuma K, Yano H, Tryk D A, Kamino T, Uchida H, Watanabe M. Effect of the state of distribution of supported Pt nanoparticles on effective Pt utilization in polymer electrolyte fuel cells. Physical Chemistry Chemical Physics, 2013, 15(27): 11236–11247

[12] Park Y C, Tokiwa H, Kakinuma K, Watanabe M, Uchida M. Effects of carbon supports on Pt distribution, ionomer coverage and cathode performance for polymer electrolyte fuel cells. Journal of Power Sources, 2016, 315: 179–191

[13] Bard A J, Inzelt G, Scholz F. Electrochemical Dictionary. London: Springer, 2008, 286

[14] Bard A J, Inzelt G, Scholz F. Electrochemical Dictionary. London: Springer, 2008, 523

[15] Perry M L, Fuller T F. A historical perspective of fuel cell technology in the 20th century. Journal of the Electrochemical Society, 2002, 149(7): S59–S67

[16] Ostwald F W. Scientific electrochemistry of today and technical electrochemistr of tomorrow. Z. fürElektrotechnik und Elektrochemie, 1894, 1: 122–125

[17] Office of Energy Efficiency and Renewable Energy. Fuel cell technologies office multi-year research, development, and demonstration plan. 2016, http://energy.gov/eere/fuelcells/downloads/fuel-cell-technologies-office-multi-year-research-development-and-22

[18] Mond L, Langer C. A new form of gas battery. Proceedings of the Royal Society of London, 46:296–304

[19] Taitelbaum I, Tobler J. StudienüberBrennstoffketten. ZeitschriftFürElektrochemie Und AngewandtePhysikalische Chemie, 1910, 16(9): 286–300

[20] Rezaei Niya S M, Hoorfar M. Study of proton exchange membrane fuel cells using electrochemical impedance spectroscopy technique – a review. Journal of Power Sources, 2013, 240(240): 281–293

[21] Stephens I E L, Bondarenko A S, Grønbjerg U, Rossmeisl J, Chorkendorff I. Understanding the electrocatalysis of oxygen reduction on platinum and its alloys. Energy & Environmental Science, 2012, 5(5): 6744–6762

[22] Schoenbein C F. On the voltaic polarization of certain solid and fluid substances. Philosophical Magazine, 1838, 14(85): 43–45

[23] Grove W R. On voltaic series and the combination of gases by platinum. Philosophical Magazine, 1839, 14(86): 127–130

[24] Grove W R. On a gaseous voltaic battery. Journal of The Franklin Institute-engineering and Applied Mathematics, 1843, 35(4): 277–280

[25] Grubb J W T. Fuel cell. U.S. Patent 2, 913, 511. 1959-11-17

[26] Grubb W T, Niedrach L W. Batteries with solid ion-exchange membrane electrolytes II. Low-temperature hydrogen-oxygen fuel cells. Journal of the Electrochemical Society, 1960, 107(2): 131–135

[27] Niedrach L W, Alford H R. A new high-performance fuel cell employing conducting-porous-teflon electrodes and liquid electrolytes. Journal of the Electrochemical Society, 1965, 112(2): 117–124

[28] Appleby A J, Yeager E B. Solid polymer electrolyte fuel cells (SPEFCs). Energy, 1986, 11(1–2): 137–152

[29] Waters R F. Fuel cell electrode and process for its manufacture: U.S. Patent 3, 405, 007. 1968–10–8

[30] Petrow H G, Allen R J. Catalytic platinum metal particles on a substrate and method of preparing the catalyst. U.S. Patent 3, 992, 331. 1976–11–16

[31] Jalan V M, Bushnell C L. Method for producing highly dispersed catalytic platinum. U.S. Patent 4, 136, 059. 1979–1–23

[32] George M, Kemp F. Sequential catalyzation of fuel cell supported platinum catalyst. U.S. Patent 3, 857, 737. 1974–12–31

[33] Martin C R, Rhoades T A, Ferguson J A. Dissolution of perfluorinated ion-containing polymers. Analytical Chemistry, 1982, 54(9): 1639–1641

[34] Raistrick I D. Modified gas diffusion electrode for proton exchange membrane fuel cells. Proceedings of the symposium on diaphragms, separation, and ion-exchange membranes. Ponnington (NJ): Electrochemical Society. 1986: 6–7

[35] Wilson M S, Gottesfeld S. Thin-film catalyst layers for polymer electrolyte fuel cell electrodes. Journal of Applied Electrochemistry, 1992, 22(1): 1–7

[36] Wilson M S, Gottesfeld S. High performance catalyzed membranes of ultra-low Pt loadings for polymer electrolyte fuel cells. Journal of the Electrochemical Society, 1992, 139(2): L28–L30

[37] Kocha S S. Principles of MEA preparation. Handbook of Fuel Cells. New Jersey: Wiley 2010

[38] W.L. Gore and Associates, Inc. PRIMEA MEAs for Transportation, 2003

[39] Debe M K, Schmoeckel A K, Vernstrom G D, Atanasoski R. High voltage stability of nanostructured thin film catalysts for PEM fuel cells. Journal of Power Sources, 2006, 161(2): 1002–1011

[40] Fuel Cell System Cost – 2015, DOE Hydrogen and Fuel Cells Program Record. 2015

[41] Lee M, Uchida M, Yano H, Tryk D A, Uchida H, Watanabe M. New evaluation method for the effectiveness of platinum/carbon electrocatalysts under operating conditions. Electrochimica Acta, 2010, 55(28): 8504–8512

[42] Borup R L, Meyers J, Pivovar B, Kim Y S, Mukundan R, Garland N, Myers D, Wilson M, Garzon F, Wood D, Zelenay P, More K, Stroh K, Zawodzinski T, Boncella J, McGrath J E, Inaba M, Miyatake K, Hori M, Ota K, Ogumi Z, Miyata S, Nishikata A, Siroma Z, Uchimoto Y, Yasuda K, Kimijima K, Iwashita N. Scientific aspects of polymer electrolyte fuel cell durability and degradation. Chemical Reviews, 2007, 107(10): 3904–3951

[43] Jahnke T, Futter G, Latz A, Malkow T, Papakonstantinou G, Tsotridis G, Schott P, Gérard M, Quinaud M, Quiroga M, Franco A A, Malek K, Calle-Vallejo F, Ferreira de Morais R, Kerber T, Sautet P, Loffreda D, Strahl S, Serra M, Polverino P, Pianese C, Mayur M, Bessler W G, Kompis C. Performance and degradation of proton exchange membrane fuel cells: state of the art in modeling from atomistic to system scale. Journal of Power Sources, 2016, 304: 207–233

[44] Strong A, Thornberry C, Beattie S, Chen R, Coles S R. Depositing catalyst layers in polymer electrolyte membrane fuel cells: a review. Journal of Fuel Cell Science and Technology, 2015, 12(6): 064001

[45] Mauritz K A, Moore R B. State of Understanding of Nafion. Chemical Reviews, 2004, 104(10): 4535–4586

[46] Wee J, Lee K, Kim S H. Fabrication methods for low-Pt-loading electrocatalysts in proton exchange membrane fuel cell systems. Journal of Power Sources, 2007, 165(2): 667–677

[47] Soboleva T, Zhao X, Malek K, Xie Z, Navessin T, Holdcroft S. On the micro-, meso-, and macroporous structures of polymer electrolyte membrane fuel cell catalyst layers. ACS Applied Materials & Interfaces, 2010, 2(2): 375–384

[48] Soboleva T, Malek K, Xie Z, Navessin T, Holdcroft S. PEMFC catalyst layers: the role of micropores and mesopores on water sorption and fuel cell activity. ACS Applied Materials & Interfaces, 2011, 3(6): 1827–1837

[49] Malek K, Eikerling M, Wang Q, Navessin T, Liu Z. Self-organization in catalyst layers of polymer electrolyte fuel cells. Journal of Physical Chemistry C, 2007, 111(36): 13627–13634

[50] Malek K, Mashio T, Eikerling M. Microstructure of catalyst layers in PEM fuel cells redefined: a computational approach. Electrocatalysis (New York), 2011, 2(2): 141–157

[51] Lopez-Haro M, Guétaz L, Printemps T, Morin A, Escribano S, Jouneau P-H, Bayle-Guillemaud P, Chandezon F, Gebel G. Three-dimensional analysis of Nafion layers in fuel cell electrodes. Nature Communications, 2014

[52] Guetaz L, Lopez-Haro M, Escribano S, Morin A, Gebel G, Cullen D A, More K L, Borup R L. Catalyst-Layer Ionomer Imaging of Fuel Cells. ECS Transactions, 2015, 69(17): 455–464

[53] Boyer C, Gamburzev S, Velev O, Srinivasan S, Appleby A J. Measurements of proton conductivity in the active layer of PEM fuel cell gas diffusion electrodes. Electrochimica Acta, 1998, 43(24): 3703–3709

[54] Makharia R, Mathias M F, Baker D R. Measurement of catalyst layer electrolyte resistance in PEFCs using electrochemical impedance spectroscopy. Journal of the Electrochemical Society, 2005, 152(5): A970–A977

[55] Liu Y, Murphy M, Baker D R, Gu W, Ji C, Jorne J, Gasteiger H A. Determination of electrode sheet resistance in cathode catalyst layer by AC impedance. ECS Transactions, 2007, 11(1): 473– 484

[56] Liu Y, Ji C, Gu W, Baker D R, Jorne J, Gasteiger H A. Proton conduction in PEM fuel cell cathodes: effects of electrode thickness and ionomer equivalent weight. Journal of the Electrochemical Society, 2010, 157(8): B1154–B1162

[57] Iden H, Ohma A, Shinohara K. Analysis of proton transport in pseudo catalyst layers. Journal of the Electrochemical Society, 2009, 156(9): B1078–B1084

[58] Hess K C, Epting W K, Litster S. Spatially resolved, in situ potential measurements through porous electrodes as applied to fuel cells. Analytical Chemistry, 2011, 83(24): 9492–9498

[59] Karan K. Determination of CL Ionomer Conductivity. ECS Transactions, 2013, 50(2): 395–403

[60] Paul D K, Fraser A, Karan K. Towards the understanding of proton conduction mechanism in PEMFC catalyst layer: conductivity of adsorbed Nafion films. Electrochemistry Communications, 2011, 13(8): 774–777

[61] Paul D K, McCreery R, Karan K. Proton transport property in supported Nafion nanothin films by electrochemical impedance spectroscopy. Journal of the Electrochemical Society, 2014, 161(14): F1395–F1402

[62] Ono Y, Nagao Y. Interfacial structure and proton conductivity of Nafion at the Pt-deposited surface. Langmuir, 2016, 32(1): 352–358

[63] Kusoglu A, Kwong A, Clark K T, Gunterman H P, Weber A Z. Water uptake of fuel-cell catalyst layers. Journal of the Electrochemical Society, 2012, 159(9): F530–F535

[64] Fukuyama Y, Shiomi T, Kotaka T, Tabuchi Y. The impact of platinum reduction on oxygen transport in proton exchange membrane fuel cells. Electrochimica Acta, 2014, 117: 367–378

[65] Suzuki T, Kudo K, Morimoto Y. Model for investigation of oxygen transport limitation in a polymer electrolyte fuel cell. Journal of Power Sources, 2013, 222: 379–389

[66] Owejan J P, Owejan J E, Gu W. Impact of platinum loading and catalyst layer structure on PEMFC performance. Journal of the Electrochemical Society, 2013, 160(8): F824–F833

[67] Liu H, Epting W K, Litster S. Gas transport resistance in polymer electrolyte thin films on oxygen reduction reaction catalysts. Langmuir, 2015, 31(36): 9853–9858

[68] Jinnouchi R, Kudo K, Kitano N, Morimoto Y. Molecular dynamics simulations on O₂ permeation through nafion ionomer on platinum surface. Electrochimica Acta, 2016, 188: 767–776

[69] Nonoyama N, Okazaki S, Weber A Z, Ikogi Y, Yoshida T. Analysis of oxygen-transport diffusion resistance in proton-exchange-membrane fuel cells. Journal of the Electrochemical Society, 2011, 158(4): B416

[70] Sabharwal M, Pant L M, Putz A, Susac D, Jankovic J, Secanell M. Analysis of catalyst layer microstructures: from imaging to performance. Fuel Cells (Weinheim), 2016, 16(6): 734–753

[71] Giner J, Hunter C. The Mechanism of operation of the Teflon-bonded gas diffusion electrode: a mathematical model. Journal of the Electrochemical Society, 1969, 116(8): 1124–1130

[72] Iczkowski R P, Cutlip M B. Voltage loss in fuel cell cathodes. Journal of the Electrochemical Society, 1980, 127(7): 1433–1440

[73] Ridge S J, White R E, Tsou Y. Oxygen reduction in a proton exchange membrane test cell. Journal of the Electrochemical Society, 1989, 136(7): 1902–1909

[74] Celiker H, Alsaleh M A, Gultekin S. A mathematical model for the performance of raney metal gas diffusion electrodes. Journal of the Electrochemical Society, 1991, 138(6): 1671–1681

[75] Broka K, Ekdunge P. Modelling the PEM fuel cell cathode. Journal of Applied Electrochemistry, 1997, 27(3): 281–289

[76] Ihonen J, Jaouen F, Lindbergh G, Lundblad A, Sundholm G. Investigation of mass-transport limitation in the solid polymer fuel cell cathode. Journal of the Electrochemical Society, 2002, 149(4): A448–A454

[77] Sun W, Peppley B A, Karan K. An improved two-dimensional agglomerate cathode model to study the influence of catalyst layer structural parameters. Electrochimica Acta, 2005, 50(16–17): 3359–3374

[78] Kamarajugadda S, Mazumder S. Numerical investigation of the effect of cathode catalyst layer structure and composition on polymer electrolyte membrane fuel cell performance. Journal of Power Sources, 2008, 183(2): 629–642

[79] Siegel N P, Ellis M W, Nelson D J, von Spakovsky M R. Single domain PEMFC model based on agglomerate catalyst geometry. Journal of Power Sources, 2003, 115(1): 81–89

[80] Yin K M. Parametric study of proton-exchange-membrane fuel cell cathode using an agglomerate model. Journal of the Electrochemical Society, 2005, 152(3): A583–A593

[81] Secanell M, Songprakorp R, Suleman A, Djilali N. Multi-objective optimization of a polymer electrolyte fuel cell membrane electrode assembly. Energy & Environmental Science, 2008, 1(3): 378–388

[82] Secanell M, Karan K, Suleman A, Djilali N. Multivariable optimization of PEMFC cathodes using an agglomerate model. Electrochimica Acta, 2007, 52(22): 6318–6337

[83] Khajeh-Hosseini-Dalasm N, Fesanghary M, Fushinobu K, Okazaki K. A study of the agglomerate catalyst layer for the cathode side of a proton exchange membrane fuel cell: modelling and optimization. Electrochimica Acta, 2012, 60: 55–65

[84] Rao R M, Bhattacharyya D, Rengaswamy R, Choudhury S R. A two-dimensional steady state model including the effect of liquid water for a PEM fuel cell cathode. Journal of Power Sources, 2007, 173(1): 375–393

[85] Eikerling M. Water management in cathode catalyst layers of PEM fuel cells a structure-based model. Journal of the Electrochemical Society, 2006, 153(3): E58

[86] Liu J, Eikerling M. Model of cathode catalyst layers for polymer electrolyte fuel cells: the role of porous structure and water accumulation. Electrochimica Acta, 2008, 53(13): 4435–4446

[87] Jain P, Biegler L T, Jhon M S. Sensitivity of PEFC models to cathode layer microstructure. Journal of the Electrochemical Society, 2010, 157(8): B1222–B1229

[88] Epting W K, Litster S. Effects of an agglomerate size distribution on the PEFC agglomerate model. International Journal of Hydrogen Energy, 2012, 37(10): 8505–8511

[89] Cetinbas F C, Advani S G, Prasad A K. A modified agglomerate model with discrete catalyst particles for the PEM fuel cell catalyst layer. Journal of the Electrochemical Society, 2013, 160(8): F750–F756

[90] Wang Q, Eikerling M, Song D. Structure and performance of different types of agglomerates in cathode catalyst layer of PEM fuel cells. Journal of Electroanalytical Chemistry, 2004, 573(1): 61–69

[91] Xia Z, Wang Q, Eikerling M, Liu Z. Effectiveness factor of Pt utilization in cathode catalyst layer of polymer electrolyte fuel cells. Canadian Journal of Chemistry, 2008, 86(7): 657–667

[92] Sadeghi E, Putz A, Eikerling M. Hierarchical model of reaction rate distributions and effectiveness factors in catalyst layers of polymer electrolyte fuel cells. Journal of the Electrochemical Society, 2013, 160(10): F1159–F1169

[93] Sadeghi E, Putz A, Eikerling M. Effects of ionomer coverage on agglomerate effectiveness in catalyst layers of polymer electrolyte fuel cells. Journal of Solid State Electrochemistry, 2014, 18(5): 1271–1279

[94] Zenyuk I V, Litster S. Modelling ion conduction and electrochemical reactions in water films on thin-film metal electrodes with application to low temperature fuel cells. Electrochimica Acta, 2014, 146(10): 194–206

[95] Kotoi R, Inoue G, Kawase M. Reaction and mass transport simulation of polymer electrolyte fuel cell for the analysis of the key factors affecting the output performance in the catalyst layer. ECS Transactions, 2016, 75(14): 385–392

[96] Perry M L, Newman J, Cairns E J. Mass transport in gas-diffusion electrodes: a diagnostic tool for fuel-cell cathodes. Journal of the Electrochemical Society, 1998, 145(1): 5–15

[97] Lin G, He W, van Nguyen T. Modelling liquid water effects in the gas diffusion and catalyst layers of the cathode of a PEM fuel cell. Journal of the Electrochemical Society, 2004, 151(12): A1999–A2006

[98] Madhusudana Rao R, Rengaswamy R. Dynamic characteristics of spherical agglomerate for study of cathode catalyst layers in proton exchange membrane fuel cells (PEMFC). Journal of Power Sources, 2006, 158(1): 110–123

[99] Harvey D, Pharoah J G, Karan K. A comparison of different approaches to modelling the PEMFC catalyst layer. Journal of Power Sources, 2008, 179(1): 209–219

[100] Das P K, Li X, Liu Z S. A three-dimensional agglomerate model for the cathode catalyst layer of PEM fuel cells. Journal of Power Sources, 2008, 179(1): 186–199

[101] Ma S, Solterbeck C H, Odgaard M, Skou E. Microscopy studies on pronton exchange membrane fuel cell electrodes with different ionomer contents. Applied Physics A, Materials Science & Processing, 2009, 96(3): 581–589

[102] Yoon W, Weber A Z. Modelling low-platinum-loading effects in fuel-cell catalyst layers. Journal of the Electrochemical Society, 2011, 158(158): B1007–B1018

[103] Jomori S, Nonoyama N, Yoshida T. Analysis and modelling of PEMFC degradation: effect on oxygen transport. Journal of Power Sources, 2012, 215: 18–27

[104] Dobson P, Lei C, Navessin T, Secanell M. Characterization of the PEM fuel cell catalyst layer microstructure by nonlinear least-squares parameter estimation. Journal of the Electrochemical Society, 2012, 159(5): B514–B523

[105] Cetinbas F C, Advani S G, Prasad A K. Three dimensional proton exchange membrane fuel cell cathode model using a modified agglomerate approach based on discrete catalyst particles. Journal of Power Sources, 2014, 250(3): 110–119

[106] Moore M, Wardlaw P, Dobson P, Boisvert J J, Putz A, Spiteri R J, Secanell M. Understanding the effect of kinetic and mass transport processes in cathode agglomerates. Journal of the Electrochemical Society, 2014, 161(8): E3125–E3137

[107] Hao L, Moriyama K, Gu W, Wang C Y. Modelling and experimental validation of Pt loading and electrode composition effects in PEM fuel cells. Journal of the Electrochemical Society, 2015, 162(8): F854–F867

[108] Weber M F, Mamicheafara S, Dignam M J. Sputtered fuel cell electrodes. Journal of the Electrochemical Society, 1987, 134(6): 1416–1419

[109] O’Hayre R, Lee S J, Cha S, Prinz F B. A sharp peak in the performance of sputtered platinum fuel cells at ultra-low platinum loading. Journal of Power Sources, 2002, 109(2): 483–493

[110] Bonakdarpour A, Stevens K, Vernstrom G D, Atanasoski R, Schmoeckel A K, Debe M K, Dahn J R. Oxygen reduction activity of Pt and Pt-Mn-Co electrocatalysts sputtered on nano-structured thin film support. Electrochimica Acta, 2007, 53(2): 688–694

[111] Gasda M D, Teki R, Lu T M, Koratkar N, Eisman G A, Gall D. Sputter-deposited Pt PEM fuel cell electrodes: particles vs. layers. Journal of the Electrochemical Society, 2009, 156(5): B614–B619

[112] Cavarroc M, Ennadjaoui A, Mougenot M, Brault P, Escalier R, Tessier Y, Durand J, Roualdès S, Sauvage T, Coutanceau C. Performance of plasma sputtered fuel cell electrodes with ultra-low Pt loadings. Electrochemistry Communications, 2009, 11(4): 859–861

[113] Schwanitz B, Schulenburg H, Horisberger M, Wokaun A, Scherer G G. Stability of ultra-low Pt anodes for polymer electrolyte fuel cells prepared by magnetron sputtering. Electrocatalysis (New York), 2011, 2(1): 35–41

[114] Schwanitz B, Rabis A, Horisberger M, Scherer G G, Schmidt T J. Sputtered cathodes for polymer electrolyte fuel cells: insights into potentials, challenges and limitations. Chimia, 2012, 66(3): 110–119

[115] Mei W, Fukazawa T, Nakano Y, Akasaka Y, Naito K. Development of alternated catalyst layer structure for PEM fuel cells. ECS Transactions, 2013, 50(2): 1377–1384

[116] Mei W, Fukazawa T, Yang T, Yoshinaga N, Kanai Y. Application of modified-ACLS electrodes on low-platinum PEFCs. ECS Transactions, 2015, 69(17): 755–759

[117] Sievers G, Mueller S, Quade A, Steffen F, Jakubith S, Kruth A, Brueser V. Mesoporous Pt–Co oxygen reduction reaction (ORR) catalysts for low temperature proton exchange membrane fuel cell synthesized by alternating sputtering. Journal of Power Sources, 2014, 268: 255–260

[118] Çögenli M S, Mukerjee S, Yurtcan A B. Membrane electrode assembly with ultra low platinum loading for cathode electrode of PEM fuel cell by using sputter deposition. Fuel Cells (Weinheim), 2015, 15(2): 288–297

[119] Stucki S, Menth A. Industrial water electrolysis. In: Srinivasan S, Salzano F J, Landgrebe A Reds. The Electrochemical Society Softbound Proceedings Series, Pennington, USA, 1978, 180–185

[120] McBreen J. Voltammetric studies of electrodes in contact with ionomeric membranes. Journal of the Electrochemical Society, 1985, 132(5): 1112–1116

[121] Tu W, Liu W, Cha C, Wu B L. Study of the powder/membrane interface by using the powder microelectrode technique I. The Pt-black/Nafion^® interfaces. Electrochimica Acta, 1998, 43(24): 3731–3739

[122] Paulus U A, Veziridis Z, Schnyder B, Kuhnke M, Scherer G G, Wokaun A. Fundamental investigation of catalyst utilization at the electrode/solid polymer electrolyte interface: Part I. Development of a model system. Journal of Electroanalytical Chemistry, 2003, 541: 77–91

[123] Jiang J, Yi B. Thickness effects of a carbon-supported platinum catalyst layer on the electrochemical reduction of oxygen in sulfuric acid solution. Journal of Electroanalytical Chemistry, 2005, 577(1): 107–115

[124] Tominaka S, Wu C, Kuroda K, Osaka T. Electrochemical analysis of perpendicular mesoporous Pt electrode filled with pure water for clarifying the active region in fuel cell catalyst layers. Journal of Power Sources, 2010, 195(8): 2236–2240

[125] Thompson E L, Baker D. Proton conduction on ionomer-free Pt surfaces. ECS Transactions, 2011, 41(1): 709–720

[126] An S J, Litster S. In Situ, ionic conductivity measurement of ionomer/binder-free Pt catalyst under fuel cell operating condition. ECS Transactions, 2013, 58(1): 427–441

[127] Debe M K, Steinbach A J. An empirical model for the flooding behavior of ultra-thin PEM fuel cell electrodes. ECS Transactions, 2007, 11(1): 659–673

[128] Steinbach A J, Debe M K, Wong J. A new paradigm for PEMFC ultra-thin electrode water management at low temperatures. ECS Transactions, 2010, 33(1): 1179–1188

[129] Steinbach A J, Debe M K, Pejsa M J. Influence of anode GDL on PEMFC ultra-thin electrode water management at low temperatures. ECS Transactions, 2011, 41(1): 449–457

[130] Debe M K, Atanasoski R T, Steinbach A J. Nanostructured thin film electrocatalysts-current status and future potential. ECS Transactions, 2011, 41(1): 937–954

[131] Kongkanand A, Dioguardi M, Ji C, Thompson E L. Improving operational robustness of NSTF electrodes in PEM fuel cells. Journal of the Electrochemical Society, 2012, 159(8): F405– F411

[132] Sinha P K, Gu W, Kongkanand A, Thompson E. Performance of nano structured thin film (NSTF) electrodes under partially-humidified conditions. Journal of the Electrochemical Society, 2011, 158(7): B831–B840

[133] Kongkanand A, Owejan J E, Moose S, Dioguardi M, Biradar M, Makharia R. Development of dispersed-catalyst/NSTF hybrid electrode. Journal of the Electrochemical Society, 2012, 159(11): F676–F682

[134] Kongkanand A, Sinha P K. Load transients of nanostructured thin film electrodes in polymer electrolyte fuel cells. Journal of the Electrochemical Society, 2011, 158(6): B703–B711

[135] Chan K, Eikerling M. A pore-scale model of oxygen reduction in ionomer-free catalyst layers of PEFCs. Journal of the Electrochemical Society, 2011, 158(1): B18–B28

[136] Petrii O A. Zero charge potentials of platinum metals and electron work functions. Russian Journal of Electrochemistry, 2013, 49(5): 401–422

[137] Zenyuk I V, Litster S. Modelling ion conduction and electrochemical reactions in water films on thin-film metal electrodes with application to low temperature fuel cells. Electrochimica Acta, 2014, 146: 194–206

[138] Frumkin A N, Petrii O A. Potentials of zero total and zero free charge of platinum group metals. Electrochimica Acta, 1975, 20(5): 347–359

[139] Garcia-Araez N, Climent V, Feliu J M. Potential-dependent water orientation on Pt(111), Pt(100), and Pt(110), as inferred from laser-pulsed experiments. electrostatic and chemical effects. Journal of Physical Chemistry C, 2009, 113(21): 9290–9304

[140] Huang J, Malek A, Zhang J, Eikerling M H. Non-monotonic surface charging behavior of platinum: a paradigm change. Journal of Physical Chemistry C, 2016, 120(25): 13587–13595

[141] Huang J, Zhang J, Eikerling M. Theory of electrostatic phenomena in water-filled Pt nanopores. Faraday Discussions, 2016, 193: 427–446

[142] Huang J, Zhang J, Eikerling M H. Particle proximity effect in nanoparticle electrocatalysis: surface charging and electrostatic interactions. Journal of Physical Chemistry C, 2017, 121(9): 4806–4815

[143] Das P K, Weber A Z. Water management in PEMFC with ultra-thin catalyst-layers. ASME 2013 11th International Conference on Fuel Cell Science, Engineering and Technology collocated with the ASME 2013 Heat Transfer Summer Conference and the ASME 2013 7th International Conference on Energy Sustainability. American Society of Mechanical Engineers, 2013

[144] Das P K, Santamaria A D, Weber A Z. Role of GDL surface wettability and operating conditions in liquid-water removal from NSTF catalyst layers. ASME 2014 International Mechanical Engineering Congress and Exposition. American Society of Mechanical Engineers, 2014

[145] Zenyuk I V, Das P K, Weber A Z. Understanding impacts of catalyst-layer thickness on fuel-cell performance via mathematical modelling. Journal of the Electrochemical Society, 2016, 163(7): F691–F703

[146] Chan K, Eikerling M. Water balance model for polymer electrolyte fuel cells with ultrathin catalyst layers. Physical Chemistry Chemical Physics, 2014, 16(5): 2106–2117

[147] Furuya Y, Iden H, Mashio T, Ohma A, Shinohara K. Effect of ionomer coverage on Pt-based catalyst on ORR activity. The Electrochemical Society, 2012 (40): 1522–1522

[148] Modestov A D, Kapustin A V, Avakov V B, Landgraf I K, Tarasevich M R. Cathode catalyst layers with ionomer to carbon mass ratios in the range 0–2 studied by electrochemical impedance spectroscopy, cyclic voltammetry, and performance measurements. Journal of Power Sources, 2014, 272: 735–742

[149] Dong B, Gwee L, Salas-de La Cruz D, Winey K I, Elabd Y A. Super proton conductive high-purity Nafion nanofibers. Nano Letters, 2010, 10(9): 3785–3790

[150] Nouri-Khorasani A, Malek K, Malek A, Mashio T, Wilkinson D P, Eikerling M H. Molecular modelling of the proton density distribution in a water-filled slab-like nanopore bounded by Pt oxide and ionomer. Catalysis Today, 2016, 262: 133–140

[151] Eslamibidgoli M J, Huang J, Kadyk T, Malek A, Eikerling M. How theory and simulation can drive fuel cell electrocatalysis. Nano Energy, 2016, 29: 334–361

[152] Muzaffar T, Kadyk T, Eikerling M. Physical modelling of the proton density in nanopores of PEM fuel cells catalyst layers. Electrochimica Acta, 2017, 245: 1048–1058

RIGHTS & PERMISSIONS

Higher Education Press and Springer-Verlag Berlin Heidelberg