Ntio- NO-- 7

Solid Polymer Fuel Cells Electrode and membrane performance studies

Steffen M0ller-Holst

Thesis submitted in partial fulfillment of the requirements for the degree of

Doktor ingeni0r

Department of Physical Chemistry Norwegian University of Science and Technology Trondheim May 1996

fi®n#BunoN of ths document « mum

1

Acknowledgments

I would like to thank my supervisor Professor Signe Kjelstrup for encouragement and valuable discussions throughout the thesis work. She has always been positive and has shown great confidence in me and my abilities. Magnar Ott0y, with whom I shared office for the first years, has been a continuous support. Thanks to him, many every-day and more fundamental problems found their solutions. He has become a friend to entrust and has been a good traveling companion on several seminars and workshops.

The employees at our workshop Nils Wasraas, Sture N0sen, Geir Solem and Arne Foss were all very helpful and understanding during the build-up of the fuel cell test-station. Rune Dahl was supportive during LabView programming for the data acquisition task. On many occasions Odd Ragnar Lerst0l assisted me with computer problems, and lately also with scanning pictures to the manuscript. Parts of the electronics of the test-station were made by Hans Lauritz Petersen. Vibeke Andersson is acknowledged for her contribution to electrode preparation. Preben J.S. Vie was helpful by preparing samples and operating the instrument during SEM-analysis. Svein Bratberg assisted me in drawing parts of the equipment, using AutoCad software. Tharald Tharaldsen was helpful photographing some of the equipment. I am also indebted to Habiba Muza Rajabu for her valuable comments to this manuscript.

I wish to thank The Norwegian Research Foundation (NFR) for financing my Ph.D. study. I am also appreciative to Nansenfondet, The Department of Physical Chemistry, NTNU, and SINTEF for financing most of the materials and equipment. The Nordic Energy Research Program is acknowledged for their invitations and travel grants to several program seminars and workshops. The companies E-TEK, Inc., MA, USA, Cabot B.V., Netherlands and Johnson Matthey, U.K. have Acknowledgment ii provided electrode materials and samples free of charge, and are hence acknowledged.

Finally, I wish to thank my wife Lise for invaluable, absolute and continuous support and encouragement, and for always believing in me and my work. My daughter Siri came into my life last November. Her smiles were precious and filled me with happiness. She made the work a lot easier and brought a wider perspective to my mind.

Trondheim, April 1996 Steffen Mpller-Holst DISCLAIMER

Portions of this document may be illegible in electronic image products. Images are produced from the best available original document.

Table Of C ontents

Acknowledgments i

Table Of C ontents m

Chapter 1 Introduction ...... l

1.1 Background ...... 1

1.2 Solid polymer fuel cells, technology development ...... 3

1.3 Problem statement ...... 5

1.4 Aim of work ...... 5

1.5 Outline of thesis ...... 6

Chapter 2 System D escription ______9

2.1 Operating principle ...... 9 2.1.1 Cell reactions...... 9 2.1.2 Water management...... 10

2.2 Cell performance characteristics ...... 12 2.2.1 Reversible cell potential ...... 12 2.2.2 Open circuit potential ...... 13 2.2.3 Cell potential and overpotentials...... 14 IV Table Of Contents

Chapter 3 Fuel C ell Efficiency ______19

3.1 Introduction ...... 19

3.2 Exergy efficiency and local heat production . ... (enclosed article)...... 23

Chapter 4 Equipment For C ell Preparation And Testing ______31

4.1 Introduction ...... 31

4.2 Test-station description ...... 32 4.2.1 Gas supply and humidification system...... 33 4.2.2 Single cell fixture...... 34 4.2.3 Instrumentation and data acquisition ...... 36 4.2.4 Operation limitations, discussion...... 38

4.3 Hot pressing equipment ...... 39

4.4 Spray deposition equipment ...... 40

4.5 Operation stability ...... 41

4.6 Conclusions ...... 45

Chapter 5 Electrode Preparation And Evaluation ______47

5.1 Introduction ...... 47 5.1.1 Electrode development...... 47 5.1.2 Aim of work...... 50

5.2 Single cells with thin film electrodes , system description ...... 51 5.2.1 Electrode structure, materials and composition...... 51 5.2.2 Membrane and electrode assembling procedures...... 53

5.3 Performance evaluation and separation of potential losses ...... 54

5.4 Experimental design ...... 54 5.4.1 A 23-screening experiment...... 55 5.4.2 Addition of Acetylene Black ...... 55 V

5.5 Materials and methods ...... 56 5.5.1 Membrane preparation ...... 56 5.5.2 Modification of commercial electrodes...... 56 5.5.3 Thin film electrodes...... 57 5.5.4 Membrane and electrode assembling...... 58 5.5.5 Cell test procedure...... 59

5.6 Results and discussion ...... 59 5.6.1 Commercial electrodes...... 59 5.6.2 Thin film electrodes, 23 -screening experiment...... 61 5.6.3 Platinum utilization, a comparison ...... 63 5.6.4 Thin film electrodes, addition of Acetylene Black ...... 64 5.6.5 Temperature and pressure dependence of cell potential ...... 66 5.6.6 Single cell performance, a comparison to literature ...... 67 5.6.7 SEM analysis ...... 69 5.6.8 Advantages and limitations of the spray deposition technique...... 70

5.7 Summary and conclusions ...... 72

Chapter 6 Ion and Water Transport in Membranes ______73

6.1 Introduction ...... 73

6.2 Possible implications for solid polymer fuel

CELL PREPARATION AND OPERATION...... 74

6.3 Membrane transference numbers from a new

emf method ...... (enclosed article)...... 75

6.4 Water and ion transport in cation exchange

membrane systems NaCl-SrCl2 and KC1-SiC12...... (enclosed article)...... 83

Chapter 7 C oncluding Remarks ______93

References . 95 VI Table Of Contents

Appendices

A. Basic Equations ...... 103

B. Compositions And Effects In The 23-Experiment .„...... 107

C. Reproducibility Of Single Cell Performance ...... 109

D. Resolution Of Potential And Current Measurements ...... 115

E. Description Of A 32-Experiment ...... 117

F. Preparation Of Nafion 117 Membranes ...... 121

G. Separation Of Potential Losses ...... 123

H. Experimental Data ...... 129 1

C hapter 1

Introduction

The purpose of this chapter is to give the background for the project. Highlights from the fuel cell technology development are reviewed. The aim of the present work is stated and an outline of the thesis is given.

1.1 Background

The increasing demands on the earth ’s limited resources, as a consequence of an expanding population and dwindling fossil reserves, make it necessary to identify and develop new technologies for more efficient generation and usage of energy. The present large scale burning of fossil fuel at very low conversion efficiencies, will eventually have to be substituted by more efficient devices for energy conversion and storage due also to increasing environmental concerns.

Fuel cells are electrochemical energy converters for production of electrical energy directly from chemical energy. Like batteries, they are discharged electrically, but in addition the fuel cells are simultaneously charged chemically by supplying a fuel (usually a gas). Due to their high efficiency, fuel cells have received great attention during the last two decades through numerous projects world-wide.

The principle of fuel cell operation includes no combustion in the conventional sense and is, thus, not limited by the Camot-efficiency 1. Theoretically, an ideal fuel cell will convert all available Gibbs energy into electricity. However, due to resistance in the materials and slow reaction rates, there are potential losses during fuel cell operation.

1 The upper limit of efficiency, t? c, for the conversion of heat into mechanical energy was shown by

Carnot to bet? C =(T2-T1)Ar2, where T2 and T, are the temperature of the heat source and heat sink, respectively. Fuel cell efficiency is dealt with in Chapter 3. 2 Chapter 1. Introduction

The lost energy is released as heat, which in turn may be converted to electricity, but then the Camot-efficiency limitation applies. Although fuel cell materials have been studied extensively, some 40-50% of the available energy in the fuel is converted to heat [1], not electricity. Comparably, the losses are larger than 60% for conventional gas power stations 1. General Electric Company in USA has recently developed a new thermal power station 2 with an efficiency of more than 60%. This is, however, close to the maximum of what is achievable from this energy conversion technology. For fuel cell systems this ideal limit is 100%, leaving large possibilities for further improvements. If the heat from the fuel cell is utilized, an efficiency of 70-80% should be obtainable.

Fuel cells are classified according to the material used as electrolyte in the cells. There are five main types: Type of cell Abbreviation Operation Temperature Alkaline Fuel Cell AFC Room temp.-200°C Solid Polymer Fuel Cell SPFC Room temp.-100°C Phosphoric Acid Fuel Cell PAFC 180-200°C Molten Carbonate Fuel Cell MCFC 600°C Solid Oxide Fuel Cell SOFC 1000°C

The electrolyte is either a liquid, e.g. an aqueous solution as in AFC and PAFC, a molten salt as in MCFC or a solid ion conductor as in SPFC and SOFC. Operation conditions vary widely, e.g. operation temperature ranges from room temperature to 1000°C. Detailed descriptions of the different fuel cell types are given by Appleby and Foulkes[3] and Kinoshita et al.[4]. In this thesis, two types of fuel cells are dealt with, i.e., the Solid Polymer Fuel Cell (SPFC) and the Solid Oxide Fuel Cell (SOFC). The SPFC is described in detail in Chapters 2 and 5, whereas the SOFC is briefly described in Chapter 3.

1 The U.S. average of fuel-to-electricity conversion efficiency was about 37% (1995) for mature baseload fossil fuel systems [2], 2 which utilizes steam to cool the blades of the gas turbine, giving extra energy recovered in the steam engine system. 1.2. Solid polymer fuel cells, technology development 3

The fuel cells may be used both as the heart of stationary power stations and as a mobile unit for transportation purposes. High temperature cells (SOFC and MCFC) are advantageous for stationary applications, where the heat produced may easily be utilized. The low temperature cells (AFC and SPFC) are better suited for mobile power production (i.e. vehicle propulsion). Legislation promoting ultra-low and zero emission vehicles, has led to large increases in the potential market for fuel cell systems. Modularity ensures that the efficiency is high even for small units of less than 100 kW. From a maintenance point of view, the fuel cell is advantageous due to its ability to convert energy with the absence of moving parts. Among the fuel cell types, the SPFC and the SOFC utilizes solid materials only. This is advantageous in terms of stable long term operation. The cell types are extremes in terms of operating temperatures, the SPFC at around 100°C and the SOFC at 1000°C.

1.2 Solid polymer fuel cells, technology development

The history of fuel cell development is covered adequately in books [5] and review articles [6,7]. Highlights of the development is given in this section with emphasis on SPFCs.

In the year of 1802 an interesting phenomenon was observed by N.Gutherot. He discovered the reversible fuel cell using H2 and Oz gases1 [8]. But, the first description of the principle of the fuel cell has (somewhat unfairly?) been attributed to Sir William Grove in the late 1830s [5]. Groves approach was, however, more advanced as he connected six cells in series and applied sulfuric acid as electrolyte [9]. Fuel cell research was scattered in the following century, until the 1930s when F.T. Bacon started to develop the modem hydrogen-oxygen fuel cell with alkaline electrolyte [5], As for many other technologies, the fuel cell development was boosted by the ‘unlimited budgets ’ of the space research organization NASA in the 1950s. Co­ production of energy and drinking water was advantageous for extraterrestrial flights * 2

1 He applied two Pt-wire electrodes, each sealed into the end of a glass tube. By immersing these in water and sending direct current form a Volta’s primary battery he was able to split water into H2 and O2. By connecting the electrodes he measured a current flowing from recombinaton of the gases to form water. 4 Chapter 1. Introduction

[6]. Low temperature fuel cells were first developed. The Solid Polymer Fuel Cell was developed by General Electric Company and was first applied as the auxiliary power source in the Gemini space missions [5]. The Gemini fuel cell unit was a 1 kW stack operating on H2 and 02. The power density of these cells was low (<50 mW/cm2) and the use of unstable, sulfonated, cross-linked polystyrene ion exchange membranes limited the cell life time [6]. These problems were the main reason for choosing the alkaline fuel cell (based on the pioneer work done by Bacon) for the Apollo program. Invention of the perfluorinated sulfonic acid membrane Nafion® (DuPont) in 1962 eventually gave SPFC its break-through some twenty years later. Nafion was developed for the chlor-alkali industry, but General Electric applied the membrane as electrolyte in SPFCs in 1966. These membranes exhibit high chemical stability and could, thus, prolong the lifetime of the SPFC dramatically. The energy crisis in 1973 was the impetus for a large number of research and development programs on fuel cells for terrestrial applications. It took, however, yet another decade before the SPFC research got its renaissance. In the 80s, some new membranes were introduced to the market. The Dow Chemical Company developed a membrane (XUS 13204.10) similar in structure to Nafion, but with lower molecular weight. Higher conductivity of the DOW membrane was due to higher density of the sulfonic acid groups and thinner membranes. Other membranes were also developed, e.g., Aciplex-S (by Asahi Chemical Industry Co., Ltd., Japan) based on carboxylic acid and Membrane C (by Chlorine Engineers, Japan). Membranes prepared by modifying a polymer substrate by irradiation grafting followed by sulfonation have been reported (see e.g.,[10]). The membranes show promising properties and the technique has been adopted by other research groups (e.g., in Finland). Large improvements have also been obtained for the electrodes as described in Section 5.1. Since the middle of the 1980s, the SPFC technology has shown large performance improvements. Today the SPFC technology can provide extremely high power densities and is rapidly approaching commercialization, especially for vehicle propulsion. 1.3. Problem Statement 5

1.3 Problem Statement

Low temperature operation makes use of noble catalysts unavoidable in SPFC. Platinum has been shown to be a catalyst material with adequate properties, but contributes to the increase of SPFC production cost. Research has shown that the amount of catalyst may be reduced through improved Pt-utilization. Prices of platinum are increasing and further reduction of Pt-loading is therefore of interest.

Electrode and membrane materials for SPFCs are usually studied separately in half cells in aqueous acidic solution. These conditions differ from the actual SPFC operation conditions. There may be changes of material properties during preparation and assembling of the electrodes and membrane to single cells. Thus, data obtained from half cell studies are not directly applicable for the single fuel cell.

Evaluation of efficiency of energy converting devices has often been based on the use of the first law of thermodynamics. This gives erroneous results when applied to e.g., electrochemical energy converters. Adequate criteria for efficiency evaluation based on the second law must be chosen to obtain information about possibilities of performance improvements for a device, and further, to provide a general basis for comparison of different energy converting devices.

1.4 Aim of work

The aim of this thesis is to study several aspects of fuel cell preparation and performance. The main emphasis is put on preparation and analysis of low Pt-loading SPFC electrodes. Important factors concerning electrode composition and preparation procedure and how these influence single SPFC performance will be addressed. By studying single fuel cells under realistic operating conditions, the changes occurring during cell preparation and assembling are taken into account. Equipment for electrode preparation and single SPFC testing shall be designed and built for such studies. Water and ion transport in membranes will be studied for some model systems. 6 Chapter 1. Introduction

Further, the efficiency of fuel cells will be studied theoretically. The aim of this part of the thesis is to present an exergy analysis method using SOFC as an example. The method is general and may be used for any energy converting device.

1.5 Outline of thesis

Most of the work presented in this thesis is published or in print. Printed articles are enclosed as reprints in separate chapters. Each chapter starts by shortly stating the purpose and contents of the chapter {italics). Information which is peripheral for the context is placed in appendices. Results from cell tests are reported either in figures or in tables. Experimental data which are used for further analysis, are included in Appendix H. References are numbered and placed in squared brackets, i.e., [n], whereas footnotes are indicated by numbers in superscript format (i.e., “).

In Chapter 2, a description of the SPFC is given. Cell performance characteristics are discussed and the major sources of potential losses are addressed.

Chapter 3 is a theoretical work on fuel cell efficiency. Expressions for efficiency calculations are discussed and the second law analysis is recommended. A second law efficiency calculation of a fuel cell is given in the enclosed article ‘Exergy efficiency and local heat production in solid oxide fuel cells'. The theoretical basis for the article was developed by my supervisor Signe Kjelstrup and G. H. Hertz. My contribution was connected to the adaptation and application of the theory to the fuel cell.

In Chapter 4, the construction of a SPFC test-station is described. This includes hardware as well as software, i.e., equipment and procedures for cell testing and data acquisition. The design and construction of the test-station has been a substantial part of the work. Some introductory experiments on cell stability are reported. Details of the equipment will be reported in a separate technical report 1.5. Outline of thesis 7

Preparation of electrodes for SPFC and evaluation of single cells (Chapter 5) constitutes the major experimental part of the thesis work. Experimental procedures are given and results of single SPFC performance tests are interpreted and discussed.

Chapter 6 contains two articles concerning ion and water transport in membranes. The method described in the first article 1 Membrane transference numbers from a new emf method’ was developed by Magnar Ott0y. I was responsible for measuring the second system (i.e., Na+/H+) and evaluating the data of this system. On the second article ‘ Water and ion transport in cation exchange membrane systems NaCl-SrCl 2 and KCl- SrClf , the experimental work was done by students co-supervised by me. A summary of the results is given and possible implications for operation of SPFCs are briefly discussed.

Overall conclusions of the theoretical and experimental work are drawn in Chapter 7. 8 Chapter 1. Introduction 9

C hapter 2

System D escription

The operating principle of the solid polymer fuel cell is described. Water transport processes in the system are briefly reviewed. Potential losses through the cell are discussed and a simplified model for the polarization curve is deduced.

2.1 Operating principle

2.1.1 Cell reactions The solid polymer fuel cell (SPFC) consists of two porous carbon supported electrodes separated by a membrane. The membrane plays the dual role as proton conducting electrolyte and gas separator. The SPFC is shown schematically in Figure

2.1. In polymer fuel cells operating on hydrogen and oxygen gases, the changes occurring in the cell are as follows:

Anode reaction iH2(g) = H+(an) + e~ (2.1) Membrane transfer

H+(an) = H+(cat ) (2.2) Cathode reaction

\02 + e~ +H+(cat) = \H20 (2.3)

Overall cell reaction (2.4)

In these equations the transfer of water across the membrane is neglected. 10 Chapter 2. System Description

Hydrogen gas is oxidized to protons, H+(an), at the anode. The protons are transferred through the membrane to the cathode, H*(cat), where they react with oxygen to form water. Equation 2.4 corresponds to the transfer of 1 mole of electrons from the anode to the cathode through the outer circuit.

ANOUfc Mh.MBRAM. Figure 2.1. Illustration of the transport and reactions occurring in the SPEC.

This is a simplified description of the changes occurring in the cell during SPEC operation. The reaction mechanisms at the cathode are complicated and not yet fully understood. As seen from Figure 2.1, water is transported through the membrane in SPFCs as discussed further in the following section.

2.1.2 Water management Water is required to facilitate the proton transport in the membrane phase. The membrane conductivity increases strongly with water content[ll]. During operation there are no extra water requirements since water is produced at the cathode[12]. Maintaining the water balance in the SPFC is, however, complicated. Water management has been subject to extensive modeling by several groups[13,14,15,16]. To maintain an adequate amount of water in the system, water is usually supplied to the cell by humidifying the reactant gases. Excessive humidification must then be avoided because it will dilute the reactant gases. Other approaches have been tried by 2.1. Operating Principle 11

HoO PRODUCTION

ELECTRO-OSMOTIC DRAG

WATER DIFFUSION

H+TRANSPORT

ANODE CATHODE NAFION MEMBRANE

Figure 2.2. A schematic illustration of the water transport and production in SPFC (from Zawodzinski et al.[ll]). incorporating fiber wicks into the membrane[17] and developing new arrangements for internal humidification[18].

Zawodzinski et al.[ll] have illustrated the water transport processes as shown in Figure 2.2. Each proton is carrying a number of water molecules from the anode to the cathode by the so-called electro-osmotic drag. The gradient in water content which arises over the membrane is counteracted by water diffusion.

The number of water molecules transferred per proton (by electro-osmotic drag) has been a topic of controversy. Xie and Okada[19] found a value of 2.6+0.1 for Nafion 117 membrane immersed in liquid water. The water content is considerably lower for the membrane when it is equilibrated in water vapor. The water transport number was found to be close to 1 for Nafion 117 in water vapor[ll]. Values for the net water transport during normal cell operation has been reported to be close to zero[14]. Evidence for the lack of net water transport has been found in three laboratories!!]. 12 Chapter 2. System Description

This indicates that the electro-osmotic drag of water is counteracted by rapid back- diffusion.

The water management depends on several parameters such as current density, temperature, gas flow rates, water partial pressures and electrode morphology. It is concluded that the water management in the SPFC is a complex topic. For a more thorough discussion of the water transport in Nation membranes, the reader is referred toOtt0y[2O].

2.2 Cell performance characteristics

2.2.1 Reversible cell potential The basic equations for the reversible cell potential are given in Appendix A. For the SPFC using pure H2/02 gases, the total cell reaction is given by Equation 2.4. By assuming that activities are equal to partial pressures for gases1 and that activity is unity for the water phase (i.e., saturated water vapor or liquid water is present), the reversible cell potential, Erev, is given by the Nemst equation:

(2.5) where £°w is the reversible cell potential when all components are in their standard states. The pressure and temperature dependence of the reversible potential is given in Appendix A. When the reactant gases are saturated at a temperature that is higher than the cell operation temperature, there will always be liquid water present in the cell. Under such conditions the water produced in the fuel cell is in the liquid phase. The gas pressures at the electrodes are then:

PA = P Ht + P*H20 (2.6 )

Pc = Po2 + PS»iO (2.7)

This corresponds to assuming that fugacities are equal to unity. 2.2. Cell performance characteristics 13 where p H*° is the saturation pressure of water at the cell temperature. The saturation pressure of water is strongly temperature dependent. By combining Equations 25-2.1, the reversible cell potential may be expressed:

-----^ln 1 (2.8) nF Sp a - p sb 2o)-(p c - p sh2oy

In the above expressions, the cell is assumed to be isothermal. Contributions to cell potentials arising from temperature gradients between the electrodes are dealt with in Chapter 3. Differences in water activity between anode and cathode may exist. The resulting potential contribution from water transport is small compared to the cell potential and was hence neglected.

2.2.2 Open circuit potential

The reversible potential (Eq.2.8) is usually not obtainable for H2/02 fuel cells when applying Pt as a catalyst For the SPFC the open circuit potential 2, Eoc, is around 1 V, which is 0.2 V lower than that of the reversible cell potential calculated from Equation 2.8. This discrepancy is, according to Bockris and Srinivasan[5], due to the presence of impurities which undergo anodic oxidation causing a so called ‘mixed potential ’. This explanation is supported by the fact that potentials close to the reversible potential were attainable by successive purification of the electrolyte[5]. Parthasarathy et al.[21] write that this potential loss can also be attributed to the very low exchange current density of the oxygen reduction reaction. The exchange current density of the oxygen reduction reaction is ca. 5-6 orders of magnitude lower than that of the hydrogen oxidation reaction[3]. Formation of some form of peroxide has also been suggested. These speculations have been doubted by Bockris and Srinivasan[5], Allen et al.[22] have studied the cathode in SPFC by in situ dispersive EXAFS spectroscopy and report the presence of chemisorbed O on Pt. The standard reduction potential of the reaction

Pt(OH)2 + 2H+ + 2e~ => Pt + 2H20 (2 o\

2 The open circuit potential is defined as the potential measured (by a high impedance voltmeter) when a negligible current is flowing through the cell. 14 Chapter 2. System Description is 0.98 V, which is close to the observed open circuit potential of the SPFC. This may allow for the presence of intermediate structures of OH on the Pt surface[22]. Cathodic oxygen reduction on noble metal and carbon electrodes was recently reviewed by Brito and Sequeira[23], The difference between reversible cell potential and open circuit potential corresponds to a loss in efficiency of 15 % already at zero current (ref. Chapter 3).

2.23 Cell potential and overpotentials

In the Solid Polymer Fuel Cell, the anode and cathode are separated by a proton conducting membrane. Oxygen gas is supplied to the cathode and hydrogen gas to the anode side of the membrane. We assume that the membrane is isotropic. When no current is flowing through the cell, there is an open circuit potential, Eoc, of around 1 volt over the cell as shown in Figure 2.3. A, C and M denote the anode, cathode and membrane, respectively. The darker gray areas in the figures represent the interface

Figure 2.3. Cross section of the single Figure 2.4. Potential losses through a cell showing the open circuit single cell at a current density, i. The potential, Eoc. A, M and C are the terms iR and 7} are the ohmic and Anode, Membrane and Cathode, non-ohmic overpotentials, respectively. respectively.

between the electrode and the membrane, i.e. the catalyst layer. During cell operation, a certain cell current density, z, is drawn from the fuel cell. This gives rise to potential losses through the cell as is illustrated in Figure 2.4. The potential losses, also called 2.2. Cell performance characteristics 15 the overpotentials, are due to transport and reaction irreversibilities. In Figure 2.4, the overpotentials are divided into ohmic and non-ohmic contributions, termed iR and T), respectively. When Figures 2.3 and 2.4 are superimposed, the potential profile through the cell during operation is obtained as shown in Figure 2.5. We see that the open circuit potential is diminished by the overpotentials and the potential profile through the cell may be written:

E = Eoc — (Ra +«a )(Re + «c) — <«m) ,2 j|Qx = Eoc-(tia - r]c) —i(RA -Rc + Rm)

#

Figure 2.5 Illustration of the overall potential profile through a single cell.

The polarization curve Three types of overpotentials are usually encountered in SPFCs. The major source of potential loss is the slow charge transfer of the oxygen reduction reaction[5,24] giving an activation overpotential. In SPFCs, porous electrodes are used to get a high three phase area (Chapter 5). This introduces ohmic overpotentials from the electrodes (both the electrolyte in the pores and the carbon support), in addition to that of the polymer membrane. Concentration overpotentials arise when the reaction is controlled by the mass transfer to the electrode surface. 16 Chapter 2. System Description

Reversible cell potential

{Open circuit potential

{Cell potential

Current Density [mA/cin ]

Figure 2.6. A typical cell-potential-current-density relation for SPFC, showing three distinct regions A,B and C.

The polarization curve (i.e., the cell potential plotted against current density) for a fuel cell is useful for many purposes of cell evaluation. A typical polarization curve is shown in Figure 2.6 (obtained in our laboratory). There are three distinct regions in this curve, each reflecting the limiting process of performance in that region. The initial steep fall in the curve (region A) is due to the activation limited oxygen reduction reaction. In the linear region B, ohmic losses in the membrane and the electrodes dominate the cell potential reduction. At high current densities, gas diffusion limitations cause the curve to bend down (region C). From the ‘transition state theory’[25] the following relationship between current density and activation overpotential is given by the Butler-Volmer equation: ( OtnF (1-cQnF i = i, exp tj -exp - (2.11) RT RT where i0 is the exchange current density and a is the symmetry factor. The anodic activation overpotential in SPFC is small due to a very high exchange current density 2.2. CELL PERFORMANCE CHARACTERISTICS 17 for the hydrogen oxidation reaction. Examination of Equation 2.11 reveals that the activation overpotential is linear with current density at low overpotentials. Then, the anodic activation overpotential is approximated by:

RT VA = (2.12) nFVo .a V where i'o ,a is the exchange current density of the anode reaction. The value for the symmetry factor for the oxygen reduction reaction on platinum (in phosphoric acid at 70°C) was close to Vi [26]. We will use a symmetry factor equal to Vi in the present work. The cathodic reduction of oxygen is subject to high overpotentials. At high cathodic overpotentials (i.e., 7]c < -5QmV), the first term on the right-hand side of Equation 2.11 is negligible compared to the second, and Equation 2.11 reduces to f jnF_ l = -In exp (2.13) RT Vc or solving explicitly for the cathodic activation overpotential, r\c: RT RT (2.14) % = T^togOo.c) - I^logCO\nF where i0,c is the exchange current density of the cathode reaction. Equation 2.14 is the Tafel equation, often written as (2-15) 7]c = a - fllog(i) where a = B log(i 0,c) is the Tafel constant and B is the Tafel slope. As discussed in Section 2.2.2, the difference between open circuit potential and reversible cell potential is usually attributed to the slow cathode reaction in SPFCs. Hence, the potential profile (Equation 2.10) may be written: (2.16) E ~ Erev -(Wa -B c) ~l(RA ~RC+Rm)

If concentrating on the first two regions (A and B) in the polarization curve, (i.e., disregarding the region of diffusion limitations), combination of Equations 2.12, 2.15 and 2.16 gives the following relation for the cell potential E = E0- Blog(i) - iR, (2.17A) where

E0 = Erev + SlOgOo.c) (2.17B) 18 Chapter 2. System Description and iR, is the sum of all linear overpotential contributions, including the linear contribution for the anodic activation overpotential (Eq.2.12). This expression was given by Srinivasan et al.[27], and has been widely used for determination of exchange current densities, Tafel slopes and ohmic potential losses by non-linear least squares fit of the cell data. The current density, i, in Equation 2.17 A is expressed in mA/cm2. Equation 2.17 is not defined at the open circuit potential (i = 0). Therefore, E0 is chosen as the potential at a low current density. In this study, we will use Eq at a current density of 2 mA/cm2. It is important to realize the limitations of the model given by Equation 2.17: • The model is only applicable at current densities where diffusion limitations are negligible. • It is assumed that the exchange current density of the oxygen reduction reaction is much higher than that of the hydrogen oxidation reaction. • The symmetry factor, a, of the oxygen reduction reaction is assumed equal to Vi. This is not generally true, but is usually a good approximation.

This model will be applied in Chapter 5 to obtain electro-kinetic parameters by least squares fit of the experimental single cell data. 19

C hapter 3

Fuel C ell Efficiency

In this chapter the first and second law efficiencies are given. Their adequacy for determining efficiency for fuel cells are discussed. A second law efficiency analysis is demonstrated for the Solid Oxide Fuel Cell.

3.1 Introduction

Efficiencies of energy converting devices are often given without reference to which expressions are used, and this causes confusion. It is common to separate the expressions of efficiency into two categories, coming from the first and second law of thermodynamics, respectively (see e.g. Bejan[28]). We will limit ourselves to discuss the efficiency of energy converters for production of work.

First law efficiency The first law efficiency may then be expressed as

where WE is the available work from the energy converter and Q is the heat supplied to the converter. This efficiency, t>, is called thermal efficiency [29]. The upper limit of the thermal efficiency for a heat engine is given by the Carnot efficiency:

^ (3.2) where J2 and T1 are the temperatures of the heat source and the heat sink, respectively. The maximum electrical work from a fuel cell (operating at isothermal conditions) is given by the Gibbs energy change (AG) of the cell reaction:

WE.rev = -AG = nFErev (3.3) 20 Chapter 3. Fuel Cell Efficiency where WE rev is the work from the converter when the process is carried out reversibly, £«, is the reversible potential of the fuel cell, n is the number of electrons involved in the reaction and F is the Faraday constant. The heat supplied is given by the enthalpy change, AH, of the reaction. Thus, the maximum thermal efficiency for the fuel cell becomes: AG AH-TAS AS AH ~ AH ~ ~ AH (3.4) by application of AG=AH-TAS.

By comparing the two upper limits of thermal efficiency for the heat engine (Eq.3.2) and the fuel cell (Eq.3.4), an interesting feature is found. While the maximum output from the heat engine increases with temperature, the electrochemical system is favored by operating at low temperatures. In this comparison the increased reaction kinetics at higher temperatures are disregarded. In contrast to the heat engine, the electrochemical energy conversion is called «cold» combustion [5]. The term ‘cold ’ may seem odd, however, when considering solid oxide fuel cells operating at around

1000°C. Environmental concerns tell us that combustion should take place at low temperatures to avoid the formation of NOx . Fuel cells produce heat by reversible and irreversible processes, like most devices for energy production. This heat may be converted to electricity in a heat engine (limited by the Carnot efficiency). If heat is produced in the cell, this heat should preferably be rejected at high temperatures (Eq.3.2).

Second law efficiency The second law efficiency is related to the destruction of work:

(3.5) where WE rev is given by Equation 3.3, and WL is the lost work. This maximum work of Equation 3.3 is also termed the available work or the exergy, indicated by the subscript E. The second law efficiency is therefore often termed the exergy efficiency[28]. Other names are thermodynamic efficiency[30] and reversibility[31]. 3.1. Introduction 21

The term reversibility is the more descriptive one, but we will adopt the term exergy efficiency, because this is more widely used.

The upper limit of the exergy efficiency is found when the process is occurring reversibly. From Equation 3.5, it can be seen that this limit is unity for all processes. The exergy efficiency indicates how efficient a machine converts the energy of a fuel to work, taking into account its operating conditions and surroundings. The electrical work produced by a fuel cell is proportional to the cell potential, E:

WE = nFE (3.6) By applying Equations 3.3 and 3.6 in Equation 3.5, the voltage efficiency arises: s = - (3.7) which then is a special case of the second law efficiency adequate for an electrochemical energy producing device. The reduction of the cell potential followed by increasing cell current in the SPEC was discussed in Chapter 2. The voltage efficiency gives the overall second law efficiency. In order to address the contributions to efficiency losses, it is, however, necessary to examine the lost work (Wfc»). This is done in the enclosed article entitled ‘Exergy efficiency and local heat production in solid oxide fuel cells'. Through this exergy analysis the potentials for improvements are clearly indicated. For any reversible process the entropy change of the universe is zero. For all other processes the entropy increases. Thus the entropy production is an adequate measure of irreversibility. The lost work, WL, is also termed dissipated energy. Dissipated energy may be calculated from the entropy changes of the system by applying irreversible thermodynamics [32].

Discussion The first law efficiency is often used in the form of Equation 3.4. In this expression the work output is referred to the heat obtainable by irreversible combustion, AH. This is an adequate reference for the heat engine where the energy to the converter is supplied as heat. In some cases, however, the entropy change, AS, of the system is positive (AS>0), and then the maximum first law is larger than 1. This is the case e.g., for the C/02 fuel cell[5]. Electrical energy may be obtained by exploiting the gradient 22 Chapter 3. Fuel Cell Efficiency in salt concentration between a river (fresh water) and the ocean, by applying cation and anion exchange membranes. The majority of the energy obtained comes from the entropy change in such a system. In this latter case the use of first law efficiency is inadequate giving an upper limit of the thermal efficiency which by far exceeds unity. Comparison of Equations 3.4 and 3.5 reveals the fundamental difference between the first and second law efficiencies. Whereas the first law efficiency refers to irreversible combustion, (denominator given by AH), the second law efficiency is referred to the work obtainable at reversible conversion. This ensures that the maximum second law efficiency always equals unity, indicating the potential for improvements. The lost work is related to irreversibilities in the process. The fuel cell is by far a more reversible energy converter than most conventional systems. In the fuel cell the chemical energy is converted directly to electrical energy. Herein lies the key to the high fuel cell efficiency. It is recommended to use the second law efficiency for all energy converting devices, including the fuel cell, because it governs a more realistic picture of the potential for improvements. Dunbar et al.[33] have studied the combination of fuel cells with fuel-fired power plants. They conclude, as we do, that the use of first law analysis is misleading for such a system, giving erroneous indications of where to save energy. These conclusions are also in general agreement with those of van den Oosterkamp et al.[34].

Conclusion We have seen that first law efficiency varies for reversible energy converting devices, whereas the second law efficiency always equals unity for such ideal energy conversion. The first law efficiency does not indicate the potentials for improvements. The second law efficiency constitutes a general basis for efficiency evaluation and is adequate for comparison of different energy converting technologies. It is concluded that the use of the first law efficiency is not appropriate for fuel cells. Exergy efficiency, which is based on the second law analysis should be used. This makes localization and quantification of energy losses in the cell possible. 3.2. Exergy Efficiency And Local Heat (enclosed article) 23

Electrochimica Acta, Voi. 38. No. 2/3. pp. 447-453, 1993 0013-4686/93 $6.00 + 0.00 Printed in Great Britain. © 1992. Pergamon Press Ltd.

EXERGY EFFICIENCY AND LOCAL HEAT PRODUCTION IN SOLID OXIDE FUEL CELLS

Signe Kjelstrup Ratkje * and Steffen Moller -Holst Division of Physical Chemistry, Norwegian Institute of Technology, University of Trondheim, N-7034 Trondheim, Norway

(Received 16 January 1992; in revised form 27 April 1992)

Abstract —The electric work method has been applied to a unit cell of the solid oxide fuel celt A new equation for the cell power is derived, which takes into account temperature gradients of the system. Local heat productions and consumptions in the cell have been calculated using new data on the trans ­ ported entropy of oxygen ions. Exergy efficiency calculations are carried out for the unit cell at 1000°C indicating the relative importance of losses due to overpotentials, ohmic resistance and cracks in the electrolyte, incomplete reactions and temperature gradients. Energy economy is obtained for direct elec­ trochemical conversion of methane in the unit cell when the overpotential at the fuel electrode is less than 0.21 V for an electric current density j = 1A cm-2. Ohmic resistance of the electrolyte plays a minor role. A natural temperature gradient of 10K across the cell reduces the work from the cell by 0.6%. The heat production in the cell is asymmetrical. A 3% gain in exergy efficiency is obtained by changing the pres ­ sure from 1 to 4 bar. The results will have a bearing on cell design and material development.

Key words: SOFC, fuel cell, exergy efficiency, electric work method, heat production.

1. INTRODUCTION externally produced from methane and further con­ verted in the cell. A detailed discussion of the energy Fuel cells are expected to play an important role in gain by cogeneration must, however, be postponed. future production of electrical energy. The solid We shall limit ourselves to the calculation of a oxide fuel cell (SOFC) converts hydrogen and unit cell. Unit cell descriptions are required for com­ oxygen to water around 1000°C. Hydrogen may be puter modelling of the system[7]. Both reversible produced outside or inside the cell before conver ­ and irreversible heat effects are taken into account, sion. The success of the SOFC relative to other fuel but losses due to diffusion will be neglected. cells or other power generating systems may depend on exploit of the heat production of the cell[l]. The heat production may be used for the reforming reac­ tion which produces hydrogen[2] from natural gas, and for cogeneration of electricity[l], Takehara and 2. PRINCIPLES coworkers show that the heat production differs between electrodes in the SOFC unit cell[3]. They Assume that hydrogen for the fuel cell is supplied also discuss how the temperature distribution in the by reforming of methane according to the reaction; cell may vary in a tubular cell construction^]. Rosen[l] concludes that exergy analyses should iCH4(g) + iH2CKg) -v iH2(g) + iC02(g). (I) be used to evaluate efficiencies of power generating Hydrogen is consumed inside the cell according to: systems. We shall give the results of an exergy analysis of the SOFC, using a new method, the elec­ iH2(g) + i02-(an) -> |H2Q(g) + e" (II) tric work method[5, 6]. Our conclusion agrees on a jO2 "(cat) -* yO2 "(an) (III) general basis with that of Rosen[l], but more details of the process are given. The characteristics of the i02(g) + e" - |02"(cat) (IV) method are detailed descriptions of causes and loca­ tion of local heat and energy changes per unit time. iH2(g) + i02(gWiH20(g) (V) The analysis extends the work of Takehara and coworkers[3, 4] on this point, because the electric where (an) and (cat) refer to the anode and cathode, work method is more accurate than previously respectively. All reactions are written for the transfer published methods. of one faraday of elementary charges and the trans ­ The analysis reveals the main factors for further ference number of O2- in the electrolyte is unity[8]. improvement of the cell efficiency. The effect of the The overall reaction of the system is the sum of reac­ reformer reaction on the exergy efficiency will be tions (I) and (V); quantified. Electrochemical conversion of methane will be compared with the process where hydrogen is gCH4(g) + i02(g) - iH2G(g) + iC02(g). (VI) We shall use the conversion ratios , ijHl, and tj0l * Author to whom correspondence should be addressed. for reactions (I), (II) and (IV), respectively.

EA 38-2/J-T 447 24 Chapter 3. Fuel Cell Efficiency

448 S. K. Ratkje and S. Moller -Holst

2.1. The emf of a unit cell with temperature gradients tracting the term iG0l(TJ we may rewrite Equation The electric work per unit time or cell power, (5) for a constant enthalpy by dW/dt, obtainable from an electrochemical cell is £,/ = -[AGfTJ given by[5], - jViSo, + + S» dT](7/F) (6) E {(dC/dt), - (dQ/dt), + dfrF/dt),} = -dtf/dt (1)

where (dt//dt), are local energy changes per unit time where AGJTJ is the Gibbs energy change of reaction (unit Is-1), (dg/dr), are local heat consumptions, (V) at the anode temperature. For Tc — T, = 0, the and d(pV/dt)i are local volume works performed by value £, from Equation (6) reduces to the emf the system per unit time. All changes are coherent according to Nernst equation for isothermal cells. In with the electric current. Summation is carried out analogy to Equation (5), we have for reaction (VI) over each volume element, where mass, temperature and volume changes occur, and total changes are EV,I = -[iGH2oTO + iGco^T^) obtained. - fGcH.tTJ — {G02(TJ For reactions (II—IV) the total energy change per unit time is: ~ [r‘(iSS=- + S3 dT](//F) Jr, S (dt//dt), = [it/H!o(TJ - it/„2(TJ = -[AG^TJ - iU02(TJi(I/n (2) - fViSo, + + S3 dT](//F). (7) where t/;(7) (units J mol-') denote the partial molar energy of component i at anode or cathode tem­ Equations (5)—(7) give the cell power for the conver ­ perature, 7% or Tc, respectively. sion of H2 and CH4, respectively, for reversible con ­ The reversible heat consumption per unit time at ditions. The contribution from the temperature the anode is: gradient has previously been neglected in expressions (dg/dt),„ = r,[iSH20(TJ - iSHj(TJ for emf for fuel cells [see eg 3, 4, 13]. According to Equation (7) the temperature gradients affect the cell - i%. - SfJd/F) (3) power in two ways; by the value of the partial Gibbs energies and by the integral of transported entropies. where SJTj is the partial molar entropy of com­ ponent i at T„, Sgz- is the transported entropy of oxygen ions, and S’, is the transported entropy of 2.2. Exergy efficiency electrons in the electrode. The transported entropies The exergy efficiency, {, of a process can be can be taken as constants over an interval of expressed by; 100°C[9]. The sign of the terms are explained as follows: Since i mole of HzO is produced at the C - («%_ - (8) anode (see reaction II), the heat iT,SHj0(TJ is con­ is the work lost in the process. This will be dis­ sumed. Oxygen ions are transported to the anode, cussed below. is the maximum available work producing the heat i^Sg,- at this location. or the available energy (unit J) ([see eg Bejan[10]). For the cathode, the heat consumption is: Theoretically, the most efficient way to produce elec­ (dg/dtL, = Tc[—*s 02(Tc) + isg2- + sy(//F). (4) tric energy from hydrocarbons is by electrochemical conversion, when the reaction is carried out If the transported entropies vary with temperature, reversibly. The maximum available work from the there is also a Thomson effect from junction and system is then that of reaction (VI). The reaction leads. This is: occurs without any changes in mole numbers, so (dg/dt)71om= -|J‘r{[idSS=- +dS53/dT}dT(//F). that volume work against the surroundings does not contribute to in this case.

The sum of Equations (3) and (4) give the 2.3. The maximum available work reversible heat changes, Ei(dg/dt),, in Equation (1). Integration of Equation (7) gives the maximum The last term on the left side of Equation (1), available work, , of the total system EidfpF/dt),, is equal to RTIdnfdt) for ideal gases. By combining this information, Equations (2-4), and -[AGv/TJ- the fundamental relation G, = t/, — TS, + pVt, we obtain the electric power, dWjdt = £VI, of cell (V) j"ViS02 + #&_ + S’) dT](I'F)t. (9) for small values of /:

EJ = -tiGH20(TJ - |GH2(TJ - 1G02(T) The integral in Equation (9) is carried out between the electrode temperatures, and (i/F)t is the number - £T=(§Sg,- + S3 d T](I/F). (5) of equivalents of charge transferred. The electric current, I, corresponds to the reaction rate (//£) = —2 dn Hz/dt. The integration of this term over time £v is the cell emf, and G, are partial molar Gibbs equals twice the number of moles H2 reacted in the energies (chemical potentials). By adding and sub ­ cell, or, at complete conversion, eight times the 3.2. Exbrgy Efficiency And Local Heat ..... (enclosed article) 25

Exergy efficiency and heat production 449 number of moles methane reacted: at the electrodes and in the electrolyte. They are heat sources. Sinks may be caused by reversible heat (I/F)t = 2(n iB - O h , = 2 An H, = 8 An c„4. (10) effects at the electrodes, the Peltier effects given by With (10) Equation (9) can be replaced by Equations (3) and (4). In addition to the Peltier effects, there are other heat changes in the half cells ^max = E ({”G}i.in- {nG }i.out)vi due to the changes in partial pressures of the gases. It is reasonable that the pressure changes occur in a + 8 + $5) dT. (11) reversible way. When the temperature is constant J\ss»- and the gases are assumed to be ideal, the heat effect is given by the entropy changes of the gases. By The summation is carried out over all reacting com­ including this effect into Equation (3) we obtain : ponents given by reaction (VI) at their temperatures and partial pressures. Some inert components are 2a* = —T, E ({"Skin - {"S};.„J,|, passing the system (ie N2 gas and excess water). These do not contribute to the maximum available - 2 An Hz + $:,). (16) work. The maximum available work as well as the avail­ The summation includes all components passing the able work are thus proportional to the amount of anode compartments. For the cathode we have simi­ moles reacted. This proportionality is an advantage larly when the exergy efficiency for fuel cells is compared 2c,i = — Tc E ({"%.!„ — {"S}i. oujall to efficiencies of other power generating machines. + 2An„ 2rc(iSS2- + SS). (17) 2.4. Lost work in SOFC with external reforming, In the cell there are also heat changes which are examples not coupled to the electric current. The heat needed The losses we consider are due to overpotentials, to warm the reactants from the reference tem­ ohmic resistance and cracks in the electrolyte and perature T° is given by: incomplete reactions. Losses due to gradients in con ­ centration and temperature are given by the dissi­ 2™.cun B = E''iCHi(T)-Hi(r)] (18) pated energy per unit time, T(dS/dt)[7,11). Reaction (I) represents a complete loss of work where the summation is carried out over the feed because this is an irreversible combustion. This is so gases CH4, H20, 02 and N2, and T is the tem­ whether reaction (I) takes place inside the fuel cell or perature of either the anode or the cathode. Reform­ externally. The loss is equal to ing of methane to H2 and C02 is an endothermic process. This irreversible combustion has a heat con ­ (12) sumption equal to the enthalpy change of reaction I, AH,. Subscript I refers to reaction I. Losses due to incom ­ plete reactions can be expressed in the same manner. 2reformer = AH,. (19) Cracks in the electrolyte may occur. Ohno and Kaga Equations (13)~(19) will be used to calculate heat assumed a leakage of 1%[12). This means that the sinks and sources in the unit cell. cell is short-circuited, some fuel reacts irreversibly with oxygen and this produces hot spots. We shall also assume that 1% of the hydrogen reacts in this way. This gives 3. CALCULATIONS 2 = 0.011 ({-«G}, - {nG}, • (13) 3.1. Procedure Irreversible heat productions due to cathode and Calculations were performed using spreadsheets anode overpotentials, and give a loss in (Quattro Pro, version 1.0 from Borland exergy: International) including a database holding the ther ­ modynamic data of the components. The following wu 3 = e, =2 a «„2 mr+m operating conditions of the cell were changed: total The value of Equation (14) depends on the current pressure, anode and cathode temperature, conver ­ density. The Joule heat evolution will be sion ratios and values for transported entropy. Although it is easy to obtain several functional «i-s.4 = Qn = 2An HlFJR (15) relationships from the spreadsheet, we have chosen to emphasize main points in tables. The values thus where R is the electrolyte resistance and IR the depend on the reference choices we have made. ohmic voltage drop across the electrolyte. We will Trends are, however, clearly seen in the tables. neglect ohmic losses in the electrodes. The lost work is the sum of Equations (12)—(15) plus the losses by incomplete reactions. 3.2. Reaction conversion ratios The maximum available work was calculated 2.5. Heat sinks and sources using complete conversion (%, = 1, i = CH4, H2, and Equations (13)—(15) give positive heat evolutions 02). In order to avoid carbon as an unwanted 26 Chapter 3. Fuel Cell Efficiency

450 S. K. Ratkje and S. Moller -Holst

byproduct in the reformer [reaction (I)], excess water Table 1, Standard thermodynamic data for the com­ is used in the reaction mixture[19]. The water/ ponents at T = 1273 K[18] carbon mole ratio in the feed is therefore 2.2. AfH? S? fl°(7>H°(T°) Oxygen is supplied from air (20mol% 02 and Component kJ mol “1 JK-'mo]"' kJmoI-1 80mol% N2). The feed to the system then consists of 1 mole of CH4, 2.2 moles of HzO, 2 moles of 02 and ch 4 -91.8 266.0 59.13 8 moles of N2. Nitrogen gas and some water vapor h 2o -249.3 243.4 37.74 pass the system as inert gasses. o 2 0 251.9 32.38 Realistic maximum values for % are not unity. For N2 0 2362 30.59 0 173.7 29.08 the reformer reaction, we may assume that ijq ,, - h 2 co 2 — 395.2 282.8 48.62 0.9[19]. Kanamura et al[4] used ij„2 = 0.9 and q0l = 0.9 for a tubular SOFC. We shall use the same = 25°C, F° = 1 bar. conversion rates in our calculation for a unit cell. Problems with the oxygen electrode are common; we where Af fff is the standard heat of formation and S, have therefore examined the effect of reducing »/0j is the entropy. The enthalpies are taken as constants below 0.9, but also the effect of changing i/H3 • in the given temperature interval. Conversion rates of H2 and Oz are inter-related. A constant pressure of 1 bar as well as a constant We have assumed that O2- is always available at temperature T = 1273 K was used for both cells and the anode for H2 to react to water. In the stationary state, the amount of 02 consumed at the cathode is reformer in the first calculation. As 02 is consumed at the cathode, work must be performed on the gas then given by the conversion of H2. The conversion mixture to maintain a constant pressure. Common ratios of CH4 and H2 give the electric current in the operating pressures are higher (4-5bar)[l]. The cell. A conversion ratio of oxygen less than one, only pressure was therefore also increased to 4 bar. Tem­ means that more 02 has to be heated. perature differences as high as 120 K have been pre­ dicted in SOFC[3] for an electrolyte thickness of 3.3. Materials and transport data 8 mm. In our calculation the temperatures of the The electrolyte which conducts 02~ usually con ­ electrodes were allowed to differ by 10 K, keeping sists of Zr02 in a mixture with 8-15 mol% Y203 the mean cell temperature equal to 1273 K. This (YSZ)[13]. We have chosen 8moi% Y203 because value for AT is chosen to illustrate the effect of a the conductivity is optimal[8], and because other temperature gradient. In a real cell, where electrolyte necessary data are available for this composition. thickness is 50-100 pm, the gradient is probably The anode is frequently a Ni-YSZ cermet[14], Dif­ smaller. The reformer was operating at the anode ferent perovskites have been used as cathode temperature. material[14, 15]. Relevant data are given for La0 6Sr0 4MnO 3[13]. The electric current density is 4. RESULTS large, typically 1.0Acm™2[7, 13], during operation of the cell. The anodic and cathodic overpotentials The maximum work obtainable (for the conditions (for the present choice of materials and current given in section 3) from the SOFC is 772 kJ. This number is for complete conversion of reactants and density, J) are estimated from Bamett[16] and Setoguchi[14], For j = 1.0 A cm-2 and temperatures serves as a 100% reference for most of the calculated around 1000°C, we have tj* n = 37 mV, if* = 50 mV exergy efficiencies. Efficiencies are given in Tables and [t)* ” + t/cal] % 0.09 V (Equation 14). The ohmic 2-4 for different reaction conversion ratios, pres­ voltage drop across the electrolyte is IR = 0.05 V, for sures, and isothermal and non-isothermal cells, the same conditions (Equation 15). The transported entropy, S*, of the anode cermet has been taken 600-r equal to that of the cathode perovskite type material. The transported entropy was calculated from the Seebeck coefficient of the material to —UK"' faraday-1. Recent data from our laboratory [26] 400 -■- indicate a value for the transported entropy of oxygen ions of 42 + 1J K "1 faraday "'. 200:

3.4. Thermodynamic data, operating conditions Standard thermodynamic data from JANAF[18] which are used in the calculation, are reproduced in Table 1. The partial molar entropy of component i is calculated from -200- Sj = S? - R \n(p jp°). (20) The reference pressure p° = 1 bar and S? is the partial molar entropy at this pressure. The partial Fig. 1. The reversible heat productions (solid lines) and molar Gibbs energy is calculated from total heat productions (stippled lines) at the electrodes in the SOFC as a function of the value of the transported G,-(T) = AtH? — TSj(T, Pj) (21) entropy of oxygen ions. 3.2. Exergy Efficiency And Local Heat (enclosed article) 27

Exergy efficiency and heat production 451

Table 2. Available work in a SOFC unit cell as a function of conversion ratios

Vau = 1-0 4ch . — 0.9 7ch , — 0.9 7ch * — 0.9 'fcM. = 0.9 7h , = 10 7h , = 0.9 ijHl = 0.9 — 0.9 4b, = 08 Conversion ratios: 7o, — 1*0 to, = 0.9 So, = 0.8 Vo* ~ 0.5 To, = 0.9

Available energy (=100%) 772 772 772 772 772

Irreversible losses: in reformer reaction -106 -105 -105 -105 -105 by incomplete reactions 0 -118 -115 -112 -172 by cracks -7 -5 -6 -6 -5 due to overpotential -69 -56 -56 -56 -50 due to joule heat -39 -31 -31 -31 —28 Work to be obtained 552 457 459 463 413 Exergy efficiency 0.71 0.59 0.59 0.60 0.54

One mole of methane is converted. The temperature is 1273 K, and the total pressure in both electrode chambers is 1 bar. The current density is 1 Acm~2. All numbers apart from efficiencies have dimen ­ sions kJ. respectively. Heat effects for some of the conditions however, not included in our analysis, and these may are shown in Table 5. The reversible heat effects at be significant]]?]. the electrodes as well as the total heat effects are Our reference values for available energy, 772 kJ, illustrated in Fig. 1. makes the exergy efficiency £ = 0.71 during power production (1A cm-2). The more realistic value, for 5. DISCUSSION Table 4. Available work in a nonisothermal SOFC unit 5.1. Exergy efficiencies cell The results of Tables 2—4 show that exergy Energy change analyses are useful in the analysis of relative impor­ tances of various cell processes, as has been con­ kJ % cluded before[l, 7,20]. Table 2 gives the losses in SOFC for p = 1 bar and 766 99.7 T = 1273 K. The major losses are due to the external +Q)dr 2 0.3 reformer and to incomplete reactions. Relative gain Available energy 768 100 in exergy efficiency will be larger by a reduction in Irreversible effects/losses: the loss from incomplete reactions than from over ­ in reformer reaction -104 potentials. This means that research should be con ­ by incomplete reactions -117 centrated on direct conversion of methane and cell by cracks -5 stack design. Efficiencies may also be increased if an due to overpotential -56 anode with a lower overpotential is found. Com­ due to joule heat -31 pared to the possible gains by these efforts, an effort Work to be obtained 454 59 of reducing the ohmic resistance seems less prom­ ising. This statement relates to the unit cell resist­ One mole of methane is converted. The conversion ratios are ^ == t?H2 = tj0i = 0.9. The temperatures are Tc = ance. It is at variance with the opinion of Setuguchi 1278 K, Ta = 1268 K, eg AT = 10K. The total pressure in et ol.[12]. Resistances of connecting materials are, both electrode chambers is 1 bar.

Table 3. Available work in a SOFC unit cell at 4 bar Table 5. Heat production in a SOFC unit cell

IfCH. = 10 7ch , — 0.9 Irreversible effects/losses 4hz = 10 4h , = 0.9 by cracks 5 Conversion ratios: 4o, = 10 4o, = 0.9 due to overpotential 56 due to joule heat 31 Available energy (= 100%) 772 772 Reversible heat productions cathode 394 Irreversible losses: anode -121 in reformer reaction -76 -78 by incomplete reactions 0 -120 Net heat production in unit cell 365 by cracks -7 —6 Gas heating needs due to overpotential -69 -56 Cathode, air supply -278 due to joule heat -39 -31 Anode -186 Work to be obtained 581 480 Heat of reaction, reformer -176 Exergy efficiency 0.75 0.62 Surplus heat -275

One mole of methane is converted. The temperature is One mole of methane is converted. The temperatures are 1273 K. The current density is 1 A cm-2. The pressure effect Te = Ta = 1273 K. The total pressure in both electrode on overpotentials has not been taken into account. All chambers is 1 bar. The conversion ratios are y;CH4 = tjHl = numbers apart from efficiencies have dimensions kJ. rj0i = 0.9. All numbers have dimensions kJ. 28 Chapter 3. Fuel Cell Efficiency

452 S. K. Ratkje and S. Moller -Holst f?i = 0.9, gives an exergy efficiency of £ — 0.59. This temperature difference may be larger than 10 K. In number corresponds to previous estimates obtained Fig. 1, we show that the heat effects depend largely from exergy analysis[l] for slightly different condi ­ on the value of Sg2-, and recent results for this tions (ie T = 1373 K). The data we have used in the value[26] give larger heat asymmetries than those calculation of £ = 0.59 have all been obtained in the calculated from [3,4] (further comments are given in laboratory, and should therefore indicate what is section 5.2). obtainable in the industry in the future. The assumption of negligible losses due to diffu­ It is seen from Table 2 that £ is almost insensitive sion is supported in the literature[7, 20]. The to variations in >]0l. This is due to our definition of problem of coupled mass transport in a porous elec­ conversion ratios (section 3.2). When t}02 decreases, trode is, however, complex, and may deserve further oxygen is consumed at a higher partial pressure attention. This is indicated by a recent article on giving a small increase in £ by reducing loss due to phosphoric acid fuel cells[21]. incomplete reactions. The hydrogen conversion, which is directly linked to the cell current, has a large impact as a reduction in ij„2 from 0.9 to 0.8 5.2. Heat effects reduces £ from 0.59 to 0.54. The main conclusion drawn from Table 4 is that Table 2 quantifies the losses caused by external there is a large asymmetry in the heat production in reforming of methane (105 kj for % = 0.9). This can the cell. Takehara et a£[3] indicated a smaller asym­ be used to develop criteria for energy economy of metry. The discrepancy between our results and electrochemical conversion of methane. Research on theirs is due to the value for the transported entropy this process is still in its early stages[20]. A suc­ of oxygen ions. So far values between 42-72 J K ™1 cessful direct conversion means that the losses in the faraday ~1 have been reported for different reformer are eliminated. Thus, the criterion is experiments[3, 4, 9, 26]. Thus a variation in Sg2- obtained that losses by overpotentials in the may be expected. Increases in the value of S$2- (see methane converting cell must be smaller than the Fig. 1), may even change the anode reaction from sum of losses in reformer and overpotentials in the endothermic to exothermic. The heat production at hydrogen fuel cell. For conversion ratios of r;i = 0.9, the cathode, however, remains positive in the given we can calculate from Table 2 that the sum of over ­ range. The overpotentials which give positive heat potentials in the cell for direct methane conversion effects do not alter these conclusions, as shown by must be < 0.26 V. If the overpotential at the oxygen the stippled lines in Fig. 1. electrode is 0.05 V (section 3.3), this means that the Figure 1 shows that a high value of the trans ­ overpotential of the anode must be <0.21 V at a ported entropy reduces the asymmetry in the heat current density of / = 1A cm-2. production. We may expect that thermal stresses in The results of Table 3 show that an elevated pres ­ the materials will be reduced when the heat evolu ­ sure is favorable for the exergy efficiency. The main tion is more evenly distributed. Efforts should there ­ reason for this is that the loss in the reformer is fore be placed on finding materials with a high value reduced. A second favorable effect may be expected. of the transported entropy. According to Inoue et a/.[21], the cathode over ­ Direct conversion of methane will cause an extra potential is reduced by increasing the pressure. The reversible heat consumption (T AS,= —71 kJ for data reported by these authors were, however, too ifi = 0.9) at the anode, giving rise to an even larger sparse to be included in the present calculation. The asymmetrical heat production (Table 5). pressure effect on overpotentials should be further The potential of cogeneration of electricity is often investigated. discussed in connection with SOFC. Knowledge The effect of a temperature difference AT = Tc about the location of heat production is essential in — 7£ = 10 K on the exergy efficiency is reported in this respect. The total heat production in the cell is Table 4. The AT is positive, as the heat production is 365 kJ for the conversion of 1 mole of methane. It is higher at the cathode than at the anode (Fig. 1). A seen from Table 5 that this is not enough to cover minor effect of the integral part in Equation (9) is the heat demand of the system, eg heat needed to seen (0.3%). The Gibbs energy change [first term on warm up the reactants and supply the heat for the the right hand side of Equation (9)] make up the reformer reaction. Even when all nitrogen is remaining 99.7%. The exergy efficiency is 0.59 in removed from the air, there is a beat deficiency. The both cases, but the work to be obtained is reduced cathode gas heating which is needed in this case by 0.6% from 457 kJ (Table 2) to 454 kJ (Table 4) in changes from —278 to — 58 kJ, making the surplus the presence of a temperature gradient, keeping the heat as defined in Table 5, vary from —275 to average cell temperature constant. We therefore con ­ —55 kJ. For a SOFC using H2 as fuel, we can disre­ clude that Equation (9) should be used when accu­ gard the heat demand of the reformer reaction. With rate results are needed. We have neglected the air as the cathode gas this cell has a surplus heat of temperature dependence of the overpotentials in -99 kj. order to emphasize the contributions from the tem­ The available energy of the exhaust gases must be perature gradient itself (through transported encountered to evaluate the possible gain of entropies). cogeneration. This heat is of the same magnitude as The difference Tc — 7£ = 10 K may seem arbitrary. the heat needed to warm the gases (because there are Here, the difference was fixed to study possible no change of mole numbers in the reaction VI, and effects of AT. In a more complete study, AT must be thus a negligible entropy change). The heat content a solution of the set of conservation equations. of the exhaust gas is around 460 kJ from Table 5. Future studies should produce this solution. The Depending on how efficient the exhaust heat can be 3.2. Exergy Efficiency And Local Heat ...... (enclosed article) 29

Exergy efficiency and heat production 453 utilized, there will be a smaller or larger heat excess The method is suited for exergy efficiency analysis. in the fuel cell system. Rosen[l] calculated an Because new information and recommendations for increase in exergy efficiency from 0.60 to 0.76, using further work can be obtained, the method should a thermal conversion ratio of 0.5, arguing that have a potential for application to other systems. thermal efficiencies tend to present overly optimistic views of performance. We have not calculated the Acknowledgements—S1NTEF and Statoil are thanked for exergy efficiency gain by the heat content of the financial support during the initial stages of this work. exhaust gas, because extensive specifications of cell design is needed to do this. We agree, however, with Rosen on a general basis. REFERENCES

5.3. Methods of calculation 1. M. A. Rosen, Int. J. HydrogenEnergy 15,267 (1990). 2 A. L. Lee, R, F. Zabransky and W. J. Huber, Ind. eng. The electric work method used in the present cal­ Chem. Res. 29,766 (1990). culation, has been applied earlier to simple thermo ­ 3. Z. Takehara, K. Kanamura and S. J. Yoshioka, J. elec­ cells, to electrolysis of aluminium and to electrolysis trochem. Soc. 136, 2506 (1989). of heavy water[22-24]. The method focuses on 4. K. Kanamura, S. Yoshioka and Z. Takehara, Proc. energy conversion directly and is thus well suited to First lnt. Symp. SOFC (Edited by S. Singhai), 89-11, describe parameters of primary engineering interest, 293. Proc. of The Electrochemical Society (1989). ie changes in mass, energy and heat in each cell com­ 5. G. H. Hertz and S. K. Ratkje, J. eieclrochem. Soc. 136, partment. The electric work method is an alternative 1698 (1988). 6. G. H. Hertz and S. K. Ratkje, Acta Chem. Scand. 44, to traditional electrochemistry which focuses on elec­ 542 (1990); 44,554 (1990). trostatic potential changes. While the electric poten ­ 7. W. R. Dunbar and R. A. Gaggioli, Comptaer Simulation tial does not depend on the extent of reaction, the of Solid Electrolyte Fuel Cells (Proc. Intersociety Energy Gibbs energy does. This shift of focus is necessary to Conversion Engineering Corf.) ASME, 2,257 (1988). compare different power production systems. 8. D. W. Strickler and W. G. Carlson, J. Am. Ceram. Soc. Another advantage of the new method is that only 47,122(1964). measurable quantities are used in the theory. On a 9. S. K. Ratkje and K. S. Forland, J. eieclrochem. Soc. general level we may then argue that the possibilities 138,2374(1991). are increased for experimental control of approx­ 10. A. Bejan, Advanced Engineering Thermodynamics, p. imations in use. Also, it is easy to include further 120. Wiley, New York (1988). 11. K. S. Forland. T. Ferland and S. K. Katkje, Irreversible complexities. Losses by gradients in concentration Thermodynamics. Theory and Applications. Wiley, may be included by adding the dissipated energy to Chichester (1988). ML, 12. Y. Ohno and Y. Kaga, Gas Permeabilities of a solid Several new predictions or recommendations have Oxide Fuel Cell Component, Proc. 1st Int. Fuel Cell been made from the present calculation, ie the rec­ Workshop (Edited by M. Watanabe, K. Ota and P. ommendation to seek for electrolytes having a large Stonehart), p. 105. Tokyo Institute of Technology, value of Sg2_, the focus on cell design as an impor­ Tokyo (1989). tant parameter. The restriction of constant tem­ 13. N. J. Maskalick and D. K. McLain, J. electrochem. Soc. 135,6(1988). perature, imposed on the system through the Nemst 14. T. Setoguchi, T. Inoue. H. Takebe, K. Eguchi, K. equation, is lifted. Morinaga and H. Arai, Solid State Ionics 37, 217 The most common expression used for fuel cell (1990). efficiency in the literature is AG/AH. This expression 15. Y. Takeda, R. Kanno, M. Noda and O. Yamamoto, is a thermal efficiency which was first derived to Bull. Inst. Chem. Res. Kyoto 64, (1986). no. 4. compare combustion engines[10]. The fuel cell has 16. S. A. Barnett, Energy 15,1 (1990). reversible combustion in contrast to the combustion 17. J. Mizusaki, J. Tabuchi, T. Matsuura, S. Yamaucbi and engine. The efficiency used in the present investiga ­ K. Fueki, J. electrochem. Soc. 136,2082 (1989). tion is therefore a measure of deviation from 18. JANAF Thermochemical Tables, 3rd ed. National Bureau of Standards, New York (1986). reversibility[10]. In agreement with others[l, 7, 23] 19. Catalyst Handbook, 2nd ed. (Edited by M. V. Twigg). we prefer the expression used here, because it focuses Wolfe Publishing, London (1989). on possible improvements in cell performance 20. R. A. Gaggioli, Proc. IEA Workshop on Mathematical directly. 6 Modelling of Natural Gas SOFC and Systems, Charmey, Switzerland, 1989. 21. T. Inoue, K. Eguchi, T. Setoguchi and H. Arai, Solid 6. CONCLUSIONS State Ionics 40/41,407 (1990). 22. M. C. Kimble and R. White, J. electrochem. Soc. 138, The solid oxide fuel cell has been analysed from a 3370(1991). 23. S. K. Ratkje, T. Ikeshoji and K. Syverud, J. electro­ new perspective by the electric work method. The chem. Soc. 137,2088 (1990). Nemst equation has been replaced by a more 24. S. K. Ratkje, Electrochim. Acta 36,661 (1991). general equation which incorporates the effects of 25. S. K. Ratkje and B. Hafskjold. J. electroanal. Chem. temperature gradients. We also describe in more 273,269(1989). detail than before, the heat production in the cell. 26. Y. Tomii and S. K. Ratkje, manuscript in preparation. 30 Chapter 3. Fuel Cell Efficiency 31

C hapter 4

Equipment For C ell Preparation And Testing

This chapter describes equipment for preparation of single solid polymer fuel cells and a test station for cell performance testing. Hardware and software for data acquisition is documented as well as the test station operation limitations. Suggestions for improvements are indicated.

4.1 Introduction

Equipment for fuel cell testing has usually been designed and constructed in research laboratories (see e.g.[35]). Lately, some manufacturers have provided complete test- stations as well as single and multi-cell test fixtures (e.g. GlobeTech, Inc., Bryan, TX, ElectroChem, Inc, Woburn, MA). The electrode manufacturer E-TEK Inc. (Natick, MA) also offers electrode tests for customers.

In this work the preparation and testing equipment were constructed and built in cooperation with the university workshop. The work of building up a SPEC test station started in the autumn of 1991. This represented the initiation of the study of Solid Polymer Fuel Cells at our laboratory. The equipment has, during these years, been in a constant state of development. Thus, all experiments have not been performed with exactly the same parts. However, when doing a series of experiments, the same equipment has been used and efforts have been made to maintain equal operating conditions. Change in experimental procedure or equipment influenced the results of different series. This was reflected in the somewhat step-wise progress in single cell performance.

When developing new fuel cell electrodes and electrolytes, the materials are tested by different electrochemical techniques. The active surface area and, thus, the degree of platinum catalyst utilization have been studied by cyclic voltammetry[36,37]. Electrolyte and electrode conductivities have been measured by impedance 32 Chapter 4. Equipment For Electrode Preparation And Testing spectroscopy[38,39], and membrane water transport by streaming potential measurements [19]. The materials are usually studied in contact with acidic aqueous solutions[40,41,42,43]. Difficulties are, however, encountered when trying to obtain good contact between polymer electrolytes and electrodes. Whereas catalyst utilization was almost 100 % in aqueous phosphoric acid [7], the corresponding value using polymer electrolytes was considerably lower (10-20 %) [44]. Catalyst utilization is further discussed in Chapter 5. During SPEC operation, water is usually supplied as vapor. Membrane properties differ significantly whether the water is supplied as liquid or as vapor[22,45]. Furthermore, properties of the materials may change during membrane and electrolyte (M&E) assembling (i.e. hot pressing)[46,47]. The assembling procedure must therefore reflect the chemical stability of the materials in use. This stresses the need of thorough and detailed testing of the cells after the M&E assembling and under realistic fuel cell operating conditions. This argument is supported by the increasing use of in situ characterization techniques[22,39].

The purpose of this chapter is to describe a test station equipped for studies of single SPFCs under realistic operating conditions and equipment for single cell preparation by hot pressing and spray depositioning. Experiments on single cell operation stability will be reported.

4.2 Test-station description

A test station for studying single SPEC at low temperatures (<150 °C) was designed. The test station has provisions for temperature, gas pressure and gas flow control. The humidity of the reactant gases was controlled by changing the temperature of the gas humidifiers. A single cell fixture was designed for rapid dismantling and reassembling. A personal computer was connected to the test station for data acquisition purposes. Computer routines were made for easy data analysis by use of a worksheet. Several parts of the test-station and the cell preparation equipment have been made in the university workshop. The different parts of the system are described below. 4.2. Test-Station Description 33

<3>

Figure 4.1. Schematical presentation of the gas flow system. ® Pressure vessels, ® Flow controllers and flowmeters, ® Manometers, ® Gas humidifiers, ©Magnetic stirrers, © Single cell fixture, © Water separators, ® Pressure controllers.

4.2.1 Gas supply and humidification system The gas flow system is shown schematically in Figure 4.1. Gases were supplied from pressure vessels of maximum 200 bar (20 MPa) equipped with pressure reduction valves. The flow of gases were controlled by Brooks Digital Mass Flow Controllers (Model 5850S) and monitored using flowmeters (Brooks, Sho-Rate 1355). Gas pressures were measured by means of manometers (Bourdon) before the gases entered the gas humidifiers. Dry gases were passed through copper tubes to the humidifiers, and humidified gases through stainless steel (316 SS) tubes, both of 1/4" outer diameter. In the gas humidifiers the reactant gases were simultaneously heated and saturated with water. The gas humidifiers were made from stainless steel (316 SS), each heated by means of an electric heating tape (Fibrox™, Thermolyne Corp.) and cooled by a water coil. The humidity of anode and cathode gases were varied individually by changing the temperatures of the humidifiers. The gases left the humidifiers saturated with water vapor and were passed through isolated tubes to the fuel cell. The humidifiers were usually operated at a higher temperature than the fuel cell, as recommended in literature [48], hence, the gases leaving the cell were oversaturated. Water leaving the fuel cell was removed from the gases by water separators (Martonair, FI 1-300- A3DD), before the gases were released to the atmosphere through Brooks Compact Pressure Controllers (Model 5866). 34 Chapter 4. Equipment For Electrode Preparation And Testing

Figure 4.2. Photograph of one half of the single cell fixture. ® Heating element, @ Gas inlet, ® Gas outlet, ® PES cell fixture, ® Gasket, © Membrane and electrode assembly.

4.2.2 Single cell fixture The single fuel cell fixture consists of two identical half cells, one of which is shown in Figure 4.2. The outer cylinder of the cell fixture was made from Polyethersulfon (Victrex™). Polyethersulfon (PES) is an amorphous plastic material with adequate properties like high electrical resistance, transparency, low water absorbance, relatively high tensile strength, thermal stability to 200 °C and high chemical resistivity, and it offers easy workability. Inside each PES cylinder a piston could move in the axial direction, to fit fuel cells with electrodes of different thicknesses. Two guided screws prevented the pistons from rotation. The pistons were made from stainless steel (316 SS) and also served as current collectors. The cell temperature was kept constant by means of two heating elements controlled by an Eurotherm PID- controller and a K-type thermo-couple. Cell design was not optimized with respect to gas flow and, thus, the fuel cells were tested at low gas utilization (customarily below 5 %). The membrane and electrode assembly is shown in detail in Chapter 5, Figure 5.3. The geometric area of the working electrode was 3.9 cm2. To ensure that the membrane and electrode assembly experienced the same compression pressure during every cell test, a pneumatic cylinder (Mannesmann Rexroth) was attached to one half of the single cell fixture as shown in Appendix C, Figure C4. The pneumatic cylinder 4.2. Test-Station Description 35 could provide a mechanical pressure of up to 25 bars over the membrane and electrode assembly.

Two generations of pistons were used (Figure 4.3). The 1st generation pistons had circularly shaped gas channels of 2 mm width and depth. Tests with the 1st generation piston showed that 3-4 times higher current densities (at the same cell potential) could be taken from the reference electrodes located in the middle of the piston, than from the working electrodes. The whole area of the reference electrode was supported on the stainless steel current collector, whereas for the working electrodes, only 25-30 % of the total area was mechanically supported. This low effective geometric area led to the design of the 2nd generation piston. The gas channels were made linear and the width and the depth was reduces to 1 and 0.5 mm, respectively. The reduced total channel volume on top of the 2nd generation piston, brought about a large increase in gas velocity at a given flow rate (ml/min). This was advantageous because it reduced the tendency of water accumulation in the channels, blocking for gas transport to the electrode.

Figure 4.3. The two types of pistons (026mm) used in the single cell fixture. ® First generation, ® Second generation with heating element attached. 36 Chapter 4. Equipment For Electrode Preparation And Testing

4.2.3 Instrumentation and Data acquisition Test station operation data (e.g., temperature, pressure and cell potential) were recorded by a PC (Commodore 386SX) using software from National Instruments (NI). The PC was also used to control parameters like the pressure, the gas flow etc.

Temperatures and voltages were measured by a Hewlett Packard 3457A Multimeter holding a 44492A multiplexer, and the cell current was measured by a Solartron 7150 plus Digital Multimeter. These instruments were connected to the PC through GPIB interface. The PC also contained a digital I/O board (NI DAQ PC-DIO-24) which controlled the cell load, the gas flows and pressures of the system. The PC-DIO-24- board was connected to a resistance switch box. By means of a relay board (NI DAQ SC-2062) the cell load could be varied from 0.6 to 127 ohm in steps of 1.0+0.2 ohm. To obtain high current densities, the resistance switch box was replaced by a Kantal resistance wire with a known specific resistance. By varying the length of the Kantal- wire and measuring the potential drop over the wire, high current densities were obtained. The minimum outer resistance obtainable was around lOQmfi. Set points to the mass flow controllers were given digitally from the PC-DIO-24- board, whereas the analogue voltage set points of the pressure controllers were obtained by D/A converting digital signals generated by the PC-DIO-24-board. The data acquisition program LabView (for Windows ver. 3.01, National Instruments) was used to control these instruments during fuel cell testing. The computer screen is shown in Figure 4.4. Some instrument drivers were available, whereas others had to be written or modified. Cell data were monitored on screen and saved to file at chosen time intervals in a Microsoft Excel worksheet format (see Appendix H) for later analysis and calculations. An overview of the cell parameters and how these were obtained is given in Table 4.1. 4.2. Test-Station Description 37

Table 4.1. An overview of the cell parameters and how these were obtained. Cell parameter Measured Set by hand Varied by computer Calculated

Cell potential X

Voltage losses (total) Anode X Cathode X

rR-loss (total) X X

Cell current X

Cell load X X

Temperatures: Cell X X Humidifiers X X

Pressures Anode Total X X Hydrogen X Water X

Cathode Total X X Oxygen X Water X

Figure 4.4. Computer screen when testing single SPFCs by applying the data acquisition program LabView 38 Chapter 4. Equipment For Electrode Preparation And Testing

4.2.4 Operation limitations, discussion The test station was designed for pressures of up to 10 bar (1 MPa), which is twice the usual operation pressure of SPFCs. Pressure differences between electrodes were of interest in order to control the water balance in SPFC[12]. A higher cathode pressure (in the range of 1-2 bar (100-200 kPa)) is customarily used in SPFC operation because it increases water diffusion towards the anode and thus counteracts the water transport by electro-osmotic drag. A large gasket area towards the membrane (Fig.4.2) makes it possible to operate with pressure differences between electrodes. Tests with Nation 117 membranes and Prototech electrodes showed that the cell can operate at pressure differences of at least 3 bar (300 kPa). The cell temperature may be varied from room temperature to 150 °C. The humidifiers can operate from room temperature to 130°C, corresponding to a water saturation pressure of psr^o = 0.03 - 2.7 bar.

When operating the fuel cell at a lower temperature than the humidifiers, the gases leaving the cell were oversaturated, giving rise to a two phase flow through the pressure controllers. Initially this caused pressure instability in the system of about +0.5 bar. Incorporation of water separators (Fig.4.1), reduced these variations to within ±0.02 bar as measured by the pressure controllers. Cell current was varied by changing the resistance of the outer circuit This was not the optimal way to obtain polarization curves, because both current and potential may vary. In some cases this caused oscillations during cell testing (Section 4.5). Recently an EG&G VersaStat (Model 253, Princeton Applied Research Corporation) was interfaced to the computer and control routines were programmed in LabView. This VersaStat was adequate to control cell current in the low current range (<250 mA). For the cell area used in the present study, 3.9 cm2, this corresponds to a current density of less than 65 mA/cm2. Much higher constant current densities can be obtained by means of a DC Electronic Load. Thus, a Hewlett Packard 6060A DC Electronic Load (max.30A) will be ordered for future work. The LabView software (NI) was useful for handling the fuel cell data acquisition task. One drawback was, however, that ordinary Windows instrument drivers were not applicable in the LabView program. Thus, one has to apply Nl-produced interface cards, and these were not always adequate for the present application. 4.3. Hot Pressing Equipment 39

4.3 Hot pressing equipment

A thermostated sample holder for pressing the electrodes to the membrane was constructed. The holder contained an insulated electric heating tape (Fibrox™, Thermolyne Corp.) and a type K-element thermo-couple connected to an Eurotherm PID-controller. The mechanical pressure was provided by a Beckman press (Scotland) as shown in Figure 4.5.

Figure 4.5. The hot pressing equipment The right figure shows the sample holder in detail. In the left figure the sample holder is inserted into the press. ® Beckmann press, ® Sample holder, with heating, (D Membrane and electrode assembly, ® Thermo-couple, ® Upper piston. 40 Chapter 4. Equipment For Electrode Preparation And Testing

4.4 Spray deposition equipment

A thin film catalyst layer was deposited directly onto the membrane by spray deposition using a Badger Airbrush Pistol (model 200). The pistol was operated at an air pressure of around 2 bar (200 kPa). Spraying time was typically between 3 and 6 seconds. The spray time was set accurately using an electronically controlled solenoid operating the pistol trigger.

Figure 4.6. The Air Brush Pistol used in the spray deposition technique. suspension, © Membrane holder. 4.5. Operation Stability 41

4.5 Operation stability

A certain operation stability 1 of the fuel cells during testing is required to obtain reliable cell data. The membrane conductivity in SPFCs depend largely on its water content. In this work, water was supplied to the cell by humidification of the reactant gases. The cell fixture design (the size and shape of the gas channels) was expected to influence on the cell humidification properties. Therefore the cell operation stability was studied at different operating conditions. The single cell fixture with 1st generation pistons was used (Figure 4.3, p.35). Since in this work detailed studies of the operation conditions (i.e. gas flow rates, cell design etc.) were not to be included, the single cells were studied at low gas utilization (i.e., high gas flow rates). Furthermore, a high reproducibility 2 is needed in order to compare single cells with different electrodes. Incorporation of a pneumatic cylinder to the cell fixture improved the reproducibility by ensuring a constant mechanical pressure over the M&E assembly. Reproducibility of the measurements is dealt with in Appendix C.

A 2?-screening experiment To investigate how critical the water supply to the membrane was, some introductory experiments were performed. All stability tests were performed using the same single cell consisting of electrodes (Prototech, 0.5 mgPt/cm2) impregnated with 5% Nation solution and hot pressed to a Nation 117 membrane. The optimal humidification temperature has been reported to be 5-10°C higher than the cell temperature[27,48] . One commonly used method of water management is to maintain a larger pressure in the cathode compartment than in the anode compartment[ 12], A screening experiment was set up as a two level factorial design with three factors. For more information about factorial design, see Chapter 5.4. The factors and their levels are given in Table 4.2.

1 Stability in terms of an stable cell potential at a given cell current density, or vice versa. 2 Cells prepared from the same materials and by the same procedure should give approximately the same cell performances. 42 Chapter 4. Equipment For Electrode Preparation And Testing

Table 4.2. The factors and levels for the screening stability experiment. LEVELS FACTORS low high Cathode humidification temperature, Tchum 50 °C 90 °C Anode humidification temperature, TAhum 70 °C 110°C Cathode gas pressure, Pc 3.5 bar 5.5 bar

Cell temperature was kept constant at 80 °C and the humidification temperatures of reactant gases were varied. The effect of back-diffusion of water (ref. Chapter 2) was studied by keeping the anodic pressure constant at 4.5 bar and varying the cathode pressure. The cell was operated at a constant resistance of the outer circuit of around 0.7 £2, corresponding to a current density of around 200 mA/cm2 at stable performance. The response of the experiments was: Stable, Meta-stable or Not stable. The criterion for a stable performance was chosen as: Constant cell potential ±5% over a period of 2 hours The meta-stable tests were those at the limits of the criterion. The order in which the experiments were run were randomized to reduce any time dependent variations.

Results and discussion of the screening stability experiment Figure 4.7 shows the result of the screening stability investigation. Both temperatures of the humidifiers (anode and cathode gas) seemed to have an impact on stability, whereas cathodic pressure was of minor importance. High humidity gave stable performance. When one or both inlet gases were undersaturated (i.e., humidified at a temperature lower than the cell temperature of 80°C), the cell operation was unstable. This was interpreted as evidence for membrane dehydration, giving increased membrane resistance. 4.5. Operation Stability 43

Only a small positive effect was observed by increasing the cathode pressure from 2.5 to 4.5 bar, indicating that the cathode pressure may be used to influence the back-diffusion of water from cathode to anode. 3.5bar Cell operation control using a constant resistance in the Figure 4.7. Results from the screening experiment outer circuit, proved to be on cell stability. # Unstable, O Stable, ©Meta ­ stable. advantageous in these studies. Unstable cells showed more or less oscillating behavior. These oscillations were explained as follows: When humidity was low, the membrane starts to dry out, especially at the anode side, due to electro-osmotic transport of water following the protons through the membrane. The membrane resistance of the cell increases due to this dehydration, causing a reduction in potential and current. The electro-osmotic drag reduces as current decreases and the diffusion of water in the membrane contributes to rehumidify the anode side of the membrane. The membrane resistance decreases allowing for a larger cell current, and the process restarts.

A second stability experiment, 3?-factorial design Based on the results from the screening experiment, a new experiment was performed using the same single cell as before. Gas pressures were held constant due to the minor effect observed in the screening experiment, whereas the cell and humidification temperatures were studied in more detail. Thus, the number of parameters was reduced to two (i.e., the cell temperature and the humidification temperature), whereas the number of levels was increased to 3, hence a 3z-factorial design (Table 4.3). 44 Chapter 4. Equipment For Electrode Preparation And Testing

Table 4.3. Factors and levels of the second cell stability experiment. LEVELS FACTORS - 0 +

Cell temperature, T^ 45 °C 70 °C 95 °C Gas humidification temperatures, Thum 60 °C 80 "C 100 °C

The response was still Stable, Meta-stable or Not Stable, but in addition the cell potential for the stable runs was used for further investigation.

Results and discussion of the second stability experiment Figure 4.8 shows the results of this investigation. Operation stability was again obtained for the oversaturated conditions only, i.e., when the humidification temperature was more than 10°C higher than the cell temperature. Tests at the lower right comer of Figure 4.8 (dashed square) gave the more stable performance and were further investigated, giving the following slopes:

dE ( dE ^ = 1.1mV / K = 0.6mV/K (4.1) Thermodynamically the cell potential decreases with temperature reflecting the entropy change of the reaction. When the reactant gases are oversaturated, the water is produced as liquid. The value of the coefficient, determined from basic thermodynamic equations (Appendix A, Equation A.6) is in this case - 0.85 mV/K. This gives an increase in cell potential (thermodynamically corrected) of l.l-(-0.85) = 2.0 mV/K. Since the cell performance was measured at a constant outer resistance, the increase in cell potential also brought about an increase in cell current. Hence, the above calculated slopes are underestimations. The improved cell performance by increasing cell and humidification temperatures may be attributed to increased reaction kinetics and reduced membrane resistance. 4.5. Operation Stability 45

480 95°C #-

100°C

Figure 4.8. Results from the screening experiment on cell stability. # Unstable, O Stable, ® Meta-stable. Cell potential [mV] is indicated adjacent to the circles.

Conclusions on cell operation stability From these experiments, the following conclusions may be drawn. The humidification temperature must be at least 10 °C higher than the cell temperature to maintain stable cell operation. This is in qualitative agreement with other investigations[27,48]. Cell potential increases with cell temperature. This more than counteracts the decrease calculated from entropy of reaction. Cell potential also increases with humidification temperature. Both effects were interpreted to be related to increased catalytic activity of the electrodes and reduced membrane resistance.

4.6 Conclusions

A test-station with temperature and pressure control systems was designed and built. The apparatus facilitates automated cell tests within a wide range of operation conditions. The equipment was shown to perform in accordance with reported literature, for variations in temperature and pressure. Incorporation of pneumatic pressure control increased the reproducibility of the single cell tests. The flexibility of the system makes it suitable for verification SPFC models. 46 Chapter 4. Equipment For Electrode Preparation And Testing 47

C hapter 5

Electrode Preparation And Evaluation

The test station, described in Chapter 4, has been applied to study single solid polymer fuel cells. Commercial electrodes have been modified and new electrodes have been synthesized. The experimental procedures, as well as the results of the single cell performance tests are given.

5.1 Introduction

Performance of the solid polymer fuel cell (SPFC) has been significantly improved during the last decades and high power densities are now attainable (>1 W/cm2) [1,37,49]. Parts of this improvement have been achieved through increased electrolyte conductivity (Section 1.2). Equally important, however, has been the improvements in properties of the electrode and its contact to the electrolyte. Development of electrodes and catalyst layers, their structure and single cell assembling procedures are described in the following sections, respectively.

5.1.1 Electrode development A major challenge for improvement of low temperature fuel cell systems has been, and still is, the reduction of cathodic activation overpotentials [5,24]. This is related to the low exchange current density of the oxygen reduction reaction[21,44]. Low operating temperature in SPFCs makes the use of noble catalyst materials unavoidable. A natural choice of electrocatalyst was platinum, showing excellent H2 oxidation kinetics, and the usage of high Pt-loading electrodes have partly solved the problem of high cathodic overpotentials. In the infancy of SPFC development, Pt-foil or platinized Ni-screen was used as a combined electrode and current collector[50]. Platinum was later supported on carbon and the introduction of these porous electrodes largely increased the Pt-utilization. The widely used electrodes produced by General Electric Company/Hamilton Standards (GE/HS) contained 4 mgPt/cm2[51], adding significantly to the total cost of the cells. 48 Chapter 5 . Electrode Preparation And Evaluation

The use of high Pt-loading electrodes represented a major impediment to commercialization of SPEC technology, because of the high cost involved. Therefore, a large amount of work was devoted to lowering the platinum loading through increased Pt-utilization(see e.g.,[51,52]), optimizing electrode morphology[38,53], and finding new catalyst materials [54], Impregnation of the electrode surface by a solution containing the polymer (Nafion) showed a large increase in the three phase area of the electrodes[27]. During the 80s, Pt-loading was successfully lowered by one order of magnitude to 0.4 mg Pt/cm2 without significant loss in performance. The contributions from Los Alomos National Laboratory[27,48] and Texas A&M University (TAMU) by Srinivasan and co-workers [49] are in this connection noteworthy. The work at Los Alomos arose from a patent by Raistrick (US Patent No. 4,876,115). Despite this successful reduction in catalyst loading, the fuel cells manufactured by Ballard Power Systems Inc. for use as power source in fuel cell buses, still applied high Pt-loading electrodes when entering the 1990s[l]. Environmental concerns have brought about an increasing demand for platinum for use in catalytic converters for automobiles. Hence, relatively limited abundance of platinum and high prices emphasizes the demand for further reduction in Pt-loading. Recent investigations show that Pt utilization may be largely enhanced by localizing the catalyst close to the electrode membrane interface. Thin electrocatalytic films consisting of carbon supported Pt and with polymer conductor as binder, show promising properties in a study by Wilson and Gottesfeld[51]. They reported that the reaction zone shrinks to some 4 pm from the membrane-electrode interface as current densities increase. Thus, by concentrating the platinum in this region, the utilization could be largely increased. Fabrication of catalyst layers and diffusion backings separately, makes it possible to formulate each structure with the properties that are most suitable for its functional] (see Section 5.2.1). Another way to decrease the platinum loading has been to alloy the Pt-catalyst. Tamizhmani and Capuano[35,55] have shown that some alloys of platinum are more electrocatalytic active than pure platinum. They studied the binary alloy Pt-Cr and the ternary alloy Pt-Cr-Cu. Of the binary alloys Pt-Ni, Pt-Cr and Pt-Co studied by Mukeijee and Srinivasan[56], especially the Pt-Cr-alloy showed promising electrocatalytic properties. Platinum has also been alloyed with Rh and Ru giving the 5.1 Introduction 49 advantage of increased activity of the electrode in the presence of CO[57]. These alloys also catalyses the CO oxidation[58]. Electrodes prepared by incorporating platinum tetramine [Pt(NH3)]2+ into the Nafion membrane, followed by precipitation of metallic platinum (by use of sodium borohydride) were reported by Millet et al.[59]. By this technique the platinum could be located close to the membrane surface. Taylor et al.[40] have developed a different technique of electrode preparation. An uncatalyzed carbon gas diffusion electrode was impregnated with Nafion, followed by electrodeposition of platinum particles through the Nafion layer. By this technique, catalyst particles were located in regions with both ionic and electronic accessibility, only. The oxygen reduction reaction for electrodes holding 0.05mgPt/cm2 was investigated and showed comparable performance to standard electrodes (0.5mgPt/cm2). This is the optimal technique from a theoretical point of view, but the lack of control of Pt-cluster size may, however, be a draw-back. The Pt4+-ions may find some favorite paths through the Nafion film and build up clusters larger than what is optimal. A novel approach has been made by Ye et al.[60], applying electrodes based on porous carbonized polyacrylonitrile foam with Pt-loading as low as 0.013 mgPt/cm2. The activity of this new material is comparable to that of bulk platinum, but its properties under real SPEC operation conditions are yet to be determined. Extensive research has been directed towards the development of oxygen reduction catalysts from organic transition metal complexes, like phtalocyanines[23]. Organic catalysts have, to my knowledge, not yet been reported in commercial use in SPFCs.

Cell life time is a critical parameter for commercialization of the SPEC technology. The loss of fluoride from a SPEC (which is the degradation product of the perfluorocarbon membrane) has been suggested as a measure of cell life time[61]. Currently, there is no general agreement as to whether the membrane or other parts of the cell constitutes the life time limiting material in SPFCs. Spectroscopic studies of phosphoric acid fuel cell electrodes (which are similar to those used in SPEC) show that platinum clusters may aggregate during cell operation forming larger particles giving a reduced performance[62] . Mukeqee and Srinivasan[56] have evaluated the 50 Chapter 5. Electrode Preparation And Evaluation life time of single SPEC with porous electrodes holding Pt, Pt-Ni and Pt-Cr catalysts. Only minor performance degradation was observed over a time period of 400-1200 hours. The ternary alloys Pt-Cr-Cu also exhibited stable performance for 300 hours[55], Wilson et al.[63] reported that special treatments were necessary to impart durability to the proton conductor in thin film catalyst layers. These treatments include the conversion of the Nation polymer to a thermoplastic form by ion exchange inclusion of large hydrophobic counterions such as tetrabutylammonium (TBA+).

Of all these efforts to improve performance and reduce the cost of SPEC electrodes, the thin film electrodes have shown to be the most promising approach. The thin film electrodes (applying platinum or platinum alloys as catalyst) show reasonable low potential losses and high catalyst utilization. Further, studies indicate that chemical stability for long term operation is obtainable by special treatments of the electrodes. The electrode preparation method should be appropriate for scaling-up purposes. The thin film electrodes are easy to prepare in large areas by screen printing, tape casting or spray depositioning.

The performance characteristics of porous electrocatalytic layers for SPEC are significantly influenced by their composition and by the way they are prepared. The number of factors influencing the electrode properties are high. Optimization of these catalyst layers has been studied by a number of investigators. In literature, it has been common to study only one parameter at a time; e.g., the effect of variation in Nation loading [48], while keeping the other factors constant. There is, however, a need to study more parameters simultaneously to investigate the interactions of the different factors. A statistical approach applying factorial design is very useful for such studies [47]. Detailed studies of this kind have scarcely been reported 1.

5.1.2 Aim of work The aim of the work presented in this chapter is to study thin film electrodes spray deposited directly onto the membrane. The electrodes will be studied in single SPFCs.

1 The development of efficient electrodes, the heart of the fuel cell, is time consuming work. It is, thus, no surprice that the recipes of novel electrodes are kept secret or are patented (e.g. US Patent # 4,610,938, # 4,876,115, #5,084,144 and #5,084,144) 5.1 Introduction 51

The influence of the electrode composition and preparation procedures on the single cell performance will be examined by applying factorial designed experiments.

5.2 Single cells with thin film electrodes, system description

5.2.1 Electrode structure, materials and composition Fuel cells containing thin film electrocatalytic layers have shown promising results and are the topic of this study. Their structure is hence described in detail, and the need for materials of highly specific properties will be explained. A sketch of the different regions within the SPFC is given in Figure 5.1. There are three distinct regions encountered in a state-of-the-art SPFC; the membrane phase, the catalyst layer and the porous electrode backing, respectively. Each region must exhibit different properties connecting to their functionality. The processes taking place in the cell will be illustrated, focusing on the cathode. However, the function of the different components of the electrodes is similar for the anode. Transport properties of the membrane phase were described in Chapter 2, and will, hence not be covered here.

Porous electrode backing The porous electrode backing facilitates gas transport to the catalyst layer, and also drains the water produced in the chemical reaction. Thus the backing must contain both hydrophobic regions (for gas transport) and hydrophilic regions (for water drainage). Poly-tetra-fluoro-ethylene (PTFE) is commonly used as hydrophobic agent. By varying the amount of the hydrophobic agent, the water transport through the electrode may be regulated[64]. With low contents of hydrophobic agents, a large portion of the electrode pores is filled with water. This may, at high current densities, result in mass transport limitation caused by low rate of gas diffusion. High contents of hydrophobic agents tend to reduce the electronic conductivity of the electrode backing and to hinder the water transport, causing electrode flooding[65]. Thus, the amount 52 Chapter 5. Electrode Preparation And Evaluation

Figure 5.1. A schematic picture of the single SPEC showing the three different regions: CD Membrane, (D Catalyst layer and d> Porous electrode backing. The zoomed circle shows schematically the catalyst layer. The white areas are pores filled with water and gas. of hydrophobic agent is a compromise between these different demands. Mosdale and Srinivasan[l] reported that the volume fraction of the cathode available for gas transport should be at least 20%, corresponding to a concentration of the hydrophobic material in the electrode of 35%. The electrode backing also functions as current collector. High surface carbon (e.g. Vulcan XC72) is commonly used for the porous backing and also as support in the catalyst layer. Vulcan XC72 has low electronic resistivity (0.2-1 ohm cm)[46].

Catalyst layer The catalyst layer constitutes the reaction zone where oxygen is reduced, where it reacts with protons from the membrane phase and produces water. In this region, the polymer and platinized carbon meet the gas, creating the so-called three phase area. 5.2. Single Cells With Thin Film Electrodes , System Description 53

The three phases are the proton conducting polymer, the electronically conducting platinized carbon and the gas phase. The key to increased Pt-utilization is to localize the platinum catalyst in contact with all three phases. Catalyst material adjacent to the three phase area will not contribute to the reaction and is thus wasted. The catalyst material is commonly Pt on Vulcan XC72 carbon support. A loading of 20 wt% has been reported to be optimal in terms of cell performance[53] . The average Pt-particle size of this material is 20A and around 30% of the atoms are surface atoms[22]. The amounts of both electronic and protonic conductors should be large enough to create a continues three dimensional network. High protonic conduction may be obtained by high fraction of polymer, at the sacrifice of electronic conduction and reduced hydrophilic properties. The composition of the catalyst layer is also the result of a series of compromises, most of which are connected to the water transport According to Bernard! and Verbrugge[12], water management and catalyst localization are two design issues of great importance in SPFC. The water management of the SPFC was addressed in Chapter 2.

5.2.2 Membrane and electrode assembling procedures Intimate contact between membrane and electrodes is required to facilitate the transport of protons from the bulk polymer phase (the membrane) to the polymer phase in the cathode catalyst layer and further to the reaction sites, (and vice versa at the anode). A good contact was achieved when the electrodes are ‘hot pressing ’ to the membrane at a temperature close to the glass transition temperature of the polymer[50]. Hot pressing is usually done at 50-100 bars and around 135°C (for the Nation membrane). An optimization study of the hot pressing procedure was reported by Srinivasan et al.[27]. A careful choice of preparation and assembling procedure must be made depending on the chemical stability of the materials in use. Certain precautions may be taken to reduce the loss of membrane conductivity. The sodium form of the polymer is chemically stable up to around 200°C[51]. Hence, replacement of the protons by Na+ ions, during electrocatalyst preparation reduce polymer degradation. Pulverization of the catalyst layer directly onto the membrane was reported by Mosdale and Stevens[46]. Techniques involving application of the catalyst directly to 54 Chapter 5. Electrode Preparation And Evaluation the membrane are expected to give good contact between the layer and the membrane, but the need for substantial drying of the membrane after spraying may, however, reduce the proton conductivity of the polymer material.

5.3 Performance evaluation and separation of potential losses

Cell performance may be evaluated by comparing cell potentials at a chosen current density or vice versa. Comparisons should be done at potentials and current densities which are realistic for use in commercial cells. For the total fuel cell area to be of an attractive size, the current density must be at least 100 mA/cm2. The single cell potential should be higher than 700 mV to avoid large losses of efficiency (Chapter 3). Other means of comparison relate to the shape of the polarization curve (see Chapter 2). This is a more sophisticated because it governs a more detailed analysis and elaborates the origin of the losses for the cell in comparison. The simple model given by Equation 2.17 shall be applied to determine electro-kinetic properties from performance data obtained for single cells. Methods for separation of potential losses are reviewed in Appendix G. In the present study, potential losses were obtained by incorporation of reference electrodes (Figure 5.3), equal to the working electrodes. At the anode side this constitutes a reversible hydrogen electrode (RHE). Ohmic losses were separated by applying the current interruption technique. Further, the mass activity of platinum (i.e., mAZmgPt) was used as an indication of the degree of catalyst utilization.

5.4 Experimental design

The complexity of the electrode in SPFCs makes a systematic approach to electrode optimization advantageous. An experiment based on factorial design: • Requires few runs per factor studied • Indicates major trends • Determines direction for further experimentation • Provides easy result evaluation. 5.4. Error ! Reference source not found . 55

Both main effects of the factors and interactions between factors are obtainable. Surprisingly few studies applying this powerful tool have been reported. Shukla et al.[43] used factorial design in optimization (of the amount of binder, the compaction load and the compaction time) of SPFC electrodes. For more information on factorial design, see Box et al.[66]. In addition, a systematic approach to the studies of electrodes is also advantageous for modeling purposes[67],

5.4.1 A 2^ screening experiment

It is expected that the amount of hydrophobic agent (i.e. Nation) in the catalyst layer will influence the cell performance, and was therefore chosen as the first factor. Further, it was shown by Zawodzinski et al.[45] that water uptake in Nation was reduced after drying at elevated temperatures, and that membrane (Nation 117) conductivity increased linearly with water content. Thus, it is expected that conductivity of the polymer is reduced by drying, and, hence drying temperature was thus varied as the second factor. It has been reported that reasonable performance can be obtained with Pt-loadings down to 0.10-0.15 mg Pt/cm2[51]. The level should be kept low to reduce electrode cost, and hence the third factor studied was a further reduction of the Pt-loading down to 0.06 mgPt/cm2. Hence, the factors studied were: 1. Nation content 2. Drying temperature 3. Pt-loading A factorial design at two levels was used in this screening experiment and thus a 23- design. This simple experiment is not capable of exploring fully a wide region in the factor space. The purpose of the screening experiment is, however, to indicate the major trends. The factors and levels are given in Table 5.1, p.62.

5.4.2 Addition of Acetylene Black Acetylene Black is a carbon with very high electronic conductivity. Its introduction into the porous carbon electrodes was suggested by Mosdale and Stevens[46] and later used by Uchida et al.[68]. The addition of Acetylene Black was studied to investigate its effect on reducing the ohmic losses of the cell. The amount of Nation in the electrocatalytic layer was kept at a low level of 15%. Aiming at low Pt-loading, this factor was kept constant at a level of 0.1 mgPt/cm2 in this experiment. To reduce 56 Chapter 5. Electrode Preparation And Evaluation degradation of the Nation polymer, the drying temperature was held at 125°C. The effect adding acetylene black was studied at two levels (0 and 10%).

This experiment was part of a larger investigation (32-factorial design) as described in Appendix E. In this experiment the amount of Acetylene Black and Nation were to be studied, each at three levels. Composition of the different catalyst materials is included in Appendix E.

5.5 Materials and methods

5.5.1 Membrane preparation

The Nation 117 (E.I. du Pont de Nemours and Company) membrane was used throughout this study. The membrane sheet was cut (in the dry state, as supplied) into circular samples of 36 mm diameter. The membranes were rinsed according to the procedure given in Appendix F, and stored in distilled water until use.

5.5.2 Modification of commercial electrodes

Commercial Prototech electrodes (0.5 mg Pt/cm2 on Vulcan XC72) were supplied by The Electrosynthesis Co., NY, USA. To enhance their performance, the electrodes were modified by vacuum deposition and impregnation as described below.

Thin Pt/Rh layer by vacuum deposition A thin layer of Pt/Rh (80/20%) was deposited on top of the electrode under high vacuum (10 5 torr) using a Coating Unit (Edwards High Vacuum, LTD, Sussex). The amount of Pt/Rh was estimated to 30+5 pg/cm 2 corresponding to a thickness of approximately 15 nm (non-porous metal). The estimation was based on the weight loss of the Pt/Rh wire during vacuum deposition.

Impregnation with Nafion solution The electrode was then impregnated (brushing by hand) with Nation 117 solution (5% mixture in lower aliphatic alcohols and water, Fluka Chemica), followed by drying in vacuum at 70°C for 2 hours. The Nation loading was determined gravimetrically to 5.5. Materials and methods 57

1.0+0.1 mg/cm2 (dry weight), close to the optimal loading recommended by

Poltarzewski et al.[38].

5.5.3 Thin film electrodes Electrode catalyst material (20 wt% Pt on Vulcan XC72, provided by Johnson Matthey Pic, UK and later by E-TEK, Inc., MA, USA), was mixed with Nation 117 solution (5 %), glycerol and water, as guided by the procedure given by Wilson and Gottesfeld[51].

The 23-screening experiment Two suspensions with different polymer (i.e., Nation) contents (15 and 25%) were prepared. To facilitate the change in polymer content while keeping the Pt-loading constant, small amounts of uncatalyzed Vulcan XC72 carbon (Cabot, Netherlands) was added to the suspension of low polymer content. The catalyst material compositions are given in Appendix B. The variation in Pt-loading was obtained by changing the thickness of the thin electrode layer.

Addition of Acetylene Black In this study, nine different suspensions were prepared as described in Appendix E. Two of these were studied and are reported here. The remaining cells will be studied and the result will be reported elsewhere. In one of the two suspensions used, 10 % of the Vulcan XC72 carbon was substituted by Acetylene Black holding the same 20% of Pt as the Vulcan carbon. The Vulcan XC72 and the Acetylene Black (Shawinigan) used in this experiment were supplied by E-TEK, Inc., MA, USA.

The suspensions were subject to extensive ultrasonic and mechanical mixing before applying to the Nation 117 membrane (Fig.5.2). A thin catalyst layer was deposited directly onto the membrane by spray deposition using a Badger Airbrush Pistol (Section 4.4). Pt-loading was determined gravimetrically from the amount of suspension supplied to the membrane. The membrane and electrode (M&E) assembly was dried in a forced convection oven and stored in cold distilled water for later performance testing. 58 Chapter 5. Electrode Preparation And Evaluation

Pt /Carbon Material' I

Nation Solution Ultrasonic Solvents Mixing * Snraving onto membrane

Drying

Single cell with Thin Film Electrode

Figure 5.2. Preparation procedure for single cells containing Thin Film Electrodes.

5.5.4 Membrane and electrode assembling The hot pressing equipment was described in Section 4.3. Prototech electrodes (as- received and modified, Section 5.5.2) were pressed onto wet Nation 117 membranes at 90 bar and 135 °C for 90 seconds. The assembling is illustrated in Figure 5.3.

Figure 5.3. Illustration of assembling of membrane and electrodes to single cell by the hot pressing procedure. 5.5. Materials and methods 59

Hot pressing was not applied for cells with thin film electrodes (Section 5.5.3),. An uncatalyzed E-TEK carbon cloth (ELAT, Vulcan XC72) was merely inserted between the M&E assembly and the piston as the cell was placed in the single cell fixture. These commercial backing electrodes (ca. 500 pm thick) are prepared by adding a mixture of carbon powder and FIFE to a carbon-cloth substrate.

5.5.5 Cell test procedure The M&E assemblies were boiled in 0.05 M H2S04 for 30 minutes and rinsed in hot distilled water before they were mounted in the single cell test fixture (Fig.4.2). Safety precautions were taken by flushing the system with N2 gas for 5-10 minutes before and after every experiment. Cell test conditions are given in figure texts. Only pure H2 and 02 gases, saturated with water at 10-15°C higher than the cell temperature, were used. The total pressures were kept either at lbar or 4.5bar at both electrodes. Gas flow rates were constant and high, corresponding to a gas utilization of at most 10% at high current densities (>900mA/cm 2). The cell was kept at moderate current densities (200-300 mA/cm2) for 10-48 hours to ensure the rehumidification of the dried protonic conductor. The time needed to rehumidify the polymer was much longer for cells with the thin film electrodes (usually >24 hours) than for cells with commercial electrodes the hot pressed to the membrane. Polarization curves were obtained when the cell had reached stable performance. The cell was kept at a constant outer resistance until stable performance, i.e., 3 minutes at low current densities and 15 minutes for current densities where mass transport limitations may occur (usually higher than 200mA/cm2).

5.6 Results and discussion

Single cell performance characteristics were obtained by applying the SPEC test- station described in Chapter 4.

5.6.1 Commercial electrodes Single cells containing commercial Prototech electrodes and Nation 117 membranes were prepared to give a reference for comparison with the thin film electrodes. The performance of single cells containing as-received and modified Prototech electrodes 60 Chapter 5. Electrode Preparation And Evaluation is shown in Figure 5.4. Polarization curve for a cell with thin film electrodes was included for later comparison. As-received materials showed poor performance, whereas impregnation (with Nation solution) of the electrodes had a tremendous positive effect on cell polarization. These results are in agreement with results reported by Ticianelli et al.[48]. The large improvement in cell performance followed by impregnation was due to an increased three-phase area, and thus a better Pt-catalyst utilization. The electrodes containing an extra layer of catalyst (applied by vacuum deposition) showed superior performance compared to the impregnated electrodes. In both cases the Nafion-solution penetrates the pores and improves the contact between the electrode and the membrane phase. Comparison of the mass activity (Table 5.2, p.63) show that the extra catalyst layer (corresponding to 0.03 mg/cm2 of Pt/Rh) contributes significantly to the reaction rate. The thin deposited layer facilitates an improved mass activity corresponding to 2500mA/mgPt. Other investigations[52] have also shown large mass activity improvements from thin layer sputtered onto the electrode surface. The deposition

"■ As-received electrodes Impregnated electrodes Vacuum deposited & impr. electrodes ~° ~ Thin film electrodes

Current density [mA/cm2]

Figure 5.4. Polarization curves for single fuel cells containing Nation 117 membrane and different electrodes. T^u = 70 °C, Thumidiriers= 85 °C, Pa„= Pcat= 4.5bar (450 kPa). 5.6. Results and Discussion 61 increased the number of active electrocatalytic sites near the membrane surface. As current density increases, the region where the reaction takes place closes up to the membrane surface. The region was in the range of 4 pm from the membrane surface at high current densities[51]. Despite the large increase in cell performance, the technique of vacuum deposition is, like the sputtering deposition technique, not considered to be an economically viable alternative for large scale fabrication of electrodes[40]. The reproducibility of the cells with thin film electrodes were examined as shown in Appendix C. From the results it was concluded that there was no significant difference in performance between cells with modified commercial electrodes and cells with thin film electrodes.

5.6.2 Thin film electrodes, 23-screening experiment Eight single cells containing thin film electrodes of different treatments were assembled and tested. The experimental design was described in Section 5.4. The cells were compared through their cell potential at a current density of 100 mA/cm2. The variation in cell potential was substantial, ranging from 225 to 675 mV. The cell potentials are shown at the comers of the cube in Figure 5.5. Calculation of the effects are shown in Appendix B. The main effects of the factors are given in Table 5.1. The reproducibility of the cell data and the significance of the effects are calculated in Appendix C. The results (Table 5.1) show that the increase in Nafion content from 15 to 25 wt% gave (in average) a gain in cell potential of 123 mV at 100 mA/cm2. This was attributed to increased proton conductivity in the catalyst layer of higher Nafion content. The lower cell potentials at high drying temperature (135°C) indicated that the polymer material experienced some chemical degradation during the 60 minutes of drying at 135 °C. This degradation may be reduced by using the polymer in the Na+ or Li+-form during drying followed by re-protonation[46] . The increase in Pt-loading had the most pronounced effect on cell potential. Cell potentials were generally low for the 0.06 mg Pt/cm2 electrodes. This was in agreement with results obtained by Wilson and Gottesfeld[51], reporting that 62 Chapter 5. Electrode Preparation And Evaluation

+145

3751 +220

IS Nafion (%) 25

Figure 5.5. A geometrical representation of the experimental factorial design. Cell potential [mV] at a current density of 100 mA/cm2 is given at the corners of the cube. reasonable performance of thin film electrodes was obtainable down to loadings of 0.10 to 0.15mgPt/cm2.

The results obtained for the low Pt-loading electrodes (0.06 mgPt/cm2) in this study

were comparable with those reported by Taylor et al.[40]. They reported polarization curves for oxygen reduction on electrodes with 0.05 mgPt/cm2 in sulfuric acid at room temperature (see Section 5.1.1, p.47). Despite the differences in test conditions, this indicate that the new technique by Taylor et al. is not superior to the thin film technique. Localization of most of the platinum at the three phase area was thus, also obtained by the thin film technique. Table 5.1. Levels and main effects on cell potential, E, of the factors in the 23 - experiment ______Levels Effect on E* Factor Low High [mV] Nafion (% by weight) 15 25 +123 Drying Temp. (°C) 125 135 -87 Pt-loading (mg/cm2) 0.06 0.1 +225 *at 100 mA/cm . 5.6. Results and Discussion 63

5.6.3 Platinum utilization, a comparison The polarization curve for the single cell of better performance containing thin film electrodes (from the 23-experiment) was included in Figure 5.4. The cell with thin film electrodes was superior to the cells with modified commercial electrodes up to around 100 mA/cm2. This indicates that the thin film electrodes were more reversible and provide better contact to the membrane. From the slope of the curves in the linear region (Figure 5.4), it is seen that the ohmic potential losses in the cells with thin film electrodes were higher than those for modified commercial Prototech electrodes. The reasons for the more steep decline for the thin film electrode may that i) overpotential was higher in thin film electrodes 1 due to the low Pt-loading, ii) ohmic loss through the thin film electrode was higher compared to the commercial carbon cloth electrode and iii) there were contact problems between the uncatalyzed carbon cloth backing and the thin film electrode. Performance data for the different electrodes are compared in Table 5.2. Mass activity for the thin film electrodes were two orders of magnitude higher than for as-received electrodes at a cell potential of 0.7 V. Comparing to both vacuum deposited and impregnated electrodes, the thin film electrodes show a fivefold higher mass activity. From this it was concluded that the thin film electrodes facilitate a more optimal localization of the catalyst.

Table 5.2. Performance comparison of single cells containing different electrodes, showing the mass activity at a cell potential of 0.7V, calculated from the current density, i, and the Pt-loading. The open circuit potential, Enr.is included. Pt-loading i* Mass Activity Pf* Eoc Type of electrode [mg cm"2] [mA cm'2] [mA mg"1] [V] As-received 0.5 4.3 8.6 0.838 Impregnated 0.5 41 82 0.944 Vacuum deposited & impr. 0.53 83 157 0.988 Thin film 0.1 85 850 0.975 * at 0.7 V

1 At these low Pt-loadings an anodic overpotential is likely to occur. This will probably be in the linear regime due to the high exchange current density of the hydrogen oxidation reaction. This will then contribute by linearly reducing cell potential. 64 Chapter 5. Electrode Preparation And Evaluation

5.6.4 Thin film electrodes, Addition of Acetylene Black The effect of substituting 10 % of the Vulcan XC72 by Acetylene Black in the thin film electrodes was studied. The two cells studied (denoted Cell 1 and 7) constitute part of a larger factorial experiment described in Appendix E. Cell 1 and cell 7 contained 0 and 10% Acetylene Black, respectively. Data from the cell tests are given in Appendix H. Polarization curves for pressurized operation (4.5 bar) were obtained as shown in Figure 5.6. The polarization curves of the cells were fitted to Equation 2.17, disregarding the part of the curve where mass transport limitation occurs. The electro- kinetic parameters obtained from the fitting are given in Table 5.3. The inner ohmic resistance of the cell was also measured by current interruption, and the results (Rci) are also included in the table. The exchange current density of the hydrogen oxidation is so high that the anodic overpotential is still in the linear regime. This was in agreement with the measured anodic overpotentials, which showed linear behavior

(- -) Cell 1

(- +) Cell 7

« 0.6

Current Density [mA/cm ] Figure 5.6. Polarization curves for cell 1 and 7 at pressurized operation (4.5 bar). The cell temperature was 70°C, and the humidification temperatures of the gases were 80°C. Platinum loading of the electrodes was 0.1 mgPt/cm2. 5.6. Results and Discussion 65

Table 5.3. Electro-kinetic properties of single cells with electrodes containing different amounts of Acetylene Black (AB). Data for Cell 7 at different temperatures are included. The data were obtained by fitting cell performance data to Equation 2.17. Ra is the ohmic resistance measured by the current interruption technique. ______E* i« Ri Rnb B Cell Conditions m (nA) r£2 cm2! [£2 cm2] [mV/dec] 1 (0% AB) 70°C, 4.5 bar 0.923 43.6 0.47 0.37 68.7 7 (10% AB) 70°C, 4.5 bar 0.960 151 0.38 0.27 68.7 7 (10% AB) 70°C, 1 bar 0.912 69.4 0.45 0.26 65.0 7 (10% AB) 50°C, 1 bar 0.893 50.9 0.63 - 72.7 7 (10% AB) 30-40°C, 1 bar 0.886 23.5 0.87 - 71.9 a at a current density of 2.0±0.3mA/cm2 b measured at 500±30mA/cm2 - not measured over the range of current densities studied (0-900 mA/cm2). In Equation 2.17, this linear activation overpotential contribution was interpreted as ohmic resistance and was included in the term z7?z. Hence, the difference between inner resistance values

(RrRCi) constitutes the anodic overpotential. This means that the non-linear overpotentials were truly attributed to the oxygen reduction reaction at the cathode. An example of a polarization curve including electrode overpotential losses is shown in Appendix G, Figure G4. The Tafel slope of the cells were close to the theoretical slope of 68.1 mV/dec at the temperature of 70°C for the two electron transfer reaction. According to Equation 2.17, the higher cell potential (at a current density of 2mA/cm2), Eo, of the cell with Acetylene Black gave this cell a higher exchange current density, z0. The reduced ohmic resistance of the cell containing Acetylene Black, was attributed to the higher electronic conductivity of this carbon type as compared to VulcanXC72. From literature values[ll], the conductivity of the Nation 117 membrane at 70°C was estimated to 0.20 £2 cm2. From the resistance values obtained by current interruption (Rci in Table 5.3), it can be seen that the ohmic resistance of the electrodes was reduced from 0.17 to 0.07 £2 cm2 by introduction of Acetylene Black. From the same data it was concluded that the conductivity of the membrane was not severely reduced from the estimated value of 0.20 £2 cm2 (for a non heat-treated membrane) during drying. The hard drying conditions (125°C for 75 minutes) of the cells caused a certain degradation seen from the discoloration of the membrane. Long 66 Chapter 5. Electrode Preparation And Evaluation rehumidification times were, however, needed to obtain peak performance for the cells (i.e., 24-48 hours at current densities of >300 mA/cm2). The difference in exchange current densities, i0, was surprising (Table 5.3). Further, the cell without Acetylene Black showed more pronounced mass transport limitations. Both effects may be related to the different quality of the two carbon blacks. The specific surface areas of VulcanXC72 and Acetylene Black were 250 and 65 m2/g[69], respectively. For the 20%Pt on VulcanXC72 the average Pt-particle size was 25A[70], while the corresponding value for Acetylene Black was not available. Bregoli[71] has shown that the activity for oxygen reduction on platinum increases with platinum particle surface area. Differences in Pt-particle size can therefore explain the difference in exchange current density. The addition of Acetylene Black obviously had a positive effect on the transport properties of the catalyst layer. As can be seen from Figure 5.6, mass transport limitations were postponed to higher current densities for the cell with Acetylene Black. This region of the polarization curves was not examin ed in detail. It may, however, be assumed that this was related to differences in pore size and morphology of the catalyst layers due to the use of two types of carbon black.

5.6.5 Temperature and pressure dependence of cell potential Cell 7 was also studied at atmospheric pressure at three different cell temperatures. Data from the cell tests are given in Appendix H. The humidification temperature was kept 10°C higher than the cell temperature. The results are shown in Figure 5.7. It was not possible to operate the cell at a temperature of 30°C (as intended) due to heat production in the cell. At high current densities the temperature increased from 30 to around 40°C with the heating element switched off. The curves were fitted to Equation 2.17 and the results are included in Table 5.3. The change in pressure from 4.5 to 1 bar brought about a reduction in exchange current density to nearly half of its value at high pressure. Further, the exchange current density decreased with temperature. Both these finding s were expected from reduced reaction kinetics. The ohmic resistance measured by current interruption, Rci, was the same at high and low pressure, whereas the ohmic resistance obtained from fitting to Equation 2.17, /?,, was higher at lower pressure. This confirms that the 5.6. Results and Discussion 67 difference i(RrRci) constitutes the anodic activation overpotential, which increases with reduces temperature. The inner resistance increased almost twofold when the temperature was decreased from 70 to 30-40°C. This was both related to increasing anodic activation overpotential and increasing internal resistance of the polymer membrane phase. Comparison of the polarization curves for Cell 7 at 70°C in Figures 5.6 and 5.7 reveal that pressurized operation was advantageous. At a current density of 500mA/cm2, the pressurized cell (4.5bar) obtained a cell potential of 615mV, whereas at atmospheric pressure the cell potential was 510mV.

30-40C > 0.8

2 0.6

U 0.4

Current Density fmA/cm ]

Figure 5.7. Polarization curves of Cell 7 (10% Acetylene Black) at atmospheric pressure, and at three different cell temperatures (70, 50 and 30-40°C). Humidification temperatures were 10°C higher than cell temperature.

5.6.6 Single cell performance, a comparison to literature The better performing single cell with thin film electrodes prepared in this study (Cell 7) showed a cell potential of 615 mV at a current density of 500 mA/cm2 (Fig.5.6).

The cell temperature was 70°C and the gas pressures were 4.5 bar at both electrodes. 68 Chapter 5. Electrode Preparation And Evaluation

Some investigations on single SPFCs with low platinum loading electrodes (<0.15 mg Pt/cm2) have been reported. Recently Wilson et al.[63] reported very good performance for single cells of 5 cm2 with 0.13-0.16 mg Pt/cm2 and Nation 117 membrane. At a current density of 500 mA/cm2 these cells gave a cell potential of around 750mV, which is 135mV higher than that of the cells from the present study (0.10 mg Pt/cm2). Part of this difference in performance can be explained by higher ohmic resistance in our cells. At 500 mA/cm2 we measured an ohmic resistance of 0.27£2cm2 (Table 5.3) whereas Wilson et al. report a high frequency resistance of around 0.180cm 2. At this current density, the higher ohmic resistance of our cells corresponds to an extra potential loss of 45mV. The rest of the discrepancy must be attributed to higher Pt-loading, higher operating temperatures and higher oxygen pressure of the study by Wilson et al.[63], A deposition technique of pulverizing the electrode material onto the membrane was applied by Stevens and Mosdale[46]. Single cells holding electrodes with 0.35mgPt/cm2 show cell potentials of around 400mV at 500mA/cm2, which is lower than the cells reported in this study despite 3-4 times higher Pt-loading. This indicates that the Pt-utilization is higher for the cells of this study. Kumar et al.[72] have recently studied cells consisting of Nation 117 membrane with electrodes holding 0.1mgPt/cm2. The catalyst layer was spray deposited to the porous carbon backing. The cells with Nation 117 membrane show excellent performance; a cell potential of 700 mV at 500 mA/cm2 (90°C, gas pressures 5 bars) was attained. This performance is somewhat better than that of the cells reported in this study (Figure 5.6), which were tested at 70°C, and 4.5 bar gas pressure. As temperature (and pressure) is lowered, the reaction rate is reduced. Hence, the somewhat lower cell potentials in our cells may be attributed to slower reaction kinetics. Cells prepared at our laboratory have not been subject to long-term stability testing. From the comparison above, it may be concluded that the cells prepared at our laboratory show performance comparable with state-of-the-art electrodes with 0.5mgPt/cm2. Our low Pt-loading electrodes also compares well to other novel low Pt- loading electrodes reported in literature. 5.6. Results and Discussion 69

The performance of the single cells in the study of Acetylene Black addition was generally better than those from the 23-screening experiment. This difference was interpreted as arising from the use of different pistons in the single cell fixture (Section 4.2.2)

5.6.7 SEM analysis A Scanning Electron Microscope (Zeiss DSM940, West Germany) was applied to study the thin film electrodes prepared by spray depositioning. The instrument was operated at 10-20kV. A single cell (Cell 7) holding 0.1 mgPt/cm2 was studied. The membrane with catalyst layers was studied in dry condition as taken from the forced convection oven. The surface of the catalyst layer is given in Figure 5.8 shows that the catalyst layer had cracks of approximately 1-3 pm width. The cracks may originate either from electrode material contraction during drying, or from the membrane expansion due to partly rehumidification when taken out of the drying oven (125°C). Numerous pores in the sub micro meter range indicate a high porosity. A cross- section of the membrane with catalyst layers was obtained by breaking the M&E assembly in the dry state, as taken from the drying oven. The membrane and catalyst

Figure 5.8. SEM-micrograph of the surface of the catalyst layer. Cracks of 1 -3pm are seen. Numerous pores in the sub micro-meter range indicate large porosity. 70 Chapter 5. Electrode Preparation And Evaluation

Figure 5.9. SEM-micrograph of the cross-section of the single cell. The catalyst layers (arrowed) are seen at each side of the Nation membrane. layers are distinguishable, as seen from Figure 5.9. The membrane and electrode thicknesses are around 180 and 7 pm, respectively. The contact between electrode and membrane looks intimate. To examine the platinum distribution in the catalyst layer, a SEM-micrograph was obtained in the Back-Scatter mode as shown in Figure 5.10. Heavy atoms appear as white spots in this mode. The picture shows that the platinum was distributed fairly evenly throughout the catalyst layer.

5.6.8 Advantages and limitations of the spray deposition technique The spray deposition technique is limited to thin electrodes when spraying the electrode directly onto the Nation membrane, using glycerol as solvent and drying in a forced convection oven at 125-135°C. Glycerol, like other alcohols, penetrates the membrane and makes it expand[46]. Thin layers of Pt/C were successfully applied to the membrane because drying is faster than the alcohol penetration. At high loadings the membrane expansion is substantial, giving electrodes of non-uniform thickness. The problem may be overcome by exposing the membrane to alcohol (to make it expand) prior to spray deposition. Such a technique was applied by Mosdale and Stevens[46]. Thicker electrode layers may be obtained by applying multiple 5.6. Results and Discussion 71

Figure 5.10. SEM -micrograph (in the Back Scatter mode) of the cross-section of a single cell holding thin film electrodes. The white dots are Pt particles, evenly distributed throughout the catalyst layer. electrocatalytic layers with intermediate drying. Due to thermal degradation of the proton conductor, the overall drying time should be minimized. Substitution of the glycerol by another solvent, either a more volatile or a less membrane-penetrating one, would make thicker uniform electrodes obtainable in one single layer. The thickness of the catalyst layer in commercial Prototech and E-TEK electrodes is about 100 pm [44]. SEM analysis at our laboratory confirmed this value. The thickness of the thin film electrodes of this work was estimated from SEM-micrographs to be 7±2 pm. The reaction zone thickness was estimated to be around 4 pm at high current densities[51]. Approximately the same density of platinum in the commercial and thin layer electrodes explains the similarity of the polarization curves in Figure 5.4. Uchida et al.[68,73] have developed a new preparation method for the electrocatalytic layer for SPEC in which the effect of different solvents were studied. This work will have implications on further development of the spray deposition technique. The technique of spraying the catalyst layer directly onto the membrane eliminates the need for hot pressing. This is advantageous for large scale production. 72 Chapter 5. Electrode Preparation And Evaluation

5.7 Summary and Conclusions

Statistical factorial design of the experiment was a useful tool for studying electrode performance. The results of this experimental study may be summarized as follows: Thin film catalyst layers were successfully applied directly onto the membrane by spray-deposition. SEM studies show intimate contact between catalyst layers and the membrane. An increase in polymer content from 15 to 25% in the catalyst layer gave better performance. Increased Pt-loading also gave improved cell performance. The drying temperature of the M&E assembly should be kept low to reduce degradation of the polymer. No severe permanent loss in membrane conductivity was encountered when drying at 125°C for 75 minutes. Thin film electrodes exhibited higher Pt- utilization than hot pressed commercial electrodes. Operation at high pressure and high temperature was advantageous in terms of cell performance. This was related to increased electrocatalytic activity of the catalyst and reduced resistance in the polymer phase. Membrane (Nation 117) resistance constitutes a large part of the total potential loss in the tested cells. The replacement of Nation 117 by thinner Nation membranes or Dow membrane will increase cell performance. 73

C hapter 6

Ion and Water Transport in Membranes

This part of the work consists of two articles published in Journal of Membrane Science. They deal with transport properties of membranes similar to those used in solid polymer fuel cells. The results of the studies are summarized and implications for solid polymer fuel cell preparation and operation are given.

6.1 Introduction

Detailed knowledge about the transport processes in membranes is necessary to understand and fully utilize the membrane properties. The aim of the first study was to demonstrate a new and accurate electro-motive-force (emf) method for determination of transference numbers in ion exchange membranes. The method was applied to study transference numbers for the cations H+, K+ and Na+ in a cation exchange membrane (Ionics CR61 AZL 389). The preferred transport of the smaller cation (IP) was demonstrated. The ratio between ion mobilities did not change with membrane composition, indicating the abscence of interactions between ions. In the second study, the divalent Sr2+-ion was introduced in order to obtain more information about the interaction between ions and between ions and the polymer membrane network. Water transference coefficients were obtained by streaming potential measurements. The results show higher water transference coefficients for the divalent Sr^-ion than for the monovalent K+ and Na+ ions. Further, water permeability was lower in the presence of the divalent ion. The latter result was interpreted as the result of a stronger interaction between the divalent ion and the polymer membrane. 74 Chapter 6. Ion And Water Transport In Membranes

6.2 Possible implications for solid polymer fuel cell preparation and operation

The Ionics membranes are similar to those applied in solid polymer fuel cells (the perfluorinated sulfonic acid type, e.g., Nation and DOW). The Ionics membranes are not perfluorinated, but sulfonic acid groups constitute the cation exchange sites. The Ionics and Nation 117 membranes have shown similar properties like water transference coefficients in proton form[ 19,74], This indicates that the water transport is related to the properties of the cation and to the ion exchange sites (sulfonic acid) involved, and to a lesser extent to the polymer membrane backbone. Insight gained from Ionics membrane studies are valuable for understanding fuel cell membrane transport processes. In SPFC operation only protons are desired in the polymer membrane. However, during preparation of the polymer, other ions may enter the structure and occupy the cation sites. Therefore, it is common to clean the membranes for organics and metallic impurities (Appendix F). From the second study it was seen that water transference coefficients were strongly influenced by the presence of divalent ions. Chapotot et al.[75] studied the Selemion membrane (Asahi Glass, Japan) equilibrated in a solution of HC1/Cu C12. The results show a large preference for the divalent ion in the membrane. They attribute the effect to the higher electrostatic attraction between the negatively charged cation exchange sites and the divalent ions. Changes in water transport properties due to the presence of other cations pin-points the need for proper cleaning of SPFC membranes. Generally, it is concluded that cations cross the ion exchange membrane at different rates and carry a different number of water molecules depending on the ion valency and size. The data reported in the articles are not directly relevant for the membranes used in solid polymer fuel cells. However, the methods demonstrated, i.e., the emf method and the streaming potential method, are general methods and can provide accurate ion and water transference coefficients, respectively. Such information about interactions between ion, water and membrane is valuable. The data may be used in models, which in turn can contribute to give a picture of the membrane structure and transport process mechanisms. 6.3. Membrane Transference Numbers From A New. ... (enclosed article) 75

Journal of Membrane Science, 74 (1992) 1-8 1 Elsevier Science Publishers B.V., Amsterdam

Membrane transference numbers from a new emf method

Magnar Ottey, Tormod Borland, Signe Kjelstrup Ratkje and Steffen Moller-Holst Division of Physical Chemistry, Norwegian Institute of Technology, University of Trondheim, N-7034 Trondheim-NTH (Norway) (Received February 10,1992; accepted in revised form June 9,1992)

Abstract

A new method for determination of the transference numbers of two ions in ion exchange membranes has been established. The method uses emf measurements in a cell with a membrane stack, and it avoids problems of concentration-polarization, diffusion and water transport. Stack thickness is about 10 mm. The method needs less time and is more precise than a corresponding Hittorf method. The method has been developed for a cation exchange membrane, with cation sites M_, in equilibrium with aqueous solutions of two electrolytes, first HC1 and KC1, next HC1 and NaCl. Transference numbers for both systems are reported with a precision of ±1%. The results imply that the ratios of the ionic mobilities, , for i=K+ and Na+ in the membrane are constant as the mole fractions vary.

Keywords: transference number; ion exchange membranes; emf measurements

Introduction change membranes when two salts are present. The method is developed for the cation ex ­ Rapid and precise methods to obtain mem­ change membrane CR61AZL 389 (from Ionics brane transference coefficients are important Inc., Watertown, MA) in equilibrium with so­ for characterization of membranes. A new the ­ lutions of HC1 and KC1. It is next applied to oretical description of transport in ion ex ­ solutions of HC1 and NaCl. change membranes has been developed in our laboratory over the last years (for a review, see Theory Ref. [1]). We are now seeking to implement the theory, and have recently reported a new General equations method for determination of water transfer ­ ence numbers and water permeabilities [ 2 ]. The following electrochemical cell is used: This work presents a new method for deter­ mination of transference numbers in ion ex- H2(g) IHCI(CHci,i)>AC1 (Cacu ) |c|

Correspondence to: Signe Kjelstrup Ratkje, Division of HC1 (CHci,n ),AC1 (CAci,n ) |H2(g) Physical Chemistry, Norwegian Institute of Technology, University of Trondheim, N-7034 Trondheim-NTH, The symbol A is used for the alkali metal, K or Norway. Na, and |c| denotes a cation exchange mem-

0376-7388/92/$05.00 © 1992 Elsevier Science Publishers B.V. All rights reserved. 76 Chapter 6. Ion And Water Transport In Membranes

2 M. Ottey et al/J. Membrane ScL 74 (1992) 1-8 brane. From irreversible thermodynamics the flux equations of cell (a) may be written [1,3]: (5b)

Jnci— -I'll Ppi —1-12 Ffh —Lis Pgz —Lu Prp (1) for i=1,2,3. A similar equation is valid for

«7aci = —Lzi Pfii —Lz2 Pth— Lss Pgs —£24 Vrp (2) transfer of H20. The frame of reference for all transference coefficients is the membrane [1 ]. «1h 20 = — L31 Vfii —Lss Pgs — LssVjis —Lsi^tf (3) The transference coefficient, tHci. gives the j= —7/4! Vfii —L42 Ffbs —L4S Vfiz — L44 Vrp (4) number of moles HC1 which are transferred from left to right in the cell, per mole electric All transport occurs in the x-direction, and F charge. The transference coefficient is related means d/dx. The gradient in electric potential to the transference number, the fraction of is Vrp—lim (Ay!Ax) for Ax-*0. The is mea­ electric charge transported by H+ through the sured with Pt|H2(g) electrodes and j is the membrane. When passing one mole electric electric current density. The units of Arp and j charge through cell (a), the following changes are J-faraday-1 and faraday-m-2-sec_1, re­ are observed: spectively. The unit faraday means one mole of Left-hand half cell: elementary electric charges (96487 C). The coefficients, Ly, are phenomenological coeffi­ 1 mole H+ — tH+ mole H+ — tA+ mole A+ = cients. The units used of Arp and j make the tA+ mole HC1—tA+ mole AC1 (6) coefficient matrix symmetric. The components of the solution, HC1, AC1 Right-hand half cell: and H20, are numbered 1,2 and 3, respectively. — 1 mole H+ + tH+ mole H+ + tA* mole A+ = Their chemical potential gradients are F/i,-. As long as the electrolyte solutions are dilute, the —tA+ mole HCl+tA+ mole AC1 (7) components of the membrane are HM, AM and From these mass balances the transference H20, where M~ denotes the cation sites in the coefficients are equal to: membrane. We assume that at any location in the membrane we can introduce an aqueous so­ tAci—tA+ (8a) lution at 1 atm in equilibrium with the mem­ brane composition. Integration across the tHci—— tA+ (8b) membrane can thus be carried out over the This means that tHci= — tAci- The value re­ aqueous solutions with known Vpi. One mem­ flects the transport properties of the membrane. brane component is chosen as the reference for In emf measurements we have j» 0. Com­ the fluxes (see Ref. [3] for further details). bined with eqn. (4) this gives:

The principles of the transference number — Vrp=yL P/iAC1+—^ P«H,,0 (9) determination By introducing eqn. (5) and the Onsager recip ­ rocal relations Ly=Lji into eqn. (9) we get: The definition of transference coefficients of the components HC1 and AC1 are: — F#>=tHclFFHCl+tAClFltAC] + tu 2o VflfiiO (10) We shall use solutions with the same ionic (6a) strength on both sides of the membrane. Thus Ppuso ~ 0. Then we can neglect the contribu- 6.3. Membrane Transference Numbers From A New. ... (enclosed article) 77

M. Ottey et ai./J. Membrane Set. 74 (1992) 1-8 3 tion to emf from the water transport. This leads tion between the cells. Half of the membranes to the following expression: are equilibrated with the reference solution, the d^>= — £Hcid|iHci—tACid^Aci (11) other half with the test solution. This gives an initially sharp concentration gradient. The and by using eqn. (8) we obtain: thickness of the stack shall prevent concentra ­ dq>=tA+ d/iHci— tA+ dfiAa tion changes in the half cells and disturbances (12) by concentration-polarization. =*a + d(/rHCI—Aaci ) The concentration gradient inside the stack or will vary during the experiment, but this vari­ ation does not affect the emf. This is because tA+ = d

Chemicals

Chemicals were analytical grade compounds from Merck, Darmstadt. The H2-gas (99.99% pure ) was from AGA Norgass AS, Norway. So­ lutions were made with an accuracy of ±0.1%.

Electrode and membrane preparations

Hydrogen electrodes were prepared accord­ ing to instructions given by Ives and Janz [4], The electrodes were stable for many weeks. Be­ fore and after each experiment, the bias poten ­ Fig. 1. The concentration cell. The cell consists of two glass flasks connected by a plexiglass membrane stack holder. tial of the electrodes was determined. If the bias The Ft electrodes are flushed with hydrogen gas. The cell was more than 0.01 mV, the electrodes were re- is placed in a thermostat. platinized. 78 Chapter 6. Ion And Water Transport In Membranes

4 M. Ottey et al./J. Membrane Sci. 74 (1992) 1-8

The membrane sheet .(CR61AZL 389, from 12 3 4 Ionics Inc., Watertown, MA) was cut into cir­ cles of 2 cm in diameter. All membrane pieces were cut from the same sheet, because previous experiments have shown variations in the properties of membranes from different sheets. Membrane thickness was 1.2 mm. The mem­ brane contains sulfonated groups in a concen­ tration of 1.6 kmol-m-3. The water content is about 48% of wet resin. The membranes were kept in a container of soaking solutions of HC1 and KC1 or of HC1 and NaCl. All solutions had Fig. 2. Details of the membrane stock holder. (1) The cell, a total salt concentration of 0.03 kmol-m-3. At (2) membranes, (3) O-rings, (4) threads for fastening the cell to the glass, (5) pressure-supporting plate, (6) grid in this concentration anions are not present in the plexiglass, (7) hollow cylinder in plexiglass, (8) threads for membrane. The electrolyte fractions of HC1 are press plate. given in Tables 1 and 2. The soaking solutions were replaced 10 times during a period of 8 cells. Hydrogen gas was supplied at a speed of weeks to ensure equilibrium. Then the solu­ 4-5 bubbles per second. tions with the membranes were kept in a water The emf was measured by a “617 Programm­ bath (25.0° C), and the solutions were replaced able Electrometer” from Keithley. The volt ­ another 3 times. This tedious procedure was meter was connected to an IBM Personal Com­ necessary to establish equilibrium between the puter PS/2-model 30 which recorded the emf membranes and the different solutions. regularly and stored the values for later analy ­ sis. Stable emf values (±0.02 mV) were usu­ ally obtained within 2 hr. Emf measurements Method development The concentration cell, illustrated in Fig. 1, is made from two glass flasks. The half-cells are The emf was first measured as a function of connected by the membrane stack holder (for stack size in order to find the minimum number details see Fig. 2). of membranes needed in the experiment. The The stack holder was first packed with ref­ time lapse before concentration-polarization of erence membranes. By reference membranes we solutions is observed, varies with stack thick ­ mean membranes in equilibrium with a 1:1 so­ ness. With a total of 4 membranes in the stack, lution of HC1/ACL The left-hand side of the polarization caused by diffusion was observed concentration cell was always used for this part after 2-3 hr as a reduction in emf (by/zV), while of the membrane stack. The rest of the mem­ 8 membranes prevented such disturbances for branes were equilibrated with the test solution. at least 5 hr. Eight membranes were used in the The solutions in the cell were thermostated to following experiments. The larger the differ­ 25.0°C±0.1°C before use. The stack was ence we have in chemical potential of the two pressed together to minimize the liquid layers solutions in the concentration cell, the larger is between the membranes. the diffusion rate. For this reason, an inter ­ The cell was thermostated at 25.0 ± 0.1 °C. mediate concentration was used for the refer­ Hydrogen electrodes, were put into the half ence solution. 6.3. Membrane Transference Numbers From A New. ... (enclosed article) 79

M. Ottey et alfJ. Membrane Set. 74(1992) 1-8 5

Different membrane packing techniques were TABLE 1 next investigated. Some membranes were wiped Transference numbers, tK+ and emf for the system HC1-KC1 off with tissues before packing, thereby remov­ for different solution compositions, x H+. Corresponding ing the liquid film between the membranes. membrane compositions, x Hm> chemical potential differ­ Others were taken directly from the storing so­ ences, Ahci -^kci . and electric potential differences, Ay, are lutions. No significant difference in the results also given were seen. The effect of different gas velocities XH+ *HM emf dtp ^HCrMKCl &k+ was then investigated. Experiments with low (mV) (kj-faraday- x ) (kJ-mol-1) gas velocity (1 bubble/sec) gave an emf value of 6.59 mV. A high velocity (more than 10 bub ­ 0.025 0.019 -55.28 -5.334 -9.052 0.899 0.050 0.039 -39.70 -3.831 -7.269 0.819 bles/sec) gave a lower value for emf (6.50 mV), 0.100 0.079 -24.80 —2.393 -5.417 0.685 probably due to cooling of the electrode. Emf is 0.150 0.120 -17.33 -1.672 -4.271 0.586 proportional to the temperature, and a reduc­ 0.200 0.161 -12.49 -1.205 -3.407 0.506 tion of 3°C will reduce the potential by 70 /zV. 0.250 0.204 -9.13 -0.881 -2.694 0.438 0.300 0.248 -6.40 -0.618 -2.071 0.378 For this reason we used the same gas velocity 0.350 0.293 -4.28 -0.413 -1.505 0.331 in all experiments (4-5 bubbles), and a tem­ 0.400 0.339 -2.66 -0.257 -0.976 0.287 perature control of the gas. 0.450 0.386 -1.18 -0.114 -0.468 0.249 Altogether 11 experiments were performed to 0.500 0.435 0.00 0.000 0.029 0.216 0.550 0.485 1.00 0.097 0.526 0.186 test reproducibility. In all these cases reprodu ­ 0.600 0.536 1.94 0.187 1.034 0.158 cibility was better than ±1%. 0.650 0.588 2.71 0.262 1.563 0.130 0.700 0.642 3.40 0.328 2.129 0.109 Results and calculations 0.750 0.698 4.04 0.390 2.752 0.0886 0.800 0.755 4.64 0.448 3.465 0.0691 0.850 0.813 5.18 0.500 4.328 0.0502 Twenty-one different test solutions were in ­ 0.900 0.874 5.61 0.541 5.475 0.0321 vestigated with HC1-KC1 solutions. Eight so­ 0.950 0.936 6.09 0.588 7.327 0.0158 lutions were examined with the system HC1- 0.975 0.968 6.30 0.608 9.110 0.0094 NaCl. Experimental results for the two sys­ tems are given in Tables 1 and 2. The difference in chemical potential of HC1 and AC1 where A = K, Na is: TABLE2

/ chemical potential differences, cr /tNaCi, and electric potential differences, Ay, arc also given

XH+ *HM emf dtp #HC|-#N*C1 Activity coefficients, y;, for the electrolyte so­ (mV) (kJ-faraday-1) (kJ-mol-1) lutions are given by Zemaitis et al. [5]. For 0.0625 0.086 -28.80 -2.779 -6.690 0.699 HCl-KCl-water solutions, we have 0.1250 0.167 -16.37 -1.580 -4.801 0.557 0.2500 0.319 -6.48 -0.625 -2.702 0.347 lny Hci= — 0.32224* 0.3648 ttihci (15a) 0.3750 0.457 -2.35 -0.226 -1.247 0.220 0.5000 0.583 0.02 0.001 0.018 0.138 lnyKci= -0.33424-0.3842 mHci (15b) 0.6250 0.700 1.48 0.143 1.283 0.0852 For solutions of HCl-NaCl the corresponding 0.7500 0.808 2.45 0.236 2.739 0.0550 0.907 3.39 0.327 expressions are: 0.8750 4.838 0.0263 80 Chapter 6. Ion And Water Transport In Membranes

6 M. Ottey et al./J. Membrane Sci. 74 (1992) 1-8

lny Hci= -0.3164+0.1698 mHCi (16a) behavior in the membrane. Results are given in Tables 1 and 2. InyNaci ——0.3260+0.3134 mHci (16b) where 7n Hci is the molality of HC1 in the solu­ tion. The error in fiHCi—jUAci is estimated to ±4 Transference numbers J. Emf is plotted (in units of kJ-faraday-1) as a function of Ahci —/*kci (in units of kJ-mol-1) in Fig. 3. The derivative of the curve in Fig. 3 gives the The following ion exchange equilibria have transference number of K+ according to eqn. been investigated by Skrede and Ratkje [6]: (13). In order to find the derivatives, the re­ sults in Fig. 3 were fitted to a polynomial. This NaCl(_, +KM+KCl(aq)+NaM (17) was done by dividing the set of data into 3 in ­ HCl(aq)+NaM^NaCl(aq) +HM (18) tervals overlapping each other. Each interval was fitted to a polynomial of degree 4. The de­ Equilibrium constants are: rivatives of the polynomials obtained in this K “nsm Okc , 0_54 matter, were used to find the transference (19) °KM°NaCl numbers. The transference number of Na+ was obtained in a similar manner. K2 °HM°NaC1 0.72 (20) Calculated results for tH+ from Tables 1 and °NaM°HCl 2 are plotted against the mole fraction of H + We may combine these expressions to give the in the membrane in Fig. 4. The error in the cal­ equilibrium constant for the KM-HM system: culated transference number of H+, tH+, is less than ± 0.005 in the presence of K+ and slightly if3=g1g2=g-qaHM (21) higher in the presence of Na+, due to fewer ex ­ aKMaHCl perimental points. Membrane compositions for the system HM- KM can be derived from Ks by assuming ideal

a -2

mol 1 Fig. 4. The transference number of H+ as a function of membrane composition, zHm, for the systems K* /H* and Fig. 3. The emf as a function of (ftra—Akci ) in cell (a). Na+/H+. 6.3. Membrane Transference Numbers From A New. ... (enclosed article) 81

M. Ottey et aL/J. Membrane Sci. 74 (1992) 1-8 7

Discussion the membrane is a function of membrane com­ position through: Transference numbers in membranes can be obtained by two methods, the Hittorf method uH+ =Uh +(1—&%m) (22a) and the emf method [7]. Hittorf measure­ u k +=wk +(1—&*hm ) (22b) ments were performed in cells with membrane stacks by Kontturi et al. [8,9]. Electric current where Uh + is the mobility of H + in a pure HM- was passed through the stack for 15-25 hr, and membrane, and Uk + is the mobility of K + in a consecutive changes in composition due to pure KM-membrane. The mobilities in the two- electric current and diffusion were analyzed. component membrane are «H+ and uK+ andxai The analysis time, which included membrane is the fraction of cationic sites held by compo ­ re-equilibration in HCl-solutions, was several nent i in the membrane. weeks. By a suitable choice of current density The physical idea behind this model is that a and number of membranes all composition reduction in the mobility of an ion is propor ­ changes were kept within the stack, and polar ­ tional to the number of neighbour ions of the ization and diffusion were controlled also in this other kind. The parameter k is a constant cho­ method. Composition analysis of the separate sen to give the best fit of the curve to the ex ­ membranes which make up the stack were, perimental data. however, time-consuming and high precision in The transference numbers were fitted to this the results were not reported [8,9]. model, using a standard computer program. The The present emf method uses a membrane calculations gave: fesslO-19 and uH+/ stack combined with a specific set of electrodes. uK+ =4.9 ±0.1 for HM-KM, and ferelO-3 and Through this method problems of diffusion and u h + /= 4.2 ± 0.2 for HM-NaM. We thus polarization phenomena are avoided. The time conclude that the interaction constant k is close taken to perform the experiments is shorter to zero in both cases. For comparison the cor­ than for the Hittorf experiment, and results are responding mobility ratios in infinite dilute obtained with a reproducibility of ±1%. In ad­ aqueous solutions are 4.8 and 7.0, respectively. dition we obtain the transference number as a With kxO we expect uNa+ /u K+ = 1.2 for the continuous function of composition. Previ­ membrane. This will be checked in future ex ­ ously, average values for a composition interval periments. The result for k is somewhat sur­ have been determined [7,8]. prising, considering the fact that this mem­ Problems with the procedure may be con ­ brane has a concentration of cationic sites of nected to the liquid layer between membranes. 1.6 kmol-m-3. Forssell et al. [9] showed that the contribution The values for the mobility ratios are reason ­ from these layers is negligible. Our investiga ­ able, given some recent results on water trans ­ tions confirm this: different packing tech ­ ference numbers in the same membranes [2], niques had no significant effect on emf. The They show a linear relationship between the water activity can be controlled experimen ­ water transference number and the transfer ­ tally. We therefore conclude that we have im­ ence number of H+. This indicates that each proved the method for transference number ion carries the same amount of water through determination. the membrane, independent of other ions pres ­ The actual data obtained can be interpreted ent. This is the same as saying that the inter ­ in terms of a model proposed by Forland et al. actions between the different cations moving [3], They suggested that the ionic mobility in through the membrane is very small. 82 Chapter 6. Ion And Water Transport In Membranes

8 M. Ottey et cd./J. Membrane Sci. 74 (1992) 1-8

Conclusions T temperature (K) Ui mobility of component i in membrane A new method for determination of transfer ­ (m2-sec_1-V_1) ence numbers in ion exchange membranes has 4* the mobility of H + in a pure HM-mem- been developed. Problems with diffusion caus­ brane (m2-sec-1-V_1) ing concentration-polarization are avoided. x coordinate axis (m) Furthermore the effect on the emf due to trans ­ x iM equivalent fraction of ion in the fer of water is eliminated. The method has been membrane established with a set of experimental results X activity coefficient for component i, on cation exchange membranes in equilibrium concentration basis with aqueous HC1/KC1 and HCl/NaCl solu­ tions. The transference numbers are given with Greek letters less than 1% uncertainty. It was found that the ratio of the ionic mo­ Hi chemical potential of component i (kJ- bilities in the membrane did not change as the mol-1) composition of the same ions in the membrane A(p electric potential (kJ-faraday-1) changed. The mobility ratios were uH+/ uK+ =4.9 ±0.1 and uH+ /«Na+ =4.2 ±0.2. References

Acknowledgement 1 K.S. Ferland, T. Ferland and S.K. Ratkje, Transport processes in electrolytes and membranes, in: S. Sien- iutycz and P. Salamon (Eds.), Advances in Nonequi ­ SINTEF is acknowledged for a grant to Stef­ librium Thermodynamics, VoL 6, Taylor and Francis, fen Moller-Holst. Dr. ing. Rune Dahl made the New York, NY, 1992, pp. 340-385. computer program which records the emf 2 T. Okada, S.K. Ratkje and H. Hancke-Olsen, Water Transport in Cation Exchange Membranes, J. Mem­ measurements. brane Sci., 66 (1992) 179-192. 3 K.S. Ferland, T. Ferland and S.K. Ratkje, Irreversible List of symbols Thermodynamics. Theory and Applications, John Wiley, Chichester, 1988. 4 D.J.G. Ives and G.-J. Janz, Reference Electrodes, Ac­ a; activity of component i (kmol-m~3) ademic Press Inc, New York, NY, 1961. C; concentration of component i (kmol- 5 J.F. Zemaitis Jr., D.M. Clark, M. Ratal and N.C. m~s) Scrivner, Handbook of Aqueous Electrolyte Ther ­ F Faraday’s constant (96487 C-eqv-1) modynamics: Theory & Application, Design Institute for Physical Property Data, sponsored by American j electric current density (faraday m ~ 2- Institute of Chemical Engineers, New York, NY, 1986. sec-1) 6 G.M. Skrede and S.K. Ratkje, Cation-exchange mem­ Ji flux of component i (mol-m-2-sec-1) branes as solid solutions with Na"*"/H + and K+/H+, k interaction parameter for ions in the Z. Phys. Chem., NeueFolge, 155 (1987) 211. membrane 7 N. Laksminarayanaiah, Transport Phenomena in Membranes, Academic Press, New York, NY, 1969, p. K thermodynamic equilibrium constant 232. Ly phenomenological coefficient related to 8 K. Kontturi, A. Ekman and P. Forssell, A method for charge and mass transfer, [mol 2/(J-m- determination of transport numbers in ion exchange sec)-1] membranes, Acta Chem. Scand., A39 (1985) 273. 9 P. Forssell, K. Kontturi and A. Ekman, The effect of m,i molality of component i (mol-kg-1) an intermediate layer on the determination of trans ­ R gas constant (8.314 J-K- ^mol-1) port numbers in ion exchange membranes, Acta Chem. ti transference coefficient of component i Scand., A39 (1985 ) 279. 6.4. Water And Ion Transport In The Cation Exchange ..... (enclosed article) 83

journal of MEMBRANE SCIENCE

ELSEVIER Journal of Membrane Science 111 (1996) 159-167

Water and ion transport in the cation exchange membrane systems NaCl-SrCl 2 and KCl-SrCl2

T. Okada a ’*, S. Kjelstrup-Ratkje b, S. M0ller-Holst b, L.O. Jerdal a , K. Friestad a , G. Xie a , R. Holmen w

‘ National Institute of Materials and Chemical Research, MIT1, Higashi I-I. Tsukuha, lharaki SOS. Japan b Division of Physical Chemistry. Norwegian Institute of Technology, N-70S4 Trondheim-NTH. Norway

Received 20 January 1995; revised 1 May 1995; accepted 10 June 1995

Abstract

Water transference coefficients, water permeabilities and ionic transference numbers were investigated for the cation exchange membrane CR61 AZL389. The membranes were equilibrated with aqueous solutions of NaCI (KCI) and SrCl2 of various compositions. No chloride ions were detected in the membranes. The water transference coefficients and the water permeabilities were determined from streaming potential measurements, whereas the ionic transference numbers were determined by a new emf method. When the membrane was in the pure KM, NaM or SrM2 form, the water transference coefficients were close to 11, 14 and 20, respectively. The water permeability is about two times larger in NaM than in SrM2. The water transference coefficient is not a linear function of the transference number of Na+ (K+) in the membrane, as found earlier. The results are explained by a variation in the membrane polymer cross-binding caused by Sr2+-binding to ionic membrane site(s). The transport behaviors in the mixture of NaM(KM) and SrM2 can all be understood by the Sr2+ ion ’s strong interaction with the polymer membrane.

Keywords: Water transport; Ion transport; Cation exchange membrane; NaCl-SrCl2; KCl-SrCI,

1. Introduction model case [1], For this system we found that the number of water molecules accompanying each We have recently reported a new analytical proce ­ cation through the membrane could be taken as dure for the accurate evaluation of water transference constant. More complicated monovalent-divalent coefficients from streaming potential measurements cation systems have not yet been investigated. In this [1]. A cation exchange membrane equilibrated with paper, we report results for water transference coeffi­ an aqueous solution of HCI and KCI was used as a cients in the same cation exchange membranes equi­ librated with mixtures of monovalent-divalent cation systems, using the technique previously reported. " Corresponding author. The same technique also provide information on 1 Present address: Norzink AS, N-5750 Odda, Norway. water permeabilities. We also report ionic transfer-

Elsevier Science B.V. SSDl 0376-7388(95)00184-0 84 Chapter 6. Ion And Water Transport In Membranes

160 7. Okada et al./Journal of Membrane Science 111 (1996) 159-167 ence numbers of the ions using another newly devel- A+ is either Na+ or K+, SrCl2 and water, may be oped technique [2]. The combined set of data can be written [6]: used to gain insight into structure and behavior of 4AC|= ~ ^12^2 ~ *- 13^3 ~ L\*v4 > (0 binary cationic mixtures of different valencies in the membrane. 2SrCU = ~L2iVFl - L-22Vb-2 ~ 2-23^3 “ L2t^4> (2) The polymer electrolyte fuel cell, which uses a A, = -f-31%1 L}2V h2 - 3 - L^V4> (3) cation exchange membrane, is predicted to be of practical importance for zero-emission vehicles [3]. j= - Li2Vp2 - L43F/i3 - LiAV4> (4) One major difficulty in the operation of this cell, is The coefficients, are phenomenological coef­ the membrane water management [4], The water ficients and we have LVi = Ly (Onsager reciprocal content at a given location is a function of the relations). The components of the solution, AC1, electro-osmotic flux of water, the diffusion of water SrCl2 and H20 are numbered 1, 2 and 3 respec ­ and water supply to the membrane from the gas tively. Their chemical potential gradients are Vp. { phase. Quantitative information on the electro- (z = 1, 2, 3). All transport occurs in the ^-direction, osmotic water transport, and water permeabilities, is so V means d/d x. The gradient in electric potential essential to control the operation of this fuel cell [4], is V = lim (A/Ax) for Ax-» 0. The A is de­ Fuel cell membranes may contain contaminations of fined for AglAgCl electrodes and j is the electric divalent cations from the production process or from current density. The dimensions of A and j are J the surroundings. The effect of these ions on mem­ F~1 and F m~2 s~!, respectively. The unit Fmeans brane operation is unknown. It may be expected that one mole of elementary electric charges (or 96 500 C the conductivity as well as the electro-osmotic water mol" 1). When a pressure difference is maintained transfer can change by introduction of such ions into across the membrane the observed electric potential the membrane. Arf>abs is related to A in Eqs. (1)—(4) by Information on water transport in the membrane • has been obtained for ion exchange membranes with Acbi = A- AVclAp (5) monovalent cations and a varying water content [5], where AVci — VA| — VA?CI is the volume change due but membranes with a mixture of cations have not to the electrode reaction, and VA?, VAsC1 are molar been studied systematically. In this work, the cation volumes of Ag and AgCl. We have A0obs = EF, exchange membranes with monovalent/divalent where E is given in volts. cation systems are investigated in order to establish a The composition anywhere in the membrane has a method for measuring and analyzing ion and water corresponding equilibrium solution with a given AC1, transport in such systems, and to compare the result SrCl2 and H20 composition. This means that one with previous results obtained for membrane component can be chosen as the reference monovalent/monovalent cation systems. Two kinds for the fluxes without changing Eqs. (1)—(4) (see [6] of systems KM/SrM2 and NaM/SrM2, where M~ for further details). Also, integration across the mem­ denotes a cation site in the membrane, are studied. brane can be carried out over the solutions in equi­ The results from this work may contribute to the librium with the membrane, over V/j., (the Scatchard systematic knowledge which gives a better back ­ assumption). The chemical potential gradient of ground for fuel cell operation. component i in an isothermal system; fyi = fyi( c) + vyp (6) 2. Irreversible thermodynamic theory has two parts, the concentration dependent part, Vp.f.c\ and the pressure dependent part, VtVp, where The theoretical background for the experimental p is the hydrostatic pressure and V; is the partial procedure has been reported previously [1,2,6], Here molar volume. For the strong electrolyte, SrCl2, we only the main points will be repeated. The flux have; equations of an electrochemical cell holding a mem­ M2(c) = /4(c) +mn(y S(Cl2cSr=.c£,-) (7) brane which allows transport of the salts AC1, where 6.4. Water And Ion Transport In The Cation Exchange ..... (enclosed article) 85

T. Okada et al./Journal of Membrane Science 111 (1996) 159-167 161

/i|(c) is the chemical potential of the standard state, A+ and Sr2*, respectively. The transference number ySrC,2 is the activity coefficient of SrCl2, cSri., ccr of the cation A+ in the membrane is then: are concentrations of the cation and anion, respec ­ V(m) = [2(d<#>/dju2) + l] *] (11) tively. In practice, the gradient between the half cells of cell (a) can be maintained by using a stack of 2.1. Transference numbers from emf measurements membranes. The composition on the left side of the cell can be kept constant, while the composition on The equations for determination of transference the right side, the test solution, is varied. The emf is numbers in the membrane with monovalent/divalent measured for different test solutions. The change in cation systems are presented here. The transference ju,2 between two separate experiments is then equal coefficient, r,, of AC1, SrCl2, or H20, is defined to the change in p2 in the test solution alone. By from the flux equations by /, = Sc_0 = Lit/LM. plotting the emf as a function of p2> we can find the The transference coefficients for AC1 and SrCl2 are derivative (d<£/dju.2) of Eq. (11), and thus the trans ­ related to the transference numbers of the cations ference number ?A.(m) as a function of composition [2,6]. For the AglAgCl electrodes we have: in the water solution.

'aci = 'a * andfsrci, = g'sr'' (8 ) 2.2. Streaming potential measurements

In the solution rA-+ rSr2. + rcr= 1, so that rAC, + The streaming potential measurements are per ­ 2fSra, < 1. We have chosen experimental conditions formed with the isothermal cell, applying a pressure so that the membrane is cation selective, rA.+ rSr2. difference over the membrane: = 1. The membrane transference numbers will be Ag(s) |AgCl(s) |AC1( x, ) ,SrCl2( x 2 ) ,p| c| denoted rA.(m) and rSrz.(m) from now on. Transfer ­ ence numbers of the ions can be obtained from cell X AC1( x, ) ,Sr€l 2(x 2),p + Ap|AgCI( s) |Ag( s) (a): (b)

Ag( s)|AgCl( s) |AC1( x ,,,) ,SiCl2(x 2,) |c I The streaming potential, defined by EF/Ap — AobJAp, is obtained by integrating Eq. (4) for X AC1( x, jj ) ,SrCl2( x 2,„ )|AgCl(s)|Ag(s) (a) Vp, = V; Vp, i.e. with identical solutions on both where |c| denotes the cation exchange membrane sides of the membrane. By introducing Eq. (5) we CR61 AZL 389 from Ionics, and x, and x 2 are obtain: equivalent fractions of the salts. For an infinitesimal EF change in emf in an isopiestic cell (d/r3 = 0), we ~AP 'SW-AVt (12) have from Eqs. (4) and (8) for j = 0: When the transference numbers and the partial 1 molar volumes of components are known, and the dobs = d<£ = -rA.(m)d^,, --rSr2-(m)d/i2 (9) streaming potential has been determined from the extrapolated emf (see below), the transference coeffi­ We may integrate this equation across solutions in cient for water, rw, can be calculated from Eq. (12). equilibrium with the membrane at each location, and The solution composition in the two half cells are take advantage of known thermodynamic data for the identical only in the start of the experiment, at time water solution. The Gibbs-Duhem ’s equation for the equal to zero. Application of a pressure difference water phase with the condition dju3 = 0 is: leads to a volume flux across the membrane. The volume flux consists mainly of water (the current -*i dMi + = 0 (10) density j is negligible), so a dilution of the salt solution occurs on the right side and a corresponding where x, and x 2 are equivalent fractions of the ions concentration occurs on the left side, if the pressure 86 Chapter 6. Ion And Water Transport In Membranes

162 Okada et a!./ Journal of Membrane Science 111 (1996) 159-167 is higher on the left side (Ap< 0). This concentra ­ membrane are: specific resistance 250 ohm cm, ion tion polarization can be corrected for as described in exchange capacity 2.6 mequiv/dry g resin, water detail in our previous work [1], We obtain content 48% of wet resin, thickness 1.2 mm. Chemi ­ cals were analytical grade from Wako Pure Chemical tM- AV^Ap- Ait (13) Industries, Tokyo, Kanto Chemical Co., Tokyo, or Merck, Darmstadt, Germany. Here A is a function of Ap, and remains constant Circular pieces of membrane were equilibrated with time if Ap is constant The emf can be plotted with aqueous solutions of KCl-SrCl2 or NaCl-SrCl2 as a function of Vr, and the time zero value is with the following equivalent fractions of alkali obtained by extrapolation. It is the irreversible char ­ metal: 0, 0.158, 0.333, 0.529, 0.75, 0.871, 0.934, acter of die phenomena which predicts this plot, and 0.967, 0.985, 0.99 and 1. The solution concentration A is directly proportional to the volume flux, Jv. The was adjusted so that the total concentration of ions time variation of the emf with an applied pressure, was 0.06 kmol m-3. This means that the concentra ­ can be used to find the water permeability of the tion was 0.02 kmol m-3 with pure SrCl2, and 0.03 membrane for the condition that (1) the pressure kmol m-3 with pure KC1. Solution mixtures of difference does not cause any salt transport in the NaCl-SrCl2 with chloride concentration of 0.01 to membrane, (2) the diffusion constants of the salts in 0.0133 kmol m-3 were also used. Membranes were the solution are known and (3) the coupling between soaked in their equilibrium solutions for 4-5 months. transports of salts in the solution is so weak that it During the equilibration period, solutions were re­ can be ignored. These conditions are reliable for placed several times. Insufficient equilibration of the solutions of one salt. The water permeability, Lp , membranes resulted in errors both in the water trans ­ with pure solutions on both sides of the membrane is ference coefficients and ionic transference numbers. then Five SrM2-membranes, equilibrated in 1.88 10"2 4 A kmol m-3 SrCl2 for several weeks, were put into (14) 0.1 kmol m~3 * *NaN0 * * 3 to check for presence of 'p Ap f Ap chloride ions. No trace of chloride was detected by where / is given below. For AC1 and SrCl2 solu­ ion chromatography after one week. tions we have, using the similar procedure as re­ The water content of the membranes was obtained ported before in a monovalent/monovalent system from the weight difference between the wet and the [1]: dry states of the membranes in the whole composi ­ tion range for KCl-SrCl2 and NaCl-SrCl2 systems. (15) The wet state membranes were weighed after they were removed from the solution and quickly wiped with filter paper. The dry state membranes were f" ~6m»-y5; weighed after drying in a vacuum oven at 1009 C for (16) 12 h. where DAC, and DSjC,2 are diffusion coefficients of 3.2. Apparatuses, procedures AC1 and SrCl2 in the solution, respectively. The apparatus for the determination of ionic trans ­ ference numbers and the streaming potentials were 3. Experimental described previously [1,2,5]. The potential between the AglAgCl electrodes was no more than 20 pCV in 3.1. Membrane preparations a HCl-solution. The test solutions were bubbled with nitrogen gas for 2-3 h prior to the experiment to The cation exchange membrane CR61 AZL389 protect the electrodes from C02 and 02 dissolved in from Ionics, Watertown, Massachusetts, was used as the solution. All measurements were carried out at our model membrane. Basic specifications of the 25°C. 6.4. Water And Ion Transport In The Cation Exchange ..... (enclosed article) 87

T. Okada et al./Journal of Membrane Science 111 (1996) 159-167 163

The emf of the concentration cell (a) was mea­ sured with the reference membrane which was equi­ librated with the solution at fixed equivalent fraction of x AC, = 0.529 (A+ = Na+ or K+). A steady state value was reached within 1-3 h. To obtain good streaming potential measurements, a pressure con ­ troller, type 250C Baratron from MKS Instruments, Massachusetts, with a pressure transducer and a con ­ trol valve, was installed in the nitrogen gas line. With cell (b), the emf was recorded (by using a PC) every second until steady state was obtained after 60-100 s.

4. Results Chemical potential of SrCl2ZK7 mole'1

4.1. Ionic transference numbers Fig. 2. The emf of cell (a) as a function of the chemical potential of SrCl2 in the test solution for the KCl-SrCl2 system, and the The emf of cell (a) is plotted against the equiva­ NaCl-StCI2 system. lent fraction of NaCl in the NaCl-SrCl2 solution in Fig. 1, and against the chemical potential of SrCl2 in Fig. 2. The corresponding results for KCI are also obtained from the slopes of Fig. 2. A polynomial of shown in these figures. Note that the reference mem­ degree 3 was used to find the derivative in the brane is selected for the solution composition x KC1 equation. It turned out that the value of the derivative or x NaC] = 0.529 in every measurement, so that the was not sensitive to the model chosen for salt activ­ curves in Fig. 1 pass through emf = 0 at this compo ­ ity coefficients. The transference numbers are shown sition. as a function of the equivalent fraction xKa or xSaCI Using Eq. (11), the transference numbers of Na+ in the solution in Fig. 3. and K+ as a function of solution composition were

W 5

0.2 0.4 0.6 0.8 1.0 Solution composition xKa , xNcCt Solution composition xxa , xXaa Fig. 3. Transference numbers of monovalent cations plotted versus Fig. 1. The emf of celt (a) as a function of equivalent fraction of equivalent fractions of monovalent salt in the solution for the KCI or NaCl in the lest solution. KC1-SiC12 system, and the NaCt-SiCI2 system. 88 Chapter 6. Ion And Water Transport In Membranes

164 T. Okadaelal./Journal of Membrane Science 111 (1996) 159-167

° 22

0.005 Solution concentration Cj^cc/kmole m"3

Fig. 4. Water transference coefficients in the membrane in equilib ­ rium with solutions of SrCl2 of different concentrations. Solution composition xKa , xNaCl

Fig. 5. Water transference coefficients as a function of the equilib ­ 42. Water transference coefficients rium solution composition for the KCl-SrCl2 system and the NaG-$rCl2 system. The dashed line represents the result for Typical plots of zero time-emf vs. applied pres ­ diluted solutions of NaC!-SrC!2 system with chloride concentra ­ sure difference have been shown before [1]. The tions of 0.01 to 0.0133 kmol m~3. water transference coefficients were calculated from Eq. (13) using such plots. The uncertainties in the streaming potentials, EF/Ap, and in the water trans ­ ference coefficients were 2% for both binary sys­ tems, but 10% for pure SrM2 when the solution concentration was less than 0.005 kmol m-3. Fig. 4 shows rw in SrM2 as a function of concen ­ tration of pure SrCl2 solution. Results for rw in binary mixtures were plotted vs. composition of the solution as solid lines in Fig. 5, and vs. the ionic transference numbers in Fig. 6. A special feature of the NaM-SrM2 system is the maximum in the curve in Figs. 5 and 6 around zNa,= 0.6 corresponding to an equivalent fraction x Na-= 0.93. To further test the significance of this observation, the measurements were repeated with the membrane equilibrated in more dilute solutions of NaCl-SrCl2. The total chloride ion concentration was then between 0.01 and 0.0133 kmol m~3. These are shown by the dashed line, and for solutions with Transference number z^(m) , ^.(m) 0.01 kmol m-3 total chloride concentration, as closed triangles in Fig. 5. The water transference coefficient Fig. 6. Water transference coefficients as a function of the trans ­ ference number of the monovalent cation in die membrane for the was shifted to higher values, in agreement with Fig. KC1-StC12 system and the NaCl-SrCl2 system. The dashed line 4, but a maximum was still regained. represents the result for diluted solutions of NaQ-SiCl2 system The water permeability was calculated for mem- with chloride concentrations of 0.01 to 0.0133 kmol m~3. 6.4. Water And Ion Transport In The Cation Exchange ..... (enclosed article) 89

T. Okada et al./Journal of Membrane Science 111 (1996) 159-167 165 branes equilibrated in 0.0150 kmol m-3 NaCl and in x ACI the water transference coefficient decreases 0.0075 kmol m* 3 SrCl2 from Eqs. (14>—(16) utiliz­ rapidly. The constant level can be explained by a ing the slope of the emf vs. square root of time membrane composition dominated by SrM2. In the curve. The results were LpNaC1 = 2.4 + 0.5 10* 14 HM-KM and HM-NaM systems [1,2], the water m2 s kg* 1 and LpS[CI2 = 6.9 ± 0.7 10* 14 m2 s transference coefficient was a linear function of the kg* 1. The water content in the membrane was mea­ ionic transference number. Such a relationship does sured to be in the range 39-43%, and did not change not apply here (see Fig. 6). A pressure build up in appreciably with the membrane composition, both in the membrane may then occur during electrbosmosis KCl-SrClj and NaCl-SrCl2 systems. [11]. The maximum observed in fw for NaM-SrM2 is the most intriguing result obtained in this investiga ­ 5. Discussion tion (Fig. 6).' McHardy et al. [13] found a similar maximum in the Zeo-Karb 315 membrane in equilib ­ 5.1. Ionic transference numbers rium with a solution of NaBr and SrBr2. One possible explanation of this result is that We have previously reported that the equilibrium chloride ions enter the membrane, when the relative constant for the exchange reaction SrCI2(aq) + 2KM amount of SrM2 increases. The membrane may then = 2KCl(aq) + SrM2 is K,h = 5.6, for a similar Ion ­ allow SrCl* to enter. We did not find significant ics type membrane (CR61 AZL386) [7]. Hie equilib ­ amounts of chloride in the membrane, however. rium constant for the reaction NaCl(aq) + KM = The presence of divalent metal ions in polymer KCl(aq) + NaM is Kth = 0.54 [8,9], in agreement membranes may increase the cross-binding of poly ­ with Kontturi and Ojala [10]. On this background we mers [14]. The water permeability was 2.4 10~14 m2 expected a preference for Sr2+ in both membrane s kg* 1 in NaM and 0.9 10* 14 m2 s kg* 1 in SrM2. systems, and a stronger preference for Sr2+ in the McHardy et al. [13] report a ratio of permeabilities of NaM-SrM2 system than in the KM-SrM2 system. more than 2. These data support an explanation of a The transference number of K+ in the membrane tighter network in SrM2 than in NaM. The increase is higher than that of Na+ for all external salt in the water activity of the solution and a parallel mixture compositions (Fig. 3). This is in agreement increase of the water activity in the membrane may with the equilibrium data and the mobility data [11], explain that rw for CSrCl! = 0.01 kmol m* 3 is larger which indicated that K+ will move more freely than than rw for CSlCI, = 0.02 kmol m~\ The maximum Na+ in a membrane with Sr2*. Kontturi et al. [12] may be masked in the KM-SrM2 system, because found a higher mobility of K* than of Na* in a the difference between the transference coefficients membrane with protons. for water in the monovalent and divalent cations are larger in this system. The amount of water in the 5.2. Water transference coefficients membrane did not change significantly in both KCl-SrCl2 and NaCl-SrCl2 systems, however. This The water transference coefficients in the mem­ altogether indicates that a tighter network in the branes equilibrated with single salt solutions agree SrM2, compared with other cationic forms, might be qualitatively with the values reported by others [5], connected with a reduction of polymer chain move ­ For the SrM2 membrane, the value of fw increases ment rather than polymer network shrinkage. to 23 as the solution concentration decreases to 0.005 kmol m~3 (Fig: 4). Due to unstable electrodes at low 5.3. Friction electricity salt concentrations, an increase beyond this number cannot be concluded from Fig. 4. The values for tv The water transference coefficient rw in a cation in NaM and KM are 14.3 and 10.7. exchange membrane has been interpreted as the In the membranes consisting of the mixture of number of water molecules transported by the ions KM (NaM) and SrM2, the change in fw with x ACI is through the membrane [Ij. NMR studies of Na+ ions small for x AC, < 0.93 (Fig. 5). For larger values of in a Nafion 117 membrane show, however, that the 90 Chapter 6. Ion And Water Transport In Membranes

166 T. Okada et al./Journal of Membrane Science 111 (1996) 159-167 value of the water transference coefficient was dif­ A diffusion coefficient of i (m2 ferent from the number of nearest neighbors of Na+ s"1) [15]. The streaming potential experiment may thus be E measured emf (volts) seen from another perspective. F Faraday constant (96487 C In this experiment, the water is forced through equiv"1) membrane pores by a pressure gradient. Ionic move ­ f factor correlating the water per­ ment is required to create electric work. A larger meability Lp and A (J sl/2 momentum is required to move an ion which is mol" ' m"') strongly bound than one which is loosely bound. j electric current density (A Then on the average relatively more water molecules cm"2) must pass, to take a strongly bound ion from one site Aci flux of the salt AC1 (mol m~2 to the next, thus the relative magnitude of the coeffi­ s"1) cients rw(SrM2) > rw(NaM) > rw(KM) [16]. 7W water flux (mol m~2 s"') volume flux (ms"1) h phenomenological coefficients 6. Conclusions 4 water permeability (m4 J"1 s"') p hydrostatic pressure (J m~3) R gas constant (8.314 J K"1 The water transport properties reported here allow mol" 1) us to conclude that Sr2+ interacts strongly with the T temperature (K) polymer membrane which contains sulfonic groups. transference coefficient of com­ It is likely that Sr2+ ion binds strongly to one site, h ponent i but also that it promotes polymer cross-binding and fj-Gn) transference number of ionic that this affects the water transport In a membrane species i+ in the membrane with two cations present, the water flow may there ­ molar volume of component i fore vary with composition across the membrane. (m3 mol" 1) Such knowledge may be of practical importance, not volume change due to the elec­ only in separation technology, but also in fuel cell trode reaction (m3 mol" 1) operation. equivalent fraction of ionic These results were obtained by application of new species i+ in the solution emf methods for determination of water and ionic activity coefficient of i transference coefficients. The usefulness of the meth ­ % chemical potential gradient of i ods were thus further demonstrated. Vh across the membrane (J m"1) In the polymer electrolyte fuel cell, a small water Tk(c) concentration dependent part of transference coefficient, but a high water permeabil ­ the chemical potential gradient ity may be a preferred combination. The streaming (J m"1) potential method may be particularly useful to find V electrical potential gradient out more about these phenomena, as the accuracy of across the membrane (J fara- this method is higher, than that given, for example, day"1 m"1) by electroosmotic experiments.

7. List of symbols Acknowledgements A slope in the emf vs. square root of time curve (J mol-1 s~1/2) Financial support by the Norwegian Institute of C; concentration of component i Technology made possible the stay of L.OJ. and (kmol m-3) K.F. in Japan. 6.4. Water And Ion Transport In The Cation Exchange ..... (enclosed article) 91

T. Okada el al. / Journal of Membrane Science 111 (1996) 159-167 167

References [9] K.S. Forland, T. Okada and S.K. Ratkje, Molten salt regular mixture theory applied to ion exchange membranes, J. Elec­ trohem. Soc., 140 (1993) 634-637. Cl] T. Okada, S.K. Ratkje and H. Hanche-Olsen, Water transport [10] T. Ojala and K. Kontturi, Transport properties of a cation in cation exchange membranes, J. Membrane Sci., 66 (1992) exchange membranes with sodium and potassium as counte ­ 179-192. rions, Acta Chem. Scand., 43 (1989) 340-344. [2] M. Ottoy, T. Forland, S.K. Ratkje and S. Meller-Holst, [11] S.K. Ratkje, T. Holt and M.G. Skrede, Cation membrane Membrane transference numbers from a new emf method, J. transport: Evidence for local validity of Nemst-Planck equa­ Membrane Sci., 74(1992) l-S. tions, Ber. Bunsenges. Physik. Chem., 92 (1988) 825-832. [3] S. Srinivasan, DJ. Manko, H. Koch, M.A. Enayetullah and [12] K. Kontturi, A. Ekman and P. Forsell, A method for determi­ AJ. Appleby, Recent advances in solid polymer electrolyte nation of transport numbers in ion exchange membranes, fuel cell technology with low platinum loading electrodes, J. Acta Chem. Scand., A39 (1985) 273. Power Sources, 29 (1990) 367. [13] WJ. McHardy, P. Meares, A.H. Sutton and J.F. Thain, [4] T. Nguyen and RJE. White, A water and heat management Electrical transport phenomena in a cation-exchange mem­ model for proton-exchange-membrane fuel cells, J. Elec- brane II. Conductance and electroosmosis, J. Colloid Inter ­ trochem. Soc., 140 (1993) 2178-2186. face Sci., 29(1969) 116-128. [5] P. Trivijitkasem and T. Ostvold, Water transport in ion [14] S.R. Caplan and K. Sollner, The influence of the characteris ­ exchange membranes, Elecirochim. Acta, 25 (1980) 171-178. tics of the activating polyelectrolyte in the preparation and on [6] K.S. Forland, T. Ferland and S.K. Ratkje, Irreversible Ther ­ the properties of interpolymer ion exchange membranes, J. modynamics. Theory and Applications, Wiley, Chichester, Colloid Interface Set., 46 (1974) 77-84. 1988. [15] R.A. Komorski and K.A. Mauri tz, A sodium-23 nuclear [7] T. Holt, T. Forland and S.K. Ratkje, Cation exchange mem­ magnetic resonance study of ionic mobility and contact ion branes as solid solutions, J. Membrane Sci., 25 (1985) 133- pairing in a perfluorosulfonate ionomer, J. Am. Chem. Soc., 151. 100 0 978) 7487-7489. [8] M.G. Skrede. S.K. Ratkje, Cation-exchange membranes as [16] E. Brendler, S.K. Ratkje and H.G. Hertz, Streaming poten ­ solid solutions with Na+/H+ and Na+/K+, Z. Phys. Chem. tials of nuclepore membranes by the electric work method, N. F., 155(1987)211-222. Elecirochim. Acta. 41 (1995) 169-176. 92 C hapter 6. Ion And Water Transport In Membranes 93

C hapter 7

C oncluding Remarks

The conclusions of the present work are summarized and perspectives of the findings for further SPFC development are indicated.

This work has shown that reasonable single SPFC performance was attainable using low Pt-loading electrodes. Cell tests were accomplished by constructing and applying a test-station, documented in Chapter 4. The test-station provides cell tests within a wide range of operating conditions (i.e., 40-150°C and 1-lObar). The cell tests were automated by applying computer controlled data-acquisition. Suggestions for further improvements were given. Reproducibility of the results was studied and found to be adequate for single cell performance comparison purposes.

Equipment for preparation and assembling of single solid polymer fuel cells was designed and built (see Chapter 4). A new technique of spraying the catalyst layer directly onto the membrane was successfully demonstrated. Low platinum loading electrodes (O.lmgPt/cm2) prepared by the new technique exhibited high degree of catalyst utilization. Single cells holding these electrodes showed performance comparable to state-of-the-art SPFCs (Chapter 5).

Single cell performance was put into a perspective in Chapter 3, where the potential losses were related to irreversibilities by application of second law efficiency analysis to the Solid Oxide Fuel Cell.

The water management in membranes was addressed in Chapter 6 for a model system. The results gave insight into membrane transport properties and have bearings on fuel cell preparation and performance. 94 Chapter 7. Concluding Remarks

The preparation of single SPFCs applying the spray deposition technique is expected to be of commercial interest, because it involves few steps and techniques that are adequate for larger scale production. The life time of the cells, holding electrodes prepared by the spray deposition technique, will, however, have to be determined. Study and development of different techniques for simple and low-cost single SPFC preparation is essential in the process towards SPFC commercialization. 95

References

1. Mosdale, R.; Srinivasan, S.; Analysis of perfonnance and of water and thermal management in proton exchange membrane fuel cells, Electrochim. Acta, 40 No.4 (1995) 413-421

2. Penner, S.S. et al.; Commercialization of fuel cells, Prog.Energy Combust.Sci., 21 (1995) 145-151

3. Appleby, A.J.; Foulkes, F.R. (1989): Fuel Cell Handbook. 1st ed. Van Nostrand Reinhold, New York. 762 pages.

4. Kinoshita, K; McLarson, F.R.; Cairns, E.J. (1988); Fuel Cells, A Handbook, DOE/METC88/6096.

5. Bockris, J.O'M.; Srinivasan, S. (1969): Fuel Cells, Their Electrochemistry. 1st ed. McGraw-Hill, New York, USA. 659 pages.

6. Srinivasan, S.; Fuel cells for extraterrestrial and terrestrial applications, J.Electrochem.Soc., 136 2 (1989) 41C-48C

7. Dhar, H.P.; On solid polymer fuel cells, J.Electroanal.Chem., 357 (1993) 237-250

8. Wakefield, E.H. (1994): History of the Electric Automobile, Battery-Only Powered Cars. 1st ed. Society of Automotive Engineers, Inc., Warrendale, PA, USA. 541 pages.

9. Grove, W.R.; On Voltaic Series and the Combination of Gases by Platinum, Phil.Mag., (1839) 127

10. Buchi, F.N.; Gupta, B.; Haas, O.; Scherer, G.G.; Study of radiation-grafted FEP-g- Polystyrene membranes as polymer electrolytes in fuel cells, Electrochim. Acta, 40 No.3 (1995) 345-353

11. Zawodzinski, T.A.J.; Springer, T.E.; Davey, J.; Jestel, R.; Lopez, C.; Valerio, J.; Gottesfeld, S.; A comparative study of water uptake by and transport through ionomeric fuel cell membranes, J.Electrochem.Soc., 140 7 (1993) 1981-1985

12. Bemardi, D.M.; Verbragge, M.W.; A Mathematical Model of the Solid-Polymer- Electrolyte Fuel Cell, J.Electrochem.Soc., 139, No.9 (1992) 2477-2491

13. Bemardi, D.M.; Verbragge, M.W.; Mathematical Model of a Gas Diffusion Electrode Bonded to a Polymer Electrolyte, AIChE J., 37, No.8 (1991) 1151-1163

14.Springer, T.E.; Zawodzinski, T.A.; Gottesfeld, S.; Polymer Electrolyte Fuel Cell Model, J.Electrochem.Soc., 138 8 (1991) 2334 96 References

15. Fuller, T.F.; Newman, J.; Water and thermal management in solid-polymer- electrolyte fuel cells, J.Electrochem.Soc., 140 5 (1993) 1218-1225

16. Nguyen, T.V.; White, R.E.; A water and heat management model for proton- exchange-membrane fuel cells, J.Electrochem.Soc., 140 8 (1993) 2178-2186

17. Watanabe, M.; Satoh, Y.; Shimura, C.; Management of the water content in polymer electrolyte membranes with porous fiber wicks, J.Electrochem.Soc„ 140, No. 11 (1993) 3190-3193 lS.Miachon, S.; Aldebert, P.; Internal hydration H2/02 polymer electrolyte membrane fuel cell, J.Power Sources, 56 (1995) 31-36

19.Xie, G.; Okada, T.; Water Transport Behavior in Nation 117 Membranes, J.Electrochem.Soc., 142 No.9 (1995) 3057-3062

20.Ott0y, M. (1996): Mass and heat transfere in ion-exchange membranes, Applicable to solid polymer fuel cells. Dr.ing. Thesis, Norwegian University of Science and Technology. EFK-rapport 1996:50,180 p. in preparation

21. Parthasarathy, A.; Srinivasan, S.; Appleby, A.J.; Electrode kinetics of oxygen reduction at carbon-supported and unsupported platinum microcrystallite/Nafion interfaces, J.Electroanal.Chem., 339 (1992) 101-121

22. Allen, P.G.; Conradson, M.S.; Wilson, S.; Gottesfeld, S.; Raistrick, I.D.; Valerio, J.; Lovato, M.; In situ structural characterization of a platinum electrocatalyst by dispersive x-ray absorption spectroscopy, Electrochim. Acta, 39 16 (1994) 2415- 2418

23. Brito, P S D.; Sequeira, C.A.C.; Cathodic oxygen reduction on noble metal and catbon electrodes, J.Power Sources, 52 (1994) 1-16

24.Springer, T.E.; Wilson, M.S.; Gottesfeld, S.; Modeling and experimental diagnostics in polymer electrolyte fuel cells, J.Electrochem.Soc„ 140 12 (1993) 3513-3526

25. Goodridge, F.; Scott, K. (1995): Electrochemical Process Engineering, A Guide to the design of Electrolytic Plant. 1st ed. Plenum Press, New York. 312 pages.

26. Essalik, A.; Savadogo, O.; Ajersch, F.; Chemical and Electrochemical Parameters for Oxygen Reduction on Pt-H2WOyCarbon Electrodes in 99% H3PO4, J.Electrochem.Soc., 142 No.5 (1995) 1368-1376

27.Srinivasan, S.; Ticianelli, E.A.; Derouin, C.R.; Redondo, A.; Advances in solid polymer electrolyte fuel cell technology with low platinum loading electrodes, J.Power Sources, 22 (1988) 359-375 97

28. Bejan, A; (1988) Advanced engineering thermodynamics, Wiley, New York, USA, 758 pages.

29. Sonntag, R.E.; Van Wylen, G.; (1982) Introduction to Thermodynamics, Classical and Statistical, 2nd.ed, Wiley, New York, USA

30. Smith, J.M.; Van Ness, H.C.; (1984) Introduction to Chemical Engineering Thermodynamics, McGraw-Hill, 3rd.ed., Singapore, 632 pages.

31. Baehr, H.D.; (1966) Thermodynamik, eine Einfurung in die Grundlagen und ihre technischen Anwendungen, 2.ed., Springer-Verlag, Berlin, 445 pages.

32. F0rland, K.S.; Fprland, T.; Ratkje, S.K.; (1988) Irreversible thermodynamics, Theory and application, Wiley, GB, 281 pages

33. Dunbar, W.R.; Lior, N.; Gaggioli, R.A.; Combining fuel cells with fuel-fired power plants for improved exergy efficiency, Energy, 16 No.10 (1991) 1259-1274

34,Oosterkamp, P.F.van den; Goorse, A.A.; Blomen, L.J.M.J.; Review of an energy and exergy analysis of a fuel cell system, J.Power Sources, 41 (1993) 239-252

35. Tamizhmani, G.; Capuano, G.A.; Improved Electrocatalytic Oxygen Reduction Performance af Platinum Ternary Alloy-Oxide in Solid Polymer Electrolyte Fuel Cells, J.Electrochem.Soc., 141 N0.4 (1994) 968-975

36. Parthasarathy, A.; Martin, C.R.; Srinivasan, S.; Investigations of the 02 Reduction Reaction at the Platinum/Nafion Interface Using a Solid-State Electrochemical Cell, J.Electrochem.Soc., 138 4 (1991) 916-921

37. Murphy, O.J.; Hitchens, G.D.; Manko, D.J.; High power density proton-exchange membrane fuel cells, J.Power Sources, 47 (1994) 353-368

38. Poltarzewski, Z.; Staid, P.; Alderucci, V.; Wieczorek, W.; Giordano, N.; Nation Distribution in Gas Diffusion Electrodes for Solid-Polymer-Electrolyte-Fuel-Cell Applications, J.Electrochem.Soc. , 139, No.3 (1992) 761-765

39.Springer, T.E.; Zawodzinski, T.A.; Wilson, M.S.; Gottesfeld, S.; Characterization of Polymer Electrolyte Fuel Cells Using AC Impedance Spectroscopy, J.Electrochem.Soc., 143 No.2 (1996) 587-599

40. Taylor, E.J.; Anderson, E.B.; Vilambi, N.R.K.; Preparation of High-Platinum- Utilization Gas Diffusion Electrodes for Proton-Exchange-Membrane Fuel Cells, J.Electrochem.Soc., 139 5 (1992) L45-L46

41. Verbrugge, M.W.; Schneider, E.W.; Conell, R.S.; Hill, R.F.; The Effect of Temperature on the Equilibrium and Transport Properties of Saturated Poly(perfluorosulfonic acid) Membranes, J.Electrochem.Soc„ 139, No.12 (1992) 3421-3428 98 References

42.Verbrugge, M.W.; Hill, R.F.; Analysis of Promising Perfluorosulfonic Acid Membranes for Fuel-Cell Electrolytes, J.Electrochem.Soc., 137, No. 12 (1990) 3770-3777

43.Shukla, A.K.; Stevens, P.; Hamnett, A.; Goodenough, J.B.; A Nafion-bound platinized carbon electrode for oxygen reduction in solid polymer electrolyte cells, J.Appl.Electrochem., 19 (1989) 383-386

44.Srinivasan, S.; Velev, O.A.; Parthasarathy, A.; Manko, D.J.; Appleby, A.J.; High energy efficience and high power density fuel cell systems, J.Power Sources, 36 (1991) 299-320

45. Zawodzinski, T.A., Jr.; Derouin, C.; Radzinski, S.; Sherman, R.J.; Smith, V.T.; Springer, T.E.; Gottesfeld, S.; Water Uptake by and Transport Through Nafion 117 Membranes, J.Electrochem.Soc„ 140 4 (1993) 1041-1047

46. Mosdale, R.; Stevens, P.; New electrodes for hydrogen/oxygen solid polymer electrolyte fuel cell, Solid State Ionics, 61 (1993) 251-255

47. Kannan, A.M.; Shukla, A.K.; Hamnett, A.; Fractional-factorial design of a porous- carbon fuel-cell electrode, J.Appl.Electrochem., 18 (1988) 149-153

48. Ticianelli, E.A.; Derouin, C.R.; Redondo, A.; Srinivasan, S.; Methods to Advance Technology of Proton Exchange Membrane Fuel Cells, J.Electrochem.Soc., 135, No.9 (1988) 2209-2214

49.Srinivasan, S.; Manko, D.J.; Koch, H.; Enayetullah, M.A.; Appleby, A.J.; Recent Advances in Solid Polymer Electrolyte Fuel Cell Technology with Low Pt Loading Electrodes, J.Power Sources, 29 (1990) 367-387

50. Grubb, W.T.; Niedrach, L.W.; Batteries with Solid Ion-Exchange Membrane Electrolytes, n Low-Temperature Hydrogen-Oxygen Fuel Cells, J.Electrochem.Soc., 107 No.2 (1960) 131-135

51. Wilson, M S.; Gottesfeld, S.; Thin-film catalyst layers for polymer electrolyte fuel cell electrodes, J.Appl.Electrochem., 22 (1992) 1-7

52. Mukeqee, S.; Srinivasan, S.; Appleby, A.J.; Effect of sputtered film of platinum on low platinum loading electrodes on electrode kinetics of oxygen reduction in proton exchange membrane fuel cells, Electrochim. Acta, 38 12 (1993) 1661-1669

53. Ticianelli, E.A.; Beery, J.G.; Srinivasan, S.; Dependence of performance of solid polymer electrolyte fuel cells with low platinum loading on morphologic characteristics of the electrodes, J.Appl.Electrochem. , 21 (1991) 597-605

54. Millet, P.; Alleau, T.; Durand, R.; Characterization of membrane-electrode assemblies for solid polymer electrolyte water electrolysis, J.AppLElectrochem., 23 (1993) 322-331 99

55. Tamizhmani, G.; Capuano, G.A.; Life test of carbon-supported Pt-Cr-Cu electrocatalysts in Solid Polymer Fuel Cell, J.Electrochem.Soc., 141 9 (1994) L132-L134

56. Mukeijee, S.; Srinivasan, S.; Enhanced electrocatalysis of oxygen reduction on platinum alloys in proton exchange membrane fuel cells, J.Electroanal.Chem., 357 (1993) 201-224

57. Ahn, J.; Holze, R.; Bifunctional electrodes for an integrated water-electrolysis and fuel cell with a solid polymer electrolyte, J.Appl.Electrochem., 22 (1992) 1167- 1174

SS.Mattera, V.D.J.; Barnes, D.M.; Chaudhuri, S.N.; Risen, W.M., Jr.; Gonzales, R.D.; A Kinetic Study of the Catalytic Oxidation of CO over Nafion-Supported Rhodium, Ruthenium, and Platinum, J.Phys.Chem, 90 (1986) 4819-4824

59. Millet, P.; Durand, R.; Dartyge, E.; Tourillon, G.; Fontaine, A.; Precipitation of Metallic Pt into Nafion Ionomer Membranes, J.Electrochem.Soc., 140 5 (1993) 1373-1380

60. Ye, S.; Vijh, A.K.; Dao, L.H.; A new Fuel Cell Electrocatalyst Based on Highly Porous Carbonized Polyacrylonitrile Foam with Very Low Platinum Loading, J.Electrochem.Soc., 143 No.l (1996) L7-L9

61. Baldwin, R.; Pham, M.; Leonida, A.; McElroy, J.; Nalette, T.; Hydrogen-Oxygen proton-exchange membrane fuel cells and electrolysers, J.Power Sources, 29 (1990) 399-412

62. Chang, J.-R.; Lee, J.-F.; Lin, S.D.; Lin, A.S.; Carbon supported platinum catalyst electrodes: Characterization by Transmission Electron Microscopy, X-ray Absorbdon Spectroscopy, and Electrochemical Half Cell Measurements on a Phosphoric Acid Fuel Cell, J.Phys.Chem, 99 (1995) 14798-14804

63. Wilson, M.S.; Valerio, J.A.; Gottesfeld, S.; Low Platinum loading electrodes for polymer electrolyte fuel cells fabricated using thermoplastic ionomers, Electrochim. Acta, 40 No.3 (1995) 355-363

64.Staiti, P.; Poltarzewski, Z.; Alderucci, V.; Maggio, G.; Giordano, N.; Fasulo, A.; Influence of electrodic properties on water management in a solid polymer electrolyte fuel cell, J.AppLElectrochem., 22 (1992) 663-667

65. Bemardi, D.M.; Water-Balance Calculations for Solid-Polymer-Electrolyte Fuel Cells, J.Electrochem.Soc., 137, No.l 1 (1990) 3344-3350

66. Box, G.E.P.; Hunter, W.G.; Hunter, J.S.; (1978) Statistics for experimenters, An introduction to Design, Data Analysis, and Model Building, Wiley, USA. 653 pages. 100 References

67. Amphlett, J.C.; Baumert, R.M.; Mann, R.F.; Peppley, B.A.; Roberge, P.R.; Harris, T.J.; Performance Modeling of the Ballard Mark IV Solid Polymer Electrolyte Fuel Cell, I Mechanistic Model Development, H. Empirical Model Development, J.Electrochem.Soc., 142 1 (1995) 1-15

68. Uchida, M.; Aoyama, Y.; Eda, N.; Ohta, A.; New preparation method for polymer- electrolyte fuel cells, J.Electrochem.Soc., 142 2 (1995) 463-468

69. Qingfeng, L.; Gang, X.; Hjuler, H.A.; Berg, R.W.; Bjerrum, N.J.; Oxygen Reduction on Gas-Diffusion Electrodes for Phosphoric Acid Fuel Cells by a Potential Decay Method, J.Electrochem.Soc., 142 No. 10 (1995) 3250-3256

70. E-TEK, Gas Diffusion Electrode and Catalyst Materials, 1995 Catalogue, E-TEK, Inc, Natick, MA, USA 1995

71. Bregoli, L.J.; The influence on platinum crystallite size on the electrochemical reduction of Oxygen in phosphoric acid, Electrochim. Acta, 23 (1978) 489-492

72. Kumar, G.S.; Raja, M.; Parthasarathy; High performance electrodes with very low platinum loading electrodes for polymer electrolyte fuel cells, Electrochim. Acta, 40 No.3 (1995) 285-290

73. Uchida, M.; Aoyama, Y.; Eda, N.; Ohta, A.; Investigation of the Microstructure in the Catalyst Layer and Effects of Both Peifluorosulfonate Ionomer and PTFE- loading Carbon on the Catalyst Layer of Polymer Electrolyte Fuel Cells, J.Electrochem.Soc., 142 No. 12 (1995) 4143-4149

74.0kada, T; Ratkje, S.K.; Hanche-Olsen, H.; Water transport in ion exchange membranes, J.Mem.Sci., 66, (1992) 179-192

75. Chapotot, A.; Pourcelly, G.; Gavach, C.; Transport competition between monovalent and divalent cations through cation-exchange membranes. Exchange isotherms and kinetic concepts, J.Mem.Sci, 96 (1994) 167-181

76. Lagergren, C.; Lindbergh, G.; Simonsson, D.; Investigation of Porous Electrodes by Current Interruption, J.Electrochem.Soc., 142 3 (1995) 787-797

77.Selman, J.R.; Lin, Y.P.; Application of ac impedance in fuel cell research and development, Electrochim. Acta, 38 14 (1993) 2063-2073

78.Buchi, F.N.; Marek, A.; Scherer, G.G.; In situ Membrane Resistance Measurements on Polymer Electrolyte Fuel Cells by Fast Auxilary Current Pulses, J.Electrochem.Soc., 142 no.6 (1995) 1895-1901 101

Appendices 102 Appendices 103

Appendix A

Basic Equations

Basic thermodynamic equations concerning electrochemical energy conversion in a fuel cell are outlined in this appendix, based on the references [5] and [25].

Reversible cell potential

When all processes occurring in the fuel cell are reversible, the reversible cell potential, Erev, is given by the difference between the reversible electrode potentials: (A.l) where subscripts C and A denote the cathode and the anode, respectively. The reversible cell potential is related to the Gibbs energy change, AG, for the reaction taking place in the fuel cell; (AJ2) where F is the Faraday constant and n is the number of electrons involved in the cell reaction. To provide electrical energy, the fuel cell reaction must be spontaneous. This corresponds to a negative Gibbs energy change, and hence a positive reversible cell potential. Consider a general reaction: aA + bB = cC + dD (A.3) For this reaction the Gibbs energy change is given by the relation

(A.4) where A(f is Gibbs energy change when all involved species are in their standard states and a A is the activity of the species A, etc. Combining Equation A.2 and A.4, the Nemst equation arises:

(A.5) 104 Appendix A where E?rev is the standard reversible potential of the cell reaction. The superscript 0 denotes that all species are in their standard state (Le. their activity is unity).

Temperature and pressure dependence of reversible cell potential

The different types of fuel cells have different operating conditions. For each cell type the conditions may be varied to obtain improved performance. The reversible cell potential dependence on temperature and pressure is given in this section.

Temperature dependence Using Equation A.2 and keeping the pressure constant, the temperature dependence of the reversible cell potential is:

(A.6)

The temperature dependence of the cell potential is given by the entropy change, AS, of the reaction. At elevated temperatures, the reversible cell potential is then:

(A.7) when entropy is constant. 7° is the reference temperature, usually 0 or 25°C. The simplification on the right hand side of Equation A.7 does not apply for large temperature differences. The integration of AS must then be carried out, taking into account the temperature variation of AS (i.e. through heat capacities).

Pressure dependence In electrochemical energy production it is adequate to encounter two types of work, the electrical work by the system, WE (the desired product) and the pressure-volume work performed on the surrounding atmosphere:

WE=W-Pw AV (A.8) where Pw is the constant pressure of the surroundings and W is the total work. Similarly to Equation A.6, the pressure dependence of the reversible cell potential may be found: Basic Equations 105

dErev ) 1 f 4AG)"!

II ->v>> (A.9) dP 1 dP J where AV is the molar volume change of the reaction. Hence, if the cell reaction is subject to a change in molar volume, the cell potential will depend on pressure. When the fuel cell operates under conditions where all species are in the gaseous phase, the pressure dependence is expressed (through the use of the ideal gas law) as:

(A. 10) where P is the total pressure. A change in mole numbers, m, this will cause a pressure dependent cell potential. Integration of Equation A. 10 gives:

(A.ll) nF po where P° is the reference pressure (usually 1 bar or 1 atm).

The cell potential, E, (at a certain current density) also varies with temperature and pressure according to the above equations. However, for the cell potential, reaction kinetics also come into account, giving more complicated relations for the temperature and pressure dependencies. 106 Appendix A 107

Appendix B

C ompositions and Effects In The 23-Experiment

The three factors varied in the 23-screening experiment (results described in Section 5.6.2) were 1. Nation content N 2. Drying temperature D 3. Pt-loading P

The levels of the components are termed - for low and + for high. Following the Yates algorithm Box et.al.[66], the calculated effects are given in Table Bl.

Table Bl. The calculation of the effects of the factors on Cell Potential, using the Yates algorithm. The + and - indicates high and low level of the factor, respectively. ______N D P Cell Potential* (1) (2) (3) Divisor Effects - - - 350 770 1370 3640 8 455 Average + - - 420 600 2270 490 4 123 N - + - 225 1225 220 -350 4 -87 D + + - 375 1045 270 100 4 25 ND - - + 550 70 -170 900 4 225 P + - + 675 150 -180 50 4 12.5 NP - + + 450 125 80 -10 4 -2.5 DP + + + 595 145 20 -60 4 -15 NDP * at 100 mA/cm

Three replicates were tested of the cell all with factors at the low level (---- ). From these replicates, the significance level for the main factors were calculated as shown in Appendix C.

In this experiment, two different compositions were studied (15 and 25% Nation). Composition of the catalyst materials is given in Table B2 below. 108 Appendix B

Table B2. Composition of the catalyst material in the ^-screening experiment Slurry 25 % Nation______#g in slurry % in slurry #g dried % dried Pt/C 0.61 3.07 0.61 75.2% Carbon 0.00 0.00 0.00 0.0% Nafion(5%) 4.05 20.25 0.2025 24.8% Water 3.07 15.34 0 0.0% Glycerol 12.27 61.35 0 0.0% Total 100.0% 100.0%

Pt in slurry 0.61 %

Slurry 15% Nation #g in slurry % in slurry #g dried % dried Pt/C 0.65 3.26 0.65 65.1% Carbon 0.20 1.00 0.20 20.0% Naflon(5%) 3.00 15.00 0.15 15.0% Water 3.24 16.18 0 0.0% Glycerol 12.94 64.72 0 0.0% Total 100.0% 100.2%

Pt in slurry 0.65 % Appendix C

Reproducibility Of Single C ell Performance

Many steps are encountered when preparing single fuel cells. In this work the steps included preparation of electrode material, rinsing of Nation membranes, impregnation of commercial electrodes or application of the electrocatalytic layer to the membrane (in the thin film technique) and assembling of membrane and electrodes to single cells. The Nation membrane itself was not homogeneous, shown for example through different expansion ratios in the direction of machining and perpendicular to this. The electrode material for the thin film electrodes was mixed extensively (both mechanically and ultrasonically) and was thus expected to be of uniform composition. The amount of electrode material applied to the membrane (thin film technique) was kept within a variation of ±3%.

Reproducibility in the 2s-experiment In the 23-screening experiment, three cells of the same composition (low level of all factors) were assembled and tested to obtain an estimation of the uncertainty. A large deviation in performance between the cells was observed. The polarization curves are shown in Figure Cl. First generation pistons (Section 4.2.2) were applied in these measurements. The cells from the 23-experiment were compared at a current density of 100 mA/cm2 (Figure 5.5). Thus, the standard deviation, o, from the three replicates was calculated at this current density. The result was a = 86 mV. The main effects of the factors in this study were calculated as differences of the form (y+-y„), where y+ and y. were the average of the cell potentials at high and low level of the factor, respectively. Since y was an average of four values, the variance of the main effects, Ve, is given by (Box et al.[66]): Ve = V(y+ - y„) = (*4 + *A) cs2

The standard deviation for the effects, ce, was calculated to be: ce = Vve = 61 mV 110 Appendix C

o 03

Current Density [mA/cm ]

Figure Cl. Polarization curve for three equal cells of the 2^ screening experiment, used to estimate reproducibility.

The calculation of the effects was based on 8 measurements (Section 5.6.2). The use of average values at high and low level decreases the degees of freedom by 2, leaving 6 degrees of freedom for the estimation of the accuracy. From the r-distribution[66], the level of significance of the effects was found as shown in table Cl. The significance levels were all over 80%, implying that the probability that the factors did have an effect on the cell potential was higher than 80%.

Table Cl. Main effects on cell potential, E, of the factors in the 23-experiment and the significance level for the respective effect. ______Factor Effect on E* [mV] Significance level Nafion (% by weight) +123 >90% Drying Temp. (°C) -87 >80% Pt-loading (mg/cm2) +225 99% *at 100 mA/cm2. Reproducibility Of Single Cell Performance 111

Improved reproducibility by incorporation of a pneumatic cylinder The reproducibility discussed above was not as high as desired, and thus, factors concerning the assembling and operation of the cells were examined. One factor that could cause variations in cell performance was the pressure by which the M&E assembly was held together during single cell performance tests. This mechanical pressure will be termed the Pressure of compaction. The single cell fixture (Figure 4.2) contains two pistons moving axially to fit electrodes of different thicknesses. The force applied to the pistons from the screws was difficult to control. A pneumatic cylinder was fitted to the cell fixture in order to provide reproducible mechanical pressures over the M&E assemblies. The pressure of compaction was measured to be in the range of 1.8-24.9 bar. A single cell containing thin film electrodes was tested at different pneumatic pressures and the results are shown in Figure C2. Second generation pistons were applied in this experiment.

The pressure of compaction was increased progressively and had a large effect on cell performance. After obtaining the polarization curve at 9.4 bar, the pressure was reduced to 1.8 bar. As can be seen from Figure C2, the performance was reduced

-x— 1.8 bar

■* —9.4 bar 1.8 bar ’

18.4 bar 24.9 bar S 0.6

0 100 200 300 400 500 600 700 800 Current Density [mA/cm2]

Figure C2. The effect of variation in pressure of compaction over the membrane and electrode assembly. * Pressure reduced from 9.4 to 1.8 bar. 112 Appendix C correspondingly, resembling the original 1.8 bar curve. After having applied 24.9 bar to the single cell, the performance did not return to the original curve upon pressure reduction. This indicated that the changes were reversible up to a certain pressure, above which the single cell was permanently modified. When the single cell was taken out from the cell fixture after applying this high pressure, the porous electrode backing adhered to the thin film catalyst layer. The porous electrode backing also showed signs of permanent compression. The high pressure of compaction resembled a mild type of hot pressing. Conditions used for hot pressing are customarily in the range of 50-100 bar and 135°C for Nation membranes.

■ Cell data ““Least square fit — 90 % confidence — interval

Current Density [mA/cm2]

Figure C3. Cell data for three cells (type Cell 7) of the same composition, containing thin film electrodes of 0.1 mg Pt/cm2. Cell temperature was 70° C, and gas pressures were atmospheric.

Reproducibility was tested again after the incorporation of the pneumatic equipment for controlling the pressure of compaction. Three equal cells were prepared and tested, and the polarization data are shown in Figure C3. The data were fitted to Equation 2.17. Comparison of Figures Cl and C3 reveals improved cell performance reproducibility. At a current density of 100 mA/cm2, the 90% confidence interval was 50 mV and the standard deviation was 31 mV. Hence the standard deviation was Reproducibility Of Single Cell Performance 113 reduced to half of the former value, by introducing the pneumatic cylinder and changing to 2nd generation piston.

Figure C4. The pneumatic cylinder attached to one half of the single cell fixture. 114 Appendix C 115

Appendix D

Resolution Of Potential And C urrent Measurements

HP3457A Multimeter

A HP 3457A Multimeter holding a 44492A multiplexer was used for measuring the potentials from the single fuel cell, the thermo-couples and the pressure controllers. The integration time was set to 200 ps and this gave a resolution of 6Vi digits. The types of potential measurements, their corresponding ranges and resolutions are show in Table Dl. The information was taken from the Instrument Operation Manual.

Table Dl. Summary on DC voltage measurements Inner Voltage Resolution Measuring Type of measurement impedance Range current (max) Thermo-couples (type K) 10 GO 0.03 V 10 nV 3 pA Cell potential 10 GQ 3 V 1 pV 300 pA Pressures (0-5V) 10 MQ 30 V 10 pV 3 nA

The internal resistance of the circuit over which the potential was measured, should be much lower than the inner impedance of the measuring device (the voltmeter). The thermo-couples have resistances of around 10Q. The fuel cell has an internal resistance in the range of 0.2-10 Q. The pressure controller potential output will not be affected by the low measuring current. From the calculated measuring currents of the HP3457A it was concluded that the use of this instrument should cause negligible systematic errors to the voltage measurements. For the fuel cell, the real potential variations related to the complex processes occurring in the cell were assumed to be much larger than the resolution of the voltmeter. 116 Appendix D

Schlumberger 7150 plus Digital Multimeter

This instrument was used to measure the current through the fuel cell. Maximum current through the instrument was 2 A. The shunt resistor in the instrument had a resistance of lOOmQ, determining the lower limit of outer circuit resistance when applying this instrument. No error was, however, introduced by this shunt resistor as it constitutes part of the total outer resistance (the cell load). The integration time was set to 40 ms giving a resolution of 4h digits. This corresponds to a resolution of lOOpA.

Currents larger than 2A were measured as potential falls (by the HP3457A) over the outer resistance, from which the cell current was calculated. Again it was concluded that there be only negligible errors in the cell current measurements. 117

Appendix E

D escription Of A 32-Experiment

Based partly on the results of the 23-screening experiment (Section 5.6.2), a new series of experiments was planned. The screening experiment indicated high ohmic resistance through the cell. Introduction of Acetylene Black, a carbon with very high electronic conductivity, was suggested by Mosdale and Stevens [46] and later used by Uchida et al.[68]. A small portion of the Vulcan XC72 carbon was substituted by Acetylene Black (Shawinigan, E-TEK, Natick, MA) in order to reduce ohmic losses. The amount of Nation in the electrocatalytic layer had a pronounced effect on cell performance and was therefore subject to further investigation. Aiming at a reduced Pt-loading, this factor was kept constant at a level of 0.1 mg Pt/cm2 in the new experiment. To reduce degradation of the Nation polymer, the drying temperature was held at 125°C. Therefore, in this experiment, two factors both concerning the composition of the electrode were to be studied, namely the Nation content and the amount of highly conductive Acetylene Black. Each factor was given three levels to be able to identify quadratic effects (Table El). Composition of the different electrodes is given in Table E2.

Table El. Factors and levels in the 32-experiment LEVELS FACTORS -1 0 4-1 Nation (% by weight) 15 25 35 Acetylene Black (% by weight) 0 5 10 118 Appendix E

Table E2. Electrode compositions of the 32-experiment:

Slurry for Cell 1 (-1,-1) ______#g in slurry % in slurry #g dried % dried Pt/C 0.455 2.60% 0.455 65.00% % AB/C Pt/AB 0 0.00% 0 0.00% 0.00% Carbon 0.14 0.80% 0.14 20.00% % Nafion Nafion 2.1 12.00% 0.105 15.00% 15.00% Glycerol 11.844 67.68% Water 2.961 16.92% Total mass TOTAL 17.5 100.00% 0.7 100.00%

dry matter 4.00% #g Solvent 16.8 Nafion 1.995 Pt in slurry 0.52%

Slurry for Cell 2 (-1,0) ______#g in slurry % in slurry #g dried % dried Pt/C 0.455 2.60% 0.455 65.00% % AB/C Pt/AB 0 0.00% 0 0.00% 0.00% Carbon 0.07 0.40% 0.07 10.00% % Nafion Nafion 3.5 20.00% 0.175 25.00% 25.00% Glycerol 10.78 61.60% Water 2.695 15.40% TOTAL 17.5 100.00% 0.7 100.00%

dry matter 4.00% #g Solvent 16.8 Nafion 3.325 Pt in slurry 0.52%

Slurry for Cell 3 (-1,+1) ______#g in slurry % in slurry #g dried % dried Pt/C 0.455 2.60% 0.455 65.00% % AB/C Pt/AB 0 0.00% 0 0.00% 0.00% Carbon 0 0.00% 0 0.00% % Nafion Nafion 4.9 28.00% 0.245 35.00% 35.00% Glycerol 9.716 55.52% Water 2.429 13.88% TOTAL 17.5 100.00% 0.7 100.00%

dry matter 4.00% #g Solvent 16.8 Nafion 4.655 Pt in slurry 0.52% 119

Table E2 (continued)

Slurry for Cell 4 (0,-1) ______#g in slurry % in slurry #g dried % dried Pt/C 0.4253 2.43% 0.42525 60.75% % AB/C Pt/AB 0.0298 0.17% 0.02975 4.25% 5.00% Carbon 0.14 0.80% 0.14 20.00% % Nafion Nafion 2.1 12.00% 0.105 15.00% 15.00% Glycerol 11.844 67.68% Water 2.961 16.92% TOTAL 17.5 100.00% 0.7 100.00%

dry matter 4.00% #g Solvent 16.8 Nafion 1.995 Pt in slurry 0.52%

Slurry for Cell 5 (0,0) ______#g in slurry % in slurry #g dried % dried Pt/C 0.4288 2.45% 0.42875 61.25% % AB/C Pt/AB 0.0263 0.15% 0.02625 3.75% 5.00% Carbon 0.07 0.40% 0.07 10.00% % Nafion Nafion 3.5 20.00% 0.175 25.00% 25.00% Glycerol 10.78 61.60% Water 2.695 15.40% TOTAL 17.5 100.00% 0.7 100.00%

dry matter 4.00% #g Solvent 16.8 Nafion 3.325 Pt in slurry 0.52%

Slurry for Cell 6 (0,+l) ______#g in slurry % in slurry #g dried % dried Pt/C 0.4323 2.47% 0.43225 61.75% % AB/C Pt/AB 0.0228 0.13% 0.02275 3.25% 5.00% Carbon 0 0.00% 0 0.00% % Nafion Nafion 4.9 28.00% 0.245 35.00% 35.00% Glycerol 9.716 55.52% Water 2.429 13.88% TOTAL 17.5 100.00% 0.7 100.00%

dry matter 4.00% #g Solvent 16.8 Nafion 4.655 Pt in slurry 0.52% 120 Appendix E

Table E2 (continued)

Slurry for Cell 7 (+1,-1) ______#g in slurry % in slurry #g dried % dried Pt/C 0.3955 2.26% 0.3955 56.50% % AB/C Pt/AB 0.0595 0.34% 0.0595 8.50% 10.00% Carbon 0.14 0.80% 0.14 20.00% % Nafion Nafion 2.1 12.00% 0.105 15.00% 15.00% Glycerol 11.844 67.68% Water 2.961 16.92% TOTAL 17.5 100.00% 0.7 100.00%

dry matter 4.00% #g Solvent 16.8 Nafion 1.995 Pt in slurry 0.52%

Slurry for Cell 8 (+1,0) ______#g in slurry % in slurry #g dried % dried Pt/C 0.4025 2.30% 0.4025 57.50% % AB/C Pt/AB 0.0525 0.30% 0.0525 7.50% 10.00% Carbon 0.07 0.40% 0.07 10.00% % Nafion Nafion 3.5 20.00% 0.175 25.00% 25.00% Glycerol 10.78 61.60% Water 2.695 15.40% TOTAL 17.5 100.00% 0.7 100.00%

dry matter 4.00% #g Solvent 16.8 Nafion 3.325 Pt in slurry 0.52%

Slurry for Cell 9 (+1,+D ______#g in slurry % in slurry #g dried % dried Pt/C 0.4095 2.34% 0.4095 58.50% % AB/C Pt/AB 0.0455 0.26% 0.0455 6.50% 10.00% Carbon 0 0.00% 0 0.00% % Nafion Nafion 4.9 28.00% 0.245 35.00% 35.00% Glycerol 9.716 55.52% Water 2.429 13.88% TOTAL 17.5 100.00% 0.7 100.00%

dry matter 4.00% #g Solvent 16.8 Nafion 4.655 Pt in slurry 0.52% 121

Appendix F

Preparation Of Nation 117 Membranes

Detailed membrane preparation prescription was given in literature [36, 38, 48]. Evaluation of these procedures led to the procedure described below. The Nation membranes were pale yellow (when received) due to impurities. The preparation procedure rendered the membranes colorless and transparent

Chemicals Hydrogen Peroxide, H 202, (3%), pa. Merck, Darmstadt Germany. Sulfuric Acid, H2SG4, (0.05 M), pa. Merck, Darmstadt Germany. Distilled water from the on-site water distillation apparatus was used.

Preparation procedure All steps were performed on an electric heater/stirrer and the magnet stirrer was set at a low speed. The solutions were heated to the correct temperature by means of a gas (propane) flame.

1. Cut the membranes to required size by applying a circular cut knife and a hammer. Be aware of and take into account the expansion when the membranes are subject to water.

2. Heat the membranes in distilled water to the boiling point.

3. Transfer them to hot 5% H2O2 and keep at 85-100°C for 30 minutes to oxidize organic impurities. This should give colorless membranes.

4. Rinse in distilled water for 10 minutes at T > 70°C.

5. Immerse in hot dilute sulfuric add (0.05 M H2SO4) at 90°C for 30 minutes to get rid of metallic impurities.

6. Rinse several times* in distilled water at T > 70°C to remove all traces of add. Magnetic stirring is recommended.

7. Store the membranes in distilled water until use.

*) renew the hot water at least twice. 122 Appendix F 123

Appendix G

Separation Of Potential Losses

Techniquesfor separationof potential losses In fuel cell research, much effort is put to the development of new materials with improved properties. Characterization of the materials is a major task in this work. In SPFC development, optimization of the electrode composition, structure and morphology is important (see Chapter 5). By separating the electrode overpotential into different contributions, it is possible to locate the origin of the potential loss. Ohmic losses occur due to limited protonic conductivity in the membrane. When porous electrodes are used, these may also give ohmic potential losses related to low electronic conductivity. High activation overpotentials are usually found at the cathode in low temperature fuel cells, due to the low exchange current density for the oxygen reduction reaction. The localization of the potential losses is essential, because the origin of the losses dictate where efforts should be put for further improvements.

Different techniques may be applied to obtain information about where losses occur in electrochemical cells. Separation of ohmic losses is usually obtained by hydrogen evolution methods, interrupter techniques or frequency response techniques[25]. The fact that hydrogen evolution displays Tafel behavior up to reasonable high current densities allows it to be used as a means of determining the ohmic correction without special equipment. Current interruption is a generally used technique for correcting measured electrode overpotentials for the ohmic potential drops in both half cells and single fuel cells. Lagergren et al. [76] reported that the use of current interruption to porous electrodes may lead to an overestimation of the ohmic resistance, due to internal currents passing in the electrode after interruption. The condition under which this overestimation may occur, is when the electronic conductivity of the electrode is in the same order of magnitude as the electrolyte conductivity. In the SPFC the electronic specific conductivity of the electrode material (Vulcan XC72) is in the range of 1-5 Scm_1[46], 124 Appendix G whereas the specific conductivity if Nafion is around 0.06Scm"1[ll]. The thickness of the electrode and the membrane is around 0.5 and 0.2 mm, respectively, giving a ratio between conductivities of 7 to 33 in the favor of the electrode. For the catalyst layer, in which a certain percentage polymer was included, the electronic conductivity is lower, and therefore the conductivities may be comparable in magnitude. Hence, an overestimation of the ohmic losses may have occurred. Large overestimations are, however, unlikely because the measured total ohmic losses of the cell were only slightly higher than the ohmic resistance estimated for the membrane alone (Section 5.6.4). This will be further examined in future. The AC impedance spectroscopy technique has only recently been applied to SPFC[77]. Springer et al.[39] report a study on SPFC characterization by in situ AC impedance spectroscopy under various experimental conditions. The impedance spectra for the cathode were related to the transport processes occurring at the electrode. Analyzing the results by applying a SPFC model, made this approach very powerful.

Biiche et al.[78] have discussed different techniques used for electrolyte resistance measurements. They conclude that, compared to the AC impedance method, current interruption offers considerable advantages. At high currents (>10 A), however, the standard interrupter technique meets difficulties related to the inductance of the DC- current loop. Therefore Biiche et al.[78] developed a fast auxiliary current pulse technique for in situ membrane resistance measurements in SPFC.

In this study the geometric area of the working electrode was 3.9 cm2. The thin film electrodes with low Pt-loading gave current densities in the range of 0-1 A/cm2, and thus a total current of the DC-loop of maximum 4 A. At these currents the standard current interruption technique is assumed to be applicable, and therefore this technique was used in this study. Separation Of Potential Losses 125

Reference electrodes In SPFC research, new materials are often studied in half cells in acidic solutions (see e.g.[40]). In these cases the Luggin capillary and a salt bridge is suitable for reference electrode measurements. Due to the compact geometry of the SPFC, it has been Figure Gl. Top view of the single cell with reference electrodes at the center of concluded that it is extremely difficult the ring shaped working electrode. to incorporate reference electrodes and still preserve the local current density[78]. When studying single fuel cells, measurements of overpotentials are commonly done by incorporating a # reference electrode at the anode compartment (see e.g.[51]). Dynamic hydrogen electrodes (DHE) have also been applied[36,38]. In this study localization and quantification of potential losses in the single cell has been achieved by incorporation of reference electrodes at both electrodes (Figure Gl). The reference electrodes were pieces of electrode identical to the working electrode. The reference electrode at Figure G2. Cross-section of half of the the anode constitutes a reversible single cell showing working (upper) and reference (lower) electrodes. hydrogen electrode (RHE). By Electrical connections for potential locating the reference electrodes at the measurements are included. center of the cell, the same operation 126 Appendix G conditions for working and reference electrode were assured. The potential differences measured between the reference electrode and the working electrode at each side are termed EA and Ec (Figure G2). These potential differences were the sum of ohmic and non-ohmic contributions. As there was no current running between the reference electrodes, the potential difference between the reference electrodes should be constant and equal to the open circuit potential. During cell testing, however, this potential decreased. This was thought to be due to the irreversibility of the reaction at the cathode reference electrode. Control measurements at open circuit conditions confirmed that the anode reference electrode was truly working as a reference electrode, whereas the cathode reference electrode potential was unstable. The electrical connections are shown in Figure G2. By applying KirchhofPs law it can be seen that E = EA + Ec (Gl)

Equipotential lines

Working Ref.electrodcs electrodes

Figure G3. Equipotential lines in the membrane indicating negligible potential loss between the reference electrodes. Separation Of Potential Losses 127

Between the working electrodes, equipotential lines are linear and parallel to the electrode surfaces (Figure G3), assuming that current density is constant over the electrode surface. In the region between the working electrodes and the reference electrodes, these lines will bend outwards as indicated. The distance from the reference electrodes to the working electrode was more than 2mm, whereas the thickness of the membrane (Nation 117) was 0.2mm. This ratio of 10, indicates that only a negligible part of the potential drop through the membrane (between working electrodes) exists between the reference electrodes. From measurements it was verified that the potential between reference electrodes varied very little with current density. The ohmic potential loss over the membrane was not attainable from this setup of potential measurements. This was estimated from literature data. An example of a polarization curve, including the measured cathodic and anodic overpotentials is shown in Figure G4. A linear fit of the anodic overpotential is included.

■* — Cell Potential ■— Cathodic overpot.

Linear (Anodic overpot.)

« 0.6

Current Density [mA/cm ]

Figure G4. Polarization curve for Cell 7 at 4.5 bar gas pressure and cell temperature of 70°C. The measured anodic overpotential and the calculated cathodic overpotential are included. 128 Appendix G

Current interruption in this study The overpotentials measured between the reference electrode and working electrode (Fig.G3) were the sum of the ohmic and non-ohmic contributions of that electrode. Ohmic losses were separated by using the current interrupter technique. This technique involved a sudden switch-off of the current which flows through the cell, after which the potential transient was detected. A flicker-free relay (Amphenol-Tuchel) was used to open the circuit and the transient was detected by means of a Tektronix 2232 (100MHz) Digital Storage Oscilloscope. Measurements of the ohmic losses at different resolutions gave the same results up to a resolution of 2|tsec. Thus, the oscilloscope was set to measure 1000 points with a resolution of 2(j.sec in the measurements. Experimental Data 129

Appendix H, Experimental D ata

Data from test of Cell 7, Cell temp. 70°C, Gas pressures 4,5 bar Cell Cell Overpot Overpot Temp Temp Temp Pressure Pressure Flow Flow Potential Current Anode Cathode Cell Anode Cath Anode Cathode Anode Cath. Density gas gas [V] [mA] [V] m [C] [C] [C] [Barg] [Barg] [mm] mm] 1.002 0.0 -0.001 -0.003 70.5 79.9 79.6 3.5 3.5 20 30 0.997 0.0 -0.001 0.001 70.6 79.9 79.5 3.5 3.5 20 30 0.996 0.0 -0.001 0.003 70.6 79.9 79.5 3.5 3.5 20 30 0.961 1.9 0.001 0.037 70.5 80.0 79.6 3.5 3.5 20 30 0.958 1.9 0.001 0.039 70.5 79.9 79.6 3.5 3.5 20 30 0.957 1.9 0.000 0.040 70.6 80.0 79.5 3.5 3.5 20 30 0.957 2.4 0.001 0.040 70.5 79.9 79.6 3.5 3.5 20 30 0.956 2.4 0.001 0.041 70.5 79.9 79.6 3.5 3.5 20 30 0.955 2.4 0.001 0.042 70.5 79.9 79.6 3.5 3.5 20 30 0.941 4.7 0.002 0.055 70.6 79.9 79.6 3.5 3.5 20 30 0.939 4.7 0.002 0.057 70.5 79.9 79.6 3.5 3.5 20 30 0.937 4.7 0.002 0.059 70.5 79.9 79.5 3.5 3.5 20 30 0.921 9.2 0.003 0.075 70.6 79.9 79.6 3.5 3.5 20 30 0.919 9.1 0.003 0.076 70.5 79.9 79.5 3.5 3.5 20 30 0.918 9.1 0.003 0.077 70.5 79.9 79.6 3.5 3.5 20 30 0.887 21.5 0.007 0.104 70.6 80.0 79.6 3.5 3.5 20 30 0.886 21.4 0.006 0.106 70.6 80.0 79.6 3.5 3.5 20 30 0.885 21.4 0.006 0.108 70.6 79.9 79.6 3.5 3.5 20 30 0.861 38.2 0.016 0.121 70.6 80.0 79.5 3.5 3.5 20 30 0.859 38.1 0.014 0.125 70.6 80.0 79.6 3.5 3.5 20 30 0.858 38.1 0.012 0.128 70.6 80.0 79.6 3.5 3.5 20 30 0.818 80.9 0.030 0.150 70.5 79.9 79.5 3.5 3.5 20 30 0.817 80.7 0.029 0.153 70.6 79.9 79.5 3.5 3.5 20 30 0.815 80.7 0.027 0.156 70.6 79.9 79.5 3.5 3.5 20 30 0.794 119.9 0.037 0.167 70.3 79.8 79.5 3.5 3.5 20 30 0.792 119.7 0.037 0.169 70.4 79.8 79.5 3.5 3.5 20 30 0.791 119.5 0.037 0.171 70.5 79.8 79.5 3.5 3.5 20 30 0.719 271.7 0.065 0.215 70.3 79.9 79.5 3.5 3.5 20 30 0.715 270.8 0.065 0.218 70.3 79.9 79.5 3.5 3.5 20 30 0.713 270.2 0.065 0.220 70.3 79.8 79.5 3.5 3.5 20 30 0.672 378.1 0.085 0.241 70.0 80.1 79.6 3.5 3.5 20 30 0.669 378.1 0.085 0.244 70.1 80.1 79.6 3.5 3.5 20 30 0.667 377.1 0.085 0.246 70.2 80.0 79.6 3.5 3.5 20 30 0.665 376.1 0.086 0.247 70.3 80.0 79.6 3.5 3.5 20 30 0.560 615.9 0.142 0.296 69.9 79.9 79.6 3.5 3.5 20 30 0.558 613.9 0.142 0.298 69.9 80.0 79.6 3.5 3.5 20 30 0.557 613.4 0.141 0.299 70.0 80.1 79.6 3.5 3.5 20 30 0.556 612.1 0.142 0.300 70.1 80.1 79.6 3.5 3.5 20 30 0.490 750.1 0.181 0.327 69.9 79.6 79.6 3.5 3.5 20 30 0.487 747.3 0.182 0.329 69.9 79.7 79.6 3.5 3.5 20 30 0.486 744.7 0.182 0.330 70.0 79.8 79.6 3.5 3.5 20 30 0.486 743.7 0.182 0.330 70.0 79.9 79.6 3.5 3.5 20 30 0.361 924.4 0.257 0.379 70.2 79.7 79.6 3.5 3.5 20 30 0.360 922.1 0.260 0.378 70.2 79.6 79.6 3.5 3.5 20 30 0.360 921.6 0.260 0.378 70.1 79.6 79.6 3.5 3.5 20 30 130 Appendix H

Data from test of Cell 7, Cell temp. 70°C, Gas pressures 1.0 bar Cell Cell Overpot Overpot Temp Temp Temp Pressure Pressure Flow Flow Potential Current Anode Cathode Cell Anode Cath Anode Cathode Anode Cath. Density gas gas m fmA] rvi rvi [Cl [Cl rq [Barg] [Barg] [mm] mml 0.989 0.0 0.001 -0.001 70.0 79.8 79.4 0.0 0.0 20 30 0.989 0.0 0.001 -0.001 70.0 79.8 79.4 0.0 0.0 20 30 0.988 0.0 0.001 -0.001 70.0 79.8 79.4 0.0 0.0 20 30 0.913 1.8 0.001 0.074 70.0 79.8 79.4 0.0 0.0 20 30 0.912 1.8 0.002 0.075 70.0 79.8 79.4 0.0 0.0 20 30 0.911 1.8 0.001 0.076 70.0 79.8 79.4 0.0 0.0 20 30 0.909 2.3 0.001 0.078 70.0 79.8 79.4 0.0 0.0 20 30 0.908 2.3 0.001 0.080 70.0 79.8 79.4 0.0 0.0 20 30 0.907 2.3 0.002 0.080 70.0 79.8 79.4 0.0 0.0 20 30 0.890 4.5 0.002 0.097 70.0 79.8 79.4 0.0 0.0 20 30 0.889 4.5 0.002 0.098 70.0 79.8 79.4 0.0 0.0 20 30 0.888 - 4.5 0.002 0.099 70.0 79.8 79.4 0.0 0.0 20 30 0.869 8.7 0.003 0.117 70.0 79.8 79.4 0.0 0.0 20 30 0.868 8.6 0.003 0.118 70.0 79.8 79.4 0.0 0.0 20 30 0.867 8.6 0.003 0.119 70.0 79.8 79.4 0.0 0.0 20 30 0.836 20.2 0.004 0.148 70.0 79.8 79.4 0.0 0.0 20 30 0.835 20.2 0.004 0.149 70.0 79.8 79.4 0.0 0.0 20 30 0.834 20.2 0.004 0.150 70.0 79.8 79.4 0.0 0.0 20 30 0.811 35.9 0.007 0.171 70.0 79.8 79.4 0.0 0.0 20 30 0.809 35.9 0.007 0.173 70.0 79.8 79.4 0.0 0.0 20 30 0.808 35.8 0.006 0.174 70.0 79.8 79.4 0.0 0.0 20 30 0.769 75.9 0.013 0.207 69.9 79.8 79.4 0.0 0.0 20 30 0.768 75.8 0.012 0.209 70.0 79.8 79.4 0.0 0.0 20 30 0.767 75.8 0.012 0.210 70.0 79.8 79.4 0.0 0.0 20 30 0.743 112.5 0.020 0.226 69.9 79.8 79.4 0.0 0.0 20 30 0.742 112.3 0.019 0.228 69.9 79.8 79.4 0.0 0.0 20 30 0.740 112.1 0.019 0.230 69.9 79.8 79.4 0.0 0.0 20 30 0.660 253.7 0.049 0.280 69.8 79.8 79.4 0.0 0.0 20 30 0.658 252.6 0.047 0.283 69.9 79.8 79.4 0.0 0.0 20 30 0.656 251.7 0.047 0.286 69.9 79.8 79.4 0.0 0.0 20 30 0.608 343.2 0.077 0.304 69.7 79.8 79.4 0.0 0.0 20 30 0.606 342.7 0.075 0.308 69.7 79.8 79.4 0.0 0.0 20 30 0.604 341.6 0.073 0.311 69.8 79.8 79.4 0.0 0.0 20 30 0.489 537.3 0.139 0.360 69.7 79.8 79.4 0.0 0.0 20 30 0.489 537.3 0.136 0.364 69.7 79.8 79.4 0.0 0.0 20 30 0.487 536.0 0.135 0.367 69.8 79.8 79.4 0.0 0.0 20 30 0.487 535.2 0.133 0.368 69.8 79.8 79.4 0.0 0.0 20 30 0.417 639.1 0.179 0.393 69.7 79.8 79.4 0.0 0.0 20 30 0.416 637.5 0.183 0.390 69.7 79.8 79.4 0.0 0.0 20 30 0.415 637.5 0.186 0.388 69.6 79.8 79.4 0.0 0.0 20 30 0.301 766.8 0.253 0.435 69.8 79.8 79.4 0.0 0.0 20 30 0.300 764.5 0.255 0.434 69.8 79.8 79.4 0.0 0.0 20 30 0.300 766.8 0.254 0.435 69.7 79.8 79.4 0.0 0.0 20 30 Experimental Data 131

Data from test of Cell 7, Cell temp. 50°C, Gas pressures 1.0 bar Cell Cell Overpot Overpot Temp Temp Temp Pressure Pressure Flow Flow Potential Current Anode Cathode Cell Anode Cath Anode Cathode Anode Cath. Density gas gas [V] [mA] [V] [V] [C] [C] [C] [Barg] [Barg] [mm] mm] 0.984 0.0 -0.002 0.002 50.4 59.7 60.1 0.0 0.0 20 30 0.984 0.0 -0.002 0.002 50.4 59.6 60.1 0.0 0.0 20 30 0.983 0.0 -0.002 0.004 50.5 59.6 60.1 0.0 0.0 20 30 0.894 1.8 0.000 0.090 50.5 59.7 60.1 0.0 0.0 20 30 0.893 1.8 -0.001 0.091 50.5 59.8 60.1 0.0 0.0 20 30 0.893 1.8 -0.001 0.092 50.4 59.7 60.1 0.0 0.0 20 30 0.890 2.3 0.001 0.093 50.4 59.5 60.2 0.0 0.0 20 30 0.888 2.2 0.000 0.095 50.4 59.6 60.2 0.0 0.0 20 30 0.887 2.3 0.000 0.097 50.5 59.7 60.1 0.0 0.0 20 30 0.868 4.3 0.003 0.113 50.4 59.2 60.1 0.0 0.0 20 30 0.866 4.3 0.003 0.115 50.5 59.3 60.1 0.0 0.0 20 30 0.865 4.4 0.002 0.117 50.5 59.4 60.2 0.0 0.0 20 30 0.844 8.4 0.006 0.133 50.4 59.4 60.2 0.0 0.0 20 30 0.843 8.4 0.006 0.135 50.4 59.3 60.1 0.0 0.0 20 30 0.842 8.4 0.005 0.137 50.4 59.2 60.2 0.0 0.0 20 30 0.809 19.6 0.011 0.165 50.4 59.8 60.2 0.0 0.0 20 30 0.807 19.6 0.010 0.166 50.4 59.7 60.2 0.0 0.0 20 30 0.807 19.5 0.010 0.168 50.4 59.6 60.2 0.0 0.0 20 30 0.782 34.6 0.015 0.188 50.3 60.1 60.2 0.0 0.0 20 30 0.780 34.5 0.015 0.189 50.3 60.0 60.2 0.0 0.0 20 30 0.779 34.5 0.014 0.190 50.4 60.0 60.2 0.0 0.0 20 30 0.733 72.4 0.024 0.227 50.4 59.7 60.1 0.0 0.0 20 30 0.732 72.3 0.024 0.228 50.4 59.8 60.1 0.0 0.0 20 30 0.731 72.2 0.024 0.229 50.3 60.0 60.2 0.0 0.0 20 30 0.702 106.4 0.031 0.250 50.1 59.1 60.2 0.0 0.0 20 30 0.700 106.0 0.032 0.253 50.2 59.3 60.1 0.0 0.0 20 30 0.698 105.7 0.032 0.254 50.3 59.4 60.1 0.0 0.0 20 30 0.599 232.2 0.062 0.323 50.2 59.0 60.2 0.0 0.0 20 30 0.594 229.5 0.063 0.327 50.2 58.9 60.2 0.0 0.0 20 30 0.591 228.4 0.063 0.329 50.2 58.9 60.2 0.0 0.0 20 30 0.546 309.5 0.080 0.358 50.0 60.0 60.2 0.0 0.0 20 30 0.541 306.7 0.080 0.362 50.1 59.7 60.2 0.0 0.0 20 30 0.538 304.9 0.081 0.365 50.1 59.4 60.2 0.0 0.0 20 30 0.428 471.2 0.128 0.429 50.1 60.5 60.2 0.0 0.0 20 30 0.425 468.6 0.127 0.431 50.1 60.3 60.2 0.0 0.0 20 30 0.423 466.3 0.128 0.433 50.1 60.2 60.2 0.0 0.0 20 30 0.372 571.0 0.155 0.457 49.9 59.6 60.2 0.0 0.0 20 30 0.368 565.8 0.154 0.461 50.0 59.7 60.2 0.0 0.0 20 30 0.367 562.7 0.154 0.463 50.0 59.6 60.2 0.0 0.0 20 30 0.365 560.4 0.155 0.464 50.0 59.6 60.2 0.0 0.0 20 30 0.364 558.6 0.155 0.465 50.1 59.6 60.2 0.0 0.0 20 30 0.362 552.7 0.156 0.466 50.2 60.4 60.2 0.0 0.0 20 30 0.361 551.9 0.156 0.467 50.2 60.5 60.2 0.0 0.0 20 30 0.269 689.5 0.197 0.518 50.1 59.6 60.2 0.0 0.0 20 30 0.269 688.4 0.198 0.517 50.2 59.6 60.2 0.0 0.0 20 30 0.268 686.9 0.199 0.517 50.4 59.6 60.2 0.0 0.0 20 30 0.268 687.4 0.201 0.515 49.1 59.7 60.2 0.0 0.0 20 30 132 Appendix H

Data from test of Cell 7, Cell temp. 30-40°C, Gas pressures 1.0 bar Cell Cell Overpot Overpot Temp Temp Temp Pressure Pressure Flow Flow Potential Current Anode Cathode Cell Anode Cath Anode Cathode Anode Cath. Density gas gas rvi [mA] m m rci [q rq [Barg] [Barg] [mm] mm] 0.991 0.0 -0.001 0.000 30.1 40.2 39.4 0.0 0.0 20 30 0.990 0.0 -0.001 0.001 30.1 40.2 39.3 0.0 0.0 20 30 0.989 0.0 -0.001 0.002 30.1 40.2 39.3 0.0 0.0 20 30 0.887 1.8 0.000 0.103 30.1 40.1 39.5 0.0 0.0 20 30 0.886 1.8 0.000 0.104 30.1 40.1 39.4 0.0 0.0 20 30 0.885 1.8 0.000 0.104 30.1 40.2 39.4 0.0 0.0 20 30 0.881 2.2 0.000 0.109 30.2 40.1 39.5 0.0 0.0 20 30 0.880 2.2 0.000 0.110 30.2 40.1 39.5 0.0 0.0 20 30 0.879 2.2 0.000 0.111 30.1 40.1 39.5 0.0 0.0 20 30 0.857 4.3 0.001 0.131 30.2 40.2 39.4 0.0 0.0 20 30 0.856 4.3 0.001 0.133 30.2 40.2 39.4 0.0 0.0 20 30 0.855 4.3 0.001 0.133 30.2 40.2 39.5 0.0 0.0 20 30 0.832 8.3 0.003 0.155 30.3 40.3 39.4 0.0 0.0 20 30 0.831 8.3 0.003 0.156 30.3 40.3 39.4 0.0 0.0 20 30 0.830 8.3 0.003 0.157 30.2 40.3 39.4 0.0 0.0 20 30 0.794 19.2 0.007 0.189 30.3 40.4 39.4 0.0 0.0 20 30 0.794 19.2 0.007 0.189 30.3 40.4 39.4 0.0 0.0 20 30 0.793 19.2 0.007 0.190 30.3 40.3 39.4 0.0 0.0 20 30 0.764 33.8 0.012 0.213 30.3 40.3 39.4 0.0 0.0 20 30 0.763 33.8 0.012 0.214 30.3 40.4 39.4 0.0 0.0 20 30 0.709 69.9 0.024 0.257 30.2 40.1 39.7 0.0 0.0 20 30 0.707 69.9 0.025 0.258 30.3 40.2 39.6 0.0 0.0 20 30 0.706 69.8 0.025 0.259 30.4 40.2 39.5 0.0 0.0 20 30 0.671 101.4 0.035 0.284 29.7 40.1 39.9 0.0 0.0 20 30 0.668 100.9 0.035 0.287 30.0 40.1 39.8 0.0 0.0 20 30 0.553 212.6 0.075 0.362 29.7 40.3 40.0 0.0 0.0 20 30 0.549 210.8 0.075 0.366 29.3 40.2 40.0 0.0 0.0 20 30 0.498 281.7 0.097 0.395 31.5 40.7 39.4 0.0 0.0 20 30 0.493 278.9 0.098 0.399 31.2 40.6 39.6 0.0 0.0 20 30 0.491 277.4 0.098 0.401 30.8 40.6 39.7 0.0 0.0 20 30 0.489 277.4 0.099 0.402 30.7 40.5 39.8 0.0 0.0 20 30 0.488 277.4 0.099 0.403 30.5 40.5 39.9 0.0 0.0 20 30 0.377 408.0 0.151 0.463 32.0 40.8 39.3 0.0 0.0 20 30 0.372 408.0 0.151 0.468 31.9 40.8 39.3 0.0 0.0 20 30 0.370 406.7 0.150 0.470 31.7 40.8 39.3 0.0 0.0 20 30 0.313 475.1 0.184 0.493 31.8 40.5 39.6 0.0 0.0 20 30 0.312 475.1 0.183 0.495 31.9 40.6 39.5 0.0 do 20 30 0.311 473.8 0.181 0.498 32.0 40.7 39.4 0.0 0.0 20 30 0.310 473.3 0.179 0.500 32.0 40.7 39.3 0.0 0.0 20 30 0.220 556.3 0.229 0.541 31.5 40.5 39.6 0.0 0.0 20 30 0.218 553.2 0.231 0.541 31.0 40.5 39.7 0.0 0.0 20 30 0.218 543.2 0.232 0.541 30.9 40.5 39.7 0.0 0.0 20 30 0.216 543.2 0.233 0.541 30.6 40.5 39.8 0.0 0.0 20 30