molecular modeling of pem fuel cell electrochemistry

MOLECULAR MODELING OF PEM FUEL CELL ELECTROCHEMISTRY

A DISSERTATION SUBMITTED TO THE DEPARTMENT OF MECHANICAL ENGINEERING AND THE COMMITTEE ON GRADUATE STUDIES OF STANFORD UNIVERSITY IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

Varun Rai March 2008

c Copyright by Varun Rai 2008

All Rights Reserved

ii

I certify that I have read this dissertation and that, in my opinion, it is fully adequate in scope and quality as a dissertation for the degree of Doctor of Philosophy.

(Heinz Pitsch) Principal Adviser


(Friedrich B. Prinz)


(Wei Cai)

Approved for the University Committee on Graduate Studies.

iii

Abstract Polymer Electrolyte Membrane fuel cell (PEMFC) is an electrochemical power-generating device that combines hydrogen (H2 ) with oxygen (O2 ) via electrochemical (electron-transfer) processes to form water. Among other features, its low operating temperatures, high theoretical efficiencies, and quick startup make PEMFC promising as a power source for several applications: laptops, small-scale power generation, automobiles, etc. Although tremendous progress has been made in PEMFC technology in the last decade or so, a number of key technological issues still remain to be addressed before PEMFC comes to the consumer markets. The two main challenges are the cost and the performance of the PEMFC technology. A major contributor to both these limitations is the catalyst used in PEMFC — used to improve the efficiency of PEMFC by enhancing the rate of the electrochemical reactions, catalyst materials form over 20% of the total system cost. And yet, about 15-20% of the theoretical efficiency of PEMFC is lost due to poor reaction rates at the catalysts. For its superior catalytic activity for the oxygen reduction reaction (ORR), which is the main electrochemical reaction in PEMFC, Pt remains the leading choice for the cathode catalyst in PEMFC. So, a solid grasp of the fundamental electrochemical processes at the Pt-electrolyte interface is necessary for the design of optimal catalysts (based on cost, activity, stability, and tolerance to contaminants) for PEMFCs. In this thesis, molecular modeling techniques were used to study the ORR on Pt, which resulted in two main contributions. First, two new computational algorithms for simulating advanced catalyst systems were developed. Second, the ORR

iv

on Pt(111) in acid solutions was studied using a combination of first-principles simulation methodologies, where both Dynamic Monte Carlo (DMC) simulations and Density Functional Theory (DFT) quantum simulations were employed. The results from this study provided several new insights about the nature of the adsorbates and the reaction intermediates involved in the ORR, and about the reaction pathways of the ORR on Pt.

v

Preface One of the early researchers and contributors to polymer electrolytes Prof. Peter V. Wright identifies the 1970s as “a decade of change from a classical period of polymer science, in which the foundations of the chemistry and physics of the subject were established alongside the consumer plastics industry, to a period in which scarcer funds became directed towards increasingly interdisciplinary and collaborative projects and in which commercial application and environmental factors became increasingly prominent.” [118] The invention of the Polymer Electrolyte Membrane (PEM) technology is accredited to the work of Thomas Grubb and Leonard Niedrach at General Electric in the early 1960s. GE announced an initial success in the mid-1960s with the development of a small fuel cell. PEM technology served as part of NASA’s Project Gemini in the early days of the U.S. piloted space program. The initial model encountered repeated technical difficulties, including cell contamination and leakage of oxygen through the membrane. Lack of reliability of the intial PEM fuel cells (PEMFC) led to the choice of batteries for the Gemini 1 through 4 missions. Later, a major setback for research in PEMFC came when Project Apollo mission-planners chose to use alkali fuel cells for both the command and lunar modules, as did designers of the Space Shuttle a decade later. This virtually put most of the fundamental R&D in PEMFC on shelf for about 20 years. The mid-1980s saw a resurgence of research in PEMFCs after a relatively dormant period. This revival of interest in PEMFCs came around when Ballard Power Systems began a 2-year contract with the Canadian Department of National Defense [90]. Since then, there has been tremendous progress in this area and a lot of promise has

vi

been demonstrated to this date. This thesis builds on the shoulders of the great work that has already been done in the field of PEMFCs. It is my sincere belief that the new insights from this thesis are worthy additions to what is already known about PEMFCs.

vii

Acknowledgements I thank Prof. Heinz Pitsch for being such a great advisor, teacher, and above all, a mentor and role model. Of late, I have come to realize how much more academic and intellectual freedom I have enjoyed with Heinz than I initially thought I had. Thanks also to Prof. Parviz Moin for his advising and support during my first year at Stanford. Many thanks to Honda R&D, Co. for their continued faith in my research. I would especially like to mention Mr. Katsunori Makino of Honda for his very helpful suggestions and discussions about my research. I also thank my reading committee members, Prof. Fritz Prinz and Prof. Wei Cai, who, in spite of their busy schedules, were available whenever I needed them. The love, care, and concern that my parents, Vidya and Shanti, have shown during my Stanford years is ineffable. Notwithstanding the enormous physical distance — they live in India — I have always found them very close to me, supporting me unconditionally during my intellectual journey at Stanford. I thank them. Thanks also to my siblings Neelam and Rahul, and to my sister-in-law Shraddha for their love and support. The years at Stanford couldn’t have been as much fun, but for my amazing roommates, Venky and Catherine, and StanfordGangs, my closest friends at Stanford. Finally, many thanks to all those who have enriched my learning experience, both academic and otherwise, over the years — my office mates Masoud and Abhishek, members of the Pitsch Group, fabulous teachers at Stanford and IIT Kharagpur (my undergrad school), Prof. Gerbrand Ceder and Byungchan Han of MIT, and other friends and well wishers — Thank you all.

viii

Nomenclature Abbreviations CGS

Complex Geometry Specification

CICV

Convergent Iterative Constraint Variation

CV

Cyclic Voltammetry

DFT

Density Functional Theory

DMC

Dynamic Monte Carlo

DMFC

Direct Methanol Fuel Cell

DOE

Department of Energy, U. S.

EA

Electron Affinity

ECR

Electrochemical Reaction

FCV

Fuel Cell Vehicle

fFRM

Fast First Reaction Method

FRM

First Reaction Method

FSB

First-Principles Site Blocking

GDL

Gas Diffusion Layer

GHG

Greenhouse Gases

HOR

Hydrogen Oxidation Reaction

ICE

Internal Combustion Engine

iid

independent and identically distributed

IP

Ionization Potential

LRC

Local Reaction Center Theory

NN

Nearest Neighbor

NNN

Next-Nearest Neighbor

ix

OCV

Open Circuit Voltage

ORR

Oxygen Reduction Reaction

PDAE

Potential-Dependent Activation Energy

PEM

Polymer Electrolyte Membrane

PEMFC

Polymer Electrolyte Membrane Fuel Cell

RDS

Rate Determining Step

RRDE

Rotating Ring Disc Electrode

RSM

Random Selection Method

rVSSM

rejection-free VSSM

SHE

Standard Hydrogen Electrode

t-b-t

top-bridge-top

t-f-b

top-fcc-bridge

UHV

Ultra-High Vacuum

VSSM

Variable Step Size Method

Symbols α

charge transfer coefficient

β

Bronsted-Hammond factor

∆Gf

change in the Gibbs free energy

∆T

time increment

interaction parameter

λ

membrane hydration number

c

configuration or state of a catalyst surface

T

temperature

ν0

pre-factor for rate coefficient

θA

surface coverage by adsorbate A

TToR,j

the minimum of the times of reaction among all enabled events of the reaction type j

E

0

EN

standard cell potential Nernst potential x

Eact

activation energy

F

Faraday constant

F ()

cumulative distribution function

H

the number of events in L (for FRM)

j0

exchange current density

j0∗

intrinsic current density

jk

kinetic current

N

number of lattice sites

nen,j

enabled events of the reaction type j

p

pressure

P ()

probability of an event

R

universal gas constant (8.314 J/K/mol)

r

the number of reaction types

t, T

time

T0

current or initial system time

TToR

time of execution of a reaction

u

a random number 0

U (U )

electrode potential with respect to the SHE (standard)

Li

event list for reaction type i

xi

Contents Abstract

iv

Preface

vi

Acknowledgements

viii

Nomenclature

ix

1 Introduction

1

1.1

Energy Scenario and the Role of Fuel Cells . . . . . . . . . . . . . . .

1

1.2

PEM Fuel Cells . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3

1.2.1

Advantages . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3

1.2.2

Challenges . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4

Scope and Organization of This Thesis . . . . . . . . . . . . . . . . .

6

1.3

2 Fundamentals of PEM Fuel Cells

10

2.1

Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

10

2.2

Thermodynamics of PEMFC . . . . . . . . . . . . . . . . . . . . . . .

12

2.3

The Triple-Phase Region . . . . . . . . . . . . . . . . . . . . . . . . .

14

2.4

The Membrane . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

14

2.5

Water Transport in PEMFC . . . . . . . . . . . . . . . . . . . . . . .

19

2.6

Catalysts for PEM Fuel Cells . . . . . . . . . . . . . . . . . . . . . .

21

2.7

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

25

xii

3 PEM Fuel Cell Electrochemistry

27

3.1

Anode Reaction: Hydrogen Oxidation . . . . . . . . . . . . . . . . . .

27

3.2

Cathode Reaction: Oxygen Reduction . . . . . . . . . . . . . . . . .

30

3.2.1

Established Features of the ORR . . . . . . . . . . . . . . . .

30

3.2.2

Mechanisms for the ORR . . . . . . . . . . . . . . . . . . . . .

33

3.2.3

Adsorbed Species During the ORR . . . . . . . . . . . . . . .

38

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

40

3.3

4 Efficient Surface Chemistry Simulations

42

4.1

Limitations of the Mean-Field Approach . . . . . . . . . . . . . . . .

43

4.2

The DMC Approach . . . . . . . . . . . . . . . . . . . . . . . . . . .

48

4.3

DMC Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

52

4.3.1

Event Lists . . . . . . . . . . . . . . . . . . . . . . . . . . . .

52

4.3.2

VSSMb . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

54

4.3.3

FRM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

55

A New DMC Algorithm: fFRM . . . . . . . . . . . . . . . . . . . . .

57

4.4.1

Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

57

4.4.2

Determination of TToR,j . . . . . . . . . . . . . . . . . . . . . .

58

4.4.3

Algorithmic Steps in fFRM . . . . . . . . . . . . . . . . . . .

60

4.5

Accuracy of fFRM . . . . . . . . . . . . . . . . . . . . . . . . . . . .

63

4.6

Performance of DMC Algorithms . . . . . . . . . . . . . . . . . . . .

64

4.7

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

67

4.4

5 DMC Simulations on Nanoparticle Surfaces

69

5.1

The Complex Geometry Specification Method . . . . . . . . . . . . .

70

5.2

Examples of Surfaces Generated with the CGS Method . . . . . . . .

73

5.3

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

74

6 Adsorbate Formation at the Pt(111) Electrode

78

6.1

The Water Discharge Reaction . . . . . . . . . . . . . . . . . . . . . .

78

6.2

Computational Approach . . . . . . . . . . . . . . . . . . . . . . . . .

79

6.3

Calculation of PDAE of Electron Transfer Reactions . . . . . . . . . .

80

xiii

6.4

6.5

6.6

6.3.1

Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

80

6.3.2

Method . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

81

Mechanism for Adsorbate Formation . . . . . . . . . . . . . . . . . .

83

6.4.1

Structure of the Water Layer . . . . . . . . . . . . . . . . . .

83

6.4.2

Adsorbate Interactions . . . . . . . . . . . . . . . . . . . . . .

85

Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . .

87

6.5.1

Activation Energies for Electrochemical Steps . . . . . . . . .

87

6.5.2

DMC Simulations . . . . . . . . . . . . . . . . . . . . . . . . .

88

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

95

7 Oxygen Reduction Reaction on Pt(111)

96

7.1

Mean-Field Model for ORR Kinetics . . . . . . . . . . . . . . . . . .

96

7.2

First-Principles Site-Blocking Model for the ORR . . . . . . . . . . .

98

7.2.1

99

7.3

Calculation of the Site-Availability Function, f (θads )

. . . . .

Applications of the FSB Model . . . . . . . . . . . . . . . . . . . . . 102 7.3.1

ORR in HClO4 . . . . . . . . . . . . . . . . . . . . . . . . . . 105

7.3.2

ORR in H2 SO4 . . . . . . . . . . . . . . . . . . . . . . . . . . 109

7.4

Reaction Intermediates in the ORR . . . . . . . . . . . . . . . . . . . 113

7.5

FSB Model for the Series2 Pathway . . . . . . . . . . . . . . . . . . . 118

7.6

Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

8 Conclusions

123

A Order Statistics in the fFRM Algorithm

125

Bibliography

128

xiv

List of Tables 44

4.1

Values of the rate constants used in the mean-field model . . . . . . .

7.1

Effect of OH(ads) on the adsorption energy of O2 . . . . . . . . . . . . . 116

7.2

Mechanistic pathways for the ORR on Pt(111) in 0.1 M HClO4 . . . . 120

xv

List of Figures 1.1

World energy outlook . . . . . . . . . . . . . . . . . . . . . . . . . . .

2

1.2

Loss mechanisms in a typical PEMFC . . . . . . . . . . . . . . . . . .

5

2.1

Basic processes and components in a PEMFC . . . . . . . . . . . . .

11

2.2

Nano-phase separation in Nafion . . . . . . . . . . . . . . . . . . . . .

15

2.3

Evolution of the membrane structure as a function of water content or hydration number . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

17

2.4

Modes of water transport in a PEMFC . . . . . . . . . . . . . . . . .

20

2.5

TEM image and model of a Pt nanoparticle . . . . . . . . . . . . . .

23

2.6

TEM micrographs for Pt-Ru/C catalysts . . . . . . . . . . . . . . . .

25

3.1

Tafel plots for ORR on Pt(111) in oxygen saturated 0.1 M HClO4 and 0.05 M H2 SO4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.2

Rotating-ring-disk electrode experiments of the oxygen reduction on Pt(hkl) in 0.05 M H2 SO4 . . . . . . . . . . . . . . . . . . . . . . . . .

3.3

32 37

Adsorbate coverage in nitrogen-saturated 0.1 M HClO4 (circles) and 0.05 M H2 SO4 (squares) solutions . . . . . . . . . . . . . . . . . . . .

39

3.4

Possible configurations for the adsorption of O2 . . . . . . . . . . . .

41

4.1

Comparison of results from DMC (stars) with the exact solution (solid lines) for the time evolution of the reaction mechanism presented in the text. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

4.2

45

Breakdown of the mean-field approach for simple surface chemistry simulations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

xvi

46

4.3

Pattern formation and kinetic phase transitions in the CO oxidation process on a Pt(111) surface with defects . . . . . . . . . . . . . . . .

4.4

Coverage-potential curve obtained by different DMC algorithms for the mechanism given by Eqs. 4.23–4.25 . . . . . . . . . . . . . . . . . . .

4.5

47 64

6

CPU time needed by different DMC algorithms for simulating 5×10

reaction steps at steady state of the chemical mechanism discussed in Section 4.6 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.6

66

Total number of function calls used by different DMC algorithms for simulating 5×106 reaction steps at steady state of the chemical mechanism discussed in Section 4.6 . . . . . . . . . . . . . . . . . . . . . .

67

5.1

Edge sharing in cubo-octahedral geometry . . . . . . . . . . . . . . .

71

5.2

Numbering of edges and calculation of site coordinates in the CGS method

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

72

5.3

Unification process for common edges . . . . . . . . . . . . . . . . . .

73

5.4

Use of the CGS method to generate THex and COG surfaces . . . . .

74

5.5

Surface of a truncated-hexagon generated using the CGS method . .

75

5.6

Surface of a cubo-octahedral generated using the CGS method . . . .

76

5.7

Examples of the application of the CGS method in DMC simulations to obtain the instantaneous species distribution on the surface of catalyst nanoparticles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . √

√

77

3R30◦ structure of water on Pt(111) . . . . . . . . . . . . . .

84

6.2

Optimized structure of Pt-Pt.OH2 [H2 O]3 . . . . . . . . . . . . . . . .

85

6.3

Potential-dependent activation energies for reactions (6.1) and (6.2) .

88

6.4

Rate coefficients for reactions (6.1) and (6.2) . . . . . . . . . . . . . .

89

6.5

Total coverage of O-containing species obtained from DMC simulations

6.1

3×

and mean-field method compared with experimental data . . . . . . . 6.6

Snapshots of the adsorbate layer on Pt(111) in 0.1 M HClO4 solution at low electrode potentials . . . . . . . . . . . . . . . . . . . . . . . .

6.7

91 93

Snapshots of the adsorbate layer on Pt(111) in 0.1 M HClO4 solution at high electrode potentials

. . . . . . . . . . . . . . . . . . . . . . . xvii

94

7.1

Site-blocking models for O2 -on-bridge . . . . . . . . . . . . . . . . . . 100

7.2

Site-blocking models for O2 -on-hollow . . . . . . . . . . . . . . . . . . 100

7.3

Coverage and structure of SO∗4 on Pt(111) . . . . . . . . . . . . . . . 103

7.4

Coverage and site-availability function f (θads ) for adsorbates in 0.1 M HClO4 and 0.05 M H2 SO4 . . . . . . . . . . . . . . . . . . . . . . . . 104

7.5

Results of the FSB model for the ORR on Pt(111) in 0.1 M HClO4 . . 107

7.6

Change in the Tafel slope

7.7

Sensitivity analysis for OH for the O2 -on-bridge case for the ORR on

. . . . . . . . . . . . . . . . . . . . . . . . 109

Pt(111) in 0.1 M HClO4 . . . . . . . . . . . . . . . . . . . . . . . . . . 110 7.8

Sensitivity analysis for O for the O2 -on-bridge case for the ORR on Pt(111) in 0.1 M HClO4 . . . . . . . . . . . . . . . . . . . . . . . . . . 111

7.9

Sensitivity analysis for OH for the O2 -on-hollow case for the ORR on Pt(111) in 0.1 M HClO4 . . . . . . . . . . . . . . . . . . . . . . . . . . 112

7.10 Results of the FSB model for the ORR on Pt(111) in 0.05 M H2 SO4 .

114

7.11 Sensitivity analysis for SO∗4 for the O2 -on-bridge case for the ORR on Pt(111) in 0.05 M H2 SO4 . . . . . . . . . . . . . . . . . . . . . . . . . 115 7.12 Effect of OH(ads) on the adsorption energy of O2 . . . . . . . . . . . . 117 7.13 Results of the FSB model for the ORR on Pt(111) in 0.1 M HClO4 with the Series2 pathway.

. . . . . . . . . . . . . . . . . . . . . . . . 121

xviii

Chapter 1 Introduction 1.1

Energy Scenario and the Role of Fuel Cells

The Energy Information Administration (U. S. Department of Energy (DOE)) projects the world marketed-energy consumption to increase by 57 percent from 2004 to 2030 (Fig. 1.1). A majority of the new demand will be in developing countries: The total energy demand in the non-OECD (Organization for Economic Co-operation and Development) countries is projected to increase by 95 percent, compared with an increase of 24 percent in the OECD countries. Such a massive increase in consumption does not augur well for reduction in the emissions of human-caused greenhouse gases (GHG). The world carbon dioxide emissions are projected to continue increasing from 26.9 billion metric tons in 2004 to 33.9 billion metric tons in 2015 and 42.9 billion metric tons in 2030. This represents an increase of 59 percent over 2004–2030 [1]. At present, the U. S. imports 60% of its oil needs, a number set to touch 70% in a few years. Of all the end uses of oil in the U.S. roughly 70% is used for transportation. Concerns about global warming and geopolitical instability in regions of maximum oil reserves combined with highest-ever gasoline prices have spurred widespread attention for alternative fuel sources. In the transportation sector, hydrogen-powered fuel-cell automobiles have been pegged as the longer-term solution to the looming energy crisis. Presently, the fuel cell technology most prominently being developed for automobiles is called the Polymer Electrolyte Membrane fuel cell (PEMFC). 1

CHAPTER 1. INTRODUCTION

2

Figure 1.1: World energy outlook. Taken from [1]. Sources: 2004: Energy Information Administration (EIA), International Energy Annual 2004 (May-July 2006), web site www.eia.doe.gov/iea. Projections: EIA, System for the Analysis of Global Energy Markets (2007). A number of aggressive R&D initiatives are being pursued to rapidly advance fuel cell technologies. In the U.S., the major initiatives are the Hydrogen Fuel Initiative and the FreedomCAR (Cooperative Automotive Research) program. Over $2 billion are planned for basic and industrial research through these programs. The goal is to enable large scale manufacturing of fuel cells, and to bring fuel cell vehicles to the customer marketplace by 2020. The national vision of America’s transition to a hydrogen economy sums the hopes from fuel cells as “Hydrogen is America’s clean energy choice. Hydrogen is flexible, affordable, safe, domestically produced, used in all sectors of the economy and in all regions of the country.” [37] Undoubtedly, commercial success in fuel cell technology will not only alleviate much of America’s energy problems, but will have a great impact on the global energy landscape with long-lasting changes to the human society at large.


1.2

3

PEM Fuel Cells

The importance and promise of PEMFCs for future-generation energy systems can be gauged from the following statement by Mark Hoberecht of Glenn Fuel Cell Technology: “PEM fuel cells are leading the way, having emerged as the leading fuel cell technology for near-term commercial applications. NASA recognizes the valuable attributes of PEM fuel cells, and is partnering with commercial vendors to adapt this technology for future space applications.” [43]

1.2.1

Advantages

Of a number of fuel cell technologies available, the wide range of power output (few Watts to hundreds of kiloWatts) and a low operating temperature make PEMFCs suitable for applications ranging from distributed power generation to powering small portable devices such as cellular phones and laptops. But probably the most important application of PEMFCs is their potential as future supplements or, in the longer-term, as replacements for internal combustion engines (ICE) for vehicular applications. Compared to conventional gasoline ICEs, some of the advantages that hydrogen-based PEMFCs offer are: • Use of renewable fuel. Hydrogen can be obtained renewably by splitting water using wind or solar energy. Different technologies for the renewable generation of hydrogen are being researched very actively along with conversion devices. • Environment friendliness. Fuel cell vehicles (FCV) running on renewable hydrogen have negligible GHG emissions. Again, best results are realized when renewable energies such as solar, wind, or biomass power are used to produce the hydrogen. • High achievable values of power density and current density. Both these features make PEMFCs particularly attractive for automobiles. • Low operating temperatures (30-80◦ C). This is advantageous for vehicles as well as portable devices like laptops, cell phones, camcoders.


4

• Structural modularity and simplicity. The solid and immobile electrolyte (the membrane) makes the overall cell inherently simple. • Lower cost of fabrication. Low operating temperatures of PEMFCs allow use of less exotic materials, making them more suitable for mass-market applications. • Quick startup. This is very important for vehicular applications. • Low noise. PEMFCs are quiet, as they do not have any moving parts. These advantages make PEMFCs one of the most researched fuel cell systems in recent times.

1.2.2

Challenges

Although tremendous progress has been made in PEMFC technology in the last decade or so, a number of key technological issues still remain to be solved before this technology arrives in the consumer markets. The two main challenges are the cost and performance of the PEMFC technology. Cost At present, estimated cost of a full fuel cell power system stands at about $100/kW. This estimate assumes current PEMFC technology with a production of 500,000 units per year. In reality, the production volumes are much smaller, running into just a few hundreds of units. So, the actual cost of such system is sometimes over $1000/kW [37, 61]. In comparison, cost of energy generated in present-day internal combustion engines (ICE) is about $10/kW. On a life-cycle basis, cost of PEMFC needs to be brought down to $30/kW before this technology can compete with ICE. Catalyst materials, mostly Platinum (Pt) or Pt-based, often contribute 20-25% to the cost of PEMFC systems [37].


5

1.23 Fuel crossover loss

Cell Potential (V)

1.00 Activation loss 0.75 Ohmic loss 0.50 Mass-transport loss 0.30 0

Current Density (A/cm2) 1.00 Figure 1.2: Loss mechanisms in a typical PEMFC.

Performance Efficiencies of present PEMFC systems range between 35-45% [61], although the theoretical limits are nearly 80% efficiency for low temperature fuel cells like PEMFC. The DOE has set a target of achieving 60% peak-efficiency for PEMFCs by 2015. The performance of fuel cells is usually described in terms of a voltage-current curve. This curve is called the V-I curve or the polarization curve. The V-I curve is obtained by measuring the voltage output of the fuel cell at different currents, or viceversa. A typical V-I curve for a PEMFC is shown in Fig. 1.2. Different performance loss mechanisms are also indicated in the same figure. Fuel crossover or internal current losses are caused when the fuel crosses over from the anode to the cathode via the membrane. Even though the ideal membrane should only be permeable to protons (in PEMFC), in reality small amounts of both


6

electrons and fuel manage to reach the cathode via the membrane. This results in losses, since the crossover electrons and fuel are just wasted. Ohmic losses are due to the flow of current through the finite resistances associated with the membrane, the cell interconnects and the bipolar plates, and is simply given by the product of the cell current and the total circuit resistance (iR). Mass-transport losses occur at high current densities (≈1 A/cm2 or above) where consumption of the fuel and oxygen results in a drop in their partial pressure in the supply streams. This is more pronounced at high current densities when the pressure drop is further aggravated by the mass transport limitations of the supply channels, which may get clogged due to excessive production of liquid water. This can dramatically reduce the operating voltage. Activation or overpotential losses are caused by the finite rate chemistry, i.e., slow reaction kinetics at the electrodes. In hydrogen-fueled PEMFCs, these losses occur most significantly at the cathode, where an active catalyst material is used to enhance the rate of the oxygen reduction reaction (ORR). But even on the best catalysts like Pt, activation losses due to slow rate of the ORR lead to an efficiency loss of around 20%.

1.3

Scope and Organization of This Thesis

For the ORR to be completed, the O–O bond in oxygen (O2 ) needs to be broken. Catalysts are used to reduce the activation barrier for breaking this bond, thereby enhancing the reaction rate. Rate of the ORR on Pt is much faster compared with other catalyst materials, but still it is quite slow compared with the electrochemical reactions at the anode. A parameter that characterizes the reaction kinetics at the electrodes is the exchange current density (j0 ); a higher value of j0 means faster kinetics. For PEMFCs with Pt catalyst, a typical value of j0 , based on real unit catalyst surface area, is 10−9 A/cm2 at the cathode and 10−3 A/cm2 at the anode [81]. Thus, the reaction kinetics at the anode are much faster compared to the cathode. Slow kinetics of the ORR is one of the major loss mechanisms in PEMFCs. Besides an inherently low ORR activity of Pt, the ORR kinetics are further slowed down by the presence of adsorbate species like OH and O. These adsorbates may slow


7

down the ORR kinetics by blocking Pt sites, thereby reducing the availability of active sites for breaking the O–O bond, and/or by altering the adsorbtion energy of reaction intermediates through electronic interactions with them. Given the bearing catalysts have both on the cost and on the performance of PEMFCs, the importance of the ORR cannot be overstated. But, notwithstanding the large body of work that has been done on the ORR, many details about the ORR mechanism, particularly the reaction intermediates and competing reaction pathways, remain unresolved. Part of the reason for this lies in that most of the present research is focused at developing novel catalysts even though very fundamental questions about the ORR have not been completely answered. Another reason is the complex nature of the problem that restricts the use of available experimental methods to probe molecular-level details at the metal-electrolyte interface. Platinum, for its superior activity for the ORR, remains the leading choice for the cathode catalyst in hydgrogen-based PEMFCs. Development of advanced catalysts for the ORR requires a solid grasp of the electrochemical processes of the ORR at the low index Pt surfaces ((111), (110), and (100)). That would set a strong platform for linking fundamental electrode processes at the molecular level to macroscopic behavior of more complex catalyst materials like Pt nanoparticles. Experimental work from the group of P. N. Ross Jr. at Lawrence Berkeley National Laboratories has been particularly outstanding in enhancing the understanding of the ORR at Pt single crystal surfaces [70, 73]. In this thesis, molecular modeling methods were used to develop a better understanding of the ORR at the Pt-electrolyte interface. Specifically, a combination of first-principles simulation methodologies employing both Dynamic Monte Carlo (DMC) simulations and Density Functional Theory (DFT) quantum simulations were used to probe the ORR on Pt(111) in acid solutions. The results of this thesis provide several new insights about the nature of the adsorbates and the reaction intermediates, and about the reaction pathways of the ORR on Pt. Chapters 2 and 3 introduce the PEMFC system and the associated electrochemistry in more detail, while Chapters 4-7 present the results obtained in this thesis. Chapter 2 introduces the basic principles of PEMFCs. The triple-phase region,


8

polymer electrolyte membrane (PEM), and water-transport issues in PEMFCs are discussed in more detail. The chapter ends with a discussion on the catalyst materials for the electrodes of PEMFCs. The electrochemical reactions at the electrodes of PEMFCs, which are the focus of this thesis, are discussed in Chapter 3. Key issues about the ORR that are investigated in the later chapters are highlighted. In Chapter 4, the mathematical and computational framework of the DMC method, which is used in Chapter 6 for DMC simulations of PEMFC electrochemistry, are laid down first. Next, the Fast First Reaction Method (fFRM) algorithm developed in this thesis is presented. fFRM is the first DMC algorithm for which the computational performance is independent of the lattice size even for systems with time-varying rate coefficients, which makes it applicable to the study of a wide range of chemical systems on large lattice structures. Chapter 5 presents the Complex Geometry Specification (CGS) method, another algorithm that was developed in this thesis, that enables to perform detailed DMC simulations for nanoparticles. In Chapter 6, the formation of adsorbates (Oads and OHads ) on Pt(111) in an acidic electrochemical environment is studied using a combination of DFT calculations and DMC simulations. Results from the DMC simulations reveal the partial coverage by the adsorbates at different potentials. OHads is shown to be the dominant adsorbate between 0.5–0.8 V, but above 0.8 V OHads and Oads co-exist. These are the first detailed first-principles simulations of the kinetics of O-containing adsorbates formed from the electrochemical discharge of water on Pt(111) in acidic solutions. The findings reported in this chapter provide crucial insights about the identity of the adsorbates and the structure of the adsorbate layer at the cathode of PEMFCs. A new first-principles based reaction kinetics model, called the First-Principles Site-Blocking (FSB) model, for the ORR on Pt(111) is presented in Chapter 7. The FSB model is based on the detailed adsorbate layer structures that were obtained in Chapter 6. Results from the FSB model provide conclusive computational evidence that the O2 intermediate in the rate determining step (rds) of the ORR on Pt(111) in acidic solutions is bridge-bonded (top-bridge-top). The same model is shown to


9

describe the ORR in two different acid solutions (HClO4 and H2 SO4 ) very well. Further, the FSB model is used to make predictions about the reaction intermediates and the reaction mechanism of the ORR on Pt(111). Based on these results, it was concluded that up to 0.9 V the ORR on Pt(111) in 0.1 M HClO4 proceeds primarily via a pathway, termed Series1, which involves the dual-site bonded O2 H intermediate. But above this potential, a pathway involving the single-site H2 O2 intermediate, called here the Series2 pathway, becomes operative. It is also shown that the most likely product of the rds for the Series1 pathway is O2 H(ads) , not O− 2(ads) , which has been suggested in the literature [99]. Finally, these results provide the insight that owing to the energetically-favorable bridge-bonded adsorbtion of oxygen on Pt, the Series1 pathway is more dominant than the Series2 pathway.

Chapter 2 Fundamentals of PEM Fuel Cells The advantages and technological challenges in the PEMFC technology were outlined in Chapter 1. This chapter presents the basic principles and processes involved in the operation of PEMFCs.

2.1

Introduction

Fuel cells are galvanic cells in which the Gibbs free energy of a chemical reaction is converted into work via an electrical current. Specifically for PEMFCs, hydrogen atoms dissociate into protons (H+ ) and electrons (e− ) at the anode. The electrons pass through an external circuit/load producing electricity while the H+ ions move across the proton-conducting membrane. On reaching the cathode, these electrons and protons react with the supplied oxygen to form water, which is the product of the fuel cell. This process is illustrated in Fig. 2.1. Half cell reactions and the total reaction of the PEM fuel cell are given as H2 → 2H+ + 2e−

U0 = 0.0 V H2 / H+ 0 cathode reaction 12 O2 + 2H+ + 2e− → H2 O UO = 1.23 V /H O anode reaction

2

total reaction

H2 + 12 O2 → H2 O

2

0 Ecell = 1.23 V.

(2.1) (2.2) (2.3)

The protons (H+ ) are usually solvated in water molecules. This solvated state of 10

CHAPTER 2. FUNDAMENTALS OF PEM FUEL CELLS

Fuel outlet

Air/water outlet

Gas diffusion layer (GDL)

H+

Cathode

Membrane

Air inlet Anode

Fuel inlet

11

Membrane

Air/O2

Catalyst (Pt)

Carbon particles

Figure 2.1: (Left) Illustration of the basic processes and components in a PEMFC. (Right) The triple-phase region. protons is called the hydronium ion and is represented as H3 O+ . The half-cell reactions exhibit standard potentials of 0.0 V for the hydrogen oxidation (U 0 ) and H 2 / H+ 0 1.23 V for the oxygen reduction (UO ). This results in a standard reversible cell 2 / H2 O 1 potential of 1.23 V. In practical systems, neither the fuel stream contains pure hydrogen nor does the oxygen stream contain pure oxygen. The oxygen stream at the cathode side is usually air, and so about 77% by mass of it is nitrogen. Further, in order to maintain adequate water content inside the membrane – a topic discussed in more detail in 1

Standard potentials are based on the standard hydrogen electrode, SHE, i.e., pH = 0, T = 298 K, preactant = 1 bar.


12

Sec. 2.5 – both streams are often humidified before entering the inlet manifolds. This kind of a dilute supply of the reactants to the cell leads to a decrease in the Gibbs free energy of the chemical reaction in the cell. This phenomenon can be described using the Nersnt equation, as discussed next.

2.2

Thermodynamics of PEMFC

The maximum electrical work (Wel ) obtainable in a fuel cell is determined by the product of the reversible cell potential E 0 and the charge Q (Q = nF ), and is thus given by a change in the Gibbs free energy at standard conditions (T = 298 K, preactant = 1 bar): Wel = ∆G0f = nF E 0 ,

(2.4)

where n is the number of electrons participating in the electrochemical reaction and F is the Faraday constant. The ∆Gf is the difference between the Gibbs free energy of the products and the Gibbs free energy of the reactants ∆Gf = Gf of products − Gf of reactants. To make the comparisons easier, it is convenient to consider these quantities in their “per mole” form. These are indicated by a bar over the lower case letter. For example, (g f )H2 O is the molar Gibbs free energy of the formation of water. Consider the basic reaction for the hydrogen/oxygen fuel cell: 1 H2 + O2 → H2 O, 2 for which we have 1 ∆g f = (g f )H2 O − (g f )H2 − (g f )O2 . 2 According to Eq. (2.4) standard open circuit voltage (OCV) E of the fuel cell is given by E0 =

−∆g 0f . 2F

(2.5)


13

For example, a hydrogen fuel cell operating at 25◦ C has ∆g 0f = −237.2 kJ/mol, so E0 =

237.2 V = 1.23 V. 2 × 96485

∆g f of a chemical reaction varies with temperature and reactant pressure. In general, then, the actual OCV is represented by the Nernst equation 

EN = E0 +

1 2



RT  pH2 .pO2  ln . 2F pH2 O

(2.6)

The efficiency of fuel cells is not limited by the Carnot efficiency, since in principle ∆g f of the cell’s reaction can be converted into electricity. As such, fuel cells represent an attractive alternative to combustion engines, for which the efficiency is limited by the Carnot efficiency. But in practice, a part of the Gibbs energy of the fuel cell reaction gets dissipated as heat via different loss mechanisms. The different loss mechanisms in a PEMFC were discussed in Sec. 1.2. For comparison, the efficiency of PEMFCs ranges between 40%-50%, while the efficiency of internal combustion engines ranges between 30%-40% [61]. The voltage of a single fuel cell is quite small, about 0.7 V when drawing a useful current. Consequently, many cells have to be connected in series in order to produce a useful voltage. Such a collection of cells in series is called a stack. A popular method of cell interconnection is through the use of bipolar plates [61, 81]. This method ensures that there is connection all over the surface of one cathode and the anode of the next cell, thereby minimizing interconnection losses. The bipolar plates also serve as a means for feeding oxygen stream to the cathode and fuel stream to the anode. Here it becomes important that the material of the bipolar plate is such that it keeps the two gas supplies strictly separated.


2.3

14

The Triple-Phase Region

The electrochemical reaction at the cathode of a PEMFC is the oxygen reduction reaction (ORR; Eq. (2.2)). For the ORR to take place, oxygen, H+ ions, and electrons all must come in close proximity. This means that there should be a region of overlap between the gas-phase pores (for the supply of oxygen), the membrane (for the supply of protons) and the electrode (for the supply of electrons). This region, called the triple-phase contact, is illustrated in Fig. 2.1. It is only in this region that the electrochemical reactions take place and where chemical species are produced or consumed [61, 81, 82]. As such, maximization of the 3-phase region is very important in fuel-cell design. Note that the water produced during the cell’s operation must be removed away from this region towards the gas supply channels. The hydrophobic Teflon often added to the porous gas diffusion layer (GDL) helps in removing this excess water.

2.4

The Membrane

Fuel cells generate electrical energy by splitting the fuel into ions and electrons at the anode using an electrolyte. The electrolyte is an electronic insulator that can conduct the generated ions very well. As such, the electrolyte acts as a species-separator that screens the electrons out, which can then be used in an electrical circuit as a source of power. In the case of PEMFC, the electrolyte is a polymeric membrane called the proton exchange or polymer electrolyte membrane (PEM). The PEM is a good conductor of protons, which are generated at the anode of a PEMFC. Thus, the role of the membrane is central to the proper working and performance of a PEMFC. The basic theme of most polymer membrane electrolytes is the use of sulphonated fluoropolymers. The backbone polymeric material (fluoropolymer) is hydrophobic, but sulphonation, i.e., the addition of sulphonic acid (HSO3 ) as side chains, creates hydrophilic regions within the polymer. For well hydrated membranes, dissociation + of sulfonic end-groups generate -SO− ions, where the former constitute 3 and H3 O

hydrophilic regions in the membrane and the latter are the major charge carriers


15

- -SO3

+ Protonic charge carrier H2O

Figure 2.2: Schematic representation of the nano-phase separation in the hydrated morphology of Nafion, derived from experiments and modeling. The figure shows the distinctions in the hydrophilic/hydrophobic separation, connectivity of the water and ion domains, and separation of SO− 3 groups, from [60]. (Fig. 2.2). R Nafion is probably the most studied and operated electrolyte for PEM fuel cells,

but other per-fluorocarbon sulfonic acid membranes, for example from Dow Chemicals and Gore and Asahi Chemicals, are also used and investigated [111]. The stability of these membranes is limited to a small temperature range and they are still under investigations by several research groups. The upper limit of temperature is dictated


16

by the need for humidification of the membrane, because liquid water is a prerequisite for proton conduction. Hydrophilic clusters in Nafion are formed through protonation of water with SO3 H groups. Experimental data of Porat [88] suggest that the negatively charged -SO− 3 groups, along with water molecules and positively charged counter-ions (H+ ) tend to aggregate and form hydrophilic clusters in the Nafion. The size of these water clusters is in the order of nanometers. Yeo [122] used small-angle X-ray measurements of Nafion and verified the existence of clusters of about 4 nm diameter with channels of 1 nm diameter and length that limit the transport properties of water and ions. The schematic and appearance of these clusters and channels are schematically shown in Fig. 2.2. Increase in membrane humidity increases the density of hydrophilic groups and decreases that of the hydrophobic groups.

Among others, James and other co-

workers [46, 47] have investigated the structure of Nafion, and have shown a microphase separation of hydrophobic and hydrophilic domains by various kinds of electrochemical and instrumental techniques. They concluded that humidity leads to a decrease of hydrophobic groups at the interface and also showed that an electrochemically active site at the metal-electrolyte interface depends on expansion of the hydrophilic site at the interface region between the materials. Protons are shielded from the Nafion chain when the hydration number (λ) — the number of moles of water per mole of sulphonic acid sites — is large enough, λ ≥ 6. Paddison [83] asserts that although the experimental investigations provide a qualitative understanding of the function of the sulfonic acid based ionomers, the specific details of how the molecular structure and the hydrated morphologies connect with the transport of protons and water through the membrane is not well understood. He used DFT calculations to find the minimum energy conformation of trifilic acid (CF3 SO3 H) with successive addition of six water molecules, and found that the separation of the proton from -SO3 H started with the addition of the third water molecule. With the addition of the sixth water molecule, he observed a complete separation of the excess proton (as a hydronium ion) from the anion. This suggests that with sufficient water, the proton is shielded from direct electrostatic interaction


17

Figure 2.3: Evolution of the membrane structure as a function of water content or hydration number λ. The circles represent the size of the water clusters. Taken from Weber and Newman [116]


18

with the sulfonate anion. His study also suggests that the ‘first’ hydration shell of the trifilate anion is made up of five water molecules. Therefore, we can conclude that at higher Nafion membrane hydration, the water in the pores is more like bulk water (Fig. 2.3), and the transport of the protons will occur through transfer from one water molecule to another, rather than interacting with sulfonic end groups. Chemical reactivity at the interface between Pt and Nafion is a function of humidity of Nafion membrane. Kanamura et al. [49] assert that in present fuel cells, not all of the Pt catalyst loaded on carbon particles fully participate in the electrochemical reactions. This may be related to the structure of the interface between Nafion and the Pt catalyst. They used AFM combined with a surface potential measurement method on the Nafion under dry conditions. Furthermore, by Fourier transform infrared measurements (FTIR) they concluded that the interfacial behavior of the Nafion membrane on the Pt electrode changes with humidity near membranes. At low humidity conditions, the interface between the Pt electrode and the hydrophilic domain is very small, while it is strongly increased at well humidified conditions. Well hydrated membranes lead to high H+ conductivity, but with decreasing water content the conductivity drops approximately in a linear fashion. So it is important to have a uniform and appropriate level of water content throughout the membrane for the fuel cell to function well. The protons (H+ ) pull water molecules along when moving from the anode to the cathode, an effect that is called the electro-osmotic drag. This becomes a major issue, especially at high current densities, when electro-osmotic drag leads to drying up of the anode and excess water production at the cathode. This causes flooding of the cathode, which leads to oxygen-transport problems. Both these conditions are extremely detrimental and limiting to the cell’s functioning. These issues are discussed further in Sec. 2.5. Improvements in membrane properties are made by producing composite membranes. This can be done in several ways, one of which consists of impregnating the membrane with a solution or with a solid powder to decrease the permeability of the reactant gas in order to decrease fuel crossover, which is particularly important for direct methanol fuel cells. Indeed, decrease of methanol crossover by modifying the structure of Nafion by incorporating Pd clusters and organosilicon compound has


19

been observed [23]. Another possibility is reinforcing the perfluorosulfonic membrane with poly-tetrafluoroethene (PTFE) components, which is the successful approach of Gore and Ashai Chemicals (though with different procedures). Novel membranes are also prepared by new techniques such as radiation grafting or plasma polymerization [20, 62]. Both these methods have been proven to induce very desirable properties in the membrane. They are promising for PEM fuel cells as long as they turn out to be mechanically and electrochemically stable.

2.5

Water Transport in PEMFC

As pointed out before, proper water content inside the membrane is of utmost importance for good functioning of the fuel cell. Too little water reduces H+ conductivity while too much water floods the cathode; both of these seriously limit the performance of a PEMFC. The water content distribution in the membrane is governed by water transport through it. Various mechanisms of water transport in the membrane are illustrated in Fig. 2.4. The supplied air removes the excess water formed at the cathode. This is simply removal by evaporation. The concentration gradient of water across the membrane, with higher concentration at the cathode, causes back diffusion of water towards the anode, thereby maintaining an adequate level of water throughout the membrane. In practice it is very difficult to achieve the desired water balance, and careful design and optimization is needed. One of the complications, as mentioned before, is the electro-osmotic drag by protons. Every proton drags 2 to 5 water molecules, and the number increases with increase in temperature of the membrane [94]. This can be very detrimental since at high current densities this might lead to the drying of the anode and flooding of the cathode. Unbalanced water content inside the membrane drastically reduces the proton conductivity and increases the ohmic resistance of the membrane, which means more performance losses. Another complication is that the water content in the membrane may be about


20

Water production at cathode

Electroosmotic drag

Removal by evaporation Back diffusion

Humidified fuel stream

H2

H2O

H2O

Air

Humidified air stream

Figure 2.4: Modes of water transport in a PEMFC. right in some parts, too dry in others, while some parts may be flooded. As an example, visualize dry air entering the cell. After passing over a portion of the electrolyte the water content in both the membrane and the air may be about right. But on moving further, the air becomes saturated with the water produced by the cathode reaction, and so it cannot remove excess water from the remaining portion of the electrolyte. This can cause the cathode in that portion to become flooded. O2 cannot reach the 3-phase region in the flooded portions of the cathode, and electrochemical activity is lost in such regions. Likewise, at relatively higher temperatures (> 60◦ C) the hot air stream may dry out the electrodes. One way to tackle this is to humidify the air stream and perhaps the fuel stream too if the operating temperatures are high.


21

This leads to remarkable improvement in cell performance. Such humidification could be external – in which case the cost, weight and complexity of the fuel cell increases – or internal, by clever internal cell design such that the water is balanced within the cell [21].

2.6

Catalysts for PEM Fuel Cells

The reactions in PEMFCs occur on the catalyst surface. Characteristics of the catalyst become very important and often times limit the performance of the cell. As such, catalysts have been a focus of intense research both in industry and academia. Preparation and characterization of the catalysts for making better use of the catalysts in fuel cells is a very active area of research. The key is to design low cost catalysts that are both reliable (high catalytic activity) and durable. Pt has the highest intrinsic activity for the ORR as well for the hydrogen oxidation reaction (HOR; the anode reaction). Consequently, the state-of-the-art commerciallyavailable catalysts use highly dispersed Pt or Pt-alloy nanoparticles as the active catalyst material. Nonetheless, there are several problems with using pure Pt as the catalyst. Pt is intolerant to CO poisoning, which can be a major problem at the anode when using a reformed fuel stream (see Sec. 3.1). It is also intolerant to methanol oxidation. This is a problem in direct methanol fuel cells (DMFC), in which the fuel methanol has a high crossover tendency from the anode to the cathode. Methanol oxidation at the cathode is very detrimental to the performance of DMFCs, as it reduces the cell potential significantly. So, ideally it is desirable to have the catalyst material at the cathode resistant to methanol oxidation. Another serious problem is Pt dissolution in the electrolyte, whereby Pt is lost progressively from the active catalyst layer. This not only reduces the effectiveness of the catalyst layer, but also leads to the degradation of the membrane. Some of these problems can be overcome by using Pt/M-alloy catalysts, where M is another metal (M= Co, Ni, Cr, Fe, Ru, Pd). For example, the CO tolerance of Pt/Ru is significantly higher than that of pure Pt, while Pt/M (M=Ni,Co,Pd) have higher ORR activity [15, 63, 73]. It is also observed that alloys with smaller Pt-Pt


22

distance than in pure Pt usually have higher ORR activity [15]. Although, the exact cause of such improvements is elusive, it is generally accepted that introduction of the second metal (M) favorably alters the electronic structure of surface Pt atoms in a way that enhances its ORR activity. Thus, depending on the particular application, it may be possible to use tailormade catalysts. But this usually involves a trade-off between activity, durability, and corrosion [40]. There is no one best catalyst for across-the-board applications. Carbon Support High dispersion results in greater availability of the catalyst surface for electrochemical activity and is achieved by using a support material, usually carbon. Carbon has a favorable pore-structure, high electrical conductivity and is corrosion-resistant. These characteristics make it a good choice for supporting the active catalyst nanoparticles. But pure carbon, being inert, cannot in and of itself act as the support. It has to be first chemically modified by a controlled preparation and heat-treatment method to functionalize its surface with oxygen-containing functional groups like anhydride, carbonyl, carboxylic, etc. It is these functional groups, then, that act as active sites for dispersing the active nanoparticles. The functionalization process greatly influences the details of catalyst dispersion and performance [15]. Catalyst Nanoparticles For highly dispersed electrocatalyst particles, the influence of the particle size and shape also becomes important. Pt nanoparticles generally assume the energetically more favorable cubo-octahedral structure consisting of Pt atoms arranged on eight (111) and six (100) crystallographic faces bounded by edge and corner atoms. Fig. 2.5 shows a model of a 1.5 nm diameter Pt particle, which consists of 201 atoms of which 122 atoms are on the surface and 60 atoms reside at the edge and corner sites [51, 73]. The role of specific adsorption sites and geometric factors of the catalyst surface on the kinetics of the ORR is widely accepted [17, 39, 51, 73]. As the electrocatalyst


23

Figure 2.5: High resolution TEM image (left) and model (right) of a 2 nm Pt catalyst nanoparticle. Taken from Markovic et al. [73]. particles become smaller, a larger fraction of the total number of atoms in the particles is associated with surface sites, and the properties of these surface atoms may differ significantly from those of surface atoms in bulk samples. The change in specific activity (current density vs. surface area of catalyst) of supported metal particles in heterogeneous catalysis as a function of particle size is commonly referred to as the particle-size effect. Many studies have reported a maximum ORR activity for 3.5 nm Pt nanoparticles [51]. On the contrary, some recent research shows that the activation energy of the ORR is independent of the particle size [120]. Besides, specific sites with special geometric arrangements (e.g. edge/corner atoms) may have enhanced activity over neighboring sites, and they are often referred to as active sites. Here again, there is no clear agreement in the literature on the actual role of the active sites. Kinoshita [51] provides a summary of a number of studies on particle-size effects for oxygen reduction on Pt. He observes: “It is evident from the observations listed that contradictory interpretations exist regarding the relationship between particle size and electrochemical activity for oxygen reduction.”


24

Characterization of Catalysts Comparison of the electrochemical performance of catalysts prepared by different methods is difficult, keeping in mind the diverse physical characteristics that they result in. It must be taken into account that, after preparation, the catalysts may be submitted to thermal treatments at different temperatures and in different atmospheres. The main objective of this procedure is to eliminate surface impurities, but the treatment may also induce modifications like an increase in particle size due to coalescence of the particles or an enrichment of the particle surface with one of the metals (in case of alloys) due to thermally-induced segregation or both. A diminution of the crystalline defects is also possible. Though the quantification of the influence of these effects on the properties of the electrocatalysts is not obvious, the influence on the performance due to thermal treatments is clear [15, 63]. It is well known that different preparation procedures lead to catalysts with different morphological characteristics. For a Pt/Ru alloy supported on carbon (Pt-Ru/C), for example, varying the composition also impacts the particle distribution. The TEM micrographs in Fig. 2.6 below clearly indicate that even with similar particle sizes (Pt-Ru/C 75:25, 5.1 nm and Pt-Ru/C 82:18, 5.5 nm), a very good distribution of the metal on the carbon support is obtained only in the latter case. Comparisons of 5.5 nm Pt-Ru/C 82:18 with commercial 3.6 nm E-TEK catalyst (Pt-Ru 50:50) at different temperatures show better efficiency in the former case, an interesting result since fundamental studies predict the opposite [63]. This indicates that the amount of Ru in the catalyst and the particle size may not be the main factors in determining the efficiency of the catalyst. So other factors, perhaps a good morphology and/or a clean method of preparation of the catalyst, may play role. In general, it is not straightforward to extrapolate results from fundamental studies (on well-characterized surfaces, like single crystals) to complex catalytic systems. Neither is it possible to evaluate the performance of a catalyst on the basis of only a few independent parameters. But whatever the situation may be, one needs to be very circumspect when comparing different systems. In the field of heterogeneous catalysis, many paradoxes have resulted not from fundamental physics-based differences, but from comparing two vastly different systems.


25

Figure 2.6: TEM micrographs for Pt-Ru/C catalysts. (A) Pt-Ru 57:25 (B) Pt-Ru 82:18. Taken from Lizcano-Valbuena et al. [63].

2.7

Summary

Fuel cells are electrochemical energy conversion devices that generate electrical energy. This is achieved by splitting the fuel into ions and electrons at the anode using a membrane. The membrane is an electronic insulator that can conduct the generated ions very well. Thus, the membrane acts as a species-separator that screens the electrons out, which can then be used in an electrical circuit as a source of power. Finally, the ions recombine with the electrons and the oxidizer to form the products at the cathode. The basic principles, processes, and technological challenges associated with PEMFCs were presented in details. We saw that it is only in the 3-phase region where the electrochemical reactions take place and where chemical species are produced or consumed. The electrolyte in PEMFCs is a polymeric material that is an electronic insulator, but can conduct protons very well. These membranes are stable in a limited range of temperature and require adequate levels of humidity to maintain high conductivity for the protons. Performance losses related to non-homogeneous membrane humidity, mass-transport limitations at high current densities, and activation


26

overpotential at low current densities were also discussed. It was stressed that even catalysts with the same composition can exhibit vastly different ORR activity owing to differences in the details of the preparation processes.

Chapter 3 PEM Fuel Cell Electrochemistry The basic principles and components of a PEMFC were discussed in Chapter 2. In this chapter, we review the electrochemical reactions at the electrodes of PEMFCs. First, the anode reaction (hydrogen oxidation) is briefly discussed. Then, through an extensive review, the main unresolved issues as regards the cathode reaction (oxygen reduction) are highlighted. Work in this thesis, presented in the following chapters, primarily focusses on addressing these issues.

3.1

Anode Reaction: Hydrogen Oxidation

In PEMFCs, the H2 fuel is oxidized at the anode with the assistance of the catalyst. In other words, the H2 fuel is split into protons and electrons at the anode (electrons are produced at the anode). This reaction is called the hydrogen oxidation reaction (HOR). Both hydrogen oxidation and evolution are very fast on Pt, i.e., these processes have high exchange current densities on Pt. In acidic environment, the HOR is given as H2 + 2H2 O 2H3 O+ + 2e−

(U 0 = 0 V).

(3.1)

Hydrogen dissociatively adsorbs on many metals, especially on those which are catalytically active. The energy required to break the H-H bond comes from the

27

CHAPTER 3. PEM FUEL CELL ELECTROCHEMISTRY

28

enthalpy of adsorption [38]. In fuel cells, the existence of a hydrogen adatom Hads on the highly electrocatalytically active Pt metal has been proven by using the cyclic voltammetry (CV) technique. Several mechanistic investigations have been carried out into the HOR mechanism that allow to partition the overall HOR into a number of steps as follows [38, 108]: 1. Mass transport of H2 to the surface and weak physisorption of molecular H2 (3.2)

H2 H2,ads 2. Dissociative chemisorption of hydrogen as atoms (“Tafel Reaction”) Pt − H

Pt + H2,ads

Pt

(3.3) Pt − H

3. Ionization and hydration. There are two possibilities here, depending upon whether oxidation of Hads or H2,ads takes place. In the first case of the so called Volmer Reaction Pt − H + H2 O Pt + H3 O+ + e− ,

(U 0 = 0 V)

(3.4)

where the overall reaction takes place twice. An alternative is the Heyrovsky Reaction given as Pt + H2,ads + H2 O Pt − H + H3 O+ + e−

(U 0 = 0 V) (3.5)

followed by the Volmer reaction. This Heyrovsky pathway is, however, insignificant in the hydrogen oxidation on Pt for which the equilibrium atomic hydrogen 0 coverage, θH , derived from cyclic voltammetry, exceeds 0.9 [108], and for which

H2 molecules also are only very weakly adsorbed. 4. Transport of the H3 O+ ions away from the electrode surface.


29

In acidic environment, hydrogen oxidation on Pt proceeds primarily through the Tafel-Volmer mechanism. It can be seen that the adsorption of hydrogen atoms plays a decisive role in the reaction, since, unless the adsorption enthalpy is sufficient, dissociative chemisorption cannot take place. Of course, the adsorption energy should not be too high, otherwise it will prove kinetically difficult to remove the adsorbed hydrogen atoms as H3 O+ . Adsorption enthalpy is calculated as the energy difference between the dissociation energy of H2 and dissociation energy of metal-hydrogen bond at the surface [38]. The performance at the anode is excellent if the fuel is pure hydrogen. However, in many systems, the hydrogen fuel may frequently come from some kind of fuel reforming process. So, in many practical applications, the fuel stream contains traces of elements or compounds such as CO, S, and NH3 . All these substances can, to varying extents, poison the anode catalysts, thereby significantly hindering the HOR. CO is one of the major poisons at the anode in low temperature fuel cells. CO has a very high affinity for Pt. This leads to the adsorption of CO species on active Pt catalyst sites so that over time the anode becomes poisoned, virtually leaving no sites for the HOR. CO concentration of 10 ppm or less usually result in satisfactory anode performance for longer time. But, presently, even the best reforming systems, without the use of CO separation units, have CO levels of the order of 1500-2500 ppm. To reactivate the surface, the CO can be oxidized to CO2 . This is achieved by addition of small quantities of oxygen in the fuel stream. But this works only if the CO concentration is low (200 ppm) to start with. Any left over oxygen that did not react with CO will react with H2 , thereby wasting fuel. Further, the oxygen addition process needs reasonably fine control, which adds to the cost and complexity of the system. Other methods to reduce CO poisoning at the anode include addition of hydrogen peroxide to the fuel stream or applying electrical pulses that increase the anode potential to levels such that CO is oxidized to CO2 . A very active area of research is development of CO-tolerant catalysts. Presently, the best CO-tolerant catalysts usually contain Pt-Ru alloy.


3.2

30

Cathode Reaction: Oxygen Reduction

At the cathode, oxygen from the air stream reacts with protons from the electrolyte (the PEM) and electrons from the electrode to form water (electrons are consumed at the cathode). This reaction is called the oxygen reduction reaction (ORR), and is given as O2 + 4H+ + 4e− → 2H2 O.

(3.6)

As pointed out in Sec. 1.3, although the ORR on Pt is much faster compared with other catalyst materials, it is quite slow compared with the HOR at the anode. The so called activation overpotential losses due to the slow kinetics of the ORR is one of the major loss mechanisms in PEMFCs. There have been a number of fundamental experimental and computational studies of the ORR at low index Pt surfaces ((111), (110), and (100)), both in acidic and basic electrolyte [25, 68, 80, 84, 101, 113], as well as at the Pt/Nafion interface [86, 107, 123]. Yet, the ORR is not well understood — some key questions about the reaction pathway of the ORR on Pt and about some of the reaction intermediates involved in the ORR still remain open. In the remainder of this chapter, we will discuss different aspects of the ORR, including its reaction mechanisms and intermediates, that are described in the literature, while highlighting the key issues about the ORR. Efficient computational methods to address these issues are presented in Chapters 4 and 5, which are then used in Chapters 6 and 7 to investigate the ORR on Pt.

3.2.1

Established Features of the ORR

Although the actual reaction mechanism for the ORR (Reaction 3.6) on Pt is still not fully determined, a few facts about the ORR are well established: • ORR on oxide-free and preoxidized Pt surfaces are qualitatively different. On (100) and (110) surfaces, O(ads) is adsorbed irreversibly on the catalyst surface at high potentials (∼0.9 V). This is evidenced from the asymmetry in the currentvoltage curve during cyclic voltammetry (CV) between 0.75–0.95 V and the large separation of Pt surface atoms. This irreversible state of O(ads) is usually


31

referred to as oxide. The oxide formation is considered as a parasitic reaction that modifies the ORR by reducing the net current output by establishing a mixed potential control [85]. On Pt(111), the CV is symmetric and relaxation of the Pt surface atoms are almost negligible up to about 1 V [71]. Thus, the adsorption of oxygen-containing species on Pt(111) is a reversible process below 1 V. As such, ORR on Pt(111) appears to be the best electrochemical system to understand the mechanistic pathway of the ORR on Pt. • Two nearly linear Tafel slopes1 are observed in CV experiments of oxygen-free Pt(111) in HClO4 with slopes of -60 mV/dec in the low current density (lcd) region and -120 mV/dec in the high current density (hcd) region, while the Tafel slope is potential-independent in H2 SO4 solutions (Fig. 3.1). In the hcd region, the rds can be unambiguously identified with a single electron transfer step [25, 86, 123]. The situation is not straightforward in the lcd region. While some studies ascribe the low Tafel slope in the lcd region to a combination of chemical and electrochemical steps, namely oxygen adsorption and its subsequent first reduction step [86, 121], others assert that the rds is a single electron transfer step throughout [73, 98, 113]. Apparently, many details about the ORR mechanism, particularly the reaction intermediates and competing reaction pathways, remain unresolved. These issues are discussed further in Sec. 3.2.2. • The catalyst surface is adsorbate-free at low potentials (below 0.6 V for 0.1 M HClO4 ), but adsorbate coverage increases with potential. The identity of adsorbates depends on the acid or the alkaline solution. For example, the adsorbates are OH (and O, as shown in Chapter 6) for the ORR in HClO4 and KOH, but in H2 SO4 the adsorbate is (bi)sulphate (SO∗4 ). Adsorbates during the ORR are discussed further in Sec. 3.2.3. 1

In Tafel plots, the potential is plotted along the y-axis and the logarithm of the kinetic current density (log |jk |) is plotted along the x-axis. jk is obtained by correcting the measured current density for mass-transport effects.


32

E (mV)

0.05 M H2SO4 0.1 M HClO4

log(|jk|) Figure 3.1: Rotating disk electrode Tafel curves for ORR on Pt(111) at 2500 rpm with a sweep rate of 50 mV/s in oxygen saturated 0.1 M HClO4 (circles) and 0.05 M H2 SO4 (squares) solutions. Different straight lines correspond to different Tafel slopes and exchange currents (inset). Taken from Wang et al. [113].

• Throughout the potential range of interest, the reaction order with respect to oxygen is 1, i.e., current density ∝ oxygen concentration [O2 ] in the solution. • Surface orientation of the catalyst can have significant effect on the ORR characteristics, even in the same solution. For e.g., ORR activity on Pt in 0.1 M HClO4 follows the order (100)