Acta Materialia 51 (2003) 5981–6000 www.actamat-journals.com

Fuel cell materials and components夽 Sossina M. Haile ∗ Department of Materials Science and of Chemical Engineering, California Institute of Technology, 138-78, Pasadena, CA, 91125, USA Accepted 31 August 2003

Abstract Fuel cells offer the possibility of zero-emissions electricity generation and increased energy security. We review here the current status of solid oxide (SOFC) and polymer electrolyte membrane (PEMFC) fuel cells. Such solid electrolyte systems obviate the need to contain corrosive liquids and are thus preferred by many developers over alkali, phosphoric acid or molten carbonate fuel cells. Dramatic improvements in power densities have been achieved in both SOFC and PEMFC systems through reduction of the electrolyte thickness and architectural control of the composite electrodes. Current efforts are aimed at reducing SOFC costs by lowering operating temperatures to 500–800 °C, and reducing PEMFC system complexity be developing ‘water-free’ membranes which can also be operated at temperatures slightly above 100 °C.  2003 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. Keywords: Fuel cells; Solid electrolytes; Electroceramics; Polymers; Platinum group metals

1. Introduction Because of their potential to reduce the environmental impact and geopolitical consequences of the use of fossil fuels, fuel cells have emerged as tantalizing alternatives to combustion engines. Like a combustion engine, a fuel cell uses some sort of chemical fuel as its energy source; but like a battery, the chemical energy is directly converted to electrical energy, without an often messy and relatively inefficient combustion step. In addition to Tel.: +1-626-395-2958; fax: +1-626-395-3933. E-mail address: [email protected] (S.M. Haile). 夽 The Golden Jubilee Issue—Selected topics in Materials Science and Engineering: Past, Present and Future, edited by S. Suresh. ∗

high efficiency and low emissions, fuel cells are attractive for their modular and distributed nature, and zero noise pollution. They will also play an essential role in any future hydrogen fuel economy. The primary components of a fuel cell are an ion conducting electrolyte, a cathode, and an anode, as shown schematically in Fig. 1. Together, these three are often referred to as the membrane-electrode assembly (MEA), or simply a single-cell fuel cell. In the simplest example, a fuel such as hydrogen is brought into the anode compartment and an oxidant, typically oxygen, into the cathode compartment. There is an overall chemical driving force for the oxygen and the hydrogen to react to produce water. Direct chemical combustion is prevented by the electrolyte that separates the fuel (H2) from the oxidant (O2). The electrolyte serves

1359-6454/$30.00  2003 Acta Materialia Inc. Published by Elsevier Ltd. All rights reserved. doi:10.1016/j.actamat.2003.08.004

5982

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

Fig. 1. Schematic of a fuel cell, comprised of an electrolyte, an anode and a cathode. The overall chemical reaction is H2 + 1/2 O2 → H2O. Anode and cathode reactions given are appropriate only for oxide ion conducting electrolytes. The reactions would be modified for electrolytes with different mobile ions, but the general principle remains unchanged.

as a barrier to gas diffusion, but will let ions migrate across it. Accordingly, half cell reactions occur at the anode and cathode, producing ions which can traverse the electrolyte. For example, if the electrolyte conducts oxide ions, oxygen will be electro-reduced at the cathode to produce O= ions and consume electrons, whereas oxide ions, after migrating across the electrolyte, will react at the cathode with hydrogen and release electrons: 1 Cathode: O2 ⫹ 2e⫺→O2⫺ 2

(1)

Anode: H2 ⫹ O2⫺→H2O ⫹ 2e⫺

(2)

1 Overall: O2 ⫹ H2→H2O 2

(3)

Analogous cathode and anode reactions for proton conducting electrolyte are: 1 Cathode: O2 ⫹ 2H+ ⫹ 2e⫺→H2O 2

(4)

Anode: H2→2H+ ⫹ 2e⫺

(5)

The flow of ionic charge through the electrolyte must be balanced by the flow of electronic charge through an outside circuit, and it is this balance that produces electrical power. Electrolytes in which protons, hydronium ions, hydroxide ions, oxide ions, and carbonate ions are mobile are all known, and are the basis for the

many categories of fuel cells under development today. Because ion conduction is a thermally activated process and its magnitude varies dramatically from one material to the next, the type of electrolyte, which may be either liquid or solid, determines the temperature at which the fuel cell is operated. State-of-the art fuel cell electrolytes are listed in Table 1, along with the mobile ionic species, temperatures of operation and fuels typically utilized. For reasons of electrode activity (which translates into higher efficiency and greater fuel flexibility), higher temperature operation is preferred, but for portable (intermittent) power applications, lower temperature operation is typically favored as it enables rapid start-up and minimizes stresses due to thermal cycling. In addition, solid electrolyte systems obviate the need to contain corrosive liquids and thus solid oxide and polymer electrolyte fuel cells are preferred by many developers over alkali, phosphoric acid or molten carbonate fuel cells. Nevertheless, each of the fuel cell types listed in Table 1 has been demonstrated in complete fuel cell systems, with alkali and phosphoric being the most mature technologies, and polymer electrolyte membrane fuel cells the most recent. Although substantial progress has been made over the last several decades in each of the fuel cell technologies listed in Table 1, only those employing solid electrolytes, PEMFCs and SOFCs,

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

5983

Table 1 Fuel cell types and selected features [1,2] Type

Temperature °C

Fuel

Electrolyte

PEMa: polymer electrolyte membrane AFC: alkali fuel cell PAFC: phosphoric acid fuel cell MCFC: molten carbonate fuel cell SOFC: solid oxide fuel cell

70–110

H2, CH3OH

Sulfonated polymers (Nafion) (H2O)nH+

100–250 150–250 500–700 700–1000

H2 H2 hydrocarbons, CO hydrocarbons, CO

Aqueous KOH H3PO4 (Na,K)2CO3 (Zr,Y)O2-d

a

Mobile ion

OH⫺ H+ CO32⫺ O2⫺

Also known as proton exchange membrane.

are the subject of the present review. Polymeric fuel cells, in particular, have garnered a tremendous amount of attention as a result of their applicability to transportation systems, and worldwide investment in PEMFCs continues to grow. Alkali fuel cells, while providing extremely high power densities, are considered by most to be impractical because of the need to remove trace CO2 from both the fuel and oxidant streams in order to prevent reaction of the electrolyte to form solid, non-conducting alkali carbonates. Nevertheless, some commercialization efforts are underway. Phosphoric acid fuel cells, the leading technology in the early 1990s, have been largely abandoned because of the inability of developers to reach high power densities, and thus meet competitive cost targets on a per watt output basis. Given their high temperatures of operation, molten carbonate and solid oxide fuel cells have their greatest applicability in stationary power generation. The former suffer from the difficulties of containing a corrosive liquid electrolyte. In particular, dissolution of NiO at the cathode and its precipitation in the form of Ni at the anode can result in electrical shorts across the electrolyte, and few US developers continue to pursue this technology. Research and development efforts in SOFCs, though perhaps not as prominent in the public eye as PEMFC R&D, continue in both industrial and academic laboratories across the world. Most, but certainly not all, of those efforts are directed at stationary power applications, as SOFCs may also serve the transportation industry as auxiliary power units. In this review we present a brief discussion of the features that determine fuel cell performance,

describe the state of the art in SOFC and PEM fuel cells, and evaluate electrolyte materials that may enable new types of fuel cells. Although electrodes/electrocatalysts are equally important to advanced fuel cells, space limitations preclude a meaningful overview of this topic. Readers are referred to an excellent overview of the many aspects of fuel cell science and technology by Carrette and coworkers for additional information [3].

2. Fuel cell characteristics The key performance measure of a fuel cell is the voltage output as a function of electrical current density drawn, or the polarization curve, Fig. 2 [4,5]. The measured voltage, E, can be written as

Fig. 2. Schematic fuel cell polarization (voltage vs. current density) and power density curves.

5984

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

E ⫽ Eeq⫺EL⫺hact⫺hiR⫺hdiff

(6)

where Eeq is the equilibrium (expected or Nernstian voltage), EL is the loss in voltage due to leaks across the electrolyte, hact is the activation overpotential due to slow electrode reactions; hiR is the overpotential due to ohmic resistances in the cell; and hdiff is the overpotential due to mass diffusion limitations. The equilibrium potential, Eeq, can, in principle, be calculated from a knowledge of the thermodynamics of the reaction in question. One first determines the change in Gibbs free energy, ⌬G, for the reaction [for example, reaction (3)] under the given conditions and Eeq is then given by ⫺⌬G/nF, where n is the number of electrons transferred in the reaction and F is Faraday’s constant. For reaction (3), the Gibbs free energy is PH2P1/2 O2 ⌬G ⫽ ⌬G (T) ⫹ RTln PH2O o

(7)

where ⌬Go(T) is the Gibbs free energy of the reaction for the case when all species are in their standard states (1 atm, pure gases) and the pressures in the second term refer to the actual pressures in the fuel cell experiment. The term ⌬Go(T) is tabulated for most reactions of interest (or can be calculated from the formation energies of the species involved). In the case of reaction (3) it is ⫺242 kJ/mol + (45.8 J/mol K) ∗ T, for all components in the vapor phase [6]. This term alone is used to define the standard potential E (T) ⫽ ⫺⌬G (T) / nF o

o

(8)

of a particular reaction. In an operational fuel cell, knowledge of the equilibrium (Nernstian) potential requires knowledge of the partial pressures of all the species involved in the reaction. In most fuel cell experiments involving hydrocarbon fuels, the partial pressures of the product gases are neither controlled nor measured, and cell potentials are simply compared to the standard potential. Under open circuit conditions (that is, no current is drawn) the measured voltage should, in principle, be exactly the Nernstian voltage. In high temperature cells, in which reaction kinetics at the electrodes are fast, deviations from this voltage are attributed to gas leaks across the membrane due to

poor sealing or cracks in the electrolyte, or to partial electronic conductivity. These two factors define EL. In low temperature cells, sluggish reaction kinetics may prevent measurement of the equilibrium potential even under otherwise ideal conditions (no electronic conductivity, no gas cross-over). Slow electrode reactions give rise to activation overpotentials at both the cathode and the anode: hact ⬵

冉冊

RT I0 ln anF I

(9)

where a, termed the transfer coefficient, is the fraction of the overpotential assisting the reaction, and I0, termed the exchange current density, is the current flowing equally in the forward and reverse directions at equilibrium (zero overpotential). As written, Eq. (9) accounts only for forward reactions (i.e. reduction in the case of the cathode and oxidation in the case of the anode), and is a simplified form of the more general Bulter-Volmer equation:

再 冋 册

I ⫽ I0 exp

冋

册冎

anFh (1⫺a)nFh ⫺ exp RT RT

(10)

which accounts for reaction rates in both directions. Alternative approximations to (10) which are more representative at the zero current limit have been presented [7]. The simple form of Eq. (9) can be further approximated by the empirical Tafel equation hact ⫽ blog(I0) ⫺ blog(I)

(11)

with b defined as the Tafel slope. Eq. (11) implies that the slow reaction kinetics of low temperature fuel cells lead to an offset in the open circuit potential, Eo, by an amount Eo⫺Eeq ⬵ blogIo.

(12)

In hydrogen/oxygen PEM fuel cells, voltage offsets on the order of 400 mV are typically observed [8], but it is unclear whether the entirety of this is due to slow electrode kinetics or in part due to hydrogen and/or oxygen diffusion through the electrolyte (i.e., contributions to EL). In any case, for all known hydrogen/oxygen fuel cells, regardless of electrolyte type, it is the cathode which is rate limiting (hydrogen electro-oxidation is

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

extremely rapid on a wide range of catalysts) and the activation overpotential is almost entirely due to the cathode. At high temperatures, even cathode activation overpotentials may be absent. When impure hydrogen or hydrocarbon fuels are utilized, however, anode kinetics may become rate-limiting at any temperature. The fourth term in Eq. (6), the ohmic overpotential of a cell, is simply given by IR, where R is the area specific resistance, and includes terms not only from the electrolyte, but also the electrodes, current collectors and lead wires in the system. The final term, hdiff, the mass diffusion term, has a form that is specific to the geometry under consideration, but it is generally established by the rate of reactants flowing to the electrolyte through the electrodes and the rate of products flowing away. Both anode and cathode can contribute to this term. As greater and greater current is drawn, the reactant depletion zone or the by-product build-up become greater and greater, and thus mass diffusion overpotentials become severe at high current densities. The impact of all the terms of Eq. (6) are highlighted in Fig. 2. The power density is simply multiple of the voltage and the current density, and, as also shown in the figure, reaches some peak value at intermediate voltages (or current density). In contrast, efficiency, which decreases with increasing overpotential, is greatest at low current densities. In the context of the polarization curve, it is evident that high power densities (and high efficiencies) result when gas diffusion and electron transport through the electrolytes are slow, electrocatalysis at the electrodes is rapid, the conductivity of each of the components, in particular, the electrolyte, is high, and mass diffusion through the porous electrodes is facile. Thus, the ideal fuel cell electrolyte is not only highly ionically conducting, but also impermeable to gases, electronically resistive and chemically stable under a wide range of conditions. Moreover, the electrolyte must exhibit sufficient mechanical and chemical integrity so as not to develop cracks or pores either during manufacture or in the course of long-term operation. The demands on fuel cell electrodes are perhaps even more extreme than those on the electrolyte.

5985

The ideal electrode must transport gaseous (or liquid) species, ions, and electrons; and, at the points where all three meet, the so-called triplepoint boundaries, the electrocatalysts must rapidly catalyze electro-oxidation (anode) or electroreduction (cathode). Thus, the electrodes must be porous, electronically and ionically conducting, electrochemically active, and have high surface areas. It is rare for a single material to fulfill all of these functions, especially at low temperatures, and consequently a composite electrode, of which the electrocatalyst is one component, is often utilized. For high-temperature solid oxide fuel cells, single component electrodes are, in principle, possible because mixed conducting (O2⫺ and e⫺) ceramics are known, and it is precisely such materials which facilitate electrocatalysis. For low temperature proton conducting polymer systems, mixed conducting systems are virtually unknown, and slow reaction kinetics require (precious metal) catalysts. Thus, composite systems are standard. Furthermore, the electrocatalyst is typically restricted to a very thin layer adjacent to the electrolyte, and another layer, the ‘gas-diffusion-layer’ (GDL) serves the role of transporting electrons and gases to and from the rest of the MEA. An image of a single-cell ceramic fuel cell and a schematic diagram of a polymeric MEA are shown in Fig. 3. Additional components in a complete fuel-cell power generator are the so-called interconnects or bipolar plates. These components serve to link up individual fuel cells in a fuel cell stack; ‘interconnect’ is the term used in the SOFC community whereas ‘bipolar plate’ is that used in the PEM community. In addition to providing electrical conduction pathways, the interconnects/bipolar plates serve to keep oxidant and fuel gases separate from one another. Thus, the ideal interconnect has high electronic conductivity (its resistance contributes to the overall IR losses in the stack), excellent impermeability, and chemical stability under both oxidizing and reducing conditions. Good mechanical properties are also required, particularly for low temperature systems in which pressure is applied to maintain gas seals. For reasons of space limitations, these components will not be discussed in any detail. Suffice it to note that tremendous efforts are being directed towards developing inex-

5986

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

detrimental reactions occur over the many tens of thousands of hours expected for fuel cell lifetimes. Moreover, the components in high temperature fuel cells, must exhibit thermo-mechanical compatibility; that is, the thermal expansion coefficients must match, and/or the materials must be tough enough to withstand mechanical stresses due to differences in thermal expansion.

3. Solid oxide fuel cells: State-of-the-art

Fig. 3. Membrane-electrode-assemblies: (a) scanning electron micrograph of a typical SOFC structure obtained in the author’s laboratory; (b) schematic of a PEM structure in which gas diffusion layers, are additionally incorporated. Together, the electrocatalyst and gas diffusion layers are often referred to as the gas diffusion electrodes.

pensive stainless steel alternatives for both the polymer and solid oxide fuel cells, and in both cases, difficulties arise from the tendency of chromium to evaporate and/or diffuse from the metal into the electrolyte. Overall, the fuel cell environment places severe demands on the properties of each of the fuel cell components. In addition to meeting those individual demands, fuel cell materials must be chemically compatible with one another, such that no

An excellent review of ceramic fuel cells and the materials from which they are constructed has been presented by Minh [9], and we only briefly summarize here the technology status for what one can term ‘conventional’ solid oxide fuel cells. Somewhat more recent, but less comprehensive, reviews have been published by Ormerod [10] and by Singhal [11]. Today’s demonstration SOFCs utilize yttria stabilized zirconia (YSZ), containing typically 8 mol% Y, as the electrolyte; a ceramicmetal composite (cermet) comprised of Ni + YSZ as the anode; and La1⫺xSrxMnO3-d, (lanthanum strontium manganite or LSM) as the cathode. Specific anode and cathode compositions are often omitted from publications, but typically x is 0.15 to 0.25 in LSM cathodes, whereas Ni-YSZ anodes are prepared from ~50:50 wt ratio NiO + YSZ mixtures which are subsequently reduced in situ to yield Ni metal particles dispersed in a porous YSZ matrix. Anode and cathode porosities are typically 25–40 vol%. The interconnect material is alkali doped LaCrO3 (lanthanum chromite), with the specific dopant (typically, Sr, Ca or Mg) and concentration being selected to best match the thermal expansion of the other fuel cell components in the geometry of interest. In addition to these four components, planar SOFCs require sealant materials to isolate anode and cathode chambers in a stacked configuration. The literature on SOFC sealants is limited, but these have been typically fabricated from various glasses or glass-ceramics. Recent advances in YSZ-based fuel cells include a move from electrolyte or cathode supported structures to ones in which the anode serves to support a thin electrolyte and thin cathode, and the use of composite, YSZ + LSM, cathodes (YSZ:LSM =

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

50:50) rather than simply LSM. These design changes, together with improved fabrication techniques, have resulted in an increase in peak power densities for laboratory planar cells operated on hydrogen from ~250 mW/cm2 @ 1000 °C in 1989 [12] to almost 2 W/cm2 @ 800 °C today [11,13,14]. Solid oxide fuel cells have shown tremendous reliability when operated continuously. For example, a 100 kW system fabricated by SiemensWestinghouse has successfully produced power for over 20,000 hrs without any measurable degradation in performance [15]. In addition, such fuel cells offer good fuel flexibility, allowing a variety of hydrocarbon fuels to be utilized. However, SOFCs are still much too costly for widespread commercialization, they function poorly under intermittent operation and the possibilities for direct utilization of hydrocarbon fuels has hardly been explored. Several of the challenges hindering SOFC technology are a consequence of the high temperatures required for their operation. High temperatures preclude the use of metals, which typically have lower fabrication costs than ceramics, for any of the non-electrochemical components of the fuel cell and also increase the likelihood of cracks developing upon thermal cycling. The interconnect material, lanthanum chromite, is particularly difficult to process because of chromia evaporation at high temperatures, which leads to poor densification. Thermal stress-induced failure at glass seals is an acute problem for planar SOFCs. Tubular SOFCs, though free of seals, are very costly on a per power output basis because the tubular design inherently produces low power densities as a consequence of the long current paths, and the manufacture of tubular ceramics has only been achieved with costly techniques such as electrochemical vapor deposition [11]. Furthermore, while SOFCs offer good fuel flexibility, allowing a variety of hydrocarbon fuels to be utilized, the less reactive fuels must typically be internally steam-reformed, that is, reacted with H2O in the anode chamber, to produce CO and H2 which can subsequently be utilized in the electrochemical reactions [16]. Although not anticipated to prevent SOFC commercialization, internal steam reforming requires recirculation of water, and its

5987

elimination simplifies fuel cell operation and reduces costs. In recognition of these challenges, there are presently major research effort ongoing to reduce the temperature of SOFC operation to 500–600 °C (although even 800 °C is often referred to as reduced temperature of operation), to develop anode materials which can directly electro-oxidize hydrocarbon fuels at these temperatures without carbon deposition, and to develop cathode materials which are active for oxygen electroreduction at these reduced temperatures. Further reductions in the temperature of operation are generally deemed undesirable because of concomitant losses in electrode activity.

4. Polymer electrolyte membrane fuel cells: State-of-the-art Polymer electrolyte membrane (PEM) fuel cells have been reviewed by Costamagna and Srinivasan [17,18] and readers are referred to that work for a more comprehensive discussion than can be provided here. The most widely implemented electrolyte in PEM fuel cells is Nafion manufactured by duPont. Nafion and related polymers are comprised of perfluorinated back-bones, which provide chemical stability, and of sulfonated side-groups which aggregate and facilitate hydration (see discussion below). It is these hydrated, acidic regions which allow relatively facile transport of protons, but also restrict PEMFCs to low temperatures of operation. As a consequence, precious metals are required for electrocatalysis. For hydrogen/air fuel cells, Pt nano-particles supported on carbon are utilized for both the anode and cathode. Proprietary Pt-based alloys may instead be used in the anode of fuel cells using ‘impure’ hydrogen, that is, hydrogen obtained by steam reforming of hydrocarbon fuels. The reformate from such processes can contain 10 s to 100 s of ppm CO and this has detrimental impact on fuel cell performance as a result of the strong adsorption of CO onto Pt. In direct methanol fuel cells PtRu alloys (typically 50:50 molar ratio) are used at the anode because of the ability of Ru to electrooxidize CO adsorbed onto Pt. The platinum content (or Pt loading) in

5988

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

hydrogen/air PEM fuel cells has been reduced dramatically over the past decade from ~4 mg/cm2 (per electrode) to less than 0.4 mg/cm2, as a result of optimal dispersion of nanoparticle Pt in the electrocatalyst layer of the fuel cell. In contrast to SOFCs, the current collector, which also serves as a gas diffusion layer, is a distinct but integral component of the PEM membrane-electrodeassembly. Typically, highly porous carbon paper treated with a hydrophobic polymer (e.g. PTFE) is used. In the fabrication of a stack, graphite bipolar plates are placed between individual MEAs. Recent advances in perfluorosulfonated polymer based fuel cells include reduction of the electrolyte thickness from ~175 to ~ 25 µm, an increase in the extent of sulfonation of the polymer to increase conductivity, control of the porosity in the electrocatalyst and gas diffusion layers, and optimization of the MEA processing to achieve excellent adhesion between the multiple layers. These development efforts have led to increases in the power densities of hydrogen/air PEM fuel cells from ~100 mW/cm2 in 1984 to over 1 W/cm2 in 2002 [1]. The long-term durability of PEM fuel cells is more of a concern than that of continuously operated SOFCs, although Ballard has reported 2000 h of operation without measurable loss in PEMFC power output. More significant at this stage than long-term durability is the hydration requirement of the polymer in order to maintain conductivity. First, because, in fact, hydronium ions, (H2O)nH+, rather than bare protons are transported across the membrane, excess water must be removed from the cathode and replenished at the anode. Second, upon hydration, sulfonated polymers swell significantly and their mechanical properties can degrade. Third, operation is generally limited to ⬍~ 90 °C, at which temperatures the precious metal anode catalysts are susceptible to CO poisoning, and electrocatalysis at both electrodes is generally sluggish. Higher temperatures of operation open up the possibility of simplifying the CO removal process, lowering Pt loadings and even the use of non-precious metal catalysts. And fourth, as a consequence of its miscibility with water, methanol easily diffuses across the hydrated polymer electrolyte from the anode to the cathode (fuel cross-over) resulting in significant efficiency losses and low power den-

sities. Indeed, to minimize cross-over, relatively thick membranes (~175 µm) and low methanol concentrations (~4% in H2O) are used in direct methanol fuel cells [19]. Despite the drawbacks associated with perfluorinated sulfonated polymers, Nafion and its close relatives continue to be the electrolyte of choice in demonstration PEMFCs because of their combination of very high conductivity and adequate mechanical properties.

5. Electrolytes The most important property of a candidate electrolyte material is, of course, the ionic conductivity. Conductivity data of a broad range of materials are summarized in Fig. 4 [20–25]. Material classes for electrolyte applications range from ceramics, to polymers to acid salts, and the mobile ion can be O2⫺, H+, or (H2O)nH+. Solids for which OH⫺ or CO3⫺ are mobile are also known, but the conductivities are not high enough to be of technological relevance. It should be noted that independent of the magnitude of the conductivity, fuel cell design inherently leads to a preference for a specific mobile species. In general, hydronium, hydroxide and carbonate ion conductors are unattractive because one must, by definition, recycle an otherwise inert species: H2O in the case of hydronium and hydroxide conductors or CO2 in the case of carbonate conductors, to maintain ion transport. Where hydrogen is the fuel, a proton conductor offers the benefits of generating the by-product H2O at the cathode, and thus the fuel does not become diluted as a function of utilization. Consequently, so long as oxygen is plentiful, cell voltages remain high. Hydronium ion conductors generally provide this same benefit, however the water recirculation requirements noted above demand delicate water management. In the case where hydrocarbons serve as the fuel, an oxygen ion conductor offers, in principle, the prospect of direct electro-oxidation: CH4 ⫹ 4O2⫺→CO2 ⫹ 2H2O ⫹ 8e ⫺

(13)

This possibility has only recently been explored, and is discussed further below. Generally, it is

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

5989

instead presumed that the high temperature of operation associated with oxygen ion conductors can be used to facilitate internal steam reforming: CH4 ⫹ H2O→CO ⫹ 3H2

(14)

with CO and H2 then used in the electro-oxidation reactions. Even in this more conservative scenario, oxygen-ion conducting electrolytes are preferred because CO can be electro-oxidized, rather than (particularly in the case of low temperature systems) poisoning the anode catalyst. Ceramic proton conductors may offer an interesting combination of benefits because of their ability to transport both protons and oxygen ions. It has been suggested [26] that water can diffuse across the electrolyte membrane, inducing steam reforming and even conversion of CO to CO2 through the water-gas shift reaction: CO ⫹ H2O → CO2 ⫹ H2.

(15)

Hydrogen generated by reactions (14) and (15) is then electrooxidized to form protons. In this case, hydrocarbons can be directly utilized and no water is produced at the anode, again avoiding dilution effects as fuel utilization increases (although CO2 dilution still occurs). 5.1. Oxygen ion conductors The classic oxygen ion conductors, stabilized (cubic) zirconia and ceria are based on the fluorite structure, Fig. 5. In order to introduce mobile oxygen vacancies into the compound, and, in the case of zirconia, stabilize the cubic structure, tri- or divalent dopants are added to the host material. The incorporation reaction can, for a typical trivalent dopant, M, be written as: M2O3 → 2M⬘Zr ⫹ 3O×o ⫹ V앫앫 O

(16)

ZrO2

Fig. 4. Conductivities of selected electrolyte materials. (a) high temperature conductors [20–22]; and (b) low temperature conductors [23–25]. PBI is polybenzymidazole where high and low refer to the acid content; PWA is phosphotungstic acid and the numeral following the acronym is the number of water molecules of hydration.

with one oxygen vacancy created for every two M atoms incorporated. For both zirconia and ceria, conductivity increases with increasing dopant concentration up to some maximum value and then decreases sharply. Similarly, the conductivity increases then decreases across the rare earth series from Yb to La. For zirconia, Sc gives rise to the highest conductivity, but Y is typically utilized for

5990

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

Fig. 5. Crystal structure of conducting oxides: (a) fluorite structure, exhibited by stabilized zirconia and by ceria; (b) perovskite structure, exhibited by oxygen ion conducting LaGaO3, and by proton conducting BaZrO3.

reasons of cost. With yttrium as the dopant, the conductivity of zirconia peaks at about 8 mole % dopant concentration. In the case of ceria, Sm [27] and Gd [28] give the highest values of conductivity, and optimal dopant concentrations are 10–20%. The strong dependence of ionic conduc-

tivity on dopant type and concentration has been explained in terms of the lattice distortions introduced by the dopant, with those that produce the least amount of strain causing the smallest variation in the potential energy landscape [29]. Overall, the ionic conductivity of ceria is approximately an order of magnitude greater than that of stabilized zirconia for comparable doping conditions. This is a result of the larger ionic radius of Ce4+ ˚ in 6-fold coordination) than Zr4+ (0.72 A ˚ ), (0.87 A which produces a more open structure through which O= ions can easily migrate. Despite its favorable ion transport properties, ceria had not, until quite recently, been considered a realistic candidate for fuel cell applications because of its high electronic conductivity. In particular, under reducing conditions, CeO2 becomes CeO2⫺x, and n-type conductivity increases with a P(O2)⫺1/4 dependence. From an analysis of relevant literature data, Steele [28] has proposed that the electrolytic domain boundary, the oxygen partial pressure at which electronic and ionic conductivities are equal, can be estimated as shown in Fig. 6(a), for 10 and 20% Gd doped ceria. In principle, one expects to be well within the electrolytic domain of ceria for fuel cells operated below 700 °C, although even at these temperatures some voltage loss is expected. Reported open circuit potentials for doped ceria, Fig. 6(b), are lower than what one would expect on the basis of the electronic conductivity of ceria and represented simply as the multiple of the ionic transference number (greater than ~0.9 [30] at 700 °C and 10⫺18 atm oxygen partial pressure) and the Nernst potential. The reasons for this discrepancy are not entirely obvious, but are likely due to electrode (in particular cathode) overpotentials, and emphasize the importance of developing electrodes compatible with ceria that enable theoretical open circuit potentials to be reached. An additional challenge lies with the chemical expansion of ceria under reducing conditions and the internal stress that result [31]. At this stage, the significance of this issue on the long-term viability of ceria-based fuel cells is unknown. It is noteworthy that planar cells experience lower stresses than tubular cells, suggesting that clever designs may alleviate possible stresses.

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

Fig. 6. Electrochemical properties of doped ceria [43]: (a) Electrolytic domain boundary at which ionic conductivity equals electronic conductivity, after [28]; (b) Open circuit potential for a hydrogen (3% H2O) // oxygen cell. The Nernst potential is that expected for an ideal electrolyte, the dotted line is approximately that expected for an electrolyte with 10% electronic conductivity and the red and blue curves are experimental data.

5991

Oxide-ion conducting perovskites have appeared in the literature for several years, but only recently have compositions with conductivities high enough for consideration in fuel cell applications appeared. The ABO3 perovskite structure, Fig. 5, is extremely amenable to tailoring via doping on both the A and B cation sites. A large variety and concentration of dopants can be accommodated in a wide range of host compounds. Introduction of divalent dopant ions, typically Sr and Mg, onto the La and Ga sites, respectively, of lanthanum gallate produces a material with a high concentration of mobile oxygen vacancies and thereby high oxygen ion conductivity. The conductivity of the particular composition La0.9Sr0.1Ga0.8Mg0.2O3⫺d was reported almost simultaneously by Goodenough and coworkers [32] and by Ishihara and coworkers [21]. The transport properties of LSGM, as it is known, are comparable to those of scandia-doped zirconia. The conductivity is entirely ionic over an extremely wide oxygen partial pressure range at temperatures as high as 1000 °C, but is not as high as that of suitably doped ceria. Thus, the conditions under which LSGM might be preferable to doped ceria appear limited to the temperature range of 700–1000 °C. Moreover, lanthanum gallate suffers from reactivity with nickel, the typical SOFC anode electrocatalyst. To address this challenge, (non-reactive) ceria buffer layers have been incorporated between the electrolyte and the anode [33]. Nevertheless, intensive research efforts to develop SOFCs incorporating lanthanum gallate continue, and recent work suggests that the ionic conductivity can be increased by further adjustments to the stoichiometry, in particular, via the addition of small concentrations of Ni or Co [34]. A sampling of recent achievements in fuel cell research using ceria and LSGM electrolytes is summarized in Table 2. The ion transport properties of bismuth oxide have received significant academic attention as a result of a rather spectacular phase transition at ~700 °C which leads to an increase in conductivity by almost three orders of magnitude. In the high temperature d phase, the compound has a cubic fluorite structure, with an extremely high (25%) oxygen vacancy content. Below the transition, the vacancies are ordered, and hence the low conduc-

5992

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

Table 2 Reduced temperature solid oxide fuel cells; materials and performance at 650 °C on hydrogen/air input gases, unless otherwise indicated Electrolyte

Thickness, µm

Anode

Cathode

Peak power density

Reference

LSGMC Ce0.9Gd0.1O1.95 Ce0.9Gd0.1O1.95 Ce0.9Gd0.1O1.95 YSZ YSZ

205 ~40 5–10 150 ~30 ~10

Ni-SDC Ni-Ru-GDC Ni-YSZ Ni-GDC Ni-YSZ Ni-YSZ

SSC Sm0.5Sr0.5CoO3-δ La0.6Sr0.4Co0.2Fe0.8O3-δ La0.6Sr0.4Co0.2Fe0.8O3-δ LSM + SDC La0.8Sr0.2FeO3-δ

240–410 770a 150 110 190 400

[35] [36] [37] [38] [39] [40]

LSGMC = La0.8Sr0.2Ga0.8Mg0.15Co0.05O3-δ. a measured at 600 °C, an even higher value can be anticipated for 650 °C.

tivity. Efforts to stabilize the high temperature phase at low temperatures have led to the development of (Bi2O3)x(Ln2O3)1⫺x materials (Ln = lanthanum metal) which show much less pronounced transition behavior, but retain overall high conductivity. Just as in the case of the Zr and Ce fluorites and the perovskites, proper matching of the dopant ionic radius to the host lattice is essential. Across the lanthanide series, the conductivity of (Bi2O3)0.75(Ln2O3)0.25 peaks for Ln = Er, and is only slightly lower for Ln = Y [41,42]. The key limitations of bismuth based compounds are their very high electronic conductivities, and tendencies to become reduced to bismuth metal in hydrogen or fuel containing atmospheres. To date, no fuel cells have been fabricated using this material, although there have been claims that placing a thin layer of bismuth oxide on the cathode side of a ceria fuel cell can increase open circuit potentials [43]. 5.2. Ceramic proton conductors In analogy to the defect chemistry of lanthanum gallate, proton transport in barium zirconate, barium cerate and related materials is achieved by first doping the material with a trivalent species (such as yttrium) on the B site so as to introduce oxygen vacancies. The dopant incorporation reaction is normally assumed to occur as per Eq. (17) (written in Kroeger-Vink notation). 2CexCe ⫹ OxO ⫹ Gd2O3⇒2Gd⬘Ce ⫹ V앫앫 O ⫹ 2CeO2.

(17)

Subsequent exposure of the material to humid atmospheres is presumed to lead to the incorporation of protons as per Eq. (18). H2O (gas) ⫹ VO앫앫 ⫹ OO⇒2OH앫O.

(18)

The protons introduced by this manner are generally not bound to any particular oxygen ion, but are instead free to migrate from one ion to the next. This easy migration results in the high proton conductivity (as high as 10⫺2 ⍀⫺1cm⫺1 at 500 °C) observed in these oxides. Much like the oxide conductors, proton conductivity peaks at intermediate dopant concentrations and with suitable matching of the dopant ionic radius to the host structure. Furthermore, proton transport dominates the overall electrical transport to temperatures of approximately 600 °C; the proton transference number of BaCe0.95Sm0.05O3, for example, is ~0.85 at this temperature [44]. At higher temperatures, both oxide ion transport and electron transport become significant. The defect chemistry of doped A2+B4+O3 perovskites is complicated by the possibility that the trivalent ion may reside on both cation sites, and not only the B4+ site as desired [45]. The consequence of partial incorporation of the dopant onto the A2+ site is that fewer oxygen vacancies than anticipated will result. The effect is exacerbated by high temperature processing which can induce BaO evaporation. A second complication arises from the highly refractive nature of the zirconate proton conductors, e.g. doped BaZrO3. In comparison to the cerates (BaCeO3 and SrCeO3), barium zirconate offers high conductivity and excellent chemical

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

stability against reaction with CO2. However, fabrication of dense electrolyte membranes from this material remains a significant challenge. Indeed, the high bulk conductivity of BaZrO3 had, for several years, remained obscured as a consequence of the material’s refractory nature, which results in fine-grained samples with high total grain boundary resistance [46,22]. In light of the reactivity of cerates with CO2 and the difficultly of fabricating dense zirconate electrolytes, it is perhaps no surprise that few complete MEAs of proton-conducing electrolytes have been constructed and characterized. Early experiments were focused simply on proof of principle and utilized Pt for both electrodes [47]. More recently, oxides have been incorporated into both the anode and cathode. In particular, La0.8Ba0.4CoO3-δ has served as the cathode and at 1000 °C, power densities as high as 200 mW/cm2 have been reached [48]. 5.3. Polymeric proton (hydronium) conductors The perfluorinated sulfonated polymers, used in the most mature PEM fuel cells, have a teflon-like back-bone with sulfonated sidegroups, and a microstructure in which hydrophobic and hydrophilic regions phase separate on the nanoscale, Fig. 7. The latter has been shown experimentally by recent SAXS (small angle X-ray scattering) measurements [49], and computationally, using molecular dynamics simulations [50]. In general, proton conductivity increases as the hydration level of the polymer increases, with H2O:SO3 ~ 15:1 and conductivity ~10⫺1 ⍀⫺1cm⫺1 under operational conditions. Materials with low equivalent weight (small y) and short distances between hydrophilic side groups (small x), Fig. 7(a), typically exhibit high conductivity. Furthermore, relative humidity has a greater impact on conductivity than does temperature, although at temperatures above ~90 °C, conductivity drops dramatically due to dehydration. As noted above, increased water content, though beneficial for the conductivity, increases swelling (resulting in dimensional instabilities), and at the highest sulfonation levels the polymer can become water soluble. Other challenges associated with the use of hydrated polymer electrolytes: methanol cross-over, difficult humidification

5993

requirements and inoperability at high temperatures, have also been noted above. In addition, the perfluorinated nature of the back-bone in polymers such as Nafion results in very high cost materials (~$700/m2) [17]. Accordingly, research efforts have targeted several areas: (1) reduction of cost through the use of entirely hydrocarbon based polymers, (2) reduction of humidification requirements to yield overall system simplifications, (3) increasing thermal stability so as to enable ‘warm’ temperature operation, particularly for automotive applications, and (4) decreasing methanol crossover. Improving mechanical properties under high levels of hydration has also received some attention. A review of all the many, many approaches described in the literature for addressing these challenges is beyond the scope of this work. Rather, we briefly highlight a few selected strategies. High chemical and thermal stability are typically features of aromatic hydrocarbon polymers that incorporate benzene into their structures. Thus, several groups are pursuing sulfonation of poly(ether ether ketone), poly(styrene) and related materials, Fig. 8, to produce high proton conductivity polymers free of fluorine [51]. Poly(styrene sulfonic acid) is, in fact, the polymer utilized in the 1960s in various NASA and other missions. The conductivities of sulfonated hydrocarbons tend to be lower that that of perfluorinated materials, but with appropriate modifications improvements may be attained. Sulfonation can be carried out directly on the polymer back-bone, as indicated in Fig. 8, or short, sulfonate-terminated side-groups can be introduced. The latter appears to provide increased thermal stability. To improve the water retention (which decreases sensitivity to humidity and enables high temperature operation) and to reduce the swelling of Nafion, others have used inorganic additives such as complex clays, silica, and phosphotungstic acid [52]. In a somewhat related approach, membranes comprised of porous teflon impregnated with Nafion have been prepared. These retain excellent mechanical properties to much thinner dimensions than Nafion alone, and has led to very high power density fuel cells [43]. Methanol cross-over can also be reduced utilizing these approaches [53].

5994

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

Fig. 7. Structure of perfluorinated sulfonated polymers (i.e. Nafion and its close relatives). (a) chemical structure; (b) nanoscale phase separated microstructure as determined by SAXS (small angle X-ray scattering), after [49]; (c) microstructure as predicted by molecular dynamics simulations, after [50].

In order to avoid completely the many difficulties associated with water-saturated polymers, others are pursuing acid-base polymer complexes, in which a strong acid is coupled to a highly basic polymer. The concept is similar to that of lithium conducting polymers, in which a lithium salt is blended with a basic polymer such as poly(ethylene oxide) [PEO]. The mobile ions, either lithium or protons, are displaced from the host salt or acid, via attraction to the oxygen ions in the polymer. Protons hop between basic sites of the polymer and/or hydrogen bond sites between

anions, depending on the acid concentration, and do not require the migration of a host species (as is the case for proton transport in hydrated polymers). A wide range of basic polymers, including PEO, PVA [poly(vinylalchohol)], Paam [poly(acrylamide)] PVP [poly(vinylpyrrolidone)] PEI poly(ethyleneimine), various poly(aminosilicates), and PBI [poly(benzimidazole)] have been examined in combination with sulfuric, phosphoric and various halide acids [54]. The proton transfer from the acid to the polymer can be sufficiently extreme so as to render the polymer host

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

5995

Fig. 8. Chemical structures of (a) polystyrene and its sulfonated analog, poly(styrene sulfonic acid) [PSSA] and (b) polyether ether ketone and its sulfonated analog, S-PEEK. The most likely site for sulfonation of the polymer host is indicated.

a cationic species that is charge-balanced by the inorganic anion (H2PO4⫺, HSO4⫺). Of the many combinations possible, PBI-H3PO4 complexes have been extensively examined, in particular, for direct methanol fuel cell applications. Much as the conductivity of hydrated polymers depends on the H2O content, the conductivity of acid-polymer blends depends heavily on the acid content, Fig. 4(b), and thus ‘acid management’ may pose longterm challenges for these materials. Nevertheless, results to date are quite promising. Early efforts produced hydrogen/oxygen fuel cells [55] (150 °C) and direct methanol fuel cells [56] (200 °C) with peak power densities of 250 and 100 mW/cm2, respectively, while more recent investigations have yielded hydrogen/oxygen cells with peak power densities as high as 1 W/cm2 at 185 °C [57]. Another route for eliminating water from proton conductors is to return to sulfonated polymers, but replace the water with a less volatile ionic liquid, in particular heterocyclic amines such as imidazole (pyrazole) and benzimidazole [58]. The nitrogen sites in these compounds can act as proton acceptors with the acidic sulfonic groups of the host polymer serving as proton donors. Proton conductivities as high as 10⫺2 S/cm at 200 °C have been reported for a sulfonated aromatic polymer

host intercalated with ~7 imidazole molecules per sulfonic group. A further modification to this approach is to link the imidazole groups directly to the polymer back-bone in order to prevent loss of the ionic liquid. Appropriate tethering ensures that the heterocyclic groups can aggregate and proton transport properties are ideally retained. Although materials with conductivities comparable to those of conventional sulfonated perfluorinated polymers have not yet been developed, Fig. 4(b), the flexibility of the approach provides many avenues for continued investigations. Care must be taken, however, that gains in eliminating water from the polymer electrolyte do not come at the expense of electrode performance, which may continue to require a hydrated polymer as part of the electrocatalyst layer so as to ensure access of fuel and oxygen to the polymer-coated catalyst particles. 5.4. Acid Salts and other inorganic proton conductors Inorganic proton conductors offer the possibility of proton transport under anhydrous conditions, and thereby circumvent many of the challenges facing polymer electrolyte based fuel cells. Of the

5996

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

inorganic proton conductors known, Fig. 4, hydrogen bonded solid acids with disordered phases show particularly high conductivities. In general, such compounds are comprised of oxyanions, for example SO4, SeO4, PO4, AsO4, or even PO3H etc., which are linked together via O–H...O hydrogen bonds. At room temperature, the structures are typically ordered and the transport properties are rather conventional. Upon slight heating, however, many in the MHXO4 and M3H(XO4) families of solid acids, where M = Cs, NH4, and X = S or Se, transform into a disordered structure and exhibit conductivities as high as 10⫺2 ⍀⫺1cm⫺1. Proton transport is facilitated by the rapid reorientation of XO4 groups in the disordered structure, Fig. 9 [59,60]. These materials are true proton conductors; no water molecules are required to serve as hosts for a vehicular transport mechanism, and the electrolyte need not be hydrated. While the proton transport properties of this class of solid acids are very attractive for fuel cell applications [61], several challenges must be addressed before they reach technological relevance. The most important of these is the tendency of sulfate and selenate based materials to become reduced under hydrogen in the presence of typical anode catalysts such as Pt [62]. The by-product of this reduction reaction, H2S (or H2Se), is an exceptional poison for the electrocatalyst, and even if membrane degradation is only slight, the impact on fuel cell performance is devastating. Other inorganic proton conductors are known and include the zirconyl phosphates, Zr(HPO4)2⫺x(OH)x앫nH2O and HUO2PO4앫nH2O

[25]. These are water-insoluble, layered compounds containing intercalated hydronium ions and have reasonable room temperature conductivity, Fig. 7(b). However, the proton transport properties are highly dependent on the humidity level of the atmosphere and thus for fuel cell applications water management remains a challenge. The complex (solid) acids H3PMo12O40앫nH2O and (also known as H3PW12O40앫nH2O heteropolyacids) exhibit exceptionally high conductivities at room temperature, ~0.17 S/cm, when 29 waters of hydration are present (n = 29) [63]. Upon slight heating, the compounds dehydrate and the conductivity drops precipitously. Moreover these materials are water soluble. As such, use of these materials in fuel cells implies the impossible requirements of retaining hydration to ensure high conductivity and removing by-product water to prevent dissolution. Although these compounds are of little value as solid state electrolytes, they may provide benefits with respect to rapid oxygen reduction kinetics when implemented as aqueous electrolytes. Indeed, a peak power density of ~700 mW/cm2 has been reported for a room-temperature, aqueous H3PW12O40앫nH2O electrolyte fuel cell operating on hydrogen/oxygen at one atmosphere [64]. 6. Electrodes and electrocatalysts 6.1. Anode electrocatalysts Electrocatalysis of hydrogen on metals such as Pt and Ni is relatively facile, and anode overpoten-

Fig. 9. Proton transport mechanism in a disordered acid sulfate compound. HSO4⫺ tetrahedra undergo rapid reorientations with the proton attached to a particular oxygen atom, left; proton transfer from one tetrahedron to the next occurs on a much slower time scale, right.

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

tials are a small contribution to the overall drop in fuel cell voltage. The rate limiting step is the adsorption of hydrogen onto the metal surface, as opposed to the subsequent reaction of that hydrogen to yield protons and electrons. In the acidic environment of a sulfonated polymer, Pt particle coarsening, particularly at the cathode (where hydration is greatest) has been observed, but has surprisingly little influence on cell voltage [65]. Challenges in PEM anode electrocatalysis arise when the hydrogen fuel contains residual CO, or when methanol is to be directly electro-oxidized. The reactions in a DMFC are: Anode: CH3OH ⫹ H2O→CO2 ⫹ 6H+ ⫹ 6e⫺ (19) Cathode: 1.5O2 ⫹ 6H+ ⫹ 6e⫺→3H2O

(20)

Overall: CH3OH ⫹ 1.5O2→CO2 ⫹ 2H2O

(21)

The most effective catalyst for methanol electrooxidation is Pt–Ru, in which the two components are typically segregated on the nanoscale. Methanol is adsorbed onto Pt clusters and then fragmented into dehydrogenation products, essentially CO. Electro-oxidation of CO, which would otherwise remain strongly absorbed onto Pt, is subsequently catalyzed by oxygen-like or hydroxyl species absorbed onto neighboring Ru sites [66,67]. Thus, anode electrocatalysts that are effective for direct methanol fuel cells, tend also to be CO tolerant. Despite significant research efforts to develop advanced Pt-alloy catalysts, none that are more effective than Pt–Ru have been identified. An alternative approach has been to introduce small amounts of oxygen to the anode chamber, the socalled ‘air-bleed’ approach, to directly oxidize CO and remove it from the Pt surface. This technique by definition causes some loss in efficiencies, and a catalyst approach would ultimately be preferable. In order for bifunctional catalysts to be effective, the two constituents must be arranged in a specific configuration on the catalyst surface. However, alloys, with their random distribution of atomic species, are ill-suited to provide precise chemical arrangements. In contrast, intermetallic compounds have highly regular and thermodynamically stable surface arrangements, and may circumvent the fundamental limitations of alloy-based electrocata-

5997

lysts. Few studies on such systems have been carried out to date, but preliminary investigations of Pt–Bi compounds by cyclic voltammetry are promising [68], and may open a new avenue of electrocatalyst research. In solid oxide fuel cell research, recent reports of direct electro-oxidation of hydrocarbon fuels have appeared [69,70]. Direct electrochemical oxidation refers to the reaction of hydrogen carbon fuels directly with oxygen ions, without intermediary reaction steps with water. As noted above, the technological consequence is that excess water need not be recirculated in the anode chamber of SOFCs [16]. While the detailed reaction steps remain to be elucidated, it is evident that incorporation of ceria into the anode cermet is essential. Moreover, minimization of the nickel content (by replacing it with a metal such as copper) is also important in order to limit carbon deposition or ‘coking’. It is quite possible that instead of direct electro-oxidation, water generated in situ in the anode compartment induces steam reforming of the hydrocarbon fuel. Establishing whether or not this occurs will be essential for understanding and ultimately optimizing the process. 6.2. Cathode electrocatalysts As in the case of CO tolerant anode catalysts, the search for effective cathode catalysts for PEM fuel cells has focused on Pt alloys. Success in developing alternatives has been only marginal. The Pt–Cr system may offer some slight advantages over Pt alone, and this has been explained in terms of both the electronic structure of the alloy (d-orbital vacancy per atom) and the crystal structure (mean Pt–Pt bond distance) [43]. This latter requirement again calls for an approach using intermetallic compounds rather than alloys. An even greater departure from the hegemony of Pt alloys is to consider oxide and other compounds, either as supports for Pt or as cathode materials directly. Such materials are expected to remain stable under the oxidizing conditions of the cathode and may offer the benefits of mixed electronic/proton conduction. Moreover, it is wellknown in the (non-electrochemical) catalysis community that the choice of catalyst support has a

5998

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

profound effect on catalyst activity, and, in particular, mixed conducting oxides such as ceria can significantly enhance reaction rates. While few oxide materials have been considered for PEM fuel cell applications, recent studies of Pt supported on hydrous, amorphous FePOx demonstrate the value of this approach [71]. Oxygen electroreduction on the surface of oxides such as LSM in solid oxide fuel cells relies on the mixed valence of the B-site cation in the transition metal perovskite. Incorporation of divalent dopants into lanthanum manganite under oxidizing conditions results in the creation of Mn4+ species, which, in turn, give rise to high electronic (p-type) conductivity via Mn3+ ↔ Mn4+ electron transfer. Furthermore, if oxygen vacancies can be retained in the perovskite, oxygen ion conductivity can be increased. In the case of LSM, oxygen ion conductivity is very low, and the material is preferably utilized in a composite cathode, in which the electrolyte material provides high ionic conductivity. For reduced-temperature solid oxide fuel cells, the search for replacements to LSM has centered around related transition metal perovskites, in which a variable valence element resides at the octahedral site. It has been known for some time that lanthanum cobaltites have higher electronic conductivities that lanthanum manganites, but they have not been suitable for YSZ-based fuel cells because they become reduced at very high temperatures. At reduced temperatures and significant Fe incorporation, however, they show excellent promise. In particular, La1⫺xSrxCo1⫺yFeyO3⫺d (LSCF, typically x ~ 0.2, y ~ 0.8) has emerged as a viable candidate. Even more recently, Sm0.5Sr0.5CoO3 has been reported to exhibit excellent cathode activity and examined by several groups. At this stage it is unclear exactly which features of the perovskite lead to low cathode overpotentials, but the tunability of the perovskite structure certainly bodes well for further improvements. As these various materials are developed, constraints regarding chemical and thermomechanical compatibility with the electrolyte will require attention. In addition, it is worth noting that lanthanide and alkaline earth containing compounds tend to form carbonates (upon reaction with CO2 in the

air stream) and this can be extremely deleterious to fuel cell performance. 6.3. Single chamber fuel cells—A cue from biology A very recent development in solid oxide fuel cells is the demonstration of single chamber fuel cells [72]. Such cells, Fig. 10, utilize mixed fuel/oxidant mixtures in a single gas inlet and rely on carefully selected anode and cathode catalysts to produce well-controlled, half-cell reactions: 1 Anode: CH4 ⫹ O2 → CO 2

(22)

⫹ 2H2 (chemical) H2 ⫹ O= → H2O ⫹ 2e⫺ (electrochemical)

(23)

CO ⫹ O= → CO2

(24)

⫹ 2e (electrochemical) ⫺

1 Cathode: O2 ⫹ 2e⫺ → O= (electrochemical) (25) 2 Ideally, simple chemical oxidation of the hydrocarbon, which would yield CO2 and H2O, does not take place. Instead, partial oxidation occurs at the anode and the products of this reaction are then consumed electrochemically, while oxygen is consumed electrochemically at the cathode. Because

Fig. 10. Schematic of a single chamber fuel cell. A hydrocarbon fuel is partially oxidized at the anode producing CO and H2 and consuming O2. The resulting oxygen partial pressure gradient drives the electrochemical reactions of the fuel cell.

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

complications due to sealing are eliminated, the SCFC greatly simplifies system design and enhances thermal and mechanical shock resistance, thereby allowing rapid start up and cool down. While it may seem optimistic, at first glance, to expect that catalysts could be sufficiently selective so as to yield good power densities, such chemical precision is routinely utilized in nature. Indeed, biofuel cells operating on aqueous glucose, that is not separated from ambient oxygen, demonstrate essentially perfect electrode selectivity [73] and may ultimately enable extremely compact fuel cell designs that are ideally suited to miniature, lowpower applications.

5999

enable operation under ‘warm’ conditions (i.e. above 100 °C) and be impermeable to methanol. Additional PEMFC research is very much directed towards the development of high activity cathode electrocatalysts and CO tolerant anode electrocatatlysts, which would furthermore be well-suited to direct methanol fuel cells. Dramatic reductions in the Pt content in PEM fuel cells have been achieved over the past 20 years, however complete elimination of Pt remains a goal. Successes in these arenas of cost and complexity reduction rely on continued advances in materials development and fabrication routes, and are essential for realizing the market and environmental potential of fuel cells.

7. Conclusions After almost a century of slow and at times almost sputtering progress, fuel cell research has exploded with activity over the past decade. The results have been tremendous, with power densities increasing by factors of two and catalyst utilization by more than an order of magnitude. These achievements have resulted from the development of new materials (e.g. La1⫺xSrxGa1⫺yMgyO3⫺d oxide ion conductors) as well as new processing techniques (e.g. electrocatalyst-layer deposition for polymer electrolyte fuel cells). Reduction of cost and system complexity remain significant challenges. Current efforts in SOFC research are aimed at (1) reducing operating temperatures to 500–800 °C to permit the use of low-cost ferritic alloys for the interconnect component of the fuel cell stack and (2) enabling the direct utilization of hydrocarbon fuels. Achieving these goals will require the development of highly active cathode materials and of highly selective anode materials that do not catalyze carbon deposition. Ceramic electrolytes that are operable at reduced temperatures are available (doped ceria and LSGM) but high performance single-cell fuel cells based on these materials have yet to be demonstrated. Current efforts in PEMFC research are focused on (1) reducing membrane cost via the use of non-fluorinated polymer electrolytes and (2) reducing system complexity via the development of ‘water-free’ electrolytes that do not require cumbersome hydration paraphernalia. Such electrolytes would additionally

Acknowledgements The electron micrograph shown in Fig. 3(a) was provided by Dr. Zongping Shao, post-doctoral scholar in the author’s laboratory and funded by DARPA, Microsystems Technology Office. Additional financial support was provided by the Department of Energy, Office of Energy Efficiency and Renewable Energy.

References [1] Hirschenhofer JH, Stauffer DB, Engelman RR, Klett MG. Fuel Cell Handbook, 4th edn. Parsons Corp., for U.S. Dept. of Energy Report No. DOE/FETC-99/1076; 1998. [2] Larmine J, Andrews D. Fuel cell systems explained. Chichester: John Wiley & Sons Ltd.; 2000. [3] Carrette L, Friedrich KA, Stimming U. Chem Phys Chem 2000;1:162. [4] Crow DR. Principles and applications of electrochemistry. 3rd ed. London: Chapman & Hall; 1988. [5] Bockris JO’M, Srinivasan S. Fuel cells: Their electrochemistry. New York City: McGraw-Hill; 1969. [6] Wagman DD, Evans WH, Parker VB, Schumm RH, Halow I, Bailey SM, Churney KL, Nutall RL. J Phys Chem Ref Data 1982;11(suppl. 2). [7] Reiss I, Goedickemeier M, Gauckler LJ. Solid State Ionics 1996;90:91. [8] Hirano S, Kim J, Srinivasan S. Electrochem Acta 1997;42:1587. [9] Minh NQ. J Am Cer Soc 1993;78:563. [10] Ormerod RM. Chem Soc Rev 2003;32:17. [11] Singhal SC. Solid State Ionics 2002;405:152–153.

6000

S.M. Haile / Acta Materialia 51 (2003) 5981–6000

[12] Singhal SC, editor. Solid oxide fuel cells. Proceedings of 1st International Symposium. Pennington (NJ): The Electrochemical Society; 1989. [13] de Souza S, Visco SJ, DeJohnge LC. J Electrochem Soc 1997;144:L35. [14] Virkar AV, Chen J, Tanner CW, Kim JW. Solid State Ionics 2000;131:189. [15] George R, Casanova AC, Veyo S. Status of Siemens Westinghouse SOFC Program. Extended Abstracts of the 2002 Fuel Cell Seminar. Washington (DC): Courtesy Associates, Inc., 2002. [16] Steele BCH. Nature 1999;400:619. [17] Costamagna P, Srinivasan S. J Power Sources, 2001;102:242. [18] Costamagna P, Srinivasan S. J Power Sources, 2001;102:253. [19] Thomas SC, Ren X, Gottesfeld S, Zelenay P. Electrochem Acta 2002;47:374. [20] Steele BCH, Mat Sci and Eng 1992;B13:79. [21] Ishihara T, Matsuda H, Takita Y. J Am Chem Soc 1994;116:380 [22] Bohn HG, Schober T. J Am Cer Soc 2000;83:768. [23] Kreuer KD, Chem Phys Chem 2002;3:771. [24] Slade RCT, Omana MJ. Solid State Ionics 1992;58:195. [25] Alberti G, Casciola M. Solid State Ionics 2001;145:3. [26] Coors G. J Power Sources 2003;118:150. [27] Eguchi K, Setoguchi T, Inoue T, Arai H. Solid State Ionics 1992;52:165. [28] Steele BCH. Solid State Ionics 2000;129:95. [29] Mogenson M, et al. 2003 (manuscript in preparation). [30] Milliken C, Guruswamy S. J Am Cer Soc 2002;85:2479. [31] Atkinson A, Ramos TMGM. Solid State Ionics 2000;129:259. [32] Feng M, Goodenough JB. Eur J Solid State Inorg Chem 1994;31:663. [33] Huang HQ, Wan JH, Goodenough JB. J Electrochem Soc 2001;148:A788. [34] Ishihara T, Shibayama T, Nishiguchi H, Takita Y. J Mat Sci 2001;36:1125. [35] Kuroda K, Hashimoto I, Adachi K, Akikusa J, Tamou Y, Komado N, Ishihara T, Takita Y. Solid State Ionics 2000;132:199. [36] Hibino T, Hashimoto A, Asano K, Yano M, Suzuki M, Sano M. Elect Solid State Letters 2002;5:A242. [37] Sahibzada M, Steele BCH, Barth D, Rudkin AR, Metcalfe IS. Fuel 1999;78:639. [38] Cheng JG, Fu QX, Liu XQ, Peng DK, Meng GY. Proceedings of Key Engineering Materials 2002;224–2:173–177. [39] Yoon SP, Han J, Nam SW, Lim TH, Oh IH, Hong SA, Yoo YS, Lim HC. Jo. Power Sources 2002;106:160. [40] Simner SP, Bonnett JF, Canfield NF, Meinhardt KD, Sprenkle VL, Stevenson JW. Electr Solid State Letters 2002;5:A173. [41] Sammes NM, Tompsett GA, Nafe H, Aldinger F. J Eur Cer So 1999;19:1801. [42] Shuk P, Weimhofer HD, Gush U, Gopel W, Greenblatt M. Solid State Ionics 1996;89:179.

[43] Wachsman ED. Solid State Ionics 2002;152–153:657. [44] Iwahara H, Yajima T, Hibino T, Ushida H. J Electrochem Soc 1993;140:1687. [45] Haile SM, Staneff G, Ryu KH. J Mat Sci 2001;36:1149. [46] Kreuer KD. Solid State Ionics 1999;125:285. [47] Taniguchi N, Yasumoto E, Gamo T. J Electrochem Soc 1996;143:1186. [48] Iwahara H, Yajima T, Hibino T, Ushida H. J Electrochem Soc 1993;140:1687. [49] Kreuer KD, J Membr Sci 2001;185:2. [50] Jang SS, Molinero V, C¸ ag˘ in, T, Merinov BV, Goddard III WA. Solid State Ionics (submitted for publication). [51] Rikukawa M, Sanui K. Prog Polym Sci 2000;25:1463– 1502. [52] Statti P, Arico AS, Baglio V, Lufrano F, Passalacqua E, Antonucci V. Solid State Ionics 2001;145:101. [53] Yamaguchi T, Miyata F, Nakaoa SI. J Membr Sci 2003;214:283. [54] Lassegues JC. In: Colomban PH, editor. Proton conductors: Solids, membranes and gels—materials and devices. Cambridge: Cambridge University Press; 1992, p. 311–328. [55] Wang JT, Savinell RF, Wainright J, Litt M, Yu H. Electrochem Acta 1996;41:193. [56] Wang JT, Wainright JS, Savinell RS, Litt M. J Appl Electrochem 1996;26:751. [57] Savadogo O, Xing B. J New Mat Electrochem Sys 2000;3:345. [58] Kreuer KD, Fuchs A, Ise M, Spaeth M, Maier J. Electrochem Acta 1998;43:1281. [59] Haile SM. Mat Today 2003:24. [60] Munch W, Kreuer KD, Traub U, Maier J. Solid State Ionics 1995;77:10. [61] Haile SM, Boysen DA, Chisholm CRI, Merle RB. Nature 2001;410:910. [62] Merle RB, Chisholm CRI, Boysen DA, Haile SM. Energy & Fuels 2003;1:21. [63] Nakamura O, Kodama T, Ogino I, Miyake Y. Chem Lett Chem Soc of Jap 1979:17. [64] Giordano N, Staiti P, Hocevar S, Arico AS. Electrochimica Acta 1996;41:397. [65] Wilson MS, Garzon FH, Sickafus KE, Gottessfeld S. J Electrochem Soc 1993;140:2872. [66] Gasteiger HA, Markivic N, Ross PN, Cairns EJ. J Phys Chem 1993;97:12020. [67] Long JW, Stroud RM, Swider-Lyons KE, Rolison DR. J Phys Chem B 2000;104:9772. [68] Casado-Rivera D, Gal Z, D’Angelo AC, Lind C, DiSalvo FJ, Abruna HD. Chem Phys Chem (accepted). [69] Murray EP, Tsai T, Barnett SA. Nature 1999;400:649. [70] Park S, Vohs JM, Gorte RJ. Nature 2000;404:265. [71] Bouwman PJ, Dmowski W, Swider-Lyons K. Ext Abstracts, 204th Meeting of The Electrochemical Society. [72] Hibino T, Hashimoto A, Yano M, Suzuki M, Yoshia S, Sano M. J Electrochem Soc 2002;149:A133. [73] Mano N, Mao F, Heller A. J Am Chem Soc 2002;124:12962.