Polymer Electrolyte (PE) Fuel Cell Systems

John F. Elter is retired from College of Nanoscale Science and Engineering

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

Glossary

Fuel Cells (FC):

Electrochemical cells that convert chemical energy from a fuel into electrical energy through the controlled transfer of electrical charge driven by the difference in electrochemical potential between two electrodes separated by an electrolyte.

Electrocatalyst:

A material that enhances the rate of an electrochemical reaction, such as the hydrogen oxidation reaction (HOR) or the oxygen reduction reaction (ORR), without itself being consumed in the reaction. In most PEM fuel cells, the electrocatalysts are nanosized materials made from the precious group metals (PGM), usually platinum, palladium, and ruthenium or alloys of these materials with nickel, cobalt, and manganese.

Polymer Electrolyte Membrane (PEM):

Cation- or anion-conducting polymer membrane that separates the two electrodes in a fuel cell. In most PEM fuel cells, the polymer is a cation (proton) conductor and in this case PEM can stand for proton exchange membrane.

Triple-Phase Boundary (TPB):

The place on the catalyst surface within the fuel cell electrodes where protons, electrons, and gaseous reactants meet in order to oxidize hydrogen or reduce oxygen.

Gas Diffusion Layer (GDL):

A component of the fuel cell used to evenly distribute reactants and electrons across the electrode surface and facilitate the removal of water.

Membrane Electrode Assembly (MEA):

A manufactured unit consisting of the electrodes, membrane and gas diffusion electrode. Sometimes the MEA can include the gasket used to seal the fuel cell. The MEA, when placed between two electrically conducting plates with channels to provide the fuel and oxygen, constitute a single cell.

Stack:

A collection of individual fuel cells, separated by electrically conducting “bipolar” plates, connected in series. Stacks can range in size from a few watts as in portable fuel cells to hundreds of kilowatts as in stationary fuel cells for combined heat and power or grid assets.

Fuel Cell System (FCS):

The combination of a fuel cell stack and the subsystems that is required to support its operation for the intended application.

Definition of Subject

This entry discusses the subject of fuel cells: how they work and how they are designed and integrated into a collection of subsystems for application in a variety of applications. Particular focus is placed upon systems utilizing polymer electrolyte membranes and how the properties of the membrane dictate the system design considerations.

Fuel cells themselves were invented over 170 years ago, with the first fuel cell demonstrated by Sir William Robert Grove in 1839. The first commercial use of a PEM fuel cell in a system occurred when General Electric developed the technology with NASA and McDonnell Aircraft for use in the Gemini space program. Since then, fuel cells have been integrated into systems intended for a wide variety of applications, from fuel cells of just a few watts for portable application up to hundreds of kilowatts for some applications in transportation.

Introduction

The vision of fuel cells running on renewable hydrogen as an energy carrier derived from water, and producing electricity with only water as a by-product, has long captured the imagination of those concerned with the environment. Since the early days of the use of fuel cells in the Gemini space program, various explorations of the potential use of fuel cells have been conducted by individuals, small firms, and large corporations. In each case, the potential of fuel cells has been validated. The long-term benefits of high efficiency and fuel flexibility are clear, especially in today’s context, wherein energy security and the need to reduce dependence on foreign oil and the need for sustainable economic development and the reduction of greenhouse gas emissions have been brought into sharp focus by war, weather, and wildlife.

Great progress has been and continued to be made in the development of fuel cells since their first commercial application roughly 50 years ago. This is particularly true of the fuel cells developed for automotive applications, which represents a very demanding application in terms of durability and cost. Progress in the automotive application has had a spillover effect to other applications, particularly for those being developed for motive power and stationary combined heat and power. This entry will attempt to lay out the status of the technology today, with a focus on the progress made in the past 5 years and where it needs to go in the future. First, a brief overview will be given of the fuel cell industry, followed by a review of some of the basic principles underlying the theory of operation. The notion of critical design parameters and specifications will be discussed as a prelude to a review of some recent advancements and to the system design process as it relates to important considerations for fuel cell systems. Finally, there will be a discussion on future directions.

The PEM Fuel Cell Industry

In reality, it is fair to say that PEM fuel cells are slowly emerging from their current status as a nascent industry. The long-term benefits of high efficiency and potential fuel flexibility are clear, but cost and durability issues have caused a slow adoption time line. In addition, the nagging issue of the infrastructure needed to provide hydrogen as a fuel has not gone away. Hydrogen availability is still a relevant issue for many of the applications that would have fuel cells as the privileged technology. Other obstacles to profitable commercialization include the excellent low cost, operational flexibility, and lifespan of the incumbent technologies which fuel cells are trying to displace. In response to these important issues, most fuel cell manufacturers, in need of revenues, are focusing on large niche markets for which the technological requirements are best suited to current fuel cell capabilities. These markets, such as backup power, forklift trucks, bus fleets, and small and large combined heat and power, offer the opportunity for fuel cell companies to manufacture sufficient quantities to gain sufficient learning and revenues to attract investors seeking the potential benefits of game-changing technologies. Even in some of these cases, however, government subsidies are needed to help the industry cope with the slow market penetration associated with product immaturity and the risk-adverse nature of the industries being targeted. Limited revenues from these markets lead to high stock volatility, making valuations difficult, limiting private investment, and further exasperating the situation.

Over the past 5 years, the PEM fuel cell industry has demonstrated a steady year-over-year increase in placements, but at a rate which is not offering the benefit of a strong learning curve and the accompanying economies of scale. On the one hand, there are large automotive players (Honda, Toyota, GM, Hyundai) who continue to make sizeable investments and steady progress in terms of cost and durability. The fuel cell-powered vehicles are now early production models that are being sold and tested in quantities that are compatible with the limited hydrogen dispensing infrastructure. These are providing important feedback on customer acceptance and validation of the fuel cell vehicle performance and reliability. On the other are those established (but still not profitable) smaller fuel cell companies focusing on niche markets (backup, forklift, other mobility applications) that are gaining commercial traction based on recognized value propositions. An example of the latter is Plug Power, which had established strong commercial agreements with both Amazon and Wal-Mart [1]. Feeding fuel cell components to these companies are corporations such as 3M, DuPont, W.L. Gore, and Asahi Glass providing the membrane electrode assemblies, as well as a host of large automotive suppliers providing “balance of plant” components such as blowers, compressors, heat exchangers, and electronics which are a critical part of the maturing supply chain necessary for industry success.

The past 5 years have also witnessed a restructuring of the field in which smaller companies are being consolidated into others as a means of acquiring fuel cell IP, know-how, and additional markets. Examples include Plug Power acquiring ReliOn and its telecom backup power business, Ballard acquiring IdaTech’s backup power assets and UTC’s PEM intellectual property for bus applications, and Hyster-Yale’s purchase of Nuvera for application in its forklift business, which is one of the fuel cell applications which is closest to profitability. In addition, China is a slumbering dragon which has begun to awaken to the possibilities of fuel cells for both motive and stationary applications, with agreements signed with Ballard, Plug Power, and others as it begins to address its need to build up expertise and its supply chain ecosystems [2].

One reason for the fragmentation of the fuel cell industry is the wide range of potential applications to which fuel cells can be applied. As mentioned, these include transportation applications, material handling, telecommunications backup, consumer electronics, as well as both residential and larger-scale cogeneration. Each of these markets has a well-defined set of operational characteristics and requirements, as defined by the customer. Another and related reason for the fragmentation of the industry is the lack of a dominant design paradigm. Each company has a preferred method of approaching the design of a fuel cell and the system into which it is embedded. Since no single approach has emerged as dominant, there is a lack of standards, de facto or otherwise, around which the industry can capture the benefits of economies of scale and of accumulated knowledge.

On the other hand, it might be argued that there is a de facto dominant design, and it is that standard which, up to now unwittingly adhered to, is limiting the true out-of-the-box thinking needed for step changes in performance at lower cost and improved reliability. In most cases, as we will see, the commercialization issues around each application are driven to a large extent by this underlying “standard” technology and its modes of failure in meeting customer requirements. In this entry, specific attention will be given to these failure modes and how they drive system complexity. In addition, some attention will be paid to the system design and how it can be simplified to enable lower costs.

Finally, as will be discussed here, there has been steady progress over the past 5 years in lowing costs and improving reliability, due in part to sustained government support of fuel cells and the persistence of the key players in the industry. The US DOE in particular has shown a steady willingness to support the technological development of fuel cells, particularly for the most demanding application, transportation. Indeed, there is hardly a technical nuance that has not received the DOE’s support. In many ways the industry is now in the position of developing engineering solutions at the material, component, and system level for minimizing the known deficiencies and failure modes intrinsic to the technology. These will be discussed below.

Still, there is “a lot of room at the bottom” for scientific and engineering innovation. Indeed, the application of nanotechnology as a means of enabling the fuel cell industry to become a sustainable ecosystem should not be underestimated. In fact, most progress within the past 5 years has been in engineering improvements at the nanoscale, thus enabling improvements in cost, performance, and durability. There is ample reason for considerable optimism around PEM fuel cell systems for the sustainable energy applications they support.

Fuel Cell Basics

Fuel cells are electrochemical devices that transform chemical energy into electrical energy, much the same way a battery converts chemical potential energy into electrical kinetic energy, except that the chemicals consumed in a fuel cell in the reaction are supplied from external sources. The fuel cell enables the generation of electricity while producing pure water by controlling the combining of hydrogen and oxygen:

$$ {\mathrm{H}}_2\; +\; {\mathrm{O}}_2\; \to \; {\mathrm{H}}_2\mathrm{O} $$

(1)

Ideally, this reaction would take place at a potential of 1.229 V at standard conditions (25 °C and 1 atm) if it were not for losses due to irreversible internal processes. In fact, the reversible potential E_r of this reaction can be expressed in terms of the change in the Gibbs free energy, ΔG:

$$ \varDelta G\; =\; -n\; F\; {E}_r $$

(2)

Here n is the number of electrons involved, and F is Faraday’s constant, which has the value of 96,485 coulombs of charge per mole of electrons. The derivation of Eq. (2) is a straightforward application of thermodynamics, recognizing that in this case, the work done is the product of the amount of charge transferred and the potential. For the case of standard conditions, we know from thermodynamics that ΔG = 237 kJ/mole when the water is produced as a liquid (corresponding to the higher heating value, HHV) and n = 2 for the H₂/O₂ fuel cell reaction. Of course, not all of the energy in the reaction is turned into electricity, and because of this, some portion of hydrogen’s HHV is converted into heat. The enthalpy change in the reaction is given by the familiar thermodynamic relationship

$$ \varDelta H\; =\; \varDelta G\; +\; T\varDelta S $$

(3)

From thermodynamic tables we find at standard conditions that ΔH = 286 kJ/mole. It follows that roughly 49 kJ/mole are converted into heat, and the theoretical efficiency η_th of a fuel cell operating at standard conditions is

$$ {\eta}_{th}\; =\; \varDelta G/\varDelta H\; =\; 237/286\; =\; 83\% $$

(4)

In addition, there are various losses which are due to irreversible processes in the cell itself. These losses, which are intrinsic to the cell and are material dependent, reduce the fuel cell efficiencies to around 40–60%, depending on the type of fuel cell and its application. There are numerous references that discuss the theory behind fuel cells in detail [3, 4] as well as the other entries in this encyclopedia. Here only a very cursory overview is given, as the focus of this entry is on the fuel cell system.

In a typical fuel cell, there are two electrodes, an anode from which electrons are removed during the oxidation of hydrogen and a cathode to which electrons flow during the reduction of oxygen, an ion-conducting electrolyte which ionically connects the anode and cathode, and an external circuit through which the electrons flow from the anode to the cathode while doing electrical work. In early fuel cells, the electrolyte was generally a liquid such as a solution of KOH, which provided for the flow of OH⁻ anions. In this case, the fuel cell was termed an “alkaline” fuel cell because of the nature of the electrolyte. In the same way, the electrolyte can be phosphoric acid, in this case providing for the flow of H⁺ cations, or protons. Such a fuel cell is called a “phosphoric acid” fuel cell, again in recognition of the electrolyte. In this entry, we will discuss polymer electrolyte fuel cell systems. In this case, the electrolyte is a polymer in the form of a membrane, hence the term “polymer electrolyte membrane” or PEM fuel cell. Polymer electrolyte fuel cells, or PEFCs, are electrochemical devices whose characteristics are determined largely by the properties of the polymer electrolyte. One of the main purposes of this entry is to illustrate the extent to which the PEMFC’s performance is in fact dominated by the properties of the polymer electrolyte. It is worth noting that in some cases, the acronym PEM is interpreted to mean proton exchange membrane, since in most cases in which the acronym PEM is used, the ion that is transported across the membrane and “exchanged” at the electrodes is the proton. However, here we will interpret the acronym to mean polymer electrolyte membrane to emphasize the point that it is the polymer electrolyte that to a large extent dictates the characteristics of the system.

In a fuel cell, the two reactants, in this case hydrogen and oxygen, are each directed macroscopically by flow channels molded into electrically and thermally conductive plates. The hydrogen on the anode side and the oxygen at the cathode side are each brought into contact with an electrode by the gas diffusion layer or “GDL” which ensures uniform flow across the cell. As discussed, a polymer membrane is provided that serves as the electrolyte, providing good ionic conductivity while maintaining the separation of these reactant gases and preventing the flow of electrons. The membrane needs to make intimate contact with the electrodes to ensure efficient charge transfer across the embedded electrode/electrolyte interface. The electrodes are generally porous to allow the diffusion of the reactant gases while providing a support for the catalysts needed to facilitate the electrochemical reactions. The electrodes also need to provide conductive pathways for the ions and electrons. Clearly, the polymer electrolyte membrane, electrodes, and GDL have a complex set of interacting functional requirements. These will be discussed in detail, especially from the point of view of their impact on system design.

In most cases, the polymer electrolyte is basically acidic, in which case the catalysts are usually based upon platinum and its alloys. At the anode electrode, hydrogen gas molecules adsorb onto the supported catalyst material which is used to help strip the hydrogen of its electron, oxidizing it to become a hydrogen cation or proton. This proton is transported through the polymer electrolyte membrane to the cathode electrode. Here, oxygen molecules are fed into cathode electrode, usually by supplying air. Like the active layer at the anode, appropriate nanosized catalyst particles are supported on electrically conducting porous materials. In the current dominant design paradigm, this support is usually a high surface area carbon. Once on the catalyst surface, the oxygen’s covalent bonds are broken, forming oxygen anions. The oxygen anions then combine with protons arriving through the membrane and electrons that have been driven through an external circuit by the difference in the anode and cathode potentials. The critical point on the catalyst surface at which protons, electrons, and reactant gases must come together is generally called the triple-phase boundary (TPB). All of the necessary components, the gas diffusion media, the porous electrodes, and the polymer electrolyte membrane, and if desired, even the conformable gasket required to provide sealing around the perimeter of the plate, are manufactured as a “membrane electrode assembly” or “MEA” by various suppliers like 3M, Gore, Johnson Matthey, and others. Since a typical PEMFC operates at a voltage of less than 1 V, multiple cells are electrically connected in series to form a “stack” of cells that can be sized for a particular application. In this case, each thermally and electrically conducting plate is shared between the anode of one cell and the cathode of another and is thus termed “bipolar.” The integrated stack design must accommodate a means for managing the products of the fuel cell reaction: electricity, water, and heat. In smaller systems, the stack can be air-cooled, but in larger systems, the plates must incorporate channels that allow for the flow of a liquid coolant.

The electrochemical equation governing the hydrogen oxidation reaction (HOR) at the anode of a single cell is

$$ {\mathrm{H}}_2\; \leftrightarrow \; 2{\mathrm{H}}^{+}\; +\; 2{\mathrm{e}}^{+}{\mathrm{E}}^{{}^{\circ}}=\; 0\; \mathrm{V} $$

(5)

This is a thermodynamically reversible reaction and, with platinum as the catalyst, is a standard reference known as the “reversible hydrogen electrode” (RHE), for which the potential is universally chosen as zero volts. The oxygen reduction reaction (ORR) at the cathode, on the other hand, is thermodynamically irreversible and is generally expressed in terms of the dominant four-electron reaction:

$$ {\mathrm{O}}_2+4{\mathrm{H}}^{+}+4{\mathrm{e}}^{-}\to 2{\mathrm{H}}_2\mathrm{O}\; \; {\mathrm{E}}^{{}^{\circ}}=1.229\; \mathrm{V}\; \left(\mathrm{vs}.\; \mathrm{RHE}\right) $$

(6)

In fact, the ORR is an extremely complicated reaction that depends on the electronic structure and number of the active sites on the catalyst surface, which in turn are influenced by many factors including the size and shape of the catalyst particle, the presence of any contaminant species competing for reaction sites, the presence of oxides which may be formed during potential cycling, the effect of the catalyst support, etc. In addition to the four-electron reaction, there is the possibility for the two-electron peroxide pathway [5]:

$$ {\mathrm{O}}_2+2{\mathrm{H}}^{+}+2{\mathrm{e}}^{-}\to {\mathrm{H}}_2{\mathrm{O}}_2\; \; {\mathrm{E}}^{{}^{\circ}}=0.695\; \mathrm{V}\; \left(\mathrm{vs}.\mathrm{RHE}\right) $$

(7)

In low-temperature PEMFCs, those that operate in the range of 65–80 °C, the peroxide pathway is problematic because if the peroxide is not quickly decomposed into water, it can form radicals that chemically attack the membrane, causing premature failure.

The ORR is a major source of efficiency loss in a PEMFC, but it is not the only source of the so-called overpotential. In general there are also resistive and mass-transport losses. The resistive losses are associated with the finite conductivities of the electrolyte and the electrodes and the contact resistance losses at the plate and GDL interfaces, whereas mass-transport losses are associated with the lack of adequate fuel or air reaching the reaction sites within the electrode. For low-temperature PEMFC systems that operate below the boiling point of water, mass-transport losses are primarily due to the buildup of liquid water in the electrode or GDL (referred to as “flooding”) that increase the tortuous pathways. Reducing all of these sources of potential losses is important for achieving the highest possible efficiency.

It is perhaps convenient at this point to briefly mention the types of PEM fuel cell that are under development. Low-temperature PEM fuel cells (LTPEMFCs) operate in the range of 65–80 °C. These systems are limited in temperature by the requirement for the membrane to be maintained in a hydrated state in order to have sufficient proton conductivity. Typical of these systems are the perfluorinated sulfonic acid (PFSA) membranes made by DuPont, the standard being Nafion®. 3M, Gore, Solvay, Asahi, and others make membranes with similar properties and characteristics. Intermediate-temperature PEMFCs that are capable of operating above the boiling point of water in the region of 120 °C are under development primarily for automotive applications. These are still in the development phase [6] and when available will have a significant impact on PEMFC performance for applications other than just automotive. Still higher-temperature PEMFCs (HTPEMFCs) are based upon phosphoric acid-doped materials such as polybenzimidazole (PBI). These proton-conducting polymer electrolyte membranes allow fuel cells to operate at temperatures in the range of 160–200 °C, thus providing for a high degree of tolerance to certain fuel cell contaminants such as CO, while at the same time offering a source of high-quality heat [7]. PBI-based systems are discussed in detail elsewhere in this encyclopedia and will be mentioned briefly below in terms of their potential to simplify system design.

For a basic understanding of PEM fuel cells, it is important to appreciate how the properties of the polymer electrolyte dictate the requirements on the fuel cell, stack, and system design. A recent comprehensive review of the properties of PFSA ionomers has been given by Kusoglu and Weber [8]. Here we will discuss just one but critically important example: water management. One of the more well-known and basic properties of the PFSA type of polymer electrolytes is that their protonic conductivity is a strong function of the level of hydration. Sulfonated fluoropolymer membranes are based on a polytetrafluoroethylene (PTFE) backbone that is sulfonated by adding a side chain ending in a sulfonic acid group (−SO₃H) to the PTFE backbone. The resulting macromolecule contains both hydrophobic regions associated with the backbone and hydrophilic regions associated with the sulfonic acid group. Thus, a hydrated PFSA membrane forms a two-phase system consisting of a water-ion phase that is distributed throughout a partially crystallized perfluorinated matrix phase. As the membrane adsorbs water, the first water molecules cause the sulfonated group to dissociate, forming hydronium (H₃O⁺) ions. The water that hydrates the membrane then forms counterions that are localized on the sulfonated end groups, which act as nucleation sites. As more water is added, the counterion clusters coalesce to form even larger clusters, until a continuous phase is formed with properties that approach those of bulk water. The level of hydration is measured in terms of a parameter λ, which is equal to the number of absorbed waters per sulfonated group. For λ = 0, there is no water, and this anhydrous form of the membrane is uncommon, since complete removal of water requires temperatures near the decomposition temperature of the polymer. Molecular dynamic simulations indicate that the primary hydration shell around the sulfonated group grows to a level of about λ equal to five waters [9]. As more water is added, for λ > 6, the added water is screened by the more strongly bound water of the primary hydration shell, and it can be considered a free phase. Saturation occurs in Nafion® at λ = 14. One of the consequences of this strong coupling is that proton transport within the membrane is accompanied by the transport of water. This effect is known as “electroosmotic drag,” and it complicates the required management of the fuel cell by-products: water and heat.

The mechanism and degree of protonic conductivity change as the level of hydration increases. Kreuer [10] has provided an in-depth review of the physics underlying the basic mechanisms of transport in proton conductors. Transport of the proton can occur by two mechanisms: structural diffusion and vehicle diffusion. The structural diffusion mechanism is associated with the proton “hopping” along water molecules (Grotthuss shuttling). Vehicular diffusion is the classical Einstein diffusive motion. The former is viewed as a discrete mechanism, the latter a continuous mechanism. In the nanosized confined hydrophilic spaces within the membrane, both mechanisms are operative, with the diffusion constant of the discrete mechanism being larger and increasing faster than that of the continuous mechanism at higher levels of hydration. At intermediate and low degrees of hydration, the proton mobility is essentially vehicular in nature. What is important here is that the underlying mechanism of transport in proton-conducting membranes changes as a function of the level of hydration, and it is fundamental to the overall behavior of the fuel cell and therefore the design of the fuel cell system.

Given this fundamental and strong dependence of protonic conductivity with water content, it is necessary to ensure that the polymer electrolyte is safeguarded against localized drying. Indeed, a situation in which the membrane is in a “drying” mode of operation may lead to a catastrophic “death spiral” failure mode, since lower conductivity leads to higher resistances, increased localized temperatures and therefore an ever increasing rate of drying, ultimately resulting in membrane failure. In addition, other failure modes, including attack by hydroxyl radicals, are also accelerated at higher temperature and drier membrane conditions, further exasperating the situation.

Hence, it is necessary to maintain the localized level of membrane hydration through “water management,” the active management of the moisture content of the membrane. One method is to design the system so that the relative humidity (RH) of the reactant streams coming into the stack is nearly saturated, an approach used by Plug Power and other fuel cell manufacturers. Another approach is to provide for water injection, an approach used by Intelligent Energy. Or one can employ porous plates to distribute the product water generated during operation, a technique previously used by United Technologies. More recently, there is also a move toward thinner membranes to enhance the back transport of water. Each of these techniques requires careful stack and system design. The design approach taken likewise needs to avoid the situation where too much water can cause localized “flooding” within the cell, which in turn results in fuel starvation and subsequent permanent performance degradation. In most cases the requirement to manage the polymer electrolyte membrane water content adds considerable system complexity and cost, as well as sources of unreliability.

Critical Parameters/Critical Specifications

It is clear that in a typical operating fuel cell, the current-voltage characteristics are dependent on the local operating temperature, pressure, flow rate, and composition of the reactant gas streams within the cell. These local conditions within the cell are critical parameters that need to be supplied by the careful design of the overall system. The system-level critical parameters around which the fuel cell system must be designed in order to achieve and maintain the desired local conditions ultimately determine the desired electrical state of the fuel cell system. Achieving the local critical parameters for adequate cell performance must take into account the design of the overall system and vice versa. It is also obvious that the performance of the fuel cell is also dependent on how well the individual functional requirements are met with the specifications on the component parts.

From a systems design perspective, it is perhaps worthwhile to describe how these terms are related. To motivate this discussion, it is important to note that a system is “technically ready” when all the required critical parameters can be simultaneously achieved through the intended interactions of the buildup of parts whose specifications can be meet with ordinary manufacturing methods. It is assumed that the critical parameters are chosen to meet the system specifications, while at the same time avoiding all known failure modes. The extent of the range in the critical parameters under which the system can operate over time, without failure, is called the “operating latitude.”

As a simple example, assume that one can write the equations of physics in the linear form

$$ {F}_i={K}_{ij}{X}_j $$

(8)

The terms F_i and X_j are termed the critical parameters, and the K_ij are termed the critical specifications that are required to meet the functional requirements of the particular component under consideration. The term “critical” is of course an indication that the specification in question is one of the significant few specifications that will be tracked all the way from the voice of the customer down to the factory floor. In a complete PEM fuel cell system, there can be hundreds of specifications that will need to be tracked during the development of the system. Of these, perhaps a few dozen will remain critical even after the product has launched and is in production.

As an example, consider the conduction of protons across the membrane and for the moment neglect the electroosmotic drag of water with the proton. In this case, Eq. (8) takes the simple form of Ohm’s law:

$$ d\varphi / dz=-\left[1/\kappa \right]\; i $$

(9)

Here dφ/dz is the potential gradient across the membrane, i is the flux of protons, and κ is the protonic conductivity. The system-level critical parameters (pressure, temperature, flow rate, and composition) establish the conditions for determining the local critical parameters (dφ/dz, i), and these are in turn determined by the functional requirement that the polymer electrolyte has a protonic conductivity, κ. Typical values lie in the range of 10^−1_10⁻² S/cm. Notice that critical parameters are generally not dimensions, but are rather forces, fields, affinities, and fluxes. They are the conditions that are required for operation. Specifications, on the other hand, are the items that an engineer can call out on a drawing: dimensions, conductivities (both electrical and thermal), moduli of elasticity, and so forth. Instead of specifying the protonic conductivity κ, which is fixed by the material chemistry, the design engineer will specify perhaps the membrane equivalent weight and its thickness, thereby affecting not only the resistive losses but also the diffusion rate of gaseous and ionic species across the membrane.

As discussed above, the protonic conductivity is known to be a strong function of the moisture content, which again will be influenced by the system-level critical parameters of flow rate, temperature, pressure, and composition. Hence, in reality, a constitutive equation of the dependence of conductivity on moisture content is needed to support larger PEM fuel cell models which are essential to account for and keep track of all the complex interactions between the critical parameters and the specifications for the components of the MEA and other critical components. An early constitutive equation defining these interactions was described by Springer et al. [11] These interactions place stringent demands upon the design of the cell, stack, and ultimately the system. Researchers and engineers have developed sophisticated models in an attempt to describe the current-voltage response of single fuel cells to the significant critical parameters through the specifications of the material properties of the cell components. These models, which are remarkably successful in describing the overall performance of a single cell, are macro-homogeneous in the sense that they use local average properties, such as electrode porosity, conductivity, etc. to describe the materials. Constitutive equations of the conductivity dependence on moisture have been developed to support the fuel cell models [12]. There are also excellent reviews of the models that have been developed to elucidate the various aspects of fuel cell behavior under a variety of conditions, for example, those by Weber and Newman [13] and Wang [14]. A comprehensive review of the mass-transport phenomenon in PEM fuel cells, with a particular focus on multiphase flow, especially on understanding interactions at embedded interfaces within the electrodes, has been conducted by Weber et al. [15]. In addition, there are specialized models that deal with the root cause of specific failures, such as platinum dissolution [16] and carbon corrosion [17] due to high potentials generated during startup and shutdown [18]. All of these have added considerable insights into how the PEM system runs and how failures are to be avoided. Finally, there are models of the fuel cell system itself. The models predict full system behavior over the range of system-level critical parameters and are useful in developing the algorithms needed in implementing model predictive control strategies.

Extensive cell-/stack-/system-level testing is also needed to assess the system’s overall performance. In this case, computational fluid dynamics (CFD) is generally used to guide the cell-/stack-level testing, aiding in the fuel cell system development process. Some of these models have been made commercially available. In particular, models for PEMFCs can be obtained from CD-adapco [19], Fluent [20], Comsol [21], and others. In fact, knowledge of the cell and stack failure modes, and their dependence on the process critical parameters and critical specifications, is fundamental to designing and developing commercially viable fuel cell systems. A more detailed discussion of failure modes will be addressed in what follows. First, however, it is important to understand the current state of the art relative to the component at the heart of the PEM fuel cell, the MEA.

Recent Advancements

The operation of the fuel cell involves complex interactions at numerous embedded surfaces within the cell, including the triple-phase boundary. These embedded surfaces can mask the ability to probe the detailed nature of these interactions. Today, processes at the nanoscale are a fertile ground for research, enabled by the tremendous advances made in computational nanoscience through the convergence of “nano” and “info” technology. These advancements in computation have allowed theoreticians to explore electrochemical and electromechanical processes at the micro-, meso-, and nanoscales, with meaningful numbers of atoms, thereby improving our theoretical understanding of the basic processes involved. Furthermore, the rapidly expanding field of nanoscience and nanoengineering is being applied in a variety of fields, to include the synthesis and characterization of nanoscale materials and structures that have direct bearing on energy generation, conversion, and storage.

Electrocatalysts

Relatively recent advancements have been made in the understanding of heterogeneous ORR catalysis using the combination of molecular- and atomic-level simulation tools, nanoscale fabrication methods, and nanoscale characterization techniques. For example, the early experimental work of Markovic [22], combined with the important theoretical work of Nørskov [23], Mavrakakis [24], Neurock [25], and others, has now given us a deeper understanding not only of the thermodynamics involved in the ORR, and the likely cause of the overpotential associated ORR, but also how variations in the electronic structure determine trends in the catalytic activity of the ORR across the periodic table. This work, coupled with the insights provided by the seminal work of Adzic [26], Stamenkovic [27], and others in the synthesis and performance characterization of Pt alloys and core-shell architectures, has significantly advanced the understanding of the state of the art of catalysts for use in both the HOR and ORR.

The exact mechanisms underlying the ORR are still debatable. However, Nørskov [28] and his group have used density functional theory (DFT) to significantly advance our understanding of the thermodynamics of the various possible reaction pathways and their dependence on surface properties. In combination with detailed DFT calculations, Nørskov and his team were able to provide a detailed description of the free-energy landscape of the electrochemical ORR over Pt as a function of applied bias. In doing so, they found that adsorbed oxygen and hydroxyl species are very stable intermediates at potentials close to equilibrium, and the calculated rate constant for the activated proton/electron transfer to these adsorbed intermediates accounts for the observed kinetics. On this basis, they were able to account for the trends in the ORR rate for a large number of different transition metals. In particular, they were able to construct “volcano” plots of maximum catalytic activity as a function of the oxygen and hydroxyl adsorption energies, describing known trends and the observed effects of alloying. It is clear that a significant amount of progress has been made in the past several years in our basic understanding of the electrocatalytic processes at the nanoscale. An excellent summary of this and other work has been given in the recent book on fuel cell catalysis by Koper [29].

Of course, the high price of platinum has naturally provided the impetus for reducing the platinum loading by improving the kinetics of the ORR reaction. Considerable experimental effort has gone into the research and development of electrodes that incorporate the advantages of new nanomaterials intended to significantly reduce precious metal catalyst loadings, while maintaining high-power densities. Chief among these are electrodes that incorporate platinum alloys with transition metals of various types as a means of increasing catalytic activity, thereby allowing a reduction in platinum loading. Currently, Pt₃Co alloys supported on carbon are achieving roughly a 2× improvement over the state-of-the-art Pt/C electrodes. Intermetallic architectures of Pt and other materials have been fabricated by a variety of techniques. Some of these, which yield so-called “core-shell” [30] structures which involve monolayers of Pt particles on Pd cores supported on carbon, are reported to have achieved very significant (≈20) increases in mass activities. The catalysts must not only have high catalytic activities but also good stability, and in this regard, the core-shell architectures offer a challenge in that Pt dissolution is known to occur during voltage cycling.

To deal with this latter problem, researchers are examining the effect nanoparticle structure on not only activity but also stability. In this regard, Adzic [31] and his coworkers have developed a promising new class of Pt-based catalysts having both high activity and high durability for the ORR. In particular, they have achieved this by depositing a Pt shell on an Au core doped with titanium oxide. They experimentally validated that the distinct microstructure of the Ti-Au core-Pt shell on a carbon support (Ti-Au@Pt/C). The nanoparticles had Au cores, Pt as the shell, with protective titanium oxide located on the low coordination edges and vertex sites of the nanoparticles. Electrochemical measurements using a rotating disk electrode indicate that the mass activity (A/mg_Pt) of the Ti-Au@Pt/C catalyst exhibits a 13-fold enhancement over that of commercially available Pt/C, whereas the specific activity (A/cm²_Pt) showed a fivefold improvement. More impressively, the catalyst exhibited excellent durability after 10,000 cycles of voltage cycling between 0.6 and 1.0 V.

Likewise improvements have been made with the development of catalysts with a “Pt-skin”-type surface structures formed upon annealing Pt-alloy nanoparticles. Stamenkovic, Markovic, and coworkers [32] have explored the activity and stability associated with this class of catalyst structure and have found increased activity and durability through the tailoring of nanomaterials in terms of size, composition, morphology, and architecture, with the objective of balancing the gains in stability with those in activity. They describe a strategy for achieving both high activity and stability as involving the de-alloying of Pt₃M nanoparticles, followed by annealing which creates a Pt-skin-like surface of several monolayers over a stable alloy-rich core, thereby achieving the desired reduction in PGM loading. Starting with PtM₃ particles, Han et al. [33] have systematically studied the effects of particle size, acid de-alloying, and post-acid treatment annealing on nanoporosity and passivation of the alloy nanoparticles and have achieved unprecedented levels of performance, exceeding the DOE 2017 performance targets in fuel cell tests using MEAs.

Underlying these demonstrated improvements in the activity and stability of Pt-alloy catalysts are important insights gained of the fundamental mechanisms at play at the atomic level. In particular, Stamenkovic, Markovic, and coworkers [34] have performed measurements of dissolution rates of surface atoms and have been able to correlate them to the kinetic rates of electrochemical reactions in real time, allowing for almost “atom-by-atom” structure-activity-stability relationships for Pt single crystals in both acidic and alkaline environments. They found that the degree of stability is strongly dependent upon the level of coordination of surface atoms, the nature of covalent and non-covalent interactions, the thermodynamic driving force for Pt ion speciation in solution, and the nature of the electrochemical reaction. All of these findings provide the opportunity to further understand and improve upon the activity and stability dependency on nanoparticle electrocatalysts.

Porous Electrodes

As discussed above, steady progress has been made experimentally in reducing the platinum loadings, particularly in the hydrogen/air fuel cells used in automotive applications. Gasteiger et al. [35] have discussed the strategies to achieve the nearly fivefold reduction in loading needed in order to achieve automotive cost targets. These strategies, which involve increasing the power density at high voltages (while reducing the Pt loading through an increase in the Pt activity and durability) and reducing mass-transport losses in the cell, should also be relevant to backup, motive, and stationary PEMFC applications. The current pathway involves optimizing the Pt-alloy nanoparticle size, composition, and morphology and process optimization, as discussed above.

Most porous electrodes are fabricated from a mixture of ionomer (for proton conduction) and nanoparticles of Pt-alloy catalyst supported on high surface area carbon (for electron conduction) that are all mixed together with a dispersion medium to form an ink-like random material. This ink is then either sprayed, doctored, or jetted onto either a microporous layer on the gas diffusion media or applied directly to the polymer electrolyte. In either case, the density and nature of the local environment at triple-phase boundaries are to a large extent left determined by the random nature of the ink, the interactions among its components, and the application process.

Recent focus has, therefore, shifted to understanding the performance loss of electrodes with low loadings at high current densities. In this case, the flux of reactants at the triple-phase boundary is necessarily high, thereby exposing limitations in mass transport at the nanoscale, including unexplained resistances that scale with Pt specific area current density. Weber [36] has discussed the nature of these unexplained transport resistances in detail and has proposed that most are caused by the ionomer thin film surrounding the catalyst sites, where confinement and substrate interactions dominate and result in mass-transport limitations. Gostick and Weber [37] have approached this problem by developing a model of the ion conduction in polymer electrolytes as a random resistor network. In the model he accounts for swelling as a function of humidity, as well as spatially varying conductivity within the nano-domains of Nafion®. Finally, he included the known confinement effect wherein a substantial drop in conductivity is apparent in films less than 50 nm. Although the model was stated as being overly simplistic, it agrees with experimental data which provide insights into some of the basic mechanisms that limit ion transport in the thin ionomer films within the porous electrode. An important and related finding with regard to the thin ionomer films is their response to annealing. Paul and coworkers [38] have studied the thickness dependence of thin films of Nafion on surface wettability and proton transport. They reported that thermal annealing induces switching of surface wettability of thin films (<55 nm) from hydrophilic to super-hydrophobic and a reduction in proton conductivity. These types of behavior obviously are important factors to consider in the fabrication of porous electrodes.

Recently, Orfanidi et al. [39] have explored the effect of ionomer distribution on oxygen mass transport and proton resistivity of the cathode catalyst layer in H2/air fuel cells by functionalizing the carbon support. The functionalization of the commercially available Vulcan XC72 carbon with (−NH_x) groups was intended to get a more uniform ionomer distribution through interaction with the ionomer’s sulfonic acid groups. Detailed experiments using MEAs prepared with differing amounts of ionomer to carbon ratio and ultralow loadings, followed by careful characterization accounting for voltage loss based upon proton resistivity and oxygen transport resistance measurements, provided insights into the major voltage loss contributions. By lowering the ionomer to carbon mass ratio, thereby reducing the effective ionomer film thickness, oxygen transport resistances were lowered, improving performance. Reducing the ratio further resulted in increased proton resistivity, indicating that low loadings require a balance between good ionomer distribution with a low ionomer/carbon ratio and adequate proton conductivity. These results are consistent with the recent Kongkanand and Mathias [40] review of the challenges of achieving high-power performance at low loadings in PEM fuel cells, in which they conclude that through a concerted effort in the design of materials and electrodes, engineers will be able to resolve the issue of “unexplained losses” at low loadings and achieve a significant improvement in cost and performance. A more detailed discussion can be found in the entry on PFSA thin films within this encyclopedia.

Nanostructured Electrodes

There is a different class of electrodes based upon highly ordered nanoscale structures that enable both increased performance and better utilization of the precious metals. These so-called thin-film nanostructured electrodes are highly structured by design and enable significant improvement in both performance and durability. To date, the most significant of these structures have been those under development by Debe [41] at 3M, wherein the structure consists of a dense forest of crystalline organic whiskers with areal densities on the order of 3 to about 10 billion whiskers per cm², highly oriented with their long axis normal to the substrate. The whiskers are nanoscale in size (roughly 50 nm by 30 nm by 300–1000 nm) and provide an interesting support for the Pt catalyst. An image of these whiskers, coated with a platinum catalyst, is shown in Fig. 1 below.

Fig. 1

3M’s Pt-coated NSTF electrode

High-resolution TEM studies of the sputter that deposited thin catalyst film on the whiskers indicate that the catalyst films grow as polycrystalline layers that expose highly oriented crystallites. The resulting catalyst “whiskerettes” growing on and at an angle of roughly 70° to the surface of crystalline whiskers represent a very interesting hierarchical nanostructure that is providing significant improvements in performance. As proposed by Debe [42], the source of the activity gain of such a system might have its contribution rooted in the nature of the metal/support interactions.

How well do these nanostructured thin-film (NSTF) electrodes meet the expected performance targets? Current DOE maturity targets for light-duty vehicle applications specify performance in terms of platinum group metal loading of <0.125 mg/cm²_MEA, a durability target of least 5000 h (at temperatures <80 °C), a mass activity of at least .44A/mg_Pt @ .9V_IRfree, and a specific activity of 720 μA/cm²_Pt @ .9V_IRfree. At a recent DOE Annual Merit Review, 3M reported that they had achieved 0.19 mg_Pt/cm²_MEA loadings on 400 cm² cells in short stacks, 5000 h of durability (20 μm, 850 EW with no stabilizers), 0.24 A/mg_Pt with PtCoMn alloy (0.40A/mg_Pt with a new Pt₃Ni₇ alloy that needs further development), and specific activity of 2100 μA/cm²_Pt. This is a substantial improvement in performance over the conventional Pt/C paradigm that has dominated fuel cell MEA designs in the past. There are, however, water management issues with the thin electrode structure, as the reduced thickness of the electrode results in flooding more easily under cold, humid conditions. 3M has addressed this by proposing the use of a unique set of operating conditions and GDL properties on the anode side with an additional cathode interfacial layer in order to reduce the impedance of water transport from the cathode to the anode as a way to improve the “operating robustness” with respect to cathode flooding [43]. In addition to this issue, the NSTF system is more susceptible to the cumulative effects of contaminants on the long-term stability at operating potentials. In particular, even though the NSTF electrodes show a 20–50X reduction in fluoride emission rates (due to their high specific activity for the ORR), they are more susceptible to poisoning by the products of membrane degradation (due to their low electrochemically active surface area ECSA) [44]; 3M is addressing this by pursuing more stable membranes.

It is important to note that in the pursuit of other types of highly ordered nanostructures, researchers are also taking aim on the issue of the catalyst support from the perspective of its stability. As will be discussed below, the usual carbon materials are used to support the catalyst nanoparticles corrode under certain operating conditions associated with startup and shutdown of the fuel cell, or conditions of localized fuel starvation, and so it is desirable to have a support which is more stable under these dynamic situations. Accordingly, researchers are developing nanostructures in which the typical carbon material is replaced with materials with greater stability. Included in this set of support materials are the use of carbon nanotubes [45] and ceramic materials such as TiO₂ [46]. It is anticipated that advanced nanostructured electrodes will try to emulate the excellent results and the lessons being learned in the case of the NSTF work underway at 3M, while avoiding problems of water management and contamination.

Low- and Intermediate-Temperature Proton-Conducting Polymer Electrolyte Membranes

There is also significant research underway addressing the development of alternative classes of polymer electrolytes for fuel cells. For example, sulfonated aromatic polymers, sulfonated polyimides, proton-conducting membranes carrying phosphonic acid groups, and polyphosphazene polymer electrolytes are all being examined as lower cost, more durable alternatives to PFSA-based membranes. There are excellent reviews of the progress being made in the development of these acid-based polymer electrolytes [47].

As noted, PFSA-based polymer electrolyte membranes used in current PEMFCs require thermal and water management systems to control temperature and keep the membrane humidified. These extra components increase the weight and volume of the fuel cell system and add complexity. The cost and complexity of the thermal and water management systems could be minimized if the fuel cell operated at higher temperatures (up to 120 °C) and at lower relative humidity (RH). In this context it is important to stress the DOE requirement for heat rejection, which reflects itself in the metric of a Q/ΔT that has recently been adopted. Currently, this metric is 1.45 kW/°C. The US Department of Energy has therefore initiated a major effort to develop new membranes that can operate at temperatures up to 120 °C without the need for humidification. While it is desired that the fuel cell membrane operates without external humidification, it is recognized that the water generated by the fuel cell itself can be utilized to provide some humidification of the membranes. It is expected that under proper operation, recycling product water from the cathode to the inlet air can provide roughly 25% RH inside a stack operating at 120 °C with inlet feeds at a water vapor partial pressure of <1.5 kPa. This reduces the burden for performance with very low water content, but operation at this low RH still remains a major challenge.

A variety of strategies have been pursued. It is known from extensive molecular level modeling of Nafion® that membrane microstructure has a substantial effect on conductivity. With block copolymers, McGrath [48] observed that as the block length increases, the performance under partially hydrated conditions increases, suggesting the presence of hydrophilic domains at higher block lengths through which protons can be transported along the sulfonic acid groups and water molecules. Utilizing block copolymers consisting of hydrophilic oligomers and hydrophobic perfluorinated oligomers, McGrath and his coworkers have prepared a block copolymer that exhibited higher proton conductivity than Nafion® 112 at 80 °C at all RH values (30–95%).

FuelCell Energy has attempted a rather sophisticated approach utilizing a four-component composite consisting of a copolymer, a support polymer, a water retention additive, and a protonic conductivity enhancer. The copolymer, which is intended to provide the basic building block for the membrane, is an advanced perfluorosulfonic acid with significantly higher conductivity than state-of-the-art polymers. The support polymer is intended to give a stable cluster structure and to enhance mechanical properties. The functionalized additives are designed to retain water at low RH conditions and to enhance the composite membrane’s proton conductivity by providing an alternate proton conduction path for the efficient transport of proton at high temperature as well as at subfreezing conditions. Progress to date has led to a membrane with conductivity about a factor of 2 better than Nafion® 112 at 120 °C [49].

3M has recently initiated a study of the use of perfluoroimide acid (PFIA) ionomers in membranes reinforced with electrospun nanofibers made of fluoropolymers [50]. These experimental membranes are exhibiting excellent properties against the DOE targets. However, these are relatively new results, and testing has revealed an unexpected performance loss during open-circuit voltage (OCV). The mechanism of this oxidative decomposition of the PFIA is not yet fully understood.

All of these approaches still rely on water for conduction. Systems utilizing phosphonic acids, heteropolyacids, protic ionic liquids, and heterocyclic bases do not rely on water for conduction. One of the major issues for these systems, however, is that the acid or base aiding proton transport is generally water soluble. The acid or base must be immobilized for use in transportation applications where condensation of liquid water under some of the operating conditions is inevitable. However, enough mobility must be retained by the active group to be able to participate in proton conduction.

In addition to these more advanced studies, progress has been made in developing improvements to the current PFSA polymer electrolyte membranes currently employed in PEMFCs, particularly with regard to degradation caused by mechanical failure due to operationally induced cyclic stresses and from peroxide generated either electrochemically or chemically as the result of oxygen crossover to the anode. With regard to the former, mechanical reinforcement is often used to improve durability as well as to enable thinner membranes. With regard to peroxide, it is generally accepted that the membrane degradation is the result of the subsequent formation of hydroxyl radicals that attack the side chains of the polymer [51,52,53]. Here again, dry conditions greatly accelerate the rate of degradation.

What is not generally agreed upon is the location and origin of the hydrogen peroxide formation. Two theories are dominant. One is that the peroxide is formed mainly at the anode due to the diffusion of oxygen across the membrane, at which point the peroxide can form either chemically or electrochemically by the reaction of oxygen with hydrogen. This hypothesis is supported by the fact that the peroxide yield is greatly enhanced in the anode potential region (<0.2 V) of the ORR, as evidenced by rotating ring disk electrode experiments [54]. The other is that the peroxide is formed at the cathode. Experimental results from several sources support this hypothesis. Liu et al. [55], Mittal et al. [56], and Miyake et al. [57] have commonly reported that the membrane degradation rate is higher for an MEA catalyzed only at the cathode side than for one catalyzed only at the anode side. This fact seems inconsistent with the hypothesis that peroxide is formed at the anode. Yu et al. [58], in an effort to identify the origin of the radicals which decompose the membrane, performed a series of experiments after which they concluded, based upon energy dispersive X-ray analysis and IR spectroscopy, that the degradation begins on the cathode and progresses inward. They also found that the formation of radicals within the membrane or near the anode was very low or absent. They postulated the formation of undesirable free radicals involving the catalyst, similar to the hypothesis of Liu and Mittal mentioned above.

Any model of the degradation process would need to propose that formation of hydrogen peroxide forms by distinct mechanisms in the cathode and anode. The commonly accepted mechanism is that the peroxide then forms radicals through Fenton reactions involving metal-ion impurities. The radicals then participate in the decomposition of reactive end groups in the membrane, to form, among other species, hydrogen fluoride, which can be detected in the product water. Higher fluoride release rates correlate with higher rates of membrane degradation. The degradation occurs through the “unzipping” of the polymer backbone and the cleavage of the polymer side chains. The two conditions which accelerate this degradation mechanism are dry conditions and high temperatures, thinning the membrane until “pinholes” form, allowing gas crossover and cell failure. Therefore, as emphasized earlier, PFSA-like polymer electrolyte materials must be kept under good temperature control and maintained hydrated.

The susceptibility to peroxide radical attack has been attributed to a trace amount of polymer end groups with residual hydrogen containing terminal bonds [59]. It is at these sites that decomposition is initiated. DuPont has reported that the number of hydrogen containing end groups can be reduced by treating Nafion® with fluorine gas. 3M has been improving their PEMs by modifying the end groups in the membrane. 3M is investigating modified polymer structures to try to control membrane morphology and has developed a new ionomer with a shorter side chain than standard perfluorosulfonic acid (PFSA) membrane ionomers without the pendant CF₃ group. They have reported that this structure provides a higher degree of crystallinity and allows for lower equivalent weight membranes with improved mechanical properties and durability under hot, dry conditions. Solvay Solexis also has a short side chain PFSA membrane, Aquivion™, that resists peroxide radical attack. In addition, companies are not only modifying the chemical structure but are also incorporating proprietary stabilizers to mitigate the effect of peroxide and the attack by its radicals.

In addition, companies are not only modifying the chemical structure but are also incorporating proprietary stabilizers to mitigate the effect of peroxide attack. Trogadas et al. [60] have investigated the efficacy of cerium oxide as a regenerative free radical scavenger. On the basis of measuring the fluoride emission rate (FER), they determined that membranes prepared with nanosized ceria particles from 0.5 to 3 wt. % lowered the FER by over one order of magnitude. Likewise, Coms et al. examined the use of cerium and manganese ions and found that the FER was reduced by factors of approximately 1000 and 100 with Ce³⁺ and Mn²⁺, respectively. As a result, companies are not only modifying the chemical structure of the PFSA but are also incorporating proprietary stabilizers to mitigate the effect of peroxide and the attack by its radicals. Recent research is even looking at additives such as microcapsules of Nafion solution that rupture to create self-healing composite membranes [61].

High-Temperature Polymer Electrolyte Membranes

Early work on the properties of proton-conducting acid polymer blends was carried out by Lassegues et al. [62, 63]. The first detailed study of the effect of doping a high-temperature polymer membrane such as polybenzimidazole with a phosphoric acid in order to achieve proton conductivity in a range suitable for use as a fuel cell was conducted at Case Western Reserve [64]. Polymer electrolyte membranes which do not rely on water as the basis of their proton conductivity offer the potential advantage of simplifying the fuel cell system, if these systems can be operated without complex external humidification schemes. Furthermore, because these systems run at high temperature, they can tolerate higher levels of CO in the reformate, thus further simplifying the system in terms of reforming and control. In addition, the higher operating temperature simplifies thermal management issues and provides a source of high-quality heat for combined heat and power applications.

Advent has recently acquired the rights to PBI-based membranes from BASF, which has restructured its fuel cell business to focus on catalysts and absorbents for fuel cell systems in its catalyst division. Consequently, Advent [65] is developing both PBI-based proton-conducting membranes for high-temperature stationary CHP fuel cell applications, as well as its own high-temperature polymer electrolyte membranes for use in HTPEM fuel cells. These latter are based on aromatic polyether polymers and copolymers containing polar pyridine moieties in the main chain which have been developed by Advent Technologies, with the idea of creating acid-base interactions in order to obtain high proton conductivity. The developed polymer products are called Advent TPS®. Such materials combine the excellent film-forming properties with high mechanical, thermal, and oxidative stability and the ability to be doped with phosphoric acid. Highly conducting polymer electrolyte membranes were produced after treatment with phosphoric acid amounts that could be controlled by varying the pyridine-based monomer content. The polar pyridine groups strongly retain the phosphoric acid molecules, due to their protonation, thus inhibiting leaching out of the phosphoric acid. The performance based on Advent TPS® MEA operating at 180 °C with pure hydrogen or reformate with different CO contents and air feed gases is shown in Fig. 2. Here again, because of the high operating temperature, this system tolerates up to 2 volume percent CO poisoning without a significant decrease of performance. The performance shown in Fig. 2 is much improved over that reported earlier by Advent, and this is due to the fact that the TPS®-based MEA is now based upon the electrode technology developed at BASF. (DeCastro E, Private communication, 2017).

Fig. 2

Polarization curves of advent TPS® (type ABM) at 180° with H₂ or reformate gas (71%H₂, 2.1%CO, and 26.9% CO₂) and air, under ambient pressure (λ_H2 = 1.2, λ_air = 2)

As is well known, and as discussed by Neyerlin and others, there is an increase in polarization losses with phosphoric acid-based fuel cells, in comparison to low-temperature PFSA-based fuel cells, and this effect is due to the presence of phosphoric acid and/or its anions that adsorb onto the surface of the catalyst [66]. Because of this, high-temperature stacks based upon phosphoric acid-based polymer electrolyte membranes are somewhat larger in order to get the same power output. The overall question is whether the benefit in system simplification overcomes the need for these larger stacks, so that there is an overall net system benefit. Reducing the effect of the adsorbed anion species would, of course, have significant benefit at the stack and system levels. One of the approaches suggested by Strmcnik and coresearchers for dealing with this effect is to utilize nanoparticles which are rationally designed to prevent anion adsorption by patterning of platinum surfaces with cyanide absorbates that can block sites for adsorption of spectator electrolyte anions [67].

A more detailed discussion of PBI-based membranes can be found in the entry on PBI in this encyclopedia.

Alkaline Polymer Electrolyte Membranes

There are also efforts underway to develop alkaline-based polymer electrolytes, which enable the replacement of precious metal catalysts for the ORR by catalysts based on inexpensive transition metals, such as Co, Fe, and Ni. This requires pursuing the development of alkaline technology by transitioning from the proton-conducting polymer electrolyte membrane and ionomer to an OH- ion-conducting membrane and ionomer. It is also argued that the alkaline ion-conducting polymer electrolyte membrane provides a much more benign chemical environment which lowers the risk of instability caused by the corrosion processes prevalent in the highly acidic environment of the PEM fuel cell.

Alkaline-based polymer electrolyte membranes are relative newcomers on the scene. It has been widely believed that the quaternary ammonium hydroxide functional group (RN₄⁺, OH⁻), which is the one used in most anion exchange membranes, is “self-destructive,” because the OH- ion is likely to attack the RN₄⁺ cation. In addition, the specific conductivity of an OH⁻-conducting ionomer was suspected of being at least a factor 3–4 lower than that of the H⁺ conducting ionomer, thereby setting a limit on power output. It was also suggested that, as the AEMs developed to date have been based on hydrocarbon, rather than fluorocarbon backbones, the preparation of effective and stable membrane/electrode assemblies would present a significantly tougher challenge vs. the case of PFSA ionomers. Historically, there has always been a concern having to do with the effect of CO₂ from the air feed on the OH- conducting ionomer. Since conversion of the OH- ion in the alkaline ionomer to bicarbonate (and/or carbonate) ion upon exposure to air, cell performance suffers from the effects on both lower ionomer conductivity and, particularly, the kinetics of electrode processes. Hence, one of the particularly interesting system applications of this technology is for operation on pure oxygen, as would be the case, for example, with a unitized AEM-based electrolyzer/fuel cell operating off of a renewable energy source [68]. (This problem is less severe in KOH systems since it does not form insoluble bicarbonate.)

Tokuyama, a Japanese company specializing for a long time in membrane technology for electrodialysis and desalination, has undertaken development of AEMs in OH- form, targeting fuel cell applications. The commercially available membranes, A201 and A901, are currently the standard against which other AEM designs are compared, much like Nafion is the gold standard for acidic membranes [69]. The ionic conductivity of the Tokuyama AEM, while quite satisfactory for many applications, is nonetheless inferior to that of Nafion. Measurements of the water uptake, ionic conductivity, and swelling properties of the A201 membrane have been measured and indicate that, even more so than Nafion®, ionic conductivity is a strong function of the hydration state of the membrane [70]. Hence, water management and water transport issues are also a predominant concern in the design of fuel cell systems based on alkaline membranes, especially since water is needed as a reactant at the cathode in order to reduce oxygen to create the OH- needed for the anode reaction [71]:

$$ \mathrm{Anode}:2{\mathrm{H}}_2+4{\mathrm{OH}}^{-}\to 4{\mathrm{H}}_2\mathrm{O}+4{\mathrm{e}}^{-}\; \; {\mathrm{E}}_{\mathrm{a}}=-828\; \mathrm{V} $$

(10)

$$ \mathrm{Cathode}:{\mathrm{O}}_2+\; 2{\mathrm{H}}_2\mathrm{O}+4{\mathrm{e}}^{-}\to 4{\mathrm{O}\mathrm{H}}^{-}\; \; {\mathrm{E}}_{\mathrm{c}}=0.401\; \mathrm{V} $$

(11)

Efforts are underway to improve upon the performance of the Tokuyama membranes in terms of its conductivity, water transport properties, and durability. Ren [72] and coworkers have reported improved performance using an approach which mimics the structure of Nafion in the sense that it has poorly ordered bi-continuous hydrophobic and hydrophilic phases. They achieved this by using a styrenic diblock copolymer with a quaternary ammonium-functionalized hydrophilic block with a cross-linkable hydrophobic block. They attributed the improved performance to the reduced tortuosity of the conductive paths, while the hydrophobic network maintains the mechanical integrity, preventing excessive water uptake. Additionally, Pivovar [73] and coworkers are pursuing the use of perfluoropolymer chemistry because of its chemical robustness, enhanced water transport, and conductivity properties compared to hydrocarbon polymers. They are pursuing different strategies to tether different cations to the perfluorosulfonyl fluoride precursor.

There is one final aspect of AEM-based fuel cells that deserves mention. Whereas in acid-based PEM systems the ORR is the predominate source of activation losses, with the hydrogen oxidation reaction being extremely facile, the same cannot be said of alkaline-based systems. For in spite of the fact that non-PGM catalysts can be used for the ORR in alkaline systems, the HOR and HER reactions prove to be problematic at the low Pt loadings normally used for the HOR and HER in acidic systems. Indeed, Durst et al. [74] have reported on the effect of PH effect on the HOR and HER for Pt, Ir, and Pd carbon-supported catalysts, with the data illustrating a 100-fold decrease in activity on all these surfaces when going from low to high pH, pointing to the need for increased understanding of the effect of the materials on binding energies.

Additional information on alkaline membrane fuel cells can be found in the entry on this subject in this encyclopedia.

Fuel Cell System Design

The process of designing of a fuel cell system follows the general design process for any commercial product in which a phase-gated, product development process (PDP) guides activities from initial concept to launch. The PDP is one of the most interesting and well-studied business processes, and usually ends up being tailored to each company’s appetite for speed and time to market, balanced by its desire to mitigate risk. Product development is very much a social science, one that integrates business, technology, engineering, and human behavior [75]. Many attempts have been made to study its characteristics and improve the manner by which products are brought to market [76]. Even today it remains a subject of research in most business schools. Before these steps are undertaken, however, usually a product strategy is developed which aligns to both the business strategy and the business model of the organization. This product strategy generally takes the form of a platform strategy, which shows the relationship between multiple products expected to be developed and launched over a 5–10-year time horizon [77].

System Requirements and Architecture

The design of the fuel cell system really begins with the “voice of the customer” (VOC). For fuel cell systems, there can be many potential customers: the end user, the OEM, or an intermediate service provider. In any case, the voice of the customer is meant that quality expectation is taken directly from the customer, properly evaluated and deployed within the product development process. One such technique for doing this is the well-known quality function deployment (QFD) process which, if properly used, enables the VOC to be deployed all the way to the factory floor. Without knowing and agreeing on the key customer requirements upfront, the fuel cell system design process cannot be successfully completed.

A very much abbreviated and simple example of one of the first steps in the process of developing fuel cell system specifications is shown in Fig. 3. This example is for a residential stationary fuel cell application, but similar requirements are relevant in most other fuel cell applications. Here is shown the high-level VOC, gathered by some technique such as the use of customer interviews or focus groups and the system attributes that will influence their achievement, arranged in order of importance. As shown, some of the key system attributes that are critical to quality (CTQs) and that influence the customer’s demand for clean, quiet, safe, and affordable energy are related to meantime between failure, major component life, load-following capability, turndown ratio, operating and maintenance cost, efficiency, startup time, audible noise, and agency compliance.

Fig. 3

Example of a limited set of attributes for a residential fuel cell system

Of course, there are a host of other attributes which need to be specified and met, including the expected ranges in ambient temperature and operating altitude, expected number starts and stops, and the quality of air and water needed for operation. All of these external requirements drive the selection of a technology set which, when optimized, needs to provide for acceptable operating latitude over the life of the system. The US Department of Energy’s Office of Energy Efficiency and Renewable Energy (EERE) has performed a great service in developing a consensus view of some of the more critical CTQs needed for PEM fuel cell systems for a variety of applications, including those for transportation as well as stationary fuel cell systems with combined heat and power. Convening OEM manufacturers, value chain suppliers, and researchers, the EERE set technical targets for 1–25 kW_e residential and light commercial combined heat and power (CHP) and distributed generation (DG) fuel cell systems operating on natural gas delivered at typical residential distribution line pressures. The most recent set are shown below in Table 1.

Table 1 DOE technical targets for fuel cell systems for stationary (CHP) applications (2015)

Along with this set of targets, the EERE developed a companion low-temperature PEM reference architecture against which it could track progress against the targets. The most recent status for 2015 is given in the table, from which it is seen that the performance in cost and reliability is still off target. It is worth noting that there is considerable “expert” opinion on whether these targets can be met with either low-temperature or high-temperature PEM systems, as the combination of cost, efficiency, and durability targets is quite demanding. In the face of this uncertainty, Japan’s Ene-Farm program supported the installation of around 200,000 small micro-CHP PEM units in the <1 kW size, mostly from Toshiba and Panasonic. It is anticipated that government subsidies will continue until 2019, during which time cost reduction efforts through system design simplification and standardization of components will continue [78].

It is clear that one of the key requirements for a successful platform strategy is a well thought out platform architecture. The platform architecture enables that product strategy to be executed. As an example, consider the high-level architecture shown in Fig. 4 below. Here one sees the functional breakdown of the system into the major components, which in turn defines requirements for each module in the system.

Fig. 4

A high-level fuel cell system architecture

This figure shows several major subsystems typical of a PEM stationary system. Included are the reactant processing module (RPM), which determines the pressure, temperature, flow rate, and composition of the reactant streams which are delivered to the power generation module or PGM. Included in the RPM are the desulfurization, reformer and shift reactors, and the required number of heat exchangers which are needed to manage the reforming process. More will be said of this below. The PGM will include the stack and its manifold, and any other subsystems needed to ensure the stack can function properly, such as the subsystems that provide for cathode recirculation and water management. Also shown in this figure is a hydrogen pump module (HPM) that can be used to purify and compress hydrogen for storage in the chemical energy storage module (CESM) for refueling a hydrogen fuel cell car and a thermal management module (TMM) that provides the methods of capturing and directing the flow of heat to either a heat load (such as the RPM) or heat sink. Electrical energy generated by the fuel cell can either be stored in the electrical energy storage module, EESM, or passed through to the power conditioning module (PCM) where it is converted to high-quality power for alternating current loads. A control module (CM) ensures that the required electrical state of the system is achieved over time, as defined by the energy management module (EMM). The CM provides the proper control of the pressure, temperature, flow rate, and composition of the fuel and air streams going into and out of the system. The result of all of this is made visible to “customer” through an appropriate user interface module (UI).

This is a high-level architecture. If the detailed lower-level subsystem requirements that support this “meta-system” are well thought out, it is possible to develop multiple products off of the same set of hardware and software components, thereby minimizing the product acquisition spending (PAS). For example, there at least four products that can be derived from this architecture, including a home refueling system, a residential CHP system (eliminating CESM and HPM), a hydrogen fuel cell system for backup applications (further eliminating the RPM system), and a stand-alone hydrogen generator (retaining only the RPM, TMM). Of course, in each case, the controls and user interface have to be modified for the intended application. Furthermore, system-level specifications need to be compatible over the ambient temperature range, etc.

Where the intended range of applications does not allow for platform architecture at the meta-system level, the platform products then need to be defined at the system level. In fact, the concept of platform-based architecture can be pursued at the component as well as the system level. Ballard®, for example, has defined its platforms at the stack level. It is currently advertising five stack platforms. Each stack platform has a different voltage range, reactant pressure (air and fuel), and cell aspect ratio, optimized for the intended application, whether it be for backup (telecommunications), motive power (forklift trucks), transportation (buses), or residential cogeneration.

While the overall architecture defines the interrelationships among form, fit, and function, systems engineering supports the development and achievement of the product specifications. The systems engineering required to enable a product platform strategy is generally quite extensive, requiring many iterations using simulations to guide the final design at the system, subsystem, and component levels. In general, the requirements for a fuel cell system are quite extensive and complex, and these need to be defined, allocated, and enforced all through the development process.

The key steps in the product development process are shown below in Fig. 5. Most activities are focused on two types: the systems engineering associated with requirements definition, ending with components specifications, and on the actual design and verification process, ending with a successful in-house design verification test (DVT), a field readiness demonstration test (FRDT), and a factory acceptance test (FAT) to validate launch readiness. A key step in the overall process is the development of the systems requirement document (SRD), which enables the development of the process and instrumentation diagram (P&ID), and its companion, the process flow diagram (PFD). The PFD, which is the output of the process engineering analysis using computational tools such as ASPEN®, captures the concepts in flow sheet format and summarizes the heat and material balances characterizing the system design. In short, the PFD, which is the link between concept and reality, defines the critical parameters at each step in the process.

Fig. 5

Product development process map

Early on in the process, it is useful to organize the systems engineering effort in pursuit of a collection of interlocking functionally important topics which are system-level attributes that need to be allocated to subsystems within the system. One example would be the allocation of efficiency targets between the various major modules. Another high-level example of such an allocation would be the “flow down” of the requirements for cost and reliability. Here the percentage of total allowed cost and failure rate is allocated among the major modules, based upon an initial test of reasonableness followed by refinement during the development process as further information becomes available. Clearly, the initial allocation of the functionally important topics is important, because it drives engineering behavior and subsequent design modifications and, eventually, time to market. The system may have other high-level requirements, such as size, weight, noise, and appearance, and these early requirements are met through systems engineering efforts. Management of the allocation process is achieved through the use of artifacts such as input/output/constraint charts (IOCs) that keep track of both functional (signal) and dysfunctional (noise) attributes of each subsystem, subject to the constraints imposed by the overall system. They track the interactions between subsystems in terms of the critical parameters, and are used to define and guide the robust design optimization process, eventually becoming the “contract” between the subsystem teams and the system engineers. Again, a combination of careful testing and simulation is required to ensure convergence to the subsystem- and system-level allocations.

An Example: Reforming Considerations

In many cases, hydrogen at required purity is not normally available, in which case it must be obtained by the reforming process of stripping hydrogen off a hydrocarbon fuel such as methane, propane, ethanol, and liquid petroleum gases (LPG) or kerosene. Generally, there exist sulfur species that are naturally occurring in the fuel (COS, H₂S) or are added as odorants (mercaptans, tetrahydrothiophene, THT). Since sulfur poisons the catalysts in the system, the first step in the reforming process generally involves the removal of harmful sulfur-containing species. The level to which sulfur needs to be removed depends upon the operating temperature, but in general, it is safe to assume that for PEM systems, the level needs to be less than about 50 parts per billion (ppb) [79]. To achieve this level of sulfur impurity, there are several available technologies. At ambient temperatures, these include physical adsorption beds using activated carbon and zeolites and chemisorption beds incorporating nickel-, copper-, and iron-based sorbents. At higher temperatures, one alternative is to use hydrodesulfurization with Co-, Mo-, and Ni-based catalysts. Another option is thermal swing absorption. Ideally, it is desirable to have a single technology in order to have a fuel flexible system that enables the fuel cell system to be sited in a variety of geographical locations. Seasonal and regional variations in sulfur content, as well as catalyst material cost and toxicity, make this a difficult engineering problem. Each of these technologies is well known, and each has its strengths and weaknesses. All add considerable capital cost and complexity to the system, not to mention annual maintenance cost which can also be a significant expense. Consequently, there is a need for continued research into the discovery and development of effective, low-cost desulfurization for fuel cell applications requiring them.

Assuming that the fuel has been adequately desulfurized, there are essentially three alternative processes for extracting hydrogen: steam reforming (SR), catalytic partial oxidation (CPO), and autothermal reforming (ATR). SR is an endothermic process. Heat must be supplied to the reactor, and this is normally accomplished by combusting an additional amount of fuel in a separate, but thermally integrated, reactor. Usually this reactor burns the unused hydrogen which flows through the anode chamber of the fuel cell stack. This component is referred to as the “anode tail-gas oxidizer” or ATO. The ATO is generally considered an integral part of the reactant processing module.

The steam reformation reaction is reversible, and the product gas is a mixture of hydrogen, carbon monoxide, carbon dioxide, and water vapor. In the case of methane, the steam reforming step is

$$ {\mathrm{CH}}_4+{\mathrm{H}}_2\mathrm{O}\to \mathrm{CO}+{\mathrm{H}}_2\mathrm{O}+2{\mathrm{H}}_2 $$

(12)

The amount of CO in reformate is a system-level critical parameter, as CO interferes with the hydrogen oxidation reaction by blocking reaction sites on the platinum catalyst surface. To remove the high levels of CO, this reformate stream is then fed to a “shift” reactor which reduces the CO concentration and makes additional hydrogen:

$$ \mathrm{CO}+{\mathrm{H}}_2\mathrm{O}+\to {\mathrm{CO}}_2+{\mathrm{H}}_2 $$

(13)

Following the “shift” reaction, the reformate is generally passed through a preferential oxidation (PROX) step to further reduce the CO content to an acceptable level so as not to poison the anode catalyst. For low-temperature PEM systems, this level is typically less than 10 ppm. To increase the tolerance to the residual CO, ruthenium is added to the anode catalyst layer, but even this is not sufficient to deal with the amount of CO in the reactant stream. In the case of low-temperature PEM systems, there is generally a small amount of “air bleed” fed into the reactant stream to oxidize any residual levels of CO that adsorb onto the catalyst. It is desirable to keep the air bleed to a minimum to reduce the formation of peroxide. Pulsed air bleed techniques have been studied to achieve this purpose [80]. Depending on the reformer design, there may be a high-temperature as well as a low-temperature shift, as well as a single- or two-stage PROX.

Unlike steam reforming, CPO is an exothermic process. It is essentially a combustion but with less than stoichiometric amount of oxygen. As in SR, the reformate gas must go through a shift reaction to produce more hydrogen, though a CPO produces less hydrogen than SR, and unlike SR the product gas contains relatively large amounts of nitrogen from the air used in the process.

Because CPO is an exothermic process, and SR is an endothermic process, these two can be combined in the “autothermal” reforming (ATR) process. The overall reaction of an autothermal process for methane, with the shift reaction included, is

$$ {\mathrm{CH}}_4+\upchi \; {\mathrm{O}}_2+\left(2-2\; \upchi \right)\; {\mathrm{H}}_2\mathrm{O}\to {\mathrm{CO}}_2+\left(4-2\upchi \right)\; {\mathrm{H}}_2 $$

(14)

In this equation, χ is the number of moles of oxygen per moles of fuel. Its value determines if the reaction (14) is exothermic, endothermic, or thermoneutral. In practice, the value is chosen to avoid the risk of carbon formation in the reactors, and the amount of steam added is usually in excess of that theoretically required. All of these process steps use specially designed catalysts to facilitate the reaction steps. Catalyst designs are in many cases proprietary, although in most cases references to catalyst materials used in similar reactions can be found in the literature.

For many applications requiring the need to reform a logistical fuel, the reactant processing module represents a significant percentage of the total system cost. This is because the reactant processing module must provide reformate with the proper level of humidity over the entire load range, the required level of CO within the constraints of a single-stage partial oxidation (PROX) step, have rapid startup and load-following capability with near-zero emissions for NOx, and this all without resulting in coke formation. To meet the efficiency requirements of the overall system, the reactant processing module must also be thermally integrated into the rest of the fuel cell system. This is a very demanding set of requirements, and the cost of the reactant processing module typically ends up being about one-third of the total system cost.

It is well known that most Japanese residential μCHP fuel cell systems use steam reformers, principally due to their high efficiency. Adachi et al. [81] have recently published the results of their effort to design an autothermal reforming fuel processor for a 1 kW μCHP residential system. They designed and bench tested an ATR system capable of achieving 80% efficiency (on a higher heating value basis) and a reformate containing 48% hydrogen (dry basis) and less than 5 ppm CO. They have studied the ATR principally to address the issues around startup time and the daily start/stop requirements.

A simplified schematic of an ATR-based reformer system is shown below in Fig. 6. Shown in this figure are the desulfurization subsystems, the ATR reactor, a low-temperature shift (LTS) reactor, PROX, ATO, low-pressure steam generators to meet the reformate due point requirements, and the appropriate heat exchanges.

Fig. 6

Simplified schematic of an ATR-based reformer subsystem

In practice, there are several critical operating requirements placed on the reformer system. First, the reformate must contain a negligible amount of carbon monoxide (and oxygen) to avoid poisoning the anode catalyst. Likewise, the fractional conversion of fuel in the reformate needs to be specified in order to minimize unburned hydrocarbon emissions. From a system control perspective, it is also desirable to deliver a predictable amount of hydrogen for a given fuel input. Finally, because of the dependence of the PEM on the level of hydration, the reformate dew point must also be controlled to a specified value. Given these specifications, Feitelberg and Rohr [82] have shown that two variables, the steam to carbon and oxygen to carbon ratios, define a unique “operating line” for the fuel processor. The operating line defines the required relationship between steam to carbon and oxygen to carbon ratio for a reformer that meets fuel conversion and reformate dew point specifications. An example of the operating line [83] for several dew points is shown in Fig. 7. Operation above this line means that the dew point is greater than that specified, implying that energy has been wasted vaporizing more water than is needed. Operation below the operating line means that the reformate dew point is lower than specified, implying that one should expect reduced MEA life or premature failure. When the PEM system is operating on the line, and the heat generated in the ATO just balances the heat consumed in the reformer, the reforming process has achieved the maximum theoretical efficiency.

Fig. 7

Reformer operating line analysis [83]

In this regard, Feitelberg and Rohr’s analysis has shown that the maximum theoretical efficiency of a SR-based system is only about one percentage point higher than the maximum theoretical efficiency of a typical ATR-based fuel processor. Hence, the ATR technology seems to be a good compromise in performance and efficiency.

Technical Readiness

A key milepost in the development of a fuel cell (or any other) system is the achievement of technical readiness; that point in the development cycle when commitments to volume manufacturing can be made with confidence. For this purpose, it is meaningful to define this state of knowledge as:

Technical readiness is achieved when all critical failure modes are known, and when the critical parameters that are required to avoid those failure modes over the required operating range can be achieved through the intended interaction of specified parts which can be manufactured with acceptable process latitude.

In discussing PEM fuel cell systems for stationary, transportation, and portable applications, it is worthwhile to do so in terms of defining those conditions needed to not only meet the intended application but also to avoid the critical failure modes.

Freezing

One of the customer-driven requirements associated with fuel cells intended for stationary and transportation applications is the ability to not only survive subfreezing temperatures but to gracefully startup from subfreezing conditions and operate over a large number of expected freeze/thaw cycles. Inasmuch as water is the product of the fuel cell reaction, it is normally carried into the stack with humidified reactants and is fundamental to the transport of protons in the membrane. Most of the work on defining the operating conditions needed for surviving freezing conditions has focused on the behavior of the water within the principle stack components: manifold, plates, flow fields, and MEA. There are three types of water in PFSA-type polymer electrolyte membranes. Water that has strong interactions with the ionic groups in the polymer can withstand temperatures well below the normal freezing point of water without freezing. Water that finds itself within the nanoscale channels of the polymer electrolyte has a freezing temperature that is dependent on the size and nature of the channel, which in turn is dependent on the degree of hydration of the polymer. This water will therefore have a range of temperatures over which it will freeze, with some of the water having its freezing temperature depressed by the presence of the hydrated ions. Finally, there is “free” water that behaves like bulk water. The result is that there is a distribution of freezing points within the PEM, with no specific “freezing point.” The PEM proton conductivity, for example, varies continuously over a wide range of temperatures, depending on the level of hydration. The proton conductivity below freezing can be an order of magnitude lower than that for normal operation [84], indicating that a large fraction of the water in the membrane is frozen. However, even in this state, the polymer electrolyte membrane is capable of carrying current, increasing with increased thawing. Also, the tightly bound water, while not freezing, has a very low conductivity, since it is not that mobile, as though it were frozen; the resulting membrane resistance is therefore key in generating heat during freeze startup.

Next to the membrane itself, the most critical components in the fuel cell stack relative to startup from freezing and cool down to subfreezing temperatures are the catalyst layer and the GDL. As previously discussed, the GDL is one of the few critical components in the system which can be designed to affect proper water management, including frozen water that exists during storage or operation from subfreezing conditions. However, only a few studies have been published on the effect of freeze/thaw cycles on the properties of the GDL and its subsequent effect on the durability of a PEM fuel cell during cold startup [85, 86]. In fact, due to the wide range of operating conditions and the difference in fuel cell design approaches, reported impacts of freeze/thaw cycles on MEA performance vary considerably.

One of the most descriptive accounts of the factors influencing the cold-start behavior of low-temperature PEM fuel cells has been given by Mao and Wang [87]. They develop a lumped analysis model of the cold startup of a PEM fuel from subfreezing conditions. The model accounts for the rate of water production, removal, and conversion during startup. The condition for a successful startup is to have the fuel cell raise its temperature beyond the freezing point before being shut down by the formation of ice in the catalyst layer. By accounting for the heat and mass balances within the catalyst layer, the authors were able to investigate the conditions under which ice forms in the catalyst layer, thus restricting the flow of oxygen to the reactions sites. In effect, the authors indicate how to calculate the critical parameter requirements for a successful startup from subfreezing conditions. Initial membrane moisture content, heating rate, and thermal masses need to be chosen so that the temperature within the catalyst layer can reach the melting point before ice can fill up the pores within the catalyst layer and the GDL. Clearly, these critical parameters will depend upon the critical specifications of the components within the MEA, as well as the thermal properties of the bipolar plates.

There are a variety of system-level strategies for dealing with freezing, including purging of the stack at shutdown to remove liquid water, heating the stack prior to or during warm-up with auxiliary energy sources, and preventing freezing conditions from occurring (standby idling, etc.). Each of these strategies, however, results in a decrease in overall system efficiency.

Start/Stop

Every fuel cell system needs to be started up and shut down in a way that meets the requirements of the customer and at the same time ensures long-term operation of the system. Frequent starting and stopping can place the system in a severe electrochemical environment. In particular, when the anode side of the fuel cell is filled with air, as would be the case in standby situation, and the system is started up with the flow of a hydrogen-rich stream through the anode, a hydrogen/air boundary forms on the anode side of the cell [88]. This boundary moves from the anode inlet to outlet and in doing so sets up a high voltage (≈1.4 V) situation across the cell. This potential sets up a reverse current which greatly accelerates the oxidation of carbon, which theoretically can proceed as

$$ \mathrm{C}+2{\mathrm{H}}_2\mathrm{O}\to {\mathrm{CO}}_2+4{\mathrm{H}}^{+}+4{\mathrm{e}}^{-}\; \mathrm{E}=0.207\; \mathrm{V}\; \; \left(\mathrm{vs}.\mathrm{RHE}\right) $$

(15)

In fact, this potential-driven “corrosion” of carbon can be quite severe, causing substantial loss of the electrochemically active surface area as the electrode degrades with the loss of the catalyst support. Enhanced fuel cell degradation can occur under the additional stress conditions associated with cold start and hot stopping [89].

In order to deal with the problem of carbon corrosion due to start/stop, it is necessary to understand the critical parameters associated with the start/stop process, as well as the critical specifications that need to be placed upon the materials. What is required to do this is to have a validated model of carbon corrosion and its dependence on the critical parameters: temperature, pressure, flow rate, and composition. Such a model has been developed, to the first order, by Meyers and Darling [90] in which the potentials driving the reaction are linked to the spatial concentrations of oxygen and hydrogen in the flow channels.

Carbon corrosion can also occur due to fuel starvation on the anode, which sometimes happens when flooding is imminent. Again, this situation sets up localized high potentials and reverse currents. Carbon as a catalyst support is problematic, and for this reason control systems need to provide algorithms to deal with the situation when a particular cell in the stack is behaving in an abnormal fashion (so-called “stack health” algorithms).

Load Following

In most applications involving fuel cells, the load is dynamic, and it is a requirement that the fuel cell system be able to respond to these load transients nearly instantaneously. This requires careful system design. This requirement translates into the need for having rapid delivery of fuel to the catalyst sites within each cell in the stack. Two transient types of operating modes were described above, startup from subfreezing and start/stop. Here focus will be on load changes during operation and on applications involving both hydrogen and reformate fuel. In the former, emphasis will be placed on the effects of rapid cycling on durability, whereas in the latter, emphasis will be placed upon achievement of the required critical parameters over the expected range in operating parameters.

Hydrogen-air fuel cells, which are used in transportation, backup power, and material handling applications, respond almost instantaneously, provided that they are fed adequate amounts of fuel. Control schemes for hydrogen-based fuel cell systems for automotive applications are discussed by Pukrushpan et al. [91] in which the primary focus of the discussion is on control of the supply of air and hydrogen. In most cases, the fuel cell system will be “hybrid” with a battery in the overall system architecture. For example, in telecom backup fuel cell systems, a small battery pack (or supercapacitor) is included to ensure the continuous delivery of power within the first millisecond after the grid drops out. In light-duty vehicles, batteries are included as a means of load leveling and energy harvesting during braking. In forklift applications, batteries offer the opportunity to alleviate the load-following requirements on the stack, thereby extending its operating life. In fact, Plug Power sets a stringent set of charging and discharging capacity requirements on the battery to maximize the performance of the system and at the same time life of the stack (Du B, Private communication, 2017). This is another example of adopting a systems engineering approach to systems design. Obviously, it is important from a systems control perspective that one of the key aspects of achieving adequate load-following capability is to develop a comprehensive stack control scheme that incorporates the relevant system-level components, critical parameters, and their linkage to the critical specifications. In most cases, this control scheme must also address both stack health and efficiency.

For reformate-based systems, the control problem becomes particularly difficult because of the strong interactions between subsystems. This is perhaps best seen when considering the function of the ATO. The ATO is designed to oxidize the unreacted anode gas into water for the reformer, as well as to achieve zero emissions from the fuel cell system. On the other hand, the ATO also serves as the heat source for steam generation. Both the stack and the ATO require an air/oxygen source. To minimize cost, improve efficiency, and increase the operating latitude over which the system can be water independent, it is desirable to use a single air blower to provide air to both the stack and the ATO. However, this shared actuator couples the controls for stack, ATO, steam generation, and reforming, thereby increasing the complexity of the interaction among the control loops.

Another complexity arises due to the lack of in situ measurement of some of the critical parameters, requiring the inference of the value of the critical parameter from the operation of the system itself. This is particularly true when, by design, the relative humidity of the input reactant streams is near or at saturation. In this situation, RH sensors become unreliable, and the RH level needs to be inferred from the system control parameters. In order to accommodate the strong interactions between the subsystems, a considerable amount of modeling and simulating of system performance is required. This can be accomplished by linking process models in ASPEN Dynamics® with controls models in Matlab®. In this way, nonlinear dependencies in the process can be linearized in a piecewise fashion and exported to Matlab, an environment in which the order of the model can be systematically reduced together with weighting factors to get time-dependent models that can provide unified descriptions of the fuel cell in both steady state and dynamic environments. An example of the result of such an exercise is shown below in Fig. 8, comparing the performance of the ATO during load transients versus the model predictions. In this particular experiment, the system was run in steady state, and the load-following requirements were simulated by suddenly manipulating the cathode air blower in pseudorandom manner. It is seen that the dynamic model is a remarkably good predictor of the ATO performance [92].

Fig. 8

Model vs. data comparison for transient operation of ATO [92]

CO Poisoning

One of the challenges facing the fuel cell system is the mismatch in the time constants of the various components in the system. The reformer has the longest time constant, followed by that of the air handling system and stack, and finally the battery. This mismatch can result in large fluctuations in the quality of fuel provided to the stack and can be the source of unwanted excursions in the amount of CO reaching the stack. With validated advanced simulation tools, it is possible to develop robust control strategies that ensure that the overall system can meet the load-following requirements during large and rapid transients. Shown below in Fig. 9 is the transient response of the Plug Power ATR-based fuel processor to large and sudden step changes in electrical load demand. Here it is seen that the control software is able to keep the CO concentration within performance specifications (<10 ppm). In this case, batteries are used to make up the load demand, while the reformer ramps up, illustrating the mismatch in time

Fig. 9

Fuel cell performance during load transients [93]

constants between the stack and reformer subsystems [93].

It is evident that advanced controls, once validated with system verification tests, can be used to maximize the system operational latitude. As shown below in Fig. 10, these controls are used to optimize the operating window for the CO concentration in the reformate stream fed to the stack. Here it is seen that there is a large operational window, in terms of PROX outlet temperature and CO concentration over which the system can operate within its specification.

Fig. 10

Prox Co operating latitude in terms of temperature and load [93]

It is perhaps important at this point to once again discuss the potential advantages of higher-temperature operation. It is well known that fuel cell systems that operate at higher temperatures can tolerate much higher levels of CO. Indeed, tests have shown that stacks at temperatures in the range of 160–200 °C can tolerate CO concentrations on the order of 10³–10⁴ ppm. In this case, it is possible to eliminate the PROX and its associated costs and sources of unreliability, while simplifying the controls requirement.

There are, of course, many other potential failure modes that need to be addressed. Only a few are mentioned here. Perhaps the most difficult are those associated with maintenance of the polymer electrolyte’s required properties over hours of operation and over a wide range of conditions. Temperature, humidity, and potential cycling place severe stress on the polymer electrolyte and its matching electrodes, and minimizing these through great attention to engineering detail is fundamental to the achievement of technical readiness.

System Cost and Reliability

Two key parameters that are obstacles to widespread fuel cell adoption are system cost and reliability. Both are interrelated through the system design, driven mainly by the complexity of the interactions between the stack and the rest of the system, which in turn is determined by the nature of the polymer electrolyte. However, system complexity is also determined by the customer-driven system specifications. System specifications can vary widely, depending upon the application, and for this reason it is nearly impossible to make specific claims about costs of fuel cell systems without reference to the detailed requirements driving the design.

The cost and reliability of PEM fuel cell systems for light vehicles are tracked quite closely as automotive manufacturers are making significant investments in developing fuel cell cars for field entry in the next 5–10 years. The US Department of Energy (DOE) has, in conjunction with industry leaders, developed cost and reliability performance target and track progress against those targets in its annual review process. In addition, the DOE has defined a baseline fuel cell system with which to model performance improvements due to advances in technology and to coordinate those findings with the assumptions in the cost projections. The latest 2017 DOE cost estimates suggest that in high-volume production, total system costs of around $45/kW_net can be achieved, with an outlook of $42/kW_net in 2020, nearly equal to the DOE 2020 target of $40/kW_net at production volumes of 500,000 units/year [94]. This improvement in cost reduction was accounted for a combination of advances made in stack power density at low loadings, based upon the data from General Motors on the use of de-alloyed PtCo catalyst on high surface carbon, as well as cost reductions in the processing of coated stainless steel bipolar plates and reduction in hydrogen sensors.

The situation for stationary fuel cells is not as clear. The current DOE performance targets for integrated stationary fuel systems of $1500/kW_e, and 60,000 h of durability and 45% electrical efficiency at rated power are difficult targets to meet with a reformate-based system. The system requirements for stationary systems can vary widely, depending upon application. In general, however, the key to cost reduction is through system simplification. One way to accomplish this is through good engineering design, including value analysis/value engineering (VA/VE). VA/VE enables one to reduce the number of parts. Another way is to avoid expensive secondary fastening operations by having parts and modules self-align to each other, eliminating adjustments, using gravity wisely, and designing for “poka-yoke.” (A poka-yoke is any mechanism in a lean manufacturing process that helps an equipment operator avoid (yokeru) mistakes (poka).) A third is to specifically employ the design for manufacturing and analysis (DFMA) techniques of Boothroyd-Dewhurst. Readers are also directed to the entry on stationary cost assessments in this encyclopedia.

Currently, stationary fuel cells include the cost burden of the reformer, which makes the achievement of $1500/kW_e difficult. As a percentage of the total system cost, the reactant processing module can be as much as 30% of the total system unit manufacturing and service cost. Whereas the stack can scale downward gracefully as a function of size (kW_e output), the reformer cannot, making low-cost small residential stationary systems even more challenging. For this reason, it is appropriate to examine alternative system architectures that decouple the reformer and provide the required hydrogen from a shared-use centralized reformer. Such architectures are currently under consideration for use in Japan. Lastly, it seems appropriate that at this stage, fuel cell systems should be developed for large niche markets that are appropriate for the current state technical readiness of the technology. Forklift applications are a prime example of matching requirements versus capability. Large stationary fuel cells that are designed for constant base loads with high combined heat and power utilization requirements are another example.

Of course, another way to reduce cost and improve reliability is through clever systems engineering using advances in technology. This is in fact the approach recently taken by Toyota engineers in their pursuit of cost reductions in the Toyota Fuel Cell System (TFCS) known as “Mirai.” In particular, the Toyota engineers pursued an approach of reducing system cost by improving stack performance in terms of power density and by enabling membrane humidification under operating conditions without the cost and complexity of an external humidification system, and in order to achieve these objectives, Toyota engineers took a broad systems perspective. First, they reduced the number of required cells in the stack by employing a newly adopted boost converter that adjusts the voltage difference between the motor and the inverter. To achieve a system without an external humidifier, Toyota engineers decided to pursue this by ensuring adequate migration of water generated at the cathode to the anode. Three interrelated modifications were made. The first was to reduce the thickness of the membrane by two-thirds, thereby increasing the diffusion of water generated in the downstream air system and improving proton conductivity. The second was to humidify the system using the moisture at the anode, thereby humidifying the cathode inlet by flowing the hydrogen fuel and the air in counter directions. By adjusting the anode operating conditions by increasing the hydrogen circulation in the anode in accordance with the driving conditions, and by reducing the anode inlet pressure, the system was made to enhance the movement of product water to the anode surface [95]. In addition to the steps taken above to eliminate the external humidifier, the Toyota engineers also improved the water management through an innovative cell flow field structure and membrane electrode assembly that enabled a compact and high-performance stack. In particular, an innovative three-dimensional fine-mesh flow field was developed to prevent water accumulation at the cathode by directing the airflow toward the MEA/GDL assembly, promoting oxygen diffusion, and drawing water generated within the MEA/GDL to the back surface of the 3D fine-mesh flow field. An optimized flow field was also developed for the anode to facilitate anode operation, with the material changed to titanium to reduce the weight and cost of the plate-coating process. In terms of the MEA, in addition to reducing the membrane thickness, the amount of platinum was reduced as well, using a de-alloyed PtCo with an improved acid treatment process, coupled with a change in the carbon support to a solid type, reducing the oxygen diffusion resistance. Along with the optimization of the ionomer thickness, the net effect of all of these synergistic modifications was to increase the current density by a factor of 2.4 and to achieve a power density of 3.1 kW/L, reportedly the highest in the industry [96].

The Mirai story described above is important because it demonstrates that a broad-based systemic approach to improving fuel cell cost and performance is within the reach of engineers engaged in the design and development of fuel cell systems. It is an example of combining a combination of clever systems engineering with careful experimentation and design innovation based upon a deep knowledge of fuel cells that has been accumulating over the past 10 years. It also demonstrates the singular importance of increasing power density in order to decrease cell count through dedicated efforts in materials and design engineering.

Fuel Cell Systems and Sustainability

The connection between fuel cell systems and sustainability is, of course, hydrogen. The lightest and most abundant element on planet Earth is destined to play a central role in the energy landscape of the future. Renewable energy sources can be used to produce hydrogen, which can then be stored as a means of addressing the natural intermittency of the renewable source. In the future, by splitting water with renewable sources of energy such as wind and solar energy and using the stored hydrogen so produced to produce electricity, one can also return the product water to its source, achieving a sustainable energy paradigm for nearly every nation on Earth. This is truly the energy “end game.” Indeed, the fact that as an energy carrier hydrogen can play a key role in the storage of energy produced from renewable resources has received renewed interest among a variety of stakeholders.

Since the fuel cell system will play a key role in the pursuit of sustainable energy, whether in electrifying the automobile for transportation, providing energy for buildings, or converting stored hydrogen generated from renewable sources into electricity, the design of the fuel cell itself should consider its impact on the environment. This encyclopedia addresses the subject of life cycle analysis elsewhere. Here it seems appropriate to note that Cooper et al. [97] have developed a method to assist in the rapid preparation of life cycle assessments of energy generation technologies and in particular polymer electrolyte fuel cell systems. The method allows one to compare different fuel cell system design approaches, using publically available and peer-reviewed life cycle assessment data, against an environmental impact weighting scheme that reflects various environmental sensitivities. Providing a fuel cell system “designed for the environment,” with the goal that no part ever ends up in a landfill, seems to be a fitting way to approach the objective of truly sustainable energy systems.

Future Directions

Progress in the development of fuel cell systems for transportation, stationary, and motive power applications will continue as long as the incentives around renewable and clean energy continue to receive public support. This support, in turn, needs to be based upon public awareness of the true cost of the incumbent technologies: environmental and social, as well as financial. The rate of progress made in the past 10 years has been accelerating due to the support of governments tending to the longer range needs of their citizens. With a view of the need for a clean energy future clearly in sight, alternative means of energy generation, conversion, and storage will play an increasing role in shaping the future for our children’s children, a time not too far away.

To meet this challenge, fuel cell research and development will need to continue to pursue a range of subjects from basic understanding of the key factors governing charge transfer at the nanoscale to enabling the simplification of fuel cell systems by developing advanced materials that can relax the constraints currently imposed by the polymer electrolyte membrane. Hence, continued examination of the property and structure relationships in electrocatalysis will be needed to point the way to improved catalyst materials. Research on improved and more stable nanostructured electrodes incorporating advanced and lower cost materials will be needed to reduce cost while improving durability. The development of new polymer electrolyte membrane materials and structures that enable broader operating latitudes at higher operating temperature and lower relative humidity will be needed to further reduce cost by enabling less complex system designs and improved control. In parallel with this, of course, innovative engineering approaches toward reducing cost while simultaneously increasing reliability will needed to close the gap between today’s cost-performance curves and those required to enable penetration of commercial markets. In other words, a systemic, broadly based, and steady attack on the issues of unit manufacturing cost, durability, and reliability at the system level is necessary to advance the state of the art around PEM fuel cell systems.