ABSTRACT
Title of dissertation: CHARACTERIZATION OF MECHANICAL
PROPERTIES AND DEFECTS OF
SOLID-OXIDE FUEL CELL MATERIALS
Patrick Owen Stanley
Doctor of Philosophy, 2018
Dissertation directed by: Professor Eric D. Wachsman
Materials Science and Engineering
Solid-oxide fuel cells (SOFCs) have the potential to help meet global energy
demands by efficiently converting fuel to electricity. The technology currently requires
high temperatures and has reliability limitations. A critical concern is the structural
integrity of the cell after redox cycling at operating temperatures. As new materials
are developed to reduce operating temperatures and improve redox stability, the
effect of the environment on the mechanical properties must be studied. Ceria-
based systems have allowed the operating temperature to be decreased to the 600 ◦C
range. For this reason, a three-point bend apparatus was developed which could test
materials up to 650 ◦C in reducing environments.
Using this apparatus, it was shown how pore geometry and amount affected
strength of porous gadolinium doped ceria (GDC) at 650 ◦C with lower aspect ratio
pores, leading to higher fracture strength due to crack tip blunting. The strength of
Ni-GDC/GDC half-cell coupons showed no dependence on loading orientation at
elevated temperatures in air, but were 47% weaker when the electrolyte was placed
in tension under H2 as compared to when the electrolyte was placed in compression.
It was also determined that a reduced Ni-GDC/GDC coupon could be exposed to air
for an extended period of time and reheated under H2 with no effect to the strength,
allowing for more options when processing and preparing cells.
A new anode material, SrFe0.2Co0.4Mo0.4O3-δ (SFCM), was investigated for
chemical expansion, oxygen non-stoichiometry, and mechanical properties. SFCM
maintains phase purity under reducing conditions, with little changes to lattice
parameter between oxidation and reduction, but under oxidation, SFCM forms
Sr2Co1.2Mo0.8O6 impurities. SFCM supports a large degree of non-stoichiometry,
up to δ = 0.176 at 600 ◦C, due to a low enthalpy of formation for oxygen va-
cancies of 44.3 kJ mol−1. Fracture toughness of SFCM was determined to be
√ √
(0.124± 0.023) MPa m in air at room temperature and (0.286± 0.038) MPa m
at 600 ◦C. The strength of SFCM-GDC half-cells increased by 31% upon heating to
600 ◦C, after which reduction decreased strength by 29%. Reduction and redox cy-
cling were shown to only decrease the characteristic strength, not alter the structural
flaw distribution, as microcracks uniformly grew.
CHARACTERIZATION OF MECHANICAL PROPERTIES AND
DEFECTS OF SOLID-OXIDE FUEL CELL MATERIALS
by
Patrick Owen Stanley
Dissertation submitted to the Faculty of the Graduate School of the
University of Maryland, College Park in partial fulfillment
of the requirements for the degree of
Doctor of Philosophy
2018
Advisory Committee:
Professor Eric D. Wachsman: Advisor, Chair
Professor Peter Sandborn: Dean’s Representative
Professor Lourdes Salamanca-Riba
Professor Sreeramamurthy Ankem
Associate Professor Isabel K. Lloyd
© Copyright by
Patrick Owen Stanley
2018
Dedication
Ad maiorem Dei gloriam
I dedicate this to my wife, Bethany, who has encouraged and supported me throughout
the process, and to my two wonderful daughters, Lucy and Madeleine.
ii
Acknowledgments
I would like to start by thanking my advisor, Dr. Eric Wachsman, for his
support during my graduate career. He allowed me the freedom to direct the details
of my research and the support to help me reach my goals. He was also supportive
of me as my family has grown through the years.
More thanks than I can express is due to my wonderful wife, Bethany, and
now our daughters, Lucy and Madeleine. When Bethany and I first met, I was still
deciding between graduate programs, and now, after many life changes, she continues
to challenge and support me in my endeavors. Lucy and Madeleine have provided
me the impetus to continue pushing through the challenges and to strike a balance
between my personal and professional pursuits.
Of course, I owe acknowledgement to my parents, Jim and Marilyn Stanley,
who raised me to be inquisitive and persistent. Thank you for helping me to become
the person that I am today.
The Materials Science and Engineering Department at UMD has been filled
with marvelous and helpful people. Dr. Kathleen Hart was one of the first people I
met in the department, seemed to help organize everything, and was always there
when needed. She has been sorely missed this past year, and I hope she is making
the most of her retirement. Kay Morris and Jenna Bishop in the business office help
the department run smoothly, if it be answering benefits and procurement questions,
or doting on my children when they visit. Of the many professors I have been able to
interact with, I would like to especially thank Dr. Isabel Lloyd, who has graciously
iii
given me her advice, helped make introductions, and passed along opportunities to
me when they come her way.
Somehow a small group of my friends from undergraduate school have accu-
mulated in the area. It has been a blessing to have them nearby, always willing to
help out when need be or to just spend time together. Dave and Katie Shahin are a
large part of what brought me to the University, and I am grateful for the continued
shared experiences with Dave. Andrew Dunkman has helped me in so many ways,
from apartment hunting a thousand miles away to programming questions, for which
I express my thanks. And finally, Angela Rudolph has been very supportive of me
and my family getting through graduate school.
I owe a good amount of thanks to the various people in lab who helped show me
the way to get started with my experiments, allowed me to bounce ideas off them, or
helped with the work. Chris Pellegrinelli and Greg Hitz were two people whom I was
always comfortable going to and asking what seemed like the most basic questions
on how to do things, helping me get started in lab. Drs. Mohammad Hussain and
Yi-Lin Huang shared their experience and knowledge of material science as I worked
through the fundamentals. Thanks to Tom Hays for being my partner as we started
mechanical testing in lab. Ian Robinson, Evans Gritton and Jon O’Neill provided
much needed help with SEM, XRD, and modeling, saving me from fumbling through
it myself.
I would like to acknowledge my friends and classmates at the University who
formed part of my support network, sharing knowledge and encouragement. Mimi
Hiebert, Doug Henderson, Michael Van Order, Beth Tennyson, Travis Dietz, Adam
iv
Pranda, Marina Pranda, and Max Lerman, thank you for the time spent in and out
of class as we worked through our problems together, laughing and having fun along
the way.
Early when starting to look into mechanical properties, I was introduced to Mr.
George Quinn at the National Institute of Standards and Technology. Mr. Quinn
was kind enough to take much of his time to look though my preliminary work and
give me feedback on testing methodology and fractography. It was a great privilege
to meet the man who literally wrote the standards that I was attempting to follow
and to have his guidance as I started to navigate my way though his world.
Thank you, as well, to Dr. Sean Bishop, who helped guide me as I made my
ambitious goals and helped me plan out the steps needed to accomplish them.
Many facilities across campus assisted in the collection and analysis of data
for this work. My thanks go to the Advanced Imaging & Microscopy Laboratory
(AIMLab), the X-Ray Crystallographic Center, the Surface Analysis Center, and Dr.
Robert Bonenberger at the Modern Engineering Materials Instructional Laboratory.
Finally, thank you to Redox Power Systems, the Department of Energy’s
National Energy Technology Laboratory, and the Maryland Industrial Partnerships
Program for supporting this work.
v
Table of Contents
Dedication ii
Acknowledgements iii
List of Tables ix
List of Figures x
List of Abbreviations xiii
1 Introduction 1
1.1 Motivation for Solid-Oxide Fuel Cell Research . . . . . . . . . . . . . 1
1.2 Solid-Oxide Fuel Cell Operation . . . . . . . . . . . . . . . . . . . . . 2
1.3 Solid-Oxide Fuel Cell Materials . . . . . . . . . . . . . . . . . . . . . 8
1.4 Mechanical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.5 Problem Statement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2 Experimental Procedures 18
2.1 Sample Preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.1 SrFe0.2Co0.4Mo0.4O3-δ Powder . . . . . . . . . . . . . . . . . . 18
2.1.2 Bars . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
2.1.3 Test Coupons . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
2.2 Density Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . 21
2.3 X-Ray Diffraction (XRD) . . . . . . . . . . . . . . . . . . . . . . . . 21
2.4 Thermogravimetric Analysis (TGA) . . . . . . . . . . . . . . . . . . . 21
2.4.1 Temperature Programmed Desorption (TPD) . . . . . . . . . 23
2.5 Conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.6 Mechanical Testing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.6.1 Development of Mechanical Test Apparatus . . . . . . . . . . 26
2.6.2 Atmospheric Treatment . . . . . . . . . . . . . . . . . . . . . 27
2.6.3 Fracture Strength . . . . . . . . . . . . . . . . . . . . . . . . . 29
2.6.4 Fracture Toughness . . . . . . . . . . . . . . . . . . . . . . . . 30
2.7 Scanning Electron Microscopy (SEM) . . . . . . . . . . . . . . . . . . 30
vi
3 High Temperature Mechanical Behavior of Porous Ceria and Ceria-Based
Solid-Oxide Fuel Cells 32
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
3.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
3.2.1 Porosity and Pore Former Choice . . . . . . . . . . . . . . . . 33
3.2.2 Flexural Stress Orientation . . . . . . . . . . . . . . . . . . . . 36
3.2.3 Reduction and Strength . . . . . . . . . . . . . . . . . . . . . 41
3.2.4 Reoxidation at Low Temperatures . . . . . . . . . . . . . . . . 44
3.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4 Defect Chemistry and Oxygen Non-Stoichiometry of SrFe0.2Co0.4Mo0.4O3-δ 51
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51
4.2 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 52
4.2.1 Phase Purity and Chemical Expansion . . . . . . . . . . . . . 52
4.2.2 Conductivity . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
4.2.3 Temperature Programmed Desorption . . . . . . . . . . . . . 57
4.2.4 Oxygen Non-Stoichiometry . . . . . . . . . . . . . . . . . . . . 58
4.2.5 Defect Equilibrium Model and Diagram . . . . . . . . . . . . . 62
4.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
5 Flexural Strength and Flaw Distributions of SrFe0.2Co0.4Mo0.4O3-δ Based
Ceramic-Supports for SOFCs at Operating Conditions 70
5.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.2 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . 70
5.2.1 Redox Cycling Stability . . . . . . . . . . . . . . . . . . . . . 70
5.2.2 Fracture Toughness . . . . . . . . . . . . . . . . . . . . . . . . 73
5.2.3 Strength of Anode Support Layer After Redox Cycling . . . . 75
5.2.4 Half-cell Strength of Anode-Supported SOFCs at Operating
Conditions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
5.3 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
6 Conclusions 89
6.1 Summary of Contributions . . . . . . . . . . . . . . . . . . . . . . . . 89
6.2 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
6.3 Publications and Presentations . . . . . . . . . . . . . . . . . . . . . 93
A Statistics 95
A.1 t-Test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95
A.2 Weibull Statistics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
B TGA Manual 102
B.1 Theory of Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
B.2 Mass Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102
B.2.1 Setup and Operation . . . . . . . . . . . . . . . . . . . . . . . 103
B.3 Heating . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
vii
B.4 Gas Delivery . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
B.5 pO2 Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109
B.6 Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
B.6.1 Gas Controls . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
B.6.2 Temperature and pO2 Measurements . . . . . . . . . . . . . . 112
B.6.3 Starting a Run . . . . . . . . . . . . . . . . . . . . . . . . . . 113
B.7 Interfacing with Other Devices . . . . . . . . . . . . . . . . . . . . . . 114
B.8 Troubleshooting . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114
C Code 116
C.1 Arduino PID Relay Furnace Controller with Serial Connectivity . . . 116
C.1.1 Hardware and Other Libraries . . . . . . . . . . . . . . . . . . 116
C.1.2 Variables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
C.1.3 Serial Communications . . . . . . . . . . . . . . . . . . . . . . 117
C.1.4 Arduino Sketch . . . . . . . . . . . . . . . . . . . . . . . . . . 118
C.2 MKS Gas Controller and Scheduler . . . . . . . . . . . . . . . . . . . 123
C.2.1 Hardware and Other Libraries . . . . . . . . . . . . . . . . . . 123
C.2.2 Program Overview . . . . . . . . . . . . . . . . . . . . . . . . 124
C.3 Temperature and pO2 Measurements . . . . . . . . . . . . . . . . . . 134
C.3.1 Hardware and Other Libraries . . . . . . . . . . . . . . . . . . 134
C.3.2 Program Overview . . . . . . . . . . . . . . . . . . . . . . . . 134
Bibliography 137
viii
List of Tables
3.1 Chi-squared values for fixed η and variable η fits of porous gadolinium
doped ceria, Gd0.1Ce0.9O3 (GDC) strength data (Figure 3.1) . . . . . 33
3.2 Summary of fit parameters for reduction of NiO-GDC/GDC half-cell
coupons under 3% H2, 3% H2O, 94% N2 at different temperatures . . 44
4.1 Refinement results of SFCM before and after exposure to reducing
and oxidizing environments . . . . . . . . . . . . . . . . . . . . . . . 54
5.1 Summary of Cycled SFCM-GDC (2:1) anode support layer (ASL)
Mechanical Properties . . . . . . . . . . . . . . . . . . . . . . . . . . 75
5.2 Summary of Weibull Fit Parameters (characteristic strength and
Weibull modulus) for Cycled SFCM-GDC (2:1) ASL . . . . . . . . . . 78
5.3 Summary of strengths for SFCM-GDC (2:1)/GDC half-cells at differ-
ent environments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 82
5.4 Summary of Weibull fitting parameters (characteristic strength and
Weibull modulus) for SFCM-GDC (2:1)/GDC half-cells at different
environments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
B.1 Cahn D200 Microbalance Specifications . . . . . . . . . . . . . . . . . 103
ix
List of Figures
1.1 Diagram showing flow of materials in the operation of an SOFC. . . . 3
1.2 Diagram showing the triple phase boundary and the importance as
being the site where incorporation and reactions occur. . . . . . . . . 7
1.3 Optical profilometer measurements of a 10 cm by 10 cm Ni-GDC/GDC
half-cell where the z-axis has been magnified 6X to highlight curvature 16
2.1 Assembled Alumina 3-Point Bend Fixture . . . . . . . . . . . . . . . 28
2.2 Design drawing and photo of mechanical testing atmospheric chamber 28
3.1 Flexural strength-porosity dependence for porous GDC at 650 ◦C and
25 ◦C using spherical poly(methyl methacrylate) (PMMA) or graphite
flake pore former: a) Measured strengths fitted using fixed geometric
constant; b) Measured strengths fitted using different geometric constants 34
3.2 Scanning electron microscope (SEM) micrographs of porous GDC bars
made using: a) PMMA pore former; b) graphite flake pore former . . 37
3.3 Temperature dependent strength of Ni-GDC anode supports and half-
cells in both air (filled data points), and reducing atmosphere (3%
H2, balance Ar, hollow data points), tested in both “electrolyte-up”
and “electrolyte-down” orientations resulting in the electrolyte layer
experiencing tension and compression, respectively . . . . . . . . . . . 39
3.4 Flexural test measurements of coupons sample sets: a) strength at
25 ◦C b) strength at 650 ◦C c) modulus at 25 ◦C d) modulus at 650 ◦C 40
3.5 SEM micrographs of unreduced half-cell fracture surface tested in air
at 25 ◦C at a) 2k magnification and b) 5k magnification . . . . . . . . 42
3.6 SEM micrographs of reduced half-cell fracture surface tested in reduc-
ing atmosphere at 650 ◦C at a) 1k magnification and b) 5k magnification 43
3.7 Thermogravimetric analysis curves for Ni-GDC/GDC half-cells show-
ing mass loss over time at temperatures ranging from 550 ◦C to 750 ◦C
in 3% H2 3% H2O balance N2, 50 sccm flow . . . . . . . . . . . . . . . 45
3.8 Thermogravimetric analysis of Ni-GDC/GDC half-cell showing no
mass gain after reduction and exposure to room temperature and
simulated air: a) Mass loss of interrupted reduction with exposure to
ambient condition; b) Temperature of sample during cycle c) Oxygen
partial pressure measured at the sample during cycle . . . . . . . . . 47
x
3.9 Flexural properties, a) strength and b) modulus, of Ni-GDC/GDC
half-cell coupons after reduction via two different methods, compared
to unreduced cells. Strength and modulus show a decrease upon
reduction but no significant difference between methods . . . . . . . . 48
4.1 a) Powder X-ray diffraction (XRD) patterns of SFCM samples taken
after synthesis, reduction, and oxidation compared to pure phase
SFCM diffraction pattern b) Crystal structure of unordered double
perovskite SFCM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
4.2 Total conductivity as pO2 changes under a) reducing and b)oxidizing
conditions at 600 ◦C . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56
4.3 Temperature programmed desorption of oxygen in SFCM with mass
loss from TGA(top) and oxygen desorption from MS (bottom) as it is
heated to 800 ◦C in N2 . . . . . . . . . . . . . . . . . . . . . . . . . . 59
4.4 Non-stoichiometry (circles, left axis) and corresponding oxygen va-
cancy concentration (diamonds, right axis) of SFCM under a) reducing
conditions and b) oxidizing conditions at 600 ◦C. Error bars are smaller
than points for non-stoichiometry and for vacancy concentration below
10−0.5 atm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
4.5 Non-stoichiometry of SFCM as pO2 changes at 400
◦C, 500 ◦C, and
600 ◦C. Error bars are smaller than data points. . . . . . . . . . . . . 63
4.6 a) Fitting results of oxygen vacancies concentration to the measured
non-stoichiometry b) Fitting results of n-type and p-type conductivity
to measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
4.7 a)Defect equilibrium diagram of SFCM based on fitted parameters.
b)Plot of natural log of external defect equilibrium versus inverse
temperature, to calculate enthalpy of formation oxygen vacancies. . . 68
5.1 a) DC conductivity of SFCM and SFCM-GDC (2:1) after cycling at
650 ◦C between 10% H2 in N2 and air. b) Mass loss of SFCM-GDC
and NiO-GDC (3:2) during reduction in 3% H2/3% H2O in N2 at 650
◦C 72
5.2 Mass changes of SFCM during cycling between 21% O2 and 3% H2/3%
H2O in N2 at 600
◦C. Each exposure was 5 hours. a) Percent mass
change with time as the gas is switched back and forth. b) Summary
of maximum and minimum changes with each cycle . . . . . . . . . . 74
5.3 a) Fracture toughness of SFCM from 20 ◦C to 600 ◦C. b) Thermal
expansion of SFCM, GDC, and SFCM-GDC (2:1) . . . . . . . . . . . 76
5.4 Results of four point bend test at ambient conditions of SFCM-GDC
(2:1) ASL with a) strength and b) modulus for samples before any
cycling and after 10 redox cycles between 10% H2 and N2 . . . . . . . 77
5.5 Fitted Weibull distributions of SFCM-GDC (2:1) ASL before any
cycling and after 10 redox cycles between 10% H2 and N2, plotted
linearly with 95% confidence intervals shaded . . . . . . . . . . . . . 79
xi
5.6 Scanning electron microscope images of epoxy filled and polished
SFCM-GDC (2:1) ASL cross sections tested a) before any cycling, b)
after 10 cycles between 10% H2 and N2 . . . . . . . . . . . . . . . . . 80
5.7 Flexural strengths of half-cell coupons made from SFCM-GDC (2:1)
anode with GDC electrolyte placed in compression, tested at 20 ◦C in
air, 600 ◦C in air, and 600 ◦C in 5% H2 . . . . . . . . . . . . . . . . . 83
5.8 Scanning electron microscope images fracture surfaces of SFCM-GDC
(2:1)/GDC half-cell cross sections tested at a) 20 ◦C, b) 600 ◦C in air . 84
5.9 Fitted Weibull distributions of SFCM-GDC (2:1)/GDC half-cells at
different environments, plotted linearly with 95% confidence intervals
shaded . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86
5.10 Scanning electron microscope images of epoxy filled and polished
SFCM-GDC (2:1)/GDC half-cell cross sections tested at a) 20 ◦C in
air, b) 600 ◦C in air and c)600 ◦C in 5% H2 . . . . . . . . . . . . . . . 87
B.1 MicroScan Acquisition with Open Method open and the current read
out of the balance displayed . . . . . . . . . . . . . . . . . . . . . . . 106
B.2 Front panel display of Gas Controller program . . . . . . . . . . . . . 111
B.3 Front panel display of Gas Scheduler program . . . . . . . . . . . . . 111
B.4 Front panel display of Temp & pO2 Measurements program . . . . . 113
C.1 MFC Controller program . . . . . . . . . . . . . . . . . . . . . . . . . 126
C.2 Gas Scheduler program . . . . . . . . . . . . . . . . . . . . . . . . . . 133
C.3 Temperature and pO2 measurement program . . . . . . . . . . . . . . 136
xii
List of Abbreviations
ASL anode support layer
CDF cumulative distribution function
csv comma separated value
DC direct current
GDC gadolinium doped ceria, Gd0.1Ce0.9O3
GFC gas correction factor
LSM lanthanum strontium manganite
MFC mass flow controller
MIEC mixed ionic-electronic conductor
MS mass spectrometer
NI National Instruments
pO2 partial pressure of oxygen
PMMA poly(methyl methacrylate)
PWM pulse width modulation
SCM Sr2CoMoO6
SEM scanning electron microscope
SFCM SrFe0.2Co0.4Mo0.4O3-δ
SFM Sr2Fe1.5Mo0.5O6
SMM Sr2MgMoO6
SOFC solid-oxide fuel cell
TGA thermogravimetric analysis
TPB triple phase boundary
TPD temperature programmed desorption
UTM universal testing machine
XRD X-ray diffraction
YSZ yttria-stabilized zirconia
xiii
Chapter 1: Introduction
1.1 Motivation for Solid-Oxide Fuel Cell Research
The world depends on fossil fuels for its daily energy needs. This is a fact that
is going to stay with us for the foreseeable future. By 2040, it is estimated that
fossil fuels will still supply 78% of total energy demand. [1] Traditionally, the annual
increase in demand is met by an increase in supply of fuel. This increase is supplied
by developments in mining and drilling operations, although recent advances have
been found to be controversial, such as hydraulic fracturing. [2, 3] These concerns do
not even account for the increased emissions caused from combustion of fossil fuels
and their effect on the global environment. [4–6]
An alternative to increasing the supply and use of fossil fuels is to increase the
efficiency of converting the fuel into useful work. Combustion generators convert
chemical energy to thermal energy, then to mechanical energy, and finally into
electrical energy. Each energy conversion step has intrinsic losses which limit efficiency.
Solid-oxide fuel cells (SOFCs) have the ability to solve this for applications where
electrical power is desired. SOFCs side step these losses by allowing for the direct
conversion of chemical energy to electrical energy. Power generation plants have
efficiencies around 30%, while a stand-alone SOFC generator can convert fuel to
1
electricity at 45 to 65% efficiency. [7, 8] SOFCs are able to run on a variety of fuel
sources, such as hydrogen, methane or even biogas. [9] This fuel flexibility and higher
efficiency positions SOFCs to bridge the gap in the energy economy as it transitions
from fossil fuels to renewable sources. However, many technical hurdles still remain
which currently limit SOFCs’ viability and must be addressed.
1.2 Solid-Oxide Fuel Cell Operation
The layout for a hydrogen-fueled SOFC is presented in Figure 1.1. Fuel is
exposed to the anode side of the cell, while an oxygen rich gas (usually air) is exposed
to the cathode side. Separating the two sides is the cell itself, with anode, electrolyte,
and cathode. The electrolyte is only conductive to oxygen ions, not electrons. A
perfect electrolyte would have no electronic conduction, allowing no leakage current
from anode to cathode and forcing all electrons to travel via an alternative route. At
the cathode, diatomic oxygen gas will disassociate into oxygen ions, each with a 2−
charge from the addition of electrons from the cathode to the oxygen. The oxygen
ions then travel though the electrolyte and react with the fuel at the anode, freeing
the previously captured electrons. To maintain charge balance, electrons must travel
in the opposite direction to the cathode. Because the electrolyte only allows oxygen
ions to diffuse though it, the electrons travel though an external circuit performing
useful work in the process. A limiting factor in this movement is the ability of oxygen
to transport into, through, and out of the cell. As the rate of oxygen transport is
increased, the ability for electrons to perform work also increases.
2
Figure 1.1: Diagram showing flow of materials in the operation of an SOFC [10]
3
The driving force for the movement of oxygen ions and electrons comes from
the chemical potential gradient across the cell. When the anode is exposed to fuel, it
creates a low partial pressure of oxygen (pO2) in comparison to the cathode, which is
exposed to air. This gradient drives the motion of oxygen through the cell from the
high pO2 of the cathode to the low pO2 of the anode and generates electric current
in the opposite direction to maintain a charge balance. The electrical potential at
open circuit conditions (voltage with no current draw) is expressed by the Nernst
equation, given in Equation 1.2 for the reaction given in Equation 1.1, where fA
is the fugacity of species A, Eo is the standard potential for the reaction, and n is
the number of charges involved in the reaction, R is the ideal gas constant and F is
Faraday’s constant. [11] Commonly, the fugacity is replaced by the partial pressure of
the species. For the case of a hydrogen-fueled SOFC, Equation 1.2 can be simplified
to Equation 1.3. [12] From these equations it can be seen that the partial pressures
of the involved species have a large role to play in the open circuit voltage of the
cell. It is critical that the fuel and air remain separated by the cell. If there is a leak
either through the cell or around it, then the electric potential across the cell will
decrease, harming performance.
aA + bB cC( + dD) (1.1)
o − RT f
c fd
E = E ln C D
nF( (1.2)fa bAfB )
o RT pH2(p
1/2
O )
E = E + ln 2 (1.3)
2F pH2O
While the electric potential is established by the partial pressure gradient, the
electric current is limited by the ionic conduction of oxygen ions from the cathode
4
to the anode. In ionic conduction, oxygen ions hop from site to site via oxygen
vacancies, locations where an oxygen atom is missing from the site it usually occupies
in the crystal lattice. Because conduction depends on vacancies, the more vacancies
that are present in the lattice, the more available sites for oxygen to move between,
the greater the conductance, and the greater power output of a cell. The relationship
between oxygen vacancy concentration and oxygen conductance is given by Equation
1.4, where σi is the ionic conductivity, [VO ] is the concentration of oxygen vacancies
in the lattice, µO is the ionic mobility of the oxygen, and Ze is the total charge
of the conducting species. [13] Thus, by adding more oxygen vacancies the ionic
conductivity increases. While additional oxygen vacancies are a benefit to the
electrical performance of the cell, the resulting changes to the interatomic bonding
can weaken bond strengths.
σi = [VO ]ZeµO (1.4)
During operation, oxygen vacancies are created by removal of lattice oxygen
from the anode as it reacts with the fuel, as shown in the half-cell reaction of Equation
1.5. This reaction takes place on the surface of the anode, but for convenience it
is written where oxygen has associated back into a diatomic state and the fuel
subsequently reacts with it. This reaction of fuel with oxygen creates an environment
with a very low pO2 at the surface of the anode. A steady state concentration of oxygen
vacancies will be reached as the cell comes into equilibrium with the environment, but
this concentration will depend on the exact environmental, material, and performance
5
conditions of the cell.
O×
1 ′
O O2 + VO + 2 e (1.5)
2
The cathode and anode materials must facilitate the incorporation of oxygen
into the cell and the reaction with the fuel. They must also serve to transport the
electrons into and out of the cell as oxygen changes state. As a result, the anode
and cathode must possess both electrical and ionic conductivity. To achieve this, a
mixture of the electrolyte material and an electronically conductive material are used
to create these structures. This mixture of materials creates specific sites where the
cell is active, known as triple phase boundarys (TPBs) and demonstrated in Figure
1.2. The TPB is the location where gas, electronic conductor, and ionic conductor
meet and oxygen can incorporate into the cell or react with the fuel. To assist gas
diffusion to the TPB and to maximize the length of TPBs, pores are added, greatly
increasing the available surface area of the electrodes. The more TPBs that are
present, the more exchange can occur between the gas and the cell at any given time
and the current output of the cell is increased. This added porosity does not come
without the disadvantage of decreasing the mechanical strength of the cell. [14, 15]
An operating SOFC would consist of a series of individual cells stacked together
for combined power output. The cells would be sealed between current collectors
and interconnects and heated to operating temperature. The elevated temperature
enables oxygen conduction though the electrolyte by increasing vacancy concentration
and mobility but decreases the theoretical voltage due to the equilibrium constant of
the H2/H2O reaction. Historically, SOFCs operated near 1000
◦C, but advances in
6
Figure 1.2: Diagram showing the triple phase boundary and the importance as
being the site where incorporation and reactions occur [7]
7
materials have allowed cells to decrease to the intermediate temperature range (650 ◦C
to 800 ◦C) and now even further to the low temperature range (below 650 ◦C). [7]
At lower operating temperatures, it becomes easier to create better seals and the
stack can be built from more readily available materials. Additionally, the cells are
subjected to less thermal expansion, reducing the stresses on the cell.
1.3 Solid-Oxide Fuel Cell Materials
Each structure of an SOFC is fabricated from combinations of materials
designed to optimize that part’s function then laminated with the other structures
to create the cell. For example, a cell can be comprised of a nickel metal-gadolinium
doped ceria, Gd0.1Ce0.9O3 (GDC) anode, a GDC electrolyte, and then a lanthanum
strontium manganite (LSM) cathode. [16, 17] The individual parts are cast into
thin flexible sheets (tapes) from bulk powders, with additives such as pore formers
where appropriate, and then laminated together and co-fired to create a single
cell. As a result, a completed cell has distinct layers in it where the material and
properties abruptly change. This composite structure can then have interdiffusion
between layers, smoothing out the abrupt changes, but creating new structures and
compositions that were not present upon lamination. [18]
The traditional material used for electrolytes in SOFCs has been yttria-
stabilized zirconia (YSZ). YSZ is very stable, has a high strength, has a ratio
of ionic conductivity to total conductivity (transference number) of approximately 1,
and good conductivity at temperatures in excess of 800 ◦C even in very low pO2s. A
8
high transference number minimizes leakage current due to electronic conduction
in the electrolyte, maintaining cell performance. As mentioned previously, GDC is
another choice for an electrolyte. Other rare-earth doped cerias have been used as
electrolytes, but GDC tends to outperform them under a variety of conditions. [19–21]
At 600 ◦C, GDC exhibits an order of magnitude higher conductivity than YSZ, but
has a smaller electrolytic domain, where the transference number is near 1. [22]
Under conditions of high temperature or very low pO2 , GDC will become electrically
conductive and start to leak electrons through the electrolyte, reducing cell perfor-
mance. As SOFC operating temperatures are reduced for improved efficiency, GDC
is favored for the electrolyte material.
The anode usually consists of a mixture of the electrolyte material, which
provides good ionic conductivity, with an electronic conductor that can facilitate
fuel oxidation reactions on the surface. Nickel has commonly been used in the
anode due to its high conductivity and catalytic properties.In combination with
an ionic conductor, the nickel creates a cermet which functions as the anode. A
challenge with nickel is that at ambient conditions it oxidizes into nickel oxide, but
at operating conditions it reduces to the desired nickel metal. There is a large lattice
parameter change between nickel and nickel oxide, such that, upon reduction it
decreases volume by over 15%. As a result, the overall porosity of the anode layer
increases as the SOFC is put into service and the anode is exposed to reducing
conditions. [23, 24] This increased porosity can greatly decrease the flexural strength
and modulus of the anode and of the overall cell. [25, 26] Redox cycling presents
an additional challenge because cycling of oxidation and reduction phases causes
9
crack growth and failure after a few cycles. [14,15,24,27–30] Additionally, nickel is
prone to poisoning from sulfur contaminants in the fuel, so alternative anodes are of
interest for use in SOFCs. [31–34]
Alternative metal-ceramic systems have been researched to serve the function
of the anode. These have included copper, cobalt, and palladium systems, but each
suffer from a significant drawback such as coking, long-term performance degradation
or cost. [35–37] A different approach is to use an all-ceramic anode, made of a
mixed ionic-electronic conductor (MIEC) where both ionic and electronic conduc-
tions occur simultaneously, such as such as La1–xSrxCr1–yMnyO3, Sr0.94Ti0.9Nb0.1O3-δ,
Ba0.98La0.02SnO3, or Sr2MgMoO6. [38–44]
Perovskite structures have yielded a number of MIECs with potential uses as
electrode materials. [45–47] Sr2MgMoO6 (SMM) and its related family of materials
(Sr2MMoO6, where M is a transition metal dopant) yield good conductivities and
catalytic actives. [48] The mixed valence state of Mo(VI)/Mo(V) provides high
electronic conduction while supporting oxygen vacancy formation. [49] If a dopant is
used which has an overlapping redox couple band, such as Fe, electronic conduction
can be further improved. Mo(VI) and Mo(V) are stable in both octahedral and
tetrahedral coordination, further adding stability to oxygen vacancy formation. [50]
SrFe0.2Co0.4Mo0.4O3-δ (SFCM) is a recently developed double perovskite MIEC with
multiple dopants on the B-site with a high conductivity of 30 S/cm at 650 ◦C. [51–53]
Due to the all-ceramic nature of these anodes, their thermal expansion coefficients
better match that of the other ceramic components of the cell, reducing sintering
stresses and improving redox cycling durability. [38] Conversely, all-ceramic anodes
10
tend to have lower conductivity and catalytic performance. [54]
This work focuses on the structures in the cell which give mechanical sup-
port, the anode and electrolyte. Specifically, Ni-GDC and SFCM-GDC with GDC
electrolytes were investigated. While YSZ systems have been studied extensively,
the mechanical properties of GDC systems have not been thoroughly explored. As
SOFCs are developed which operate at lower temperatures, GDC is the preferred
choice for an electrolyte material due to its high conductivity. Ni-GDC serves as a
reliable starting place to characterize the mechanical properties of the anode, focusing
on the changes which occur during reduction as NiO reduces to Ni metal. SFCM
being a new material has no available data on the mechanical properties, let alone
the non-stoichiometry which occurs in the material under reducing conditions.
Various cathode materials exist which assist in the incorporation of oxygen
into the lattice. The cathode must have many similar properties to that of the anode
but be stable in oxidizing environments instead of reducing. As a result, perovskite
structures again are found in many of the materials. LSM is one of the most
commonly used materials, but others include doped SrCoO3, La0.6Sr0.4Co0.2Fe0.8O3,
BaxSr1-xTi1-yFeyO3-y/2+δ. [55–58] The cathode does not tend to play a structural role
in the fabrication of a cell, and thus is not investigated or developed in this work.
Outside of the cell itself, several materials are needed to create a functioning
stack. Metal interconnects support individual cells, holding them together, creating
connections for the current to flow through current collectors and gas to flow through
channels to the cell. [59] In intermediate and high temperature stacks, expensive
metal alloys such as Crofer 22 APU treated with conductive ceramics like LaCrO3
11
are needed to survive the extreme heat. [60,61] As temperatures are lowered to below
600 ◦C, more common metals can be used for interconnects, such as steel. Sealing
materials are another critical component to assembly of an SOFC stack. Glasses
with transition temperatures near the operating range are used at high temperatures
as they can soften and flow into place creating gas tight seals. At lower temperatures
new options are available, such as alumina felt or mica gaskets. [62, 63]
1.4 Mechanical Properties
Ideally, an SOFC is as thin as possible to minimize diffusion path lengths and
ionic resistances in the cell. [64] Realistically, the cell must be able to withstand
the stresses of being manufactured, sealed, heating, and use. This means that a
compromise must be made as to how much the cell is supported and which components
do the supporting. Traditionally, electrolyte supported cells were used with YSZ
electrolytes where the thick dense electrolyte provides the structural support for
the cell. Recently cells with an anode support layer (ASL) made from GDC have
been able to provide lower resistances due to the thinner electrolyte while adequately
supporting the cell with a thicker anode which fuel can diffuse further into. [65,66]
Now the anode layer with its porosity must support the majority of the stresses the
cell is subjected to.
Ceramic materials used in SOFCs mechanically fail when stresses concentrate
at a pre-existing microstructural flaw, overcoming the strength of the material,
propagating a crack though the cell. Stresses can be intrinsic to the cell, left from
12
lamination, sinter or cooling, or they can be applied externally during the sealing
or operation of the cell. Microstructural flaws could be intentional with the added
porosity or unintended flaws such as inclusions, contamination or grain agglomeration.
Stresses are concentrated most by large, sharp featured flaws. As a result, in a
uniform stress field, the largest flaw will usually cause failure. Ceramic materials
will tend to fail where the sample is under tensile load and at a location near the
point of maximum applied stress. Samples, even if processed together from the same
raw materials, will each have a random sampling of flaws from the total batch, and
a particular one which causes failure. No matter the care and attention paid to
processing, some distribution of flaws will exist. The analysis of the distribution of
flaws from a batch of samples is known as Weibull analysis and is explained in detail
in Appendix A.2. To perform Weibull analysis, a large number of samples must be
tested, but as a result the overall size/shape distribution of flaws can be obtained.
This allows for the observation of the flaw distribution and if it changes at different
points or if another phenomenon is occurring. One unique benefit of Weibull analysis
is that the strength of a sample can be scaled up to a larger piece, as long as the
distributions of flaws, or the Weibull modulus, is known.
Most of the materials used in SOFCs have crystal structures which promote the
generation of oxygen vacancies. As oxygen vacancies are produced, the inter-atomic
bonding of the structure changes, which can decrease the elastic modulus and fracture
toughness of the crystal. [67, 68] If the strength of the atomic bonds weakens on
average, due to added vacancies, it then follows that the elastic modulus and fracture
toughness of the crystal would also decrease. This relationship has been shown to fit
13
for single grains of GDC using nanoindentation, but this does not necessarily hold
true for an actual cell. [69] Grain boundaries can play a large role in the mechanical
properties of a bulk sample as they can impede or enhance the propagation of a
crack. For this reason, microstructure combined with environmental conditions can
play a large role on the overall mechanical properties of a fuel cell.
Another parameter which can change based on environment is the fracture
toughness of the material. The fracture toughness is how much stress it takes for a
crack to grow and propagate once initiated, or in other words the material’s ability
to resist crack growth. Just as modulus changes upon heating or reduction, so too
can fracture toughness change. Equation 1.6 demonstrates how fracture toughness
and flaw geometry impact the strength of a sample, where σ is the strength, KIc is
the fracture toughness, Y is the shape factor for the flaw which caused failure and
a is the size of the flaw. [70] The fracture toughness and flaw distributions play a
direct role in the overall strength of the cell.
K√Icσ = (1.6)
Y a
The first external stress applied to the cell, and the one most likely to result
in failure, is the sealing of an individual cell into a stack, creating a gas tight seal
between the anode and cathode sides. In this process, multiple cells are sandwiched
with gasket seals and interconnects between them and compressed until a gas tight
seal is achieved. While it is a compressive force that is applied, the cells are never
perfectly flat, resulting in flexural stresses as the cell is pressed flat. Figure 1.3
highlights this fact by using an optical profilometer to measure the flatness of a
14
10 cm by 10 cm Ni-GDC/GDC half-cell. It can be seen that the cell curves by over
a millimeter, mostly at edges, where the seal would be taking place. It is for this
reason that the study of the failure of SOFC materials utilizes flexural testing.
1.5 Problem Statement
Much of the study of fracture in ceramic materials has been done on technical
ceramics for medical applications and for coating metal components. [71–78] These
materials are optimized for fracture toughness and durability and very rarely ex-
perience temperatures above a few hundred degrees Celsius. The fracture surface
analysis and correlations between microstructure and strength described for these
materials are a valuable starting point for investigating fuel cell materials. However,
there is a lack of extensive investigation into the properties of low temperature SOFC
materials at their expected operating temperatures and environments.
Most of the previous research into the mechanical properties of SOFCs has
focused on YSZ based devices. [24,79,80] This material has been the standard for
the field and has desirable mechanical properties but requires high temperatures to
function well. Nakajo et al. conducted a wide ranging study of materials used in anode
supported SOFCs which included some attention to temperature and atmosphere
effects but did not go beyond YSZ. [80] As efforts are made to lower SOFC operating
temperatures, a shift to ceria-based electrolytes has occurred. Less attention has been
given to the mechanical behavior of doped ceria materials across SOFC operating
conditions. Flexural strength and Young’s modulus measurements for GDC have
15
Figure 1.3: Optical profilometer measurements of a 10 cm by 10 cm Ni-
GDC/GDC half-cell where the z-axis has been magnified 6X to highlight
curvature
16
been carried out in ambient conditions by Yasuda et al. [81] They characterized the
effects of sintering temperature on density and the resulting mechanical properties.
Further testing of this material system at elevated temperatures and anode gas
reducing environments must be done to fully understand the mechanical behavior of
GDC.
Additionally, most research has focused on measuring and comparing the
Young’s moduli of different materials as it changes with temperature or after reduction
treatments at room temperature. [23,82–84] While the modulus can be correlated to
the fracture strength of the cell, direct measurement of flexural strength remains the
best means to characterize the strength of the cell.
To help develop SOFCs into a viable technology which can stretch the gap in the
energy economy, research is needed to improve the reliability of the cells during man-
ufacturing and operation. This work aims to develop the understanding of fracture
mechanics of real-world structures, by understanding how temperature and partial
pressure of oxygen affect the strength, fracture toughness, and flaw distributions
of SOFC components. This was done by analyzing the flexural properties of bulk
bars and half-cell coupons at ambient conditions and at conditions similar to those
found in an SOFC. Microscopy was used to look for changes in microstructure after
various treatments in addition to Weibull analysis. Additionally, thermogravimetry,
conductivity, and temperature programed desorption were used to understand the
atomic defect structure of a new material to relate it to the mechanical properties.
17
Chapter 2: Experimental Procedures
2.1 Sample Preparation
2.1.1 SrFe0.2Co0.4Mo0.4O3-δ Powder
SFCM was created from stoichiometric amounts of strontium carbonate (SrCO3,
Sigma-Aldrich), iron oxide (Fe2O3, Sigma-Aldrich), cobalt oxide (Co2O3, Inframat
Advanced Materials), and molybdenum oxide (MoO3, Alfa-Aesar) using conventional
solid-state methods. The constituents were ball milled for 24 hours in ethanol and
dried using a 100 ◦C oven. Afterwards the powder was calcined at 1100 ◦C for four
hours.
2.1.2 Bars
Porous GDC bars were created by uniaxially pressing a mixture of GDC and
pore former in a rectangular die. GDC powder (Gd0.1Ce0.9O2-δ, fuelcellmaterials)
was mixed with the varying amounts of of poly(methyl methacrylate) (PMMA)
microspheres (5 µm diameter) or graphite flake with 0.1wt% of polyvinyl butyral
added to act as a binder during pressing and a drop of fish oil to act as a dispersant
during milling. Different pore geometries were created from the flake and spherical
18
pore formers. The PMMA or graphite was burned out by pre-sintering at 400 ◦C,
which was followed by sintering at 1500 ◦C for 4 hours. Percent porosity was measured
using the Archimedes density technique and samples were polished with 600 grit
sandpaper before testing.
Dense SFCM bars were used to maximize the total mass and mass changes
during thermogravimetric testing. Samples were made by combining SFCM powder
with 0.6 wt% polyethylene glycol 600, 1.8 wt% ethylene glycol, and 0.6 wt% glycerol
in isopropyl alcohol and ball milling overnight. After drying at 100 ◦C, the powder
was ground by mortar and pestle, then pressed uniaxially into rectangular bars at
30 MPa, then isostatically pressed at 30 MPa. Bars were sintered by heating to 400 ◦C
for one hour and then 1340 ◦C for four hours, using a 3 ◦C min−1 heating and cooling
rate. This produced bars with 97% theoretical density by Archimedes’ technique.
To create fracture toughness test samples, dense bars were cut and sanded to final
dimensions of 3 x 4 x 25 mm. The chevron notch required for testing was cut using the
jig described by Jenkins, Chang and Okura following the ASTM procedure. [85, 86]
2.1.3 Test Coupons
Tape casting was used to create ASL and half-cell coupons used in flexural
testing. A 6:4 weight ratio of NiO and GDC was ball milled with ethanol, toluene,
and Menhaden fish oil for 24 hours. Afterwards, polyvinyl butyral and benzyl butyl
phthalate were added as binder and plasticizer and the mixture was ball milled for
an additional 24 hours. The slurry was degassed using a vacuum and then tape
19
cast using a 700 µm doctor blade height. After drying overnight, the tapes were cut
into 12 cm by 12 cm squares. Three squares were layered on top of each other and
laminated using a hot press at 49 ◦C for 30 minutes at 2 tons. After lamination, the
green tapes were cut into rectangular coupons which measured 25 mm by 10 mm.
The sample coupons were fired at 1400 ◦C for 4 hours, after which the average final
dimensions were 8.16 x 24.15 x 2.98 mm.
A GDC electrolyte layer was cast using a 40 µm blade height for the creation
of half-cell coupons. Three ASL layers were laminated as before, then the additional
single electrolyte layer as before was laminated on for 2 hours. Half-cell coupons
were sintered at the same parameters as the ASL coupons.
Tape casting was also used to create test coupons of porous SFCM-GDC ASL
and SFCM-GDC/GDC half-cells. Using ethanol as a solvent, SFCM-GDC was ball
milled with polyvinyl butyral, benzyl butyl phthalate, 12 µm PMMA (16 wt% with
respect to SFCM-GDC), and Menhaden fish oil. The tape was cast to a thickness
of 110µm on Mylar then laminated using a hot press to a final thickness of 660µm.
Dense GDC was casted to 30 µm and laminated to the top of the SFCM-GDC to
create half-cells. Individual coupons were cut from the green tape, sintered at 1200 ◦C
for four hours, with holds at low temperatures to burn out organic binder and pore
former. The final thicknesses were measured to be 400µm for the SFCM-GDC ASL
and ∼20 µm thick GDC electrolyte.
Test coupons had their edges sanded to remove defects left from cutting,
following ASTM standards. [87] Top and bottom surfaces were not sanded to preserve
possible defects left from tape casting procedure, which would be representative of
20
industrially manufactured SOFCs.
2.2 Density Measurements
Density of test bars was determined using the Archimedes method following
Equation 2.1, where ρ is the density of the sample, ρliq is the density of the liquid
used, wdry is the weight of the sample before submersion and wsub is the apparent
weight after submersion. Isopropyl alcohol was used as the liquid for measurements
and samples were placed in a vacuum chamber while submerged to ensure all open
porosity was filled before measuring the submerged weight.
wdry
ρ = ρliq (2.1)
wdry − wsub
2.3 X-Ray Diffraction (XRD)
X-ray diffraction (XRD) was used to confirm phase purity of the SFCM after
synthesis and during the testing process. A Bruker D8 Advance with LynxEye
was used with a Cu Kα source. A step size of 0.02° was used with a dwell of 0.8 s.
GSAS-II was used to perform the Rietveld refinement calculations and VESTA was
used to visualize the unit cell. [88, 89]
2.4 Thermogravimetric Analysis (TGA)
Thermogravimetric analysis (TGA) of samples during oxidation or reduction
was measured using a Cahn D200 microbalance. The samples were placed in a
crucible suspended from a platinum wire or suspended from the wire itself. The wire
21
attached to the microbalance and was enclosed by an alumina tube inside a furnace.
Heating control was achieved with a PID loop and temperature measurement done
by a K-type thermocouple placed immediately below the sample inside the furnace
tube. A heating and cooling rate of 10 ◦C min−1 was used. Mass, temperature, and
pO2 measurements were taken at fixed intervals.
Gas flow was controlled at consistent 50 sccm by mass flow controllers, which
mixed dry nitrogen, oxygen, hydrogen, and humidified nitrogen to obtain the pO2
desired. For general oxidizing or reducing conditions, 21% O2 in N2, or 3% H2, 3%
H2O in N2 was used. Measurements of the pO2 were taken by a calibrated YSZ
oxygen sensor at 800 ◦C located before the sample. For glsSFCM, intermediate pO2
ranges were not able to be tested due to SFCM’s incompatibility with the species
required to create that environment.
Further details of the operation of the TGA and its various components are
given in Appendix B.
To prepare a sample for non-stoichiometry measurements, it was pre-weighed,
wrapped in platinum wire and suspended from the balance, placed in the furnace
with simulated air (21% O2, 79% N2) flowing. Once the mass had stabilized, the
furnace was heated to 800 ◦C to allow for any organic contamination to burn off.
After the mass stabilized at this elevated temperature, the sample was introduced
to various environments. The mass of the sample would be noted only after steady
state had been reached for a condition. After testing a sample, a bar of alumina was
cut to the same dimensions as the sample and the process was repeated to obtain a
blank which could be subtracted from the measurements to remove buoyancy effects.
22
Oxygen non-stoichiometry was calculated from thermogravimetry data using
Equation 2.2, where δ is the degree of non-stoichiometry, MWSFCM is the molecular
weight of SFCM (208.74 g mol−1), MW is the atomic weight of oxygen (16.0 g mol−1O ),
wsample is the weight of the sample, and ∆w is the weight change as recorded by the
TGA. The oxygen vacancy concentration ([VO ]) is calculated using Equation 2.3,
where ρ is the density of SFCM, and NA is Avogadro’s number. For this work, based
on the plateau present in the data at oxidizing conditions, it is assumed that in a
pure oxygen environment all oxygen vacancies are filled with no oxygen interstitial
or surface species (δ = 0 at log(pO2) = 0). [90–92]
MWSFCM
δ = ∆w (2.2)
MWO wsample
δρNA
[VO ] = (2.3)
MWSFCM
2.4.1 Temperature Programmed Desorption (TPD)
The effluent from the TGA was used as the inlet to a mass spectrometer (MS)
to perform temperature programmed desorption (TPD). The sample was prepared
as before and heat treated to remove any carbon contaminants but was allowed to
cool under simulated air. It was then heated to 800 ◦C at 5 ◦C min−1 under a 50 sccm
flow of nitrogen as the MS measured the 32 m/z signal which corresponded to O2
desorption. Additional m/z signals were monitored to observed for other species.
23
2.5 Conductivity
Electric conductivity across pO2 was measured using the four-wire technique and
a Stanford SR 830 lock-in amplifier. A bar shape sample was used with dimensions
of 6.46 mm× 3.3 mm× 1.3 mm. Gold paste was used as a current collector, and
the current range was between 0.005 to 0.05 A. A YSZ oxygen sensor operating at
800 ◦C was used to monitor the changes in oxygen partial pressures. Intermediate
pO2 ranges were not tested due to an incompatibility between SFCM and the CO
and CO2 required to obtain those pO2 .
Conductivity during redox cycling was measured using rectangular bars of
SFCM and SFCM-GDC (2:1) composite. The samples were connected to a Keithley
2400 source meter by silver paste and wire. Using an in-house built reactor, the
sample could be heated and the gas environment could be controlled. An initial
measurement was taken after heating and 50 hours of exposure to 10% H2 in N2,
then the sample exposed cycled between air and reducing conditions over a period of
14 days.
2.6 Mechanical Testing
Measurements of mechanical properties were collected using a Tinius Olsen
10ST universal testing machine (UTM) equipped with a 250 N load cell. Experiments
where any samples would be tested at elevated temperatures or under reducing
environments were conducted using a custom built three point flexural test fixture
24
placed inside a gas-tight chamber and furnace. Otherwise, a fully-articulating four
point flexural test fixture was used. Samples tested at ambient conditions were
placed on the appropriate testing fixture and loaded until failure. For samples tested
at elevated temperatures, the sample was heated in the test chamber at 10 ◦C min−1,
allowed to equilibrate for 20 minutes, then tested. Samples to be reduced were placed
in the chamber and exposed to reducing conditions as described below.
All strength tests conducted in the UTM were done at a rate of 0.2 mm min−1
with a 20 mm lower span. After measuring the force at fracture, stress was calculated
using Equation 2.4 for 3-point flexural of rectangular samples, where σf is the stress
at failure, F is the load at failure, L is the span of the fixture, b is the width
of the sample and d is the thickness of the sample. At room temperature, the
coupons or bars were loaded into the fixture and tested in batches of 5 per condition.
Occasionally samples would break or be damaged before testing, reducing the sample
set. For sample sets at elevated temperatures in ambient atmosphere, the samples
were loaded into the front of the chamber, acting as a staging area, prior to heating
the on fixture in the center. Upon transferring the next coupon from the staging
area to the fixture, a 20 minute waiting period was used to ensure that the coupon
had reached thermal equilibrium prior to testing.
3FL
σf = (2.4)
2bd2
Following the completion of the set, all coupon pieces were cooled at a rate of
10 ◦C/min. For testing in reducing atmosphere, each coupon was reduced and tested
individually. Half-cell coupons were tested in two orientations, “electrolyte-up” and
25
“electrolyte-down.” These orientations cause the dense electrolyte layer to experience
compressive or tensile stress. The “electrolyte-up” orientation subjects the electrolyte
to compressive stress and vice-versa.
2.6.1 Development of Mechanical Test Apparatus
To develop a testing apparatus capable of simulating the various conditions
experienced by operating SOFCs, appropriate materials were chosen based on thermal
and chemical stability criteria. Alumina was chosen to build the bend fixture due
to its chemical stability and high hardness. The sample rests crossways on two
stationary 6.35 mm diameter rods which are placed in troughs separated by 20 mm.
The upper half of the fixture consists of a 6.35 mm rod used to apply stress to the
sample from the UTM crosshead. Both pieces are attached using silica-based cement
to 300 mm rods attached at the anchor points of the UTM. Assembly and alignment
is assisted with the use of a 3D printed jig to ensure repeatability. The complete
alumina, 3-point bend fixture is seen in Figure 2.1 with the two bottom rollers, a
sample, and the top roller making contact, applying flexural stresses to the sample.
To create the atmosphere control system, a combination of standard and custom
vacuum system parts were used to enclose the bend fixture. For the main body
of the chamber, a 3-inch inner diameter, stainless steel tee was made with QF80
and QF50 flanges. Welded NPT ports on top and bottom allow for gas inlet and
outlet and the introduction of a thermocouple through a Swagelok Ultra-Torr fitting.
Figure 2.2 shows the engineering drawings and photo of the custom pieces from
26
A&N Corp. Flexible bellows allow for the motion of the cross-bar to be translated
into the fixture, requiring the subtraction of the spring force to be removed during
analysis. The alumina fixture was then inserted through each end and the chamber
incorporated into the furnace with QF50 to 1/2 inch tube adapters that fit over the
alumina rods.
To heat the fixture and chamber, a 1 ft. cube was assembled from steel
plates with cutouts on top, bottom and front. Steel wire mesh was used to create
a framework inside to hold nickel-chromium alloy heating elements. Silica-based
wool insulation was packed between the center cavity and the case walls. A K-type
thermocouple was inserted into the chamber and used with an external PID controller
to cycle the heating elements. The furnace was placed on a supporting scaffold
to hold it in position during tests. The test fixture, atmosphere chamber, and
furnace were affixed together as one unit and inserted into the UTM. The design was
compared and qualified to a regular steel 3-point bend apparatus to ensure accurate
results at room temperature.
2.6.2 Atmospheric Treatment
To test individual coupons under reducing environments, samples were loaded
into the test chamber at 400 ◦C, after previous pre-heating in the staging area. After
20 minutes, the temperature would be increased to the desired set-point of 650 ◦C at
a ramp rate of 10 ◦C min−1 and the gas would be switched first to N2 to flush the
chamber, then to humidified argon containing 3% H2. It would then be allowed to
27
Figure 2.1: Assembled Alumina 3-Point Bend Fixture
Figure 2.2: Design drawing and photo of mechanical testing atmospheric
chamber
28
sit for 18 hours before testing to allow reduction. After testing, the chamber would
be flushed again with N2 and cooled to 400
◦C before opening to change samples and
repeat the process.
If samples had been previously reduced as a batch by exposure to hydrogen in
a tube furnace and cooled under hydrogen, the atmospheric treatment was shortened,
as supported by TGA results. Pre-reduced samples were loaded into the chamber at
or below 400 ◦C, the chamber flushed N2, then with H2 in Ar. After this the chamber
would be heated to 650 ◦C, the sample tested after a 30 minute period and cooled
back down to 400 ◦C. At this point, the chamber would be flushed with N2. While
the chamber maintained a positive pressure of N2, it could be opened, a new sample
loaded, closed, and allowed to flush. This procedure allows for the batch reduction
of many test coupons and replaced the 18 hour reduction time with a shorter 30
minute period.
2.6.3 Fracture Strength
A loading rate of 0.2 mm/min was used for all strength measurements. Strength
was calculated from the maximum force measured before failure according to Equation
2.5 or 2.4 depending on if the four point or three point fixture is used respectively,
where σ is the strength, P is the maximum force, L is the span width of the test
fixture, b is the width of the sample and d is the thickness of the sample. Equation
2.5 is for a fixture where the top span is 1/2 the width of the bottom span. [87]
3PL
σ = (2.5)
4bd2
29
2.6.4 Fracture Toughness
The fracture toughness of chevron notched samples were measured using a
loading rate of 0.001 mm/min. Fracture toughness was calculated from the maximum
force using Equation 2.6 or 2.7 for four point or three point fixtures. [86, 93] Y ∗min is
the shape factor as calculated by Equation 2.8, So and Si are the outer and inner
spans, B is the width of the sample, W is the height of the sample, a0 is the distance
from the tip of the chevron notch to the bottom of the sample, and a1 is the average
distance from the side of the chevron notch to the bottom of the sample. Each
sample was measured after failure, but in this study the approximate values were
So = 40mm, Si = 20mm, B = 3.0mm, W = 4.0mm, a0 = 0.80mm, a1 = 3.8mm.
Fracture toughness was measured only under air at room temperature and up to
600 ◦C. Under reduction the fracture toughness bars of SFCM would spontaneously
fracture, preventing measurements under that condition.
[ ]
∗ P [So − Si]K −6Ivb = Ymin 10 (2.6)[BW 3/2]
P
K ∗Ivb = Ymin 10
−6 (2.7)
BW 1/2
∗ 0.38742− 3.0919(a0/W ) + 4.2017(a1/W )− 2.3127(a 21/W ) + 0.6379(a1/W )3Ymin = 1.0000− 2.9686(a0/W ) + 3.5056(a0/W )2 − 2.1374(a 30/W ) + 0.0130(a1/W )
(2.8)
2.7 Scanning Electron Microscopy (SEM)
After fracture, scanning electron microscope (SEM) images were taken of the
fracture surfaces. This was to observe the nature of the crack as it propagated,
30
determining if it was transgranular or intergranular fracture, to find any flaws on
the fracture which would lead to failure, and to observe pore geometry. Images
were captured on an Hitachi SU-70 Schottky field emission SEM. Fracture surface
images were mounted vertically and GDC samples were sputter coated with gold.
Observations of porosity were taken using epoxy infiltrated samples and polished
using down to a 0.25 µm diamond solution.
31
Chapter 3: High Temperature Mechanical Behavior of Porous Ceria
and Ceria-Based SOFCs
3.1 Introduction
This study presents the results of a broad range of flexural tests involving the
materials used in ceria-based anode supported SOFCs along with the assembled
half-cells. Using a purpose-built temperature controlled environmental chamber
installed in a UTM, porous doped ceria bars, ASLs composed of nickel and doped
ceria cermet, and half-cells composed of an ASL and a doped ceria electrolyte were
tested. The various test conditions used included expected operating temperatures
(450 ◦C–650 ◦C), and both reducing and oxygenated atmospheres. These variations in
test conditions are important because these cells must maintain their integrity from
when they are first placed in a sealed stack to when they reach operating conditions.
In particular, this study was intended to determine at what point in their life SOFCs
are most vulnerable to mechanical failure and the mechanisms involved. Additionally,
the effect of the anode-electrolyte interface on flexural strength was explored.
32
3.2 Results
3.2.1 Porosity and Pore Former Choice
As expected, GDC bars with greater porosity displayed lower flexural strength
values as compared to less porous samples. The relationship between porosity and
strength followed an exponential trend. This behavior is well established for porous
ceramics and is described by Equation 3.1, where σf is the flexural strength with
porosity, σo is the flexural strength without porosity, η is a geometric constant
dependent on the system, and P is the volume percent porosity of the sample. [94]
σf = σ
−ηP
oe (3.1)
The data was fit using this equation and both a fixed geometric constant, η, of
0.047 (Figure 3.1a) and a variable geometric constant (Figure 3.1b). Chi-squared
values, indicating the divergence of the fit from the measured data for each sample
set, were calculated for the two fitting methods and are displayed in Table 3.1.
The deviance from the fit curve for the fixed η method is quite large relative
Table 3.1: Chi-squared values for fixed η and variable η fits of porous GDC
strength data (Figure 3.1)
Sample Set χ2 (Fixed η) χ2 (Variable η)
GDC PMMA 650 ◦C 3.406 0.062
GDC Graphite 650 ◦C 0.668 0.008
GDC Graphite 25 ◦C 3.575 5.063
33
Figure 3.1: Flexural strength-porosity dependence for porous GDC at 650 ◦C
and 25 ◦C using spherical PMMA or graphite flake pore former: a) Measured
strengths fitted using fixed geometric constant; b) Measured strengths fitted
using different geometric constants
34
to the variable η fit for both sample sets tested at 650 ◦C. Both fitting methods are
fairly poor in the case of the room temperature graphite set. The poor fitting results
in this case are due to the >60% porosity samples. The flatter pore geometry caused
by the graphite likely leads to comparatively greater strength loss at high porosity
due to greater pore connectivity. The comparison between the high temperature
data sets is the most useful, given this, and the fact that SOFCs operate at this
temperature. The poor fit of fixed η models suggests that a material-specific η value
may be an oversimplification of the porosity-stress relationship in ceramics.
Samples tested at higher temperatures displayed a slightly higher flexural
strength. Generally, it is expected that increasing temperature lowers strength
of dense ceramics once plastic deformation processes become activated at high
temperatures. [95] Below the temperature at which plastic deformation starts to
occur, temperature has a minimal effect on flexural strength of dense ceramics. For
dense GDC, it has been demonstrated that the strength decreases by 19% between
room temperature and 800 ◦C. [13] Porosity changes this behavior by essentially
creating a composite material, keeping the rate of change constant or increasing with
temperature. [82] For porous GDC, thermal expansion places a compressive stress
on any surface flaws that are sites for crack initiation and propagation. Additionally,
the porosity helps reduce bulk stresses caused by thermal expansion. These two
effects increase the material’s fracture strength more than any decrease from plastic
deformation at 625 ◦C.
Cross-sectional SEM of the fracture surfaces of the porous GDC samples, both
PMMA and graphite, is shown in Figure 3.2. Figure 3.2a shows individual spherical
35
pores left by the PMMA while Figure 3.2b has an inter-connected network of long
pores left by the graphite flakes.
The differences in quality of fit between Figure 3.1a and Figure 3.1b shows that
the geometric factor is microstructurally dependent. Samples with porosity formed
using PMMA spheres showed significantly greater strength as compared to samples
with graphite-formed porosity. This can be explained by the effect of pore geometry
on crack initiation and propagation in the ceramic. If a crack enters a pore, the pore
can now be considered the new crack tip. The energy required to advance the crack
is highly dependent on the geometry of the tip. For a spherical pore, this geometric
factor is maximized and results in higher resilience to fracture. This is summarized
by Equation 3.2 where σo is the stress at the crack tip, c is the length of the crack
tip and ρ is the radius of curvature of the crack tip. [96]
√
σo ρ
σf = (3.2)
2 c
Based on the mechanical behavior of these samples, porous ceramics should
be designed and constructed such that the pore geometry is as low aspect ratio as
possible to maximize strength.
3.2.2 Flexural Stress Orientation
Figure 3.3 shows the measure flexural strengths of a number of SOFC coupon
samples. Anode support coupons were tested along with half-cells consisting of anode
supports and electrolyte layers at temperatures ranging from 25 ◦C to 650 ◦C. The
highest temperature test was repeated with coupons under reducing conditions (3%
36
Figure 3.2: SEM micrographs of porous GDC bars made using: a) PMMA
pore former; b) graphite flake pore former
37
H2, balance Ar).
All unreduced sample types showed increased strength at elevated temperatures
with little difference between types at a given temperature. The high-temperature
reduced coupons displayed large differences in strength depending on the orientation
of the sample. Tests in which the dense electrolyte layer was placed in compression
resulted in the highest strength values, while the samples were weakest when the
electrolyte was placed in tension. In reduced samples, the anode support layer
becomes a ceramic-metal composite and is therefore somewhat elastic while the
electrolyte remains a brittle ceramic. The electrolyte-in-compression condition
maximizes the mechanical performance of the coupon by placing the layers in their
preferred stress state.
Figure 3.4 shows the box plots for the unreduced SOFC coupon samples. A
Student’s t-test was used to statically determine if the means of samples tested at
different conditions were equal based on 95% confidence. Details on the use of the
test are given in Appendix A.1. There is no discernible difference in the strength
between the three sample types at each temperature. At elevated temperature there
was a significant difference in the modulus of ASL only samples and samples with
electrolyte in tension with a p-value of 0.0091. This difference was not present at
room temperature. As the materials reach elevated temperature, small differences
in elasticity become magnified due to different thermal effects on dense and porous
layers.
Additionally, SEM analysis of the fractured half-cells showed very good adhesion
between layers (Figures 3.5 and 3.6). Delamination is a common failure mode in
38
Figure 3.3: Temperature dependent strength of Ni-GDC anode supports and
half-cells in both air (filled data points), and reducing atmosphere (3% H2,
balance Ar, hollow data points), tested in both “electrolyte-up” and “electrolyte-
down” orientations resulting in the electrolyte layer experiencing tension and
compression, respectively
39
Figure 3.4: Flexural test measurements of coupons sample sets: a) strength at
25 ◦C b) strength at 650 ◦C c) modulus at 25 ◦C d) modulus at 650 ◦C
40
layered ceramics and one that would be particularly damaging to SOFCs due to
resulting ionic conductivity loss between layers. [97] In the half-cell coupons, it
was clear that the fracture plane contained mixed transgranular and intergranular
fracture. Some grains were sheared through while others remained whole. The
striations visible in Figures 3.5b and 3.6b are characteristic of fracture proceeding
through a grain, while other grains remained whole.
3.2.3 Reduction and Strength
Mass loss of SOFC half-cell coupons exposed to reducing atmosphere at various
elevated temperatures is shown in Figure 3.7. Mass loss is attributed to the reduction
of NiO, used as a precursor in fabrication, to Ni metal, which serves as the catalyst
for fuel oxidation and electronic conductor, and the reduction of ceria to CeO2-δ.
Ni-GDC/GDC half cells showed the expected trend of increasing reduction rate at
higher temperatures. Reduction curves were fit to an exponential decay and the
summary of parameters is shown in Table 3.2. At 650 ◦C and 700 ◦C the reduction
occurs very quickly, reaching steady state values after 18 hours. 550 ◦C showed a
much slower mass loss than even 575 ◦C. At temperatures lower than 550 ◦C, the
kinetics are slow enough to tolerate a brief exposure to oxygen without re-oxidizing
the sample. Of note in the reduction is that each temperature appears to approach
a different asymptote, showing that the amount of NiO and GDC reduced at steady
state is dependent on the temperature of the cell. This will affect the mechanical
properties of the cells as it changes both the porosity and amount of nickel metal in
41
Figure 3.5: SEM micrographs of unreduced half-cell fracture surface tested in
air at 25 ◦C at a) 2k magnification and b) 5k magnification
42
Figure 3.6: SEM micrographs of reduced half-cell fracture surface tested in
reducing atmosphere at 650 ◦C at a) 1k magnification and b) 5k magnification
43
Table 3.2: Summary of fit parameters for reduction of NiO-GDC/GDC half-cell
coupons under 3% H2, 3% H2O, 94% N2 at different temperatures
Temperature (◦C) Exponential Rate (h−1) Asymptote (Mass %)
550 0.0463 89.78
575 0.135 89.09
600 0.182 88.36
650 0.512 87.75
700 0.821 87.87
the samples.
3.2.4 Reoxidation at Low Temperatures
Ni-GDC/GDC fuel cells are sensitive to re-oxidation once reduced. Re-oxidation
causes fracture of cells due to the large volume difference between NiO and Ni. [80]
Based on the results from the TGA, the reaction rates below 300 ◦C are sufficiently
slow to allow for the exposure of a reduced cell to oxygen without detrimental
results. [98] To confirm this, a sample was measured by TGA to ensure no mass gain
during oxygen exposure and the strength of samples were tested to ensure there was
no discernible difference between reduction methods.
For TGA analysis, a piece of a half cell was heated to 650 ◦C with a 10 ◦C min−1
ramp rate and reduced. It was then cooled while still under reducing atmospheres.
Once below 100 ◦C, it was exposed to simulated air (21% O2) for 18 hours, placed back
44
Figure 3.7: Thermogravimetric analysis curves for Ni-GDC/GDC half-cells
showing mass loss over time at temperatures ranging from 550 ◦C to 750 ◦C in
3% H2 3% H2O balance N2, 50 sccm flow
45
into reducing atmosphere, and heated back up to 650 ◦C. Figure 3.8 shows the change
in mass overlaid onto the temperature and oxygen partial pressure experienced by
the sample. Following this treatment, the cell showed no mass gain during the oxygen
exposure and continued to reduce at the same rate as before once it was returned
to the initial conditions. This shows that cells which are cooled appropriately do
not re-oxidize and could be handled in between a batch reduction of cells, and their
assembly into a stack configuration.
Box plots of flexural strength and modulus of the in-situ reduced and batch
reduced coupons are shown in Figure 3.9. Mechanical strength of half-cell coupons
which had been reduced in-situ with an 18 hour reduction time showed no difference
in strength when compared with coupons which had been previously batch reduced,
cooled, and reheated under reducing environments. Both of these sample types
showed a dramatic decrease in strength and Young’s modulus compared to the un-
reduced samples, but no statistically significant difference between treatments. The
decrease in strength and modulus is due to the increase in porosity and conversion of
NiO to metallic nickel in the cells. The decrease in variation between measurements
in reduced cells is likely due to the increased number of large voids that form during
NiO reduction.
The statistical similarity of the properties of the samples reduced via each
method indicates that planar, Ni-GDC/GDC based SOFC ASL and half-cells are
able to be safely reduced, cooled, and handled in ambient conditions without leading
to damaging re-oxidation of the nickel anode material. This resilience could enable
some degree of large-batch reduction of cell anodes prior to SOFC stack assembly,
46
Figure 3.8: Thermogravimetric analysis of Ni-GDC/GDC half-cell showing no
mass gain after reduction and exposure to room temperature and simulated
air: a) Mass loss of interrupted reduction with exposure to ambient condition;
b) Temperature of sample during cycle c) Oxygen partial pressure measured at
the sample during cycle
47
Figure 3.9: Flexural properties, a) strength and b) modulus, of Ni-GDC/GDC
half-cell coupons after reduction via two different methods, compared to unre-
duced cells. Strength and modulus show a decrease upon reduction but no
significant difference between methods
48
leading to more rapid startup and a greater degree of sealing control.
3.3 Conclusions
A temperature and atmosphere controlled three-point bend fixture was designed
and built for use in a universal testing machine. SOFC coupons and component
materials were evaluated for flexural strength at room temperature and SOFC oper-
ating temperatures. In addition, the effects of porosity percent and pore geometry on
flexural strength in GDC were investigated. Furthermore, the impact of temperature
on the reduction rate of NiO in NiO-GDC SOFC anodes was examined along with
the resilience to re-oxidation at ambient conditions of this SOFC component.
Pore geometry had a significant impact on the flexural strength of GDC,
with spherical pores showing the greatest resistance to fracture. This supports the
concept of pores acting as the new crack tip once a crack has advanced to the pore.
Additionally, samples tested at 650 ◦C were stronger than those at room temperature.
This is likely due to localized compressive stresses from thermal expansion of the
material. This hypothesis is further supported by the results of testing NiO-GDC
anode support coupons and half-cells. Coupons tested at 25 ◦C, 450 ◦C, and 650 ◦C
displayed a linear strength dependence with temperature. There was no statistical
difference in strength between anode support layers and half-cells composed of anode
support and electrolyte at a given temperature in air. Half-cells in which NiO
was reduced to Ni by exposure to H2 at 650
◦C displayed significant differences in
strength when the electrolyte layer was subjected to compressive stress as opposed to
49
tensile stress. Placing the ceramic electrolyte in compression and the metal-ceramic
composite anode in tension resulted in the highest strength.
The reduction temperature of NiO-GDC/GDC half-cells was shown to have
an effect on the rate of NiO reduction and amount of NiO reduced. At lower
temperatures, the oxidation rate of Ni-GDC is slow enough that the anode can be
exposed to air for significant periods below 100 ◦C. Any re-oxidation, combined
with the cooling and re-heating of a cell back to 650 ◦C, showed no effect on the
mechanical properties when compared to cells which had been reduced in-situ at
650 ◦C. These results indicate that it is possible to reduce and cool cell components
to an extent without any additional effects to mechanical properties, allowing for
more flexibility during cell manufacturing, stack assembly, and with quality control
screenings.
This work leads to three important conclusions for the mechanical properties of
GDC-based SOFCs using Ni anodes. Porous GDC used in anode supported SOFCs
should be fabricated such that the pore geometry is spherical as this maximizes
energy required to advance a crack through the ceramic. Care should be taken
in stack construction to ensure any out-of-plane cells are placed to compress the
electrolyte and to place the reduced anode in tension so as to lower the chance of
fracture. Finally, it is possible to reduce the anodes of Ni-GDC SOFCs and then
handle them at ambient conditions for quality control and stack assembly. This will
remove a degree of variability from the manufacture of cells and stacks.
50
Chapter 4: Defect Chemistry and Oxygen Non-Stoichiometry of Dou-
ble Perovskite SrFe0.2Co0.4Mo0.4O3-δ
4.1 Introduction
SFCM is a new material that is of particular interest because of its high
conductivity and stability in hydrocarbon fuels. [51] SFCM takes advantage of the
perovskite structure and multivalent cations similar to SMM, but the combination of
both Fe and Co further increase the conductivity of the material. It has been shown
that an SOFC using SFCM as an anode support material is redox stable up to 30
cycles at 600 ◦C between H2/3%H2O and nitrogen. [52,53]
This work expands upon the fundamental knowledge of SFCM and aims to
understand the defect chemistry which leads to its high conductivity as an MIEC.
XRD and Rietveld refinement show the phase stability and lattice parameter changes
of SFCM across pO2s. Direct current (DC) conductivity is measured as a function
of pO2 to determine changes electrical conductivity types. TPD spectroscopy was
used to characterize the oxygen desorption as it occurs from the lattice. The non-
stoichiometry of SFCM was measured under oxidizing and reducing environments at
various temperatures via thermogravimetric analysis and a defect equilibrium model
51
and diagram is proposed from the data with results compared to SMM and other
perovskite materials.
4.2 Results and Discussion
4.2.1 Phase Purity and Chemical Expansion
The XRD patterns obtained after reduction in 1.9% H2, 2.9% H2O at 600
◦C
and oxidation in pure oxygen at 600 ◦C, in addition to the fresh as synthesized powder
and the pattern obtained from Rietveld refinement, are presented in Figure 4.1a.
A tetragonal, unordered B-site, double perovskite structure was used as a starting
point for Rietveld refinement based on the structures of Sr2CoMoO6, Sr2NiMoO6
and Sr2Fe0.75Co0.25MoO6. [48, 99] Results from the Rietveld refinement calculations
are given in Table 4.1. The refined unit cell is presented in Figure 4.1b and was used
to create the theoretical XRD pattern for comparison.
SFCM proved to be stable under reducing conditions, being pure phase based on
the Rietveld refinement. Overall phase purity remains at greater than 76% as synthe-
sized, 85% under oxidizing conditions and at 100% after reduction. Sr2Co1.2Mo0.8O6
was found in SFCM under ambient and oxidizing conditions. This secondary phase
reversibly disappears when returning to reducing conditions after oxidation.
Changes in SFCM’s lattice parameters are attributed to the oxidation states of
the B-site cations and the concentration of vacancies in the sample after exposure to
different pO2s. Reduction of SFCM from an as synthesized state causes a 0.025%
decrease in a and the volume of the unit cell and a 0.025% increase in c. These are
52
Figure 4.1: a) Powder XRD patterns of SFCM samples taken after synthesis,
reduction, and oxidation compared to pure phase SFCM diffraction pattern b)
Crystal structure of unordered double perovskite SFCM
53
Table 4.1: Refinement results of SFCM before and after exposure to reducing
and oxidizing environments
Condition Oxidized Fresh Reduced
Space Group I 4m I 4m I 4m
a (A) 5.5636(2) 5.5558(3) 5.5544(2)
c (A) 7.9310(2) 7.9184(3) 7.9204(2)
3
Volume (A ) 245.49 244.41 244.35
Sr-O1 (A) 2.794 2.790 2.790
Sr-O2 (A) 2.789 2.785 2.784
M1-O1 (A) 1.993 1.990 1.990
M1-O2 (A) 1.782 1.780 1.780
M2-O1 (A) 1.950 1.948 1.947
M2-O2 (A) 2.183 2.180 2.180
Phase Purity (% SFCM) 85.24 76.08 100.0
Rw (%) 14.277 15.519 17.324
54
a result of the bond lengths between strontium and oxygen in site 2 (Sr-O2) and
the distance between metal site 2 and oxygen site 1 (M2-O1) decreasing. Oxidizing
the SFCM results in a 0.14% increase in a and a 0.16% increase in c resulting an
0.44% increase in cell volume. This is the result of all bond lengths increasing by
0.002 to 0.004 A. These small lattice parameter shifts are in line with the general
nature of perovskites to display little to no change due to chemical expansion. [67]
Additionally, when changes do occur, perovskites have been shown to have anisotropic
chemical expansion, as such with Sr2Fe1.5Mo0.5O6 (SFM) where a increases upon
reduction while c decreases. [100] In summary, SFCM shows very little change to
lattice parameters as a result of changing metal oxidation states or vacancy creation
as it under goes reduction or oxidation.
4.2.2 Conductivity
SFCM has a conductivity dependence typical of MIEC conductors, as shown
in Figure 4.2. At high pO2 ranges (10
−1.5 atm to 1 atm) it shows p-type conductivity,
where conductivity increases with pO2 . Down to 10
−1 atm, the change in conductivity
appears to be linear with a slope of 0.089 (log-log scale). Below 10−1 atm the slope
changes away from the previous trend, suggesting a change in the predominant
charge carrier or phase of the system. At low pO2 ranges (10
−24 atm to 10−19 atm)
the conductivity behaves linearly with n-type conductivity, increasing at lower pO2 ,
with a slope of −0.11 (log-log scale). The conductivity measured in this work is
comparable to the 30 S/cm at 600 ◦C as reported by Pan et. al. [51]
55
Figure 4.2: Total conductivity as pO2 changes under a) reducing and b)oxidizing
conditions at 600 ◦C
56
Based on the slopes in the high pO2 region, at least two defect regimes exist in
that area transitioning near 10−1 atm. Both high and low pO2 regions have slopes
much lower than +1/4 or -1/4 respectively, which indicates that the relationship
between electronic charge carriers and oxygen is dependent of the other defect species
in the material. In fact, the lower than 1/4 slope suggest that the electronic charge
carries are taken up by the other species, decreasing the expected number based
on pO2 change. At low pO2 , n-type conductivity is promoted by the generation of
oxygen vacancies, freeing electrons as negative electronic charge carriers to maintain
charge balance. Some of these electrons are consumed by the reduction of B-site
metals to lower oxidation states. At high pO2 , oxygen vacancies are filled, consuming
electrons and allowing for the creation of holes, promoting p-type conductivity. The
different slopes between high and low pO2 are the result of different defect species
interacting with the electronic carriers at different points in the reduction.
4.2.3 Temperature Programmed Desorption
TPD was performed in conjunction with TGA to determine the extent of
oxygen desorption from SFCM. Figure 4.3 presents the data from the TGA on top
with the mass and rate of mass change, while the 32 m/z signal from oxygen TPD
in the mass spectrometer is on bottom. The cyclic noise present in both mass and
MS signals is due to transients of the PID controller on the TGA furnace, but the
noise remained much lower than the measured signal. The rate of mass loss matches
the oxygen signal from the MS, confirming that oxygen is generated from SFCM
57
when heated under N2 and is the cause for mass loss in the sample. SFCM shows
two maxima for the rate of oxygen loss during heating. The first, α, occurs at 405 ◦C
with the second peak, β, occurring near 800 ◦C. The MS monitored for other species,
such as carbon (CO, CO2) and water, but no significant amounts were observed.
Perovskite materials have oxygen desorption classified based on the temperature
at which it occurs. Low temperature desorption (referred to as α), occurs due to the
reduction of metals and desorption of oxygen near the surface. High temperature
desorption (referred to as β), takes place when oxygen is able to diffuse through the
bulk and be released. [101] SFCM shows both α and β desorption, indicating that
both the surface and lattice can generate oxygen vacancies due to the reduction of
the different B-site cations and that those vacancies are mobile to move from the
bulk to the surface. SMM shows only β desorption above 800 ◦C, while SFM has
α desorption from 400 ◦C to 800 ◦C and Sr2CoMoO6 (SCM) desorbs starting below
400 ◦C. [102,103] In the case of SFM and SCM, the use of the multivalent Fe or Co
allows for reduction at a lower temperature than SMM. For SFCM, the same effect
is seen through the combined use of Fe and Co. This demonstrates SFCM’s ability
to begin to create oxygen vacancies at 400 ◦C and exchange oxygen with the bulk
lattice starting as low as 450 ◦C and peaking near 800 ◦C.
4.2.4 Oxygen Non-Stoichiometry
Oxygen non-stoichiometry and oxygen vacancy concentrations are shown in
Figure 4.4 at 600 ◦C for high and low pO2 regions. At high pO2 , SFCM has two linear
58
Figure 4.3: Temperature programmed desorption of oxygen in SFCM with
mass loss from TGA(top) and oxygen desorption from MS (bottom) as it is
heated to 800 ◦C in N2
59
regions, changing between them near 10−0.75 atm. The transition between regions in
oxygen stoichiometry occurs at a similar pO2 to where the conductivity changes in
Figure 4.2. This change is the result of a change in defect regime regions. In the low
pO2 region, a linear trend occurs down to 10
−21 atm where the non-stoichiometry
rate increases. Due to the pure phase XRD under reducing conditions, it is not likely
that the increase in non-stoichiometry is due to phase decomposition and instead is
a change in regimes where more oxygen vacancies are present.
SFCM shows similar trends in non-stoichiometry to SMM and SFM, but with
the added complexity expected from the additional B-site species. SMM approaches
a plateau in oxygen content as it reached 1 atm of oxygen the same as SFCM but
possesses a constant slope of -1/6 when plotted on a log-log scale. [104] In comparison,
SFCM has multiple slopes associated with the degree of vacancy formation, as shown
in Figure 4.4b. These different slopes can be attributed to the reduction of the three
different B-site cations at different pO2 . As an oxygen vacancy forms, the electrons
required to maintain charge balance can be accommodated by the reduction of Fe,
Co, or Mo, each with an independent equilibrium constant, whereas SMM only has
Mo reduction to accommodate electrons from oxygen vacancy generation. The high
slopes of Figure 4.4b show how SFCM can generate more oxygen vacancies at lower
pO2 because of the Fe and Co cations. SFM and SMM show similar amounts of
non-stoichiometry at the same pO2 but at 1000
◦C. [105] SFCM shows almost twice
the non-stoichiometry at a much lower temperature of 600 ◦C. SFCM’s ability to
create vacancies at higher pO2 results in a larger accumulated vacancy concentration
at low pO2 .
60
Figure 4.4: Non-stoichiometry (circles, left axis) and corresponding oxygen
vacancy concentration (diamonds, right axis) of SFCM under a) reducing
conditions and b) oxidizing conditions at 600 ◦C. Error bars are smaller than
points for non-stoichiometry and for vacancy concentration below 10−0.5 atm.
61
TGA non-stoichiometry measurements were performed on the same sample at
400 ◦C, 500 ◦C, and 600 ◦C which are temperatures in the low temperature-SOFC
range. Figure 4.5 shows the non-stoichiometry measurements for the sample tested
at three temperatures in the low pO2 (a) and high pO2 (b) regions. As expected,
lowering the temperature reduces the non-stoichiometry at a given pO2 and decreases
the pO2 at which transitions between regimes occur. At 400
◦C oxygen vacancy
formation decreases considerably compared to 500 ◦C and 600 ◦C This aligns with
the results of the TPD (Figure 4.3) where SFCM does not start desorbing oxygen
until 350 ◦C to 400 ◦C.
4.2.5 Defect Equilibrium Model and Diagram
To best understand what is occurring in SFCM a defect equilibrium diagram and
model need to be created to relate the changes in conductivity and non-stoichiometry
to the concentration of the various defects present in the material. Defect reactions
with their corresponding equilibrium equations are given in Equations 4.1–4.4, which
use Kröger–Vink notation, where a species is denoted by the site it sits on (subscript)
and the relative charge to the site’s expected valency (superscript). For example,
VO is a vacancy in an oxygen site with a 2+ charge and B
′
B is a B-site metal in a
B-site with a 1− charge. K represents the equilibrium, or mass-action, constant for
the reaction. To simplify the set of equations, Equation 4.4 represents the combined
reduction of any B-site cation (Fe, Co, or Mo). β is the ratio of A-site vacancies to
62
Figure 4.5: Non-stoichiometry of SFCM as pO2 changes at 400
◦C, 500 ◦C, and
600 ◦C. Error bars are smaller than data points.
63
B-site vacancies from Schottky defects.
∅ e′ + h Ki = np (4.1)
∅ 3 V + V r′′ + V′′′′O S B Ks = [VO ]3[V ′′Sr ][V′′′′] = [V ]3[V′′′′]β+1B O B (4.2)
O×
1 1/2
O O2 + VO + 2 e
′ K = p [V ][O×]−1 2r O O O n (4.3)2 2
B×B + e
′ B′ ′ −1B KB = [BB]n (4.4)
The Duncan-Wachsman approach allows for the solution of defect equations
across pO2 regions using dominant defect triads, compared to the dominant defect
pairs of the Brouwer approach which lead to an inaccurate results near transition
regions. [106] This approach is needed in this work to model the defect equilibria
because the measured pO2 range crosses various Brouwer regions, as indicated by the
non-linearity of Figure 4.4. Using this method, the defect equations can be solved to
give models of the concentration across pO2 . The equations used are taken directly
from Duncan et. al paper. To apply this approach to SFCM, the interactions of
electronic defects with B-site cations as given by Equation 4.4 were ignored for the
time being. It was assumed that all electrons generated from vacancy formation
remained in the conduction band. It was also assumed that a portion of B-site
cations were reduced at all conditions, acting as a constant dopant to the structure,
where regardless of pO2 . The degree of constant B-site reduction could then be fit
from the model.
Using the TGA and conductivity data, the reaction equilibrium constants can
be fitted for the defect model. The results of the fit are shown in Figure 4.6. Starting
with the oxygen vacancies least squares fitting approach, the parameters are fitted in
64
the low pO2 region first, adding terms and parameters from the Duncan-Wachsman
method until the entire equation had been fit. A similar process was followed to fit
the conductivity data to the model. Separate equations were used to fit n-type and
p-type conductivities, based on their respective predominant defects.
The defect equilibria values were fitted and calculated from thermogravimetry
and found to be to be 1.1× 10−15 atm0.5/F.U.2 for Kr, 5.43× 10−7 /F.U.2 for Ki,
and 1× 10−25 /F.U.5 for Ks, where F.U. is the formula unit of SFCM. The minimum
concentration of B′B defects (M
′
B) was determined to be 0.108 /F.U. and β was 100.
A deviation from the model is observed at oxidizing conditions and can be attributed
to the phase impurity which forms at those conditions. The parameters relating to
the Schottky defects (β and Ks) did not have a significant impact on the fit of either
the oxygen vacancies conductivity, and thus are not considered to be well fit. This
is because Schottky defects are not expected in high concentration except at high
temperatures and very high pO2 , neither of which were tested in this work. The defect
mobilities were calculated from the fitted data at 600 ◦C to be 1.432 cm2 V−1 s−1
for electrons 0.4887 cm V−1 s−1 for oxygen vacancies and 0.0806 cm V−1 s−1 for holes.
A deviation in slope is again visible from the model for the conductivity data at
oxidizing conditions, which can be attributed to the impurity phase. This difference
in slope adds uncertainty for the value of the hole mobility. The electron mobility
is significantly greater than the value reported for small polaron defect mobility in
SMM. [104]
The complete defect equilibrium diagram for electronic defects and vacancies
based on the fitted values is shown in Figure 4.7a. The model predicts an electrolytic
65
Figure 4.6: a) Fitting results of oxygen vacancies to the measured non-
stoichiometry with the parameters K = 1.1× 10−15r atm0.5/F.U.2, Ki =
5.43× 10−7 /F.U.2, Ks = 1× 10−25 /F.U.5, [B′B] = 0.108 /F.U., β = 100. b)
Fitting results of n-type and p-type conductivity to measurements based on
previous parameters, µe = 1.43 cm V
−1 s−1, µV = 0.4887 cm V
−1 s−1, and
O
µh = 0.0806 cm V
−1 s−1.
66
domain in the mid-pO2 region. In comparison to the traditional Brouwer regions, the
measured pO2 only cover the boundary between Region I and IIa up to the lower
portion of Region IIb.
To determine the enthalpy of oxygen formation, the natural log of the equlib-
rium constant for the external reaction (Equation 4.3) was plotted against the
reciprocal temperature. The enthalpy was found to be 44.3 kJ mol−1 in SFCM under
reducing conditions. SFCM has an enthalpy of VO formation lower than that of
comparable materials. SFM and SMM have much higher enthalpies of 253 kJ mol−1
and 110 kJ mol−1 respectively. [104, 105] SrFeO3− δ has a more comparable, but still
greater, enthalpy of 80 kJ mol−1. [107] This low enthalpy of formation of oxygen
vacancies in SFCM allows for a greater range of non-stoichiometry and thus SFCM’s
higher performance.
4.3 Conclusions
SFCM has potential as an electrode material in a number of electrochemical
device applications. Specifically, it has a potential to replace nickel-based anodes in
SOFCs due to its high conductivity and redox stability, creating an all-ceramic anode.
SFCM forms impurity phases when heated at 600 ◦C under oxidizing environments
but becomes pure phase with reduction. It also exhibits very small chemical expansion
with changing pO2 , in line with its perovskite structure. SFCM is an improvement
over previous double perovskite materials because conductivities above 40 S/cm
in reducing conditions below 600 ◦C. This is in part because SFCM supports the
67
Figure 4.7: a)Defect equilibrium diagram of SFCM based on fitted parameters.
b)Plot of natural log of external defect equilibrium versus inverse temperature,
to calculate enthalpy of formation oxygen vacancies.
68
desorption of oxygen from both surface and lattice positions shown by overlapping α
and β oxygen desorption, starting as low at 350 ◦C and continuing up until above
800 ◦C. Thermogravimetry at lower temperatures demonstrated SFCM’s activity
down to 400 ◦C at which the amount of oxygen vacancy formation slows dramatically.
Oxygen vacancies form more rapidly than other MIEC materials with decreasing
pO2 as combination of B-site cations enables reduction though a large range of pO2
resulting in a low enthalpy of formation for oxygen vacancies at 44.3 kJ mol−1. Defect
equilibrium and mobility values were calculated from the themogravimetric and
conductivity measurements and a defect equilibrium diagram for SFCM at 600 ◦C
was created.
69
Chapter 5: Flexural Strength and Flaw Distributions of
SrFe0.2Co0.4Mo0.4O3-δ Based Ceramic-Supports for
Solid-Oxide Fuel Cells at Operating Conditions
5.1 Introduction
This work characterizes the changes which occur in fracture toughness and
flaw distributions of SFCM, SFCM-GDC ASL, and SFCM-GDC/GDC half-cells
when exposed to oxidizing and reducing conditions up to 600 ◦C by use of flexural
testing and Weibull analysis. Additionally, electrical conductivity, thermogravimetric
analysis, and thermal expansion are used to determine the causes for changes in
mechanical properties under these conditions.
5.2 Results and Discussion
5.2.1 Redox Cycling Stability
SFCM has a high initial conductivity of 35 S/cm but decreases after 3 cycles
to 19 S/cm and again at 9 cycles to 7.9 S/cm, shown in Figure 5.1a. The addition of
GDC to create a SFCM-GDC composite decreases the initial conductivity to 15 S/cm
70
but increases stability. At 4 cycles, SFCM-GDC decreases conductivity and stabilizes
to 8 S/cm for 19 cycles. The addition of the GDC improved the redox cycling stability
by providing a very stable, contiguous framework for the SFCM. [13,68,108]
Figure 5.1b shows the mass loss of SFCM-GDC as it is reduced in 3% humidified
H2. For comparison, another popular anode material, nickel oxide GDC, is plotted
along with it. The reduction of SFCM-GDC results in a less than 1% reduction of
mass from the formation of oxygen vacancies in the SFCM and GDC, and it reaches
steady state in under two hours. NiO-GDC on the other hand loses a large amount
of mass from the reduction and phase change of NiO to Ni metal taking over 30
hours to reach steady state.
Further cycling of SFCM in the TGA was performed and results are given
in Figure 5.2 with 5.2a showing the mass loss of SFCM with time as it is cycled
between air and hydrogen and 5.2b being a summary of the total mass changes
that occur with each cycle. With each cycle, less than 1% total mass change is
observed. Reduction occurs over a time period of five hours while oxidation happens
at a faster rate, approximately one hour. Fitting the reduction and oxidation to
exponential functions gives a biexponential with decay rates of 0.438 and 27.89 h−1
for reduction and an oxidation rate of 2.90 h−1. Previous experiments have shown
that SFCM generates an impurity phase of Sr2Co1.2Mo0.8O6 under ambient and
oxidizing conditions. [109] This impurity phase does not contribute to the gain or
loss of oxygen as seen in the TGA. Sr2Co1.2Mo0.8O6 has a cubic structure compared
to SFCM’s tetragonal structure, leading to additional stress and strain created by
its generation impacting the mechanical and electrical properties between oxidizing
71
Figure 5.1: a) DC conductivity of SFCM and SFCM-GDC (2:1) after cycling
at 650 ◦C between 10% H2 in N2 and air. b) Mass loss of SFCM-GDC and
NiO-GDC (3:2) during reduction in 3% H ◦2/3% H2O in N2 at 650 C
72
and reducing environments.
5.2.2 Fracture Toughness
√
Fracture toughness was found to be (0.124± 0.023) MPa m at room tempera-
ture, increasing with temperature up to 600 ◦C, as shown in Figure 5.3a. From room
√
temperature to 500 ◦C, the rate of increase in KIc is low at 4.3× 10−5 MPa m/◦C.
From 500 ◦C to 600 ◦C the fracture toughness increases by 90%. The increase in
fracture toughness is most likely a result of the thermal expansion of the material.
During process of creating the bar, the sample was cooled from 1340 ◦C to room
temperature. This cooling and associated contraction induces stresses in the bar
sample, which combine with the stresses added during testing to fracture the sample.
As the sample is heated, the residual stresses are relaxed, increasing the amount of
stress which must be applied externally to cause the bar to fracture. Any change
in fracture toughness due to the weakening of bonds is masked by this effect. The
increase in fracture toughness from 500 ◦C to 600 ◦C matches the temperature where
the rate of thermal expansion also increases, as shown in Figure 5.3b.
Figure 5.3b also presents the thermal expansion of GDC and a SFCM-GDC
composite. At temperatures below 550 ◦C the expansion of SFCM-GDC matches that
of pure SFCM, greater than that of pure GDC. Above 550 ◦C, the rate of expansion
slows in comparison to SFCM and matches the rate of expansion for GDC. Thus,
below 550 ◦C, SFCM-GDC expands the same as SFCM and above 550 ◦C it expands
the same as GDC. We would then expect that SFCM-GDC composites have the same
73
Figure 5.2: Mass changes of SFCM during cycling between 21% O2 and 3%
H2/3% H2O in N2 at 600
◦C. Each exposure was 5 hours. a) Percent mass
change with time as the gas is switched back and forth. b) Summary of
maximum and minimum changes with each cycle
74
Table 5.1: Summary of Cycled SFCM-GDC (2:1) ASL Mechanical Properties
Mean Modulus Mean Strength
Number
Number Modulus Std Dev Strength Std Dev
of Cycles
(GPa) (GPa) (MPa) (MPa)
0 14 24.5 5.01 34.3 7.23
10 20 27.3 3.60 22.4 4.94
increase in fracture toughness as SFCM up to 550 ◦C due expansion and relaxation
of intrinsic stresses, but not further increase the rate above 500 ◦C, as SFCM does.
5.2.3 Strength of Anode Support Layer After Redox Cycling
To understand how redox cycling as the result of long term use affects the
strength of SFCM based SOFCs, SFCM-GDC ASL test coupons were cycled for 10
times between nitrogen and 10% H2 environments. Figure 5.4 shows the strength
and modulus from the samples which were not cycled (0 cycles) and those cycled 10
times, tested using a four-point bend at ambient conditions. Table 5.1 summaries
the means and standard deviations of the test results. The overlapping Student’s
t circles in Figure 5.4b demonstrate that cycling did not significantly change the
elastic modulus of the samples, while the strength did significantly decrease. This is
the result of changes in microstructural flaws, rather than the change of an intrinsic
material property.
The strengths of the cycled and un-cycled cells were fitted to Weibull distri-
75
Figure 5.3: a) Fracture toughness of SFCM from 20 ◦C to 600 ◦C. b) Thermal
expansion of SFCM, GDC, and SFCM-GDC (2:1)
76
Figure 5.4: Results of four point bend test at ambient conditions of SFCM-GDC
(2:1) ASL with a) strength and b) modulus for samples before any cycling and
after 10 redox cycles between 10% H2 and N2
77
Table 5.2: Summary of Weibull Fit Parameters (characteristic strength and
Weibull modulus) for Cycled SFCM-GDC (2:1) ASL
Char. Weibull
Number α α β β
Strength Modulus
of Cycles CI Lower CI Upper CI Lower CI Upper
(α, MPa) (β)
0 37.1 33.7 40.8 5.75 4.06 9.85
10 24.3 22.3 26.5 5.39 4.00 8.28
butions to observe how flaw distributions changed with redox cycling. Figure 5.5
presents the fitted data and distributions and Table 5.2 summarizes the fit parameters
of the Weibull modulus with 95% confidence intervals. The Weibull moduli between
the two data sets are very similar, indicating that the flaw distribution is the same
between them. The decrease in characteristic strength is then due to the growth
of flaws with cycling. SEM images in Figure 5.6 are of a sample which had not
been cycled (a) and one which had been cycled 10 times (b) after being epoxy filled
and polished. The sample which had been cycled 10 times has a greater number of
pores at larger sizes compared to the un-cycled sample. This indicates that the finer
microstructure of the pores is susceptible to changes during redox cycling due to
chemical expansion, but the larger flaws which lead to complete cell failure do not
change as the result of redox cycling. [110]
78
Figure 5.5: Fitted Weibull distributions of SFCM-GDC (2:1) ASL before any
cycling and after 10 redox cycles between 10% H2 and N2, plotted linearly with
95% confidence intervals shaded
79
Figure 5.6: Scanning electron microscope images of epoxy filled and polished
SFCM-GDC (2:1) ASL cross sections tested a) before any cycling, b) after 10
cycles between 10% H2 and N2
80
5.2.4 Half-cell Strength of Anode-Supported SOFCs at Operating
Conditions
Figure 5.7 gives box plots of the results of three point flexural testing of SFCM-
GDC/GDC half-cells, and Table 5.3 summaries the strengths. The circles on the
right of 5.7 represent the results of a Student’s t-test, to determine if the means of the
different data sets are the same. The more the circles overlap the more similar the
data sets are. In this case, there is no overlap between circles at the 95% confidence
interval, highlighting that the strengths tested at each condition are unique from
each other.
At ambient conditions the strength of the half-cells was measured to be
57.7 MPa. Upon heating, the strength increases to 74.2 MPa. This result corre-
lates with the increase in fracture toughness discussed earlier, which is expected
based on Equation 1.6. Another contributing factor to the increase in strength is the
effect thermal expansion has on pre-existing cracks and flaws. Any expansion of the
material helps push closed a surface crack which would lead to failure, decreasing its
size and increasing the stress that would be required to cause that crack to open
and propagate.
Upon reduction of the half-cell at 600 ◦C, the fracture strength decreases,
below that of the ambient strength, to 39.9 MPa. During this transition to reducing
conditions, the impurity phase reduces back into SFCM, causing a decrease in
mechanical properties from the change in lattice parameters. Additionally, bulk
mechanical properties have changed as SFCM becomes pure phase, oxygen vacancies
81
Table 5.3: Summary of strengths for SFCM-GDC (2:1)/GDC half-cells at
different environments
Temperature Mean Std Dev
Atmosphere Number
(◦C) (MPa) (MPa)
20 Air 17 57.7 6.71
600 Air 5 74.2 2.98
600 5% H2 14 39.9 5.89
are generated and B-site transition metals are reduced.
The fracture surfaces of samples tested at 20 ◦C and 600 ◦C in air are shown in
Figure 5.8. The low temperature fracture (5.8a) shows intergranular fracture though
the majority of the ASL. Near the interface, the electrolyte posses transgranular
fracture, which transitions back to intergranular fracture away from the interface. At
higher temperature (5.8b) the fracture preferred the grain boundary more resulting
in only intergranular fracture except for a few isolated spots in the electrolyte at the
interface.
Weibull analysis was performed to compare the distribution of flaws between
the ambient tested half-cells and the in-situ reduced half-cells. Figure 5.9 displays
the fitting of Weibull distributions to the flexural strength of half-cell coupons tested
at ambient conditions and during reduction at 600 ◦C. Table 5.4 summarizes the
fitting parameter for the Weibull modulus with 95% confidence intervals using the
maximum-likelihood method. Between the two conditions, there is an appreciable
82
Figure 5.7: Flexural strengths of half-cell coupons made from SFCM-GDC
(2:1) anode with GDC electrolyte placed in compression, tested at 20 ◦C in air,
600 ◦C in air, and 600 ◦C in 5% H2
83
Figure 5.8: Scanning electron microscope images fracture surfaces of SFCM-
GDC (2:1)/GDC half-cell cross sections tested at a) 20 ◦C, b) 600 ◦C in air
84
change in the Weibull modulus of the two samples. The Weibull modulus decreases
from 10.4 at ambient conditions to 7.29 when at reducing conditions. This change in
Weibull modulus is the result of the reduction of SFCM and GDC, as the materials
and any impurities phases change, the bulk properties of the materials change,
resulting in a wider distribution of flaws with more variability.
To confirm any changes in flaw distribution, fractography would need to be
performed on every sample, identifying the flaw, but this is difficult due to the porous
nature of the fracture surface which obscures the fracture origin. Scanning electron
microscopy was performed on fracture surfaces and on epoxy filled samples from each
condition to look for any changes in crack propagation or microstructure. Figure 5.10
shows the epoxy filled and polished samples with no discernible difference between
samples. No difference is observed because the likely cause of failure, microcracks,
would exist along grain boundaries and are obscured by the grain boundaries and
pores. [24, 111]
5.3 Conclusions
In this work the mechanical properties of SFCM and SFCM-GDC structures are
characterized as the environment is changed from ambient to the working conditions
of SOFCs and with redox cycling. The electrical conductivity of SFCM-GDC is
stable up to 19 cycles due to the reversibility and phase stability of SFCM-GDC
in the environment range. The fracture toughness of SFCM was found to be
√
(0.124± 0.023) MPa m at room temperature and increases with temperature, due
85
Figure 5.9: Fitted Weibull distributions of SFCM-GDC (2:1)/GDC half-cells at
different environments, plotted linearly with 95% confidence intervals shaded
Table 5.4: Summary of Weibull fitting parameters (characteristic strength and
Weibull modulus) for SFCM-GDC (2:1)/GDC half-cells at different environ-
ments
Cha. α α Weibull β β
Temp.
Environment Strength Lower Upper Modulus Lower Upper
(◦C)
(α, MPa) CI CI (β) CI CI
20 Air 60.6 57.4 63.6 10.4 6.89 14.7
600 Hydrogen 42.5 39.2 45.8 7.29 4.79 10.3
86
Figure 5.10: Scanning electron microscope images of epoxy filled and polished
SFCM-GDC (2:1)/GDC half-cell cross sections tested at a) 20 ◦C in air, b)
600 ◦C in air and c)600 ◦C in 5% H2
87
to relaxation of residual stresses. The flexural strength of SFCM-GDC/GDC half-
cells increases by 23.9% upon heating from room temperature to 600 ◦C. This is due
to thermal expansion pushing cracks closed during heating, requiring additional stress
to propagate surface cracks and the increase in fracture toughness. Once exposed to
reducing conditions, the flexural strength decreases by 29.4% of the original strength,
as impurities reduce and SFCM and GDC generate oxygen vacancies. Reduction
was shown to increase the size and variability of the distribution of flaws which lead
to failure. Redox cycling led to the uniform increase in critical flaw size and changed
the fine microstructural porosity, decreasing strength from 34.3 MPa to 22.4 MPa.
These findings show that SFCM, when combined with GDC, make a suitable
anode material for SOFCs. SFCM-GDC is a stable material system, both in terms
of conductivity and mechanically, with redox cycling. Reduction does increase the
likelihood of failure by decreasing characteristic strength and the distribution of flaws,
but repeated cycling only decreases characteristic strength and does not change the
Weibull modulus.
88
Chapter 6: Conclusions
6.1 Summary of Contributions
SOFCs are an interesting technology with the promise of helping meet growing
energy demands by improving fuel conversion efficiency. The mass deployment and
adoption of SOFCs are limited by their high operating temperature and reliability
concerns. New materials have been developed to help lower the operating temperature
of SOFCs to less than 600 ◦C, but an understanding of how these materials affect
reliability needed to be developed. This work has done that, investigating factors
which impact mechanical strength and thus the structural integrity of the cell.
To properly investigate the impacting factors on mechanical properties, a test
apparatus was built which was capable of testing materials at conditions which are
common in an SOFC. Previously, only the temperature or pO2 effects on strength
or modulus could be measured. This new apparatus not only can test the flexural
properties of materials at temperatures in excess of 600 ◦C, but it also allows for the
control of the gaseous environment. The interdependent effects of temperature and
pO2 are now able to be evaluated using this setup.
The use of GDC has enabled the lowering of SOFC temperatures, with Ni-
GDC commonly performing as the anode. As a result of this trend, the mechanical
89
properties of GDC, Ni-GDC, and Ni-GDC/GDC half-cells were investigated. It was
demonstrated how the amount and choice of pore-former can impact the mechanical
strength due to total porosity percent and shape factor of the pores providing crack
tip blunting. Coupons of ASL Ni-GDC and half-cell Ni-GDC/GDC linearly increased
strength upon heating with no dependence on half-cell orientation for strength. Upon
reduction, all cells showed significant decrease in strength as NiO reduced to Ni,
increasing porosity and as GDC reduced. After reduction, the highest strength
was shown by samples which had the electrolyte in compression. Finally, it was
demonstrated that reduced Ni-GDC/GDC half-cells could exposed to air for extended
periods of time when below 100 ◦C without oxidation of the Ni or significant impact
to the flexural strength upon reheating under reducing atmospheres.
To develop a reliable, redox stable system, the phase purity and oxygen non-
stoichiometry of a new ceramic anode material, SFCM, was measured under oxidizing
and reducing conditions at temperatures in the low temperature-SOFC range. Under
reducing conditions SFCM is pure phase, but forms Sr2Co1.2Mo0.8O6 as an impurity
under oxidizing conditions. SFCM itself has very little lattice parameter change at
different pO2 but the impurity Sr2Co1.2Mo0.8O6 has a cubic structure with a very
different lattice parameter. Oxygen desorbs from SFCM starting at 350 ◦C and
continues in excess of 800 ◦C, enabling it to serve as an MIEC below 600 ◦C. SFCM
posses a low enthalpy of formation of oxygen vacancies (44.3 kJ mol−1), allowing a
large number of oxygen vacancies to form, up to a δ = 0.176 at 600 ◦C in 97% H2.
The Duncan-Wachsman method was then applied to create a defect equilibrium
diagram to match the observed behavior.
90
This information on the phase and point defect structure of SFCM was used
in understanding the mechanical properties of SFCM and SFCM-GDC composite
√
half-cells. SFCM fracture toughness was measured to be (0.124± 0.023) MPa m
√
at room temperature, increasing to (0.286± 0.038) MPa m at 600 ◦C, due to the
relaxation of residual stresses from sintering. SFCM-GDC/GDC half-cells increased
strength by 23.9% after heating from room temperature to 600 ◦C due to the increase
in fracture toughness combined with thermal expansion applying compressive stresses
to pre-existing flaws. After reduction, the half-cells showed a decrease in strength as
impurities decreased and oxygen vacancies were generated in SFCM and GDC. Redox
cycling was found to have no impact on the flaw distribution of SFCM-GDC/GDC
half-cells, but the cells did decrease the overall characteristic strength due to the
growth of microcracks from the changing amount of SFCM impurities.
Combined, this work evaluates the impact of both temperature and pO2 on
the flexural strength of a current, state of the art electrolyte and anode and a new
potential anode material, based on microstructure and point defects. These results
can feed back into the design and fabrication of SOFC cells and stacks to improve
reliability. With the knowledge that GDC based cells only decrease strength upon
reduction, not heating, pre-reduction steps or a gentler reduction process may be
developed which reduce the likelihood of failure. Additionally, this work shows how,
while an improvement compared to Ni-GDC, SFCM cells do weaken during redox
cycling due to impurity formation.
91
6.2 Future Work
• Develop models using finite element analysis incorporate results into a model
for full-sized cells
• Test mitigation strategies to reduce crack propagation such as crack tip defection
or crack pinning
• Optimize ASL thickness to maximize strength and minimize defects
• Measure η values for other pore formers and materials systems
• Study cause and effects of temperature dependance on extent of NiO/Ni
reduction
• Perform neutron diffraction on SFCM to determine B-site ordering
• Test other compositions of SFCM to create pure phase material across pO2
• Perform X-ray photoelectron spectroscopy on SFCM samples quenched from
various pO2 ’s to measure changes in B-site oxidation states
• Determine a means to measure properties of SFCM in intermediate and very
high pO2 ’s
• Measure effect of thermal cycling on flaw distributions in SFCM-GDC/GDC
half-cells
• Measure fracture toughness of GDC and SFCM/GDC mixtures at various
environmental conditions
92
• Measure thermal and chemical expansions of different SFCM/GDC mixtures
6.3 Publications and Presentations
Publications
• Stanley P, Hays TH, Langdo T, Wachsman ED. Mechanical Properties of
SOFC Anode Support Materials at Operating Conditions. ECS Trans 2017
May 30;78(1):2285–91.
• Stanley P, Hays T, Langdo T, Gore C, Eric D. High Temperature Mechanical
Behavior of Porous Ceria and Ceria-Based SOFCs. J Am Ceram Soc. (accepted)
• Stanley P, Hussain AM, Huang Y, Gritton JE, Neill JAO, Wachsman ED.
Defect chemistry and oxygen non-stoichiometry in SrFe0.2Co0.4Mo0.4O3-δ for
solid-oxide fuel cells. (in progress)
• Stanley P, Hussain AM, Hays T, Robinson I. Flexural strength and flaw distri-
butions of SrFe0.2Co0.4Mo0.4O3-δ solid-oxide fuel cells at operating conditions.
(in progress)
Presentations
• Stanley P, Hays TH, Langdo T, Wachsman ED. Mechanical Properties of SOFC
Anode Support Materials at Operating Conditions. SOFC-XV, Hollywood, FL
July 2017
93
• Stanley P, Hays TH, Wachsman ED. Mechanical Characterization of SOFC
Anode Support Materials at Operating Conditions. ECS Fall Meeting. National
Harbor, MD Oct 2017
94
Appendix A: Statistics
A.1 t-Test
The Student’s t-test, or particularly the two sample t-test, is a statistical test
commonly used to determine if two sets of data are significantly different from each
other. A standard t-test creates a confidence interval around a mean for in which the
true mean is thought to exist. By comparing the differences of two means and using
the null hypothesis test to determine if the difference is zero, it can be determined if
the two means are distinct, with a particular confidence. In the case of randomly
dividing samples to receive separate treatments, this constitutes an unpaired test, as
compared to a before-after sample set in which measurements are paired. The t-test
thus gives a way to evaluating two sets of data, say flexural strength of reduced and
unreduced samples, and to calculate if the reduction had any significant effect.
Like all statistical tests, this relies on a few base assumptions about the
population. First, it is assumed that the population has a normal distribution, which
reasonably fits for the samples which were studied. This assumption can be tested
by other statistical tests, such as the Shapiro–Wilk test or Pearson’s chi-squared
test, or can be confirmed visually by comparing a histogram of the data to a normal
distribution plot. The Student’s t-test is not overly sensitive to the fit a normal
95
distribution, but cases such as bimodal distributions will give inaccurate results. [112]
Second, for the t-test as described by Student, the two populations should have the
same variance, meaning that the spread of the data compared to the mean should
be equivalent, as determined by the F test. In the event where the variances are
not equal, a variation of the t-test may be used known as the t-test for unequal
variances or the Welch’s t-test. [113] Finally, the two populations should be sampled
independently from each other. The t-test works reasonably well with small sample
sets, allowing analysis of set with 5 or more samples each. [114]
The t-test begins by calculating the t-score or t-value for a data set. The
broadest and most applicable means of doing this is given by the Welch’s t-test,
Equation A.1, which is for sample sets with unequal variances and unequal sample
sizes. X is the mean of the sample set, s2 is the variance, and n is the number of
samples in the sample set. Because the true variance of the population is not known,
the variance of the sample set is used which is the square of the standard deviation.
√X1 −X2t = (A.1)
s21 s
2
+ 2
n1 n2
Once the t-score is calculated it is compared to the critical t-score based on
the confidence level desired and degrees of freedom in the measurements. Typically,
a confidence level of 95% is used, but 99% is also frequently found. This can also be
referred to the level of significance, which is 1 minus the confidence level, so 0.05 or
0.01. In the simple case of a single set of data, the degrees of freedom is n− 1. This
is because for a sample set to have a given mean, all but one of the values could be
any number (and thus are free), but the last value, must have a value which causes
96
the mean to be what is expected. [115] For two sample t-tests where the sets have
the same number of samples the degrees of freedom become 2n − 2. But, for the
case where the variances are unequal and measurement number may be unequal like
Equation A.1, the pooled degrees of freedom must be calculated from Equation A.2,
known as the Welch–Satterthwaite equation. The degrees of freedom calculated by
Equation A.2 is then rounded down to (the next)integer. [116]
s2 s2
2
1 + 2
dof = ( n)1 n(2 ) (A.2)
s2
2 2
1 s
2
2
n1 n2
n1− +1 n2−1
With the desired confidence level and degrees of freedom calculated, the critical
t-value is then looked up in a table. This table gives the pre-calculated t-values
between which 95% or 99% of all values fall for the degrees of freedom in the
distribution. If the calculated t-value is closer to zero than the critical t-value, it is
determined that the two sample sets do not have a significant difference between
them. This is because the difference between the sets falls within the range where
95% of any random selection of two data sets from the same distribution would fall.
Conversely, if the calculated t-value is further from zero than the critical t-value, it
can be said there is a significant difference between the two and the true population
means of the two sets are different.
Many software packages will automatically calculate these values given an
input dataset. The work in this thesis used JMP by SAS Institute to perform the
calculations, but Minitab, SPSS, R, Python, Matlab, or Excel are all capable of
performing the calculations with built-in functions.
97
A.2 Weibull Statistics
Many things in life do not have one true value but are randomly distributed
across a range of possible values, creating a distribution. This distribution can be
modeled and analyzed to determine the likelihood of a particular outcome given the
total possible number of outcomes. Just as there are many things which can fit a
distribution, there are many different types of distributions which can represent the
real world case. The most commonly known distribution is the normal distribution
but others are frequently used and are variations on each other, such as Student’s t,
Lorentz, exponential, Chi-squared. In fracture mechanics, the Weibull distribution
is used to represent the distribution of the strength of individual samples. The
cumulative distribution function (CDF) of a Weibull distribution is given in Equation
A.3, where P is the probability of failure having occurred, σ is the applied stress
which is greater than zero, σmin is the minimum stress needed for failure, σθ is a
scaling factor known as the characteristic strength, and m is the Weibull modulus,
or the distribution shape factor. [117] Commonly there is considered no minimum
stress for failure to occur, resulting in σmin = 0.
( )
− σ−σ
m
min
P = 1− e σθ (A.3)
The Weibull distribution was first proposed by Waloddi Weibull as an empirical
fit to the observations he had been making. In a curious case, years later, his Weibull
distribution was backed up by theory as extreme value distributions and weakest
link theory was developed in the study of fracture mechanics. [118] In weakest link
98
theory, it is the result of only the largest flaw which causes failure. If a material
has a normal, symmetric distribution of flaw sizes, then the cases of the largest flaw
will be naturally skewed to the right, above the mean size and following the tail of
the normal distribution. Of the functions which were developed for extreme values,
Weibull’s offered the most versatility in terms of bounds, fit and confidence intervals.
The Weibull distribution given in Equation A.3 has two parameters, the
characteristic strength and the Weibull modulus. The characteristic strength is very
similar in concept to the mean of a distribution but will always be slightly higher.
The mean of a distribution is defined as the point where the total probability has
reached 50%. In contrast, the characteristic strength is defined as the point where
the total probability has reached 63.2%, or otherwise the strength which makes the
exponential term of Equation A.3 go to −1. The Weibull modulus is a measure of
the width of the distribution and thus the variability of flaw sizes and strength of
the material. The value of the Weibull modulus depends highly on the material
and the processing which occurred, but typically ranges in the 1 to 20 range. From
an engineering perspective, a high Weibull modulus is desired because it makes
the failure point more predictable. If multiple flaw types exist, each with their
own distribution, the application of the Weibull distribution become complicated.
Failure of each sample must be assigned the different types via fractography, and
the contribution of each distribution attributed to that of the whole.
The CDF can be plotted linearly by rearranging Equation A.3 to take the form
of Equation A.4, where the slope of the line is m and the intercept is −m ln(σθ).
This allow for fitting of the Weibull parameters and a visual estimation of if the data
99
fits a Weibull distribution. The preferred fitting method is the maximum likelihood
method, rather than the linear regression method, as it results in smaller confidence
intervals for the estimated parameters.
ln(− ln(1− P )) = m ln(σ)−m ln(σθ) (A.4)
To achieve meaningful results with small confidence intervals, a large number
of samples must be tested. It is best practice that for Weibull parameters to be
estimated, in excess of 30 samples should be tested. [119] While sometimes it is not
practical to test such numbers of samples, values reported with smaller sample sets
should have the confidence intervals examined.
One unique advantage of the Weibull distribution is the ability to scale the
stress applied to a sample to a different size. [120] When applying stress to a relatively
small sample, due to the low volume of the stressed sample, there is a low probability
that a large critical flaw will be found which causes failure. If, one was to repeat
the test on a larger sample, the added volume means a greater chance of having a
critical flaw cause failure at a lower strength. As a result, the larger volume which
is placed under stress, the weaker it will be. Equation A.5 shows this relationship
where VE is the effective volume which was placed under stress. For a pure tensile
test, VE is the entire volume of the sample, but for flexural tests VE is much less.
In addition to being able to scale the overall size of the sample up or down, this
size scaling allows for the comparison between test methods, such as 3-point and
4-point bending. Using the known effective volumes from 3 and 4-point bending
with similar sample dimensions and bottom spans, Equation A.6 allows the results
100
of the two tests to be related to each other, as long as m is constant between the two
test methods. As a result, the 3-point bend testing will always result in a greater
strength compared to 4-point (as long as m > 1), due to the smaller volume which it
engages when testing. ( ) 1
σ1 V mE2
= (A.5)
σ2 (VE1 ) 1
m+ 2 m
σ3pt = σ4pt (A.6)
2
101
Appendix B: TGA Manual
B.1 Theory of Operation
Thermogravimetry works on the principle of measuring small changes in mass
as the environment changes. Changes in environment can come from changing
temperature or gas composition. This system uses a Cahn microbalance to measure
mass, a furnace to heat the sample up to 1100 ◦C, and mass flow controllers (MFCs)
to control the gaseous environment.
B.2 Mass Measurement
The TGA uses a Cahn D200 microbalance for the mass sensing capability.
The microbalance works by measuring the current required to maintain balance
between the sample and a counter balance. The sample can be hung at two different
points on the arm of the balance, Loop A or Loop B. The choice of loops determines
the capacity and sensitivity of the measurements, as shown in Table B.1. Full
specifications are given in the microbalance manual. Loop A is used most commonly
for its increased sensitivity.
The microbalance has an independent flow of nitrogen gas to it to maintain
102
Table B.1: Cahn D200 Microbalance Specifications
Loop A Loop B
Capacity (g) 1.5 3.5
Maximum Weight Change (mg) 150 750
Sensitivity (µg) 0.1 1
Repeatability 10% total load on both pans
a consistent atmosphere. Only a small amount of flow is required (5 to 10 sccm) to
keep positive pressure, but if the protective cover is removed additional time will be
required to purge the balance.
B.2.1 Setup and Operation
Alignment of the balance is key to producing a low noise signal. Ideally, the
hang-down wire should run through the center of the tubes and at minimum it must
not touch anywhere or allow the sample pan to touch the walls of the furnace. The
hang-down wire is made from platinum or a platinum-rhodium mixture and is in two
pieces, to facilitate the changing out of the bottom portion in case of contamination
and degradation. Two successful methods of straightening the wire are to lay it flat
on a surface and roll it out from the center, similar to working dough, or to hang it
with a weight and soften it using heat allowing it to stretch straight. Hooks on the
ends are created using tweezers and a razor blade for a sharp bend.
With the hang-down wire straight and in place, the furnace tube can be adjusted
103
to place the sample pan in the middle, equidistance from all walls. Moving the tube
can be achieved by adjusting the supporting clamp at the bottom of the furnace or
adjusting the insulation holding it in place at the top. Alignment usually requires
looking straight down the tube from above the microbalance. A light can be placed
at the bottom of the tube, replacing the thermocouple, to give illumination as to the
location of the sample pan in the tube. Alternatively, the furnace can be heated to
800 ◦C so that the tube and pan glow, giving a better light, but adjustments must
wait for the furnace to cool. Usually, both methods are used, the flashlight for a first
alignment and the hot furnace to check afterwards.
With everything aligned, the counterbalance needs to be properly adjusted.
Place on the sample pan any appropriate crucible which will be used and open the
microbalance cover and remove the cover from the counterbalance by gently pulling.
At this point, the microbalance is turned off using the switch on the back of the
microbalance controller unit. Counterbalance weights are added or removed to have
the empty sample pan and crucible balance with the counter. A light touch to the
arm of the balance may be needed to off-set static and friction but ensure this is
done with clean tweezers. Once the system balances, or does so as closely as possible,
the covers may be replaced and the balance turned back on. Static, especially in
the winter time, commonly causes the counterbalance to swing wildly as the cover
is placed back over it. Using an un-gloved hand or wiping the inside down with
deionized water helps reduce the static.
The thermocouple in the bottom of the furnace should be placed as close to
the sample as possible. To achieve this, the thermocouple is incrementally raised
104
while observing the mass readout from the microbalance. Once the balance becomes
unstable or jumps, the thermocouple should be backed out 0.5 to 1 cm and tightened
in place. The thermocouple is covered with a quartz sheath to act as a thermowell
and protect the thermocouple. If changing between materials systems or significant
degradation has occurred to the thermowell, it should be replaced.
At this point, the system should be burned out to remove any contamination
on the sample pan or hang-down wire, and time is needed for it to stabilize. The
furnace should be heated to 1000 ◦C for 4 hours in an oxygenated environment. This
also provides time for the system to stabilize, as monitored by creating a run with
the MicroScan Acquisition software. It can take up to 24 hours for it to stabilize. If
random jumps occur in the mass, recheck the alignment and thermocouple placement.
After stabilization, a sample can be loaded in to a crucible and sample pan.
While the microbalance is calibrated to measure the mass of the sample added, it is
good practice to measure and record the mass using a separate balance beforehand.
Crucibles may be reused where appropriate. After the sample is placed, the furnace
is raised slowly, to ensure not altering the alignment, and connected to the fittings
above. The furnace should be at the maximum height position when finished. Time
should then be given to allow the mass reading to stabilize before starting a test.
The program MicroScan Acquisition controls the TGA and records data. After
opening, select “Balance > Establish Connection” to connect to the D200. If an
error occurs, power cycle the balance, restart the program, and check the serial
cable leading to the balance. Upon successful connection, the current readout of
the balance will be displayed. The balance may be tared before adding a sample
105
Figure B.1: MicroScan Acquisition with Open Method open and the current
read out of the balance displayed
by selection “Balance > Tare.” Starting a new run is achieved by selecting “File
> Open Method.” Figure B.1 shows the “Open Method” screen with the current
balance readout from MicroScan Acquisition.
From here, the total length of the run is defined along with the sample rate. The
sample rate will need to match the rate given in the other programs. Additionally,
the run title can be defined and the loop can be changed if needed. Once the run
length and sample rate are entered, selecting “Ok” brings up the menu where the run
can be started. Selecting “Start Run” causes MicroScan Acquisition to prompt for
106
the location to save the data in a proprietary format, and once the file path is given,
it then starts collecting data. During the run, collected data is plotted and can be
viewed using the controls at the top of the plot. After a run is complete, the saved
file is converted to a comma separated value (csv) by opening it with MicroScan
Analysis and selecting “File > Export... > Comma/Tab Separated...”
B.3 Heating
Heating is done by a furnace surrounding the alumina or quartz tube and
sample. It is controlled by a Eurotherm controller and readings are taken from a
thermocouple placed inside the furnace tube. As a result, there is a lag in the system
between the furnace elements heating and the thermocouple sensing the heat. The
controller can be programmed with up to 8 segments in the temperature profile and
programming can be performed from the front panel of the unit or by the Eurotherm
iTools software. The iTools software must be closed before attempting to use the
LabVIEW program which reads the temperature from the controller.
B.4 Gas Delivery
Gas flows are controlled by the MKS Type 647C Multi-channel Flow Ratio/-
Pressure Controller and the individual MKS Type M100B MFCs. Each MFC controls
one gas by measuring the flow with a thermal mass sensor and adjusting it using a
valve to meet the desired set point. Each MFC has a rated size and must have the
zero and gas correction factor (GFC) properly set. The zero can be set physically
107
using a screw on the MFC if a large adjustment is needed or electronically using
the controller. The zero should be set with both sides of the MFC open to the
atmosphere to ensure no leak from a leftover pressure differential. The GFC can
be theoretically calculated based on the heat capacity of the gas compared to a
standard. This assumes that the MFC is properly calibrated, which may not be a
safe assumption. As such, it is best to calibrate the MFC by adjusting the GFC so
that it measures correct flow rates as given by a bubble meter.
If an MFC leaks despite being set to off, the tension spring for the valve
located inside the MFC may need adjustment. The manual for the MFC outlines the
procedure. If the tension is correct and the valve still leaks, service is advised. An
alternative is to add an external shut-off valve, which can be closed either manually
or automatically. This has been implemented for one of the current MFCs.
Different gases can be supplied to the MFCs as needed for the experimental
setup. If the gas to an MFC is changed, a new GFC will need to be obtained. Gases
can humidified after the MFC with the use of a bubbler, but they should not be
humidified beforehand for risk of damaging the MFC. Care should also be taken when
mixing or changing gas compositions to avoid potentially dangerous combinations,
such as mixing fuels and oxidizers. When changing between a reducing and oxidizing
environment, it is good practice to allow sufficient time for a sweep gas (such as N2)
to reduce the risk of the different atmospheres mixing.
The MFCs can be controlled by a computer with LabVIEW. They can either
be manually set or given a scheduled program to follow. Typically, a constant total
flow rate will be used among all tests. A good starting point for the total flow rate
108
is 50 sccm. After turning on the MKS controller, the master flow control must be
turned on by pushing “ON” followed by “0/All.”
B.5 pO2 Measurement
pO2 of the gas flow is measured by a home-built YSZ sensor located after the
MFCs combine into one flow. Gas flows down a two-bore alumina tube along with a
platinum wire to the end of a YSZ tube. The wire is lightly connected to the inside
of the YSZ tube using platinum paste. The gas then flows in the reverse direction,
between the alumina and YSZ tubes to the outlet. Outside the YSZ tube at the
end, a separate piece of platinum wire is connected to the end of the tube. This
setup is then placed in a furnace and heated to 800 ◦C where the difference of partial
pressure can be measured according to the Nernst equation. Voltage potentials from
the two wires are measured by a Keithley 2000 digital multimeter.
Once the system is setup, very little needs to be done to maintain the sensor.
If the sensor needs to cooled in anticipation of a power outage or in order to be
worked on, a ramp rate of 1 ◦C min−1 is to be used to avoid damaging the YSZ tube.
For accurate pO2 measurements, after any modification a calibration curve needs to
be created using a reference meter. Voltage measurements and pO2 calculations are
recorded by the computer using a LabVIEW program.
109
B.6 Controls
Almost all of the components of the TGA system are computer controllable.
As previously mentioned and explained, the microbalance is controlled by the
MicroScan Acquisition program. LabVIEW programs interface with the Keithley,
MKS controller, and Eurotherm temperature controller.
B.6.1 Gas Controls
Gas flows can be controlled with either the Gas Controller or Gas Scheduler
programs. Gas Controller allows for the immediate control of any gas flow channel
while Gas Scheduler changes the gas flows at predetermined times allowing for
the automated changing of composition. Figure B.2 shows the front panel of Gas
Controller, Figure B.3 is the front panel of Gas Scheduler, and the code for both
programs is explained in Appendix C.2
To start using Gas Controller, ensure proper channels are set for the MKS
Controller Port, Arduino Port, and Servo Channel. On and Off position adjust the
position of the external shut-off valve controlled by the Arduino. These values should
only change if changing USB to serial adapters or with new hardware. Starting
the LabVIEW program immediately begins the control to gasses to the specified
values. The radio buttons on the left turn channels on or off and set them to the set
point. To end the program, push the STOP button within the program front panel.
The program displays the current reading for the flow to ensure flow is occurring as
specified.
110
Figure B.2: Front panel display of Gas Controller program
Figure B.3: Front panel display of Gas Scheduler program
111
The Gas Scheduler requires the input of a tab delimited text file, referred to as
a schedule. An example schedule comes with the program. In this schedule, the time
for the desired gas flow is given in seconds, followed by the set point of each channel.
The program ends on the row with a negative time value, retaining those set points.
Before starting the program, the file path to the schedule must be given by
clicking the folder icon next to the Path to Program File field. Again, ensure that
the com ports and channels are correct. The program records the measured flows at
a set interval, as defined by the Recording Period (in seconds) and records them to
a csv. Additionally, comments can be added to the csv in the optional comments
field. After starting the program, the start button needs to be pressed.
B.6.2 Temperature and pO Measurements2
The program Temp & pO2 Measurements reads the values from the Eurotherm
controller and the Keithley multimeter, plots them, and records the results to a csv.
Figure B.4 shows the front panel of the program with the user controls.
Before running the program, ensure that the com ports and channels are
correct. The Measurement Period should match that given to MicroScan Acquisition.
Enabling the timer provides for the program to stop at after the designated time.
Comments can be added to the csv file by filling in the File Comments field. When
the program starts running, it will ask for the location to save the csv. As a note,
the timer for the program starts as soon as the program does, before defining the
csv location, but the first measurement does not occur until afterwards.
112
Figure B.4: Front panel display of Temp & pO2 Measurements program
B.6.3 Starting a Run
Multiple programs needs to be started simultaneously for a run of the TGA.
Because each program starts differently, the following is a recommendation for the
order to start the programs. Time values from MicroScan should then be used
in subsequent analysis. This procedure minimizes the amount of time difference
between measurements on the TGA, temperature and pO2 .
To begin, enter all values into the various programs. In MicroScan Acquisition,
follow the full setup, including clicking Start Run, but stop immediately afterwards
with the file browser up. In Temp & pO2 Measurements, follow the full setup
including designating where to save the data. As soon as the file path to save data
113
is submitted, switch to MicroScan Acquisition and complete the setup, submitting
the file path also. If using Gas Scheduler, start that program as normal followed by
the temperature program on the Eurotherm.
B.7 Interfacing with Other Devices
The effluent of the TGA can be fed to other devices for advanced analysis, such
as an MS. When doing this, one thing to note is physical distance between the TGA
and the other instrument. The longer the gas line connecting the two instruments,
the longer it will take gas to travel from one instrument to the other. During this
time period, diffusion will occur up and down the line, broadening any signal to the
second instrument. This delay and broadening effect will need to be investigated for
each experimental setup.
B.8 Troubleshooting
• Unsteady mass: check alignment and thermocouple placement
• No gas flow on any channel: check master flow control
• No gas flow on one channel: check if the channel is on and manual valves
are open
• MFC leaks when off: Check zero, adjust spring tension as described in MFC
manual, add external shutoff valve or return for repairs
114
• No or noisy voltage from pO2 sensor: Check wire connections from sensor,
rebuild sensor with new platinum paste
• Communications Error: Check that no other program is communicating
with the hardware, restart the hardware and unplug then plug-in the USB
connection
115
Appendix C: Code
The programs referenced in this appendix are published as open-source software
and are in the public domain. The code and latest versions of all parts are available on
GitHub at https://github.com/patrickstanley/ with the provided repository names
and contributions are welcome.
C.1 Arduino PID Relay Furnace Controller with Serial Connectivity
The purpose of this Arduino sketch is to power a heating furnace (controlled
by a solid-state relay) with PID controls. In addition, serial communications are
used to monitor, log, and interact with the controller. A running average is used to
reduce thermocouple measurement noise for the derivate component of the controls.
This sketch is available at https://github.com/patrickstanley/arduino-pid.
C.1.1 Hardware and Other Libraries
This was built to be used on a Arduino Uno R3, but should be easily run on a
variety of different boards. For temperature sensing, an Adafruit MAX31856 is used
with the provided library. Also used is the PID controller library by br3ttb. This
sketch pieces the two libraries together and adds serial communications functionality.
116
C.1.2 Variables
The following is a list of control variables which may be changed per setup.
• RelayPin: The physical pin the relay is connected to.
• Adafruit MAX31856(CS, DI, DO, CLK): Pins for the MAX31856
• WindowSize: Length of on/off cycle, needed to account for AC power
• printdelay: Delay time in milliseconds for printout on serial
• MaxOP: Maximum operating power, used to extend life of elements
• myPID(&Input, &Output, &workingSet,P,I,D, DIRECT): PID values need to
be adjusted for the particular setup. If inverse behavior is experienced (e.g.
heats when it should cool) switch DIRECT.
• workingSet and Setpoint: Initial setpoint for power on
• ramprate: Sets a ramp rate for the working setpoint. Measured in milliseconds
for 1C change.
C.1.3 Serial Communications
Baud rate is set to 9600. Output is tab delimitated of Setpoint, Working
Setpoint, Thermocouple Temperature, and Output Power. To change Setpoint, send
a new number over serial and press return. The newline character of the serial
monitor may have to changed send \n as end of line.
117
C.1.4 Arduino Sketch
1 #inc lude <Adafruit MAX31856 . h> // loads l i b r a r y f o r thermocouple reader
2 #inc lude <PID v1 . h> // loads PID l i b r a r y
3 #inc lude <RunningAverage . h> // loads c l a s s f o r Running Avgerage
↪→ c a l c u l a t i o n s
4 #de f i n e RelayPin 2 // s e t pin number which i s connected to power r e l ay
↪→ f o r dev i c e
5
6 const byte numChars = 32 ;
7 char rece ivedChars [ numChars ] ; // an array to s t o r e the r e c e i v ed data
8
9 boolean newData = f a l s e ; // t r i g g e r to determine i f data needs to be
↪→ read
10
11 Adafruit MAX31856 max = Adafruit MAX31856 (4 , 5 , 6 , 7) ; // Software SPI ,
↪→ Pin number f o r : CS, DI , DO, CLK
12
13 //Def ine Var iab l e s
14 double workingSet , Setpoint , Input , Output ;
15 i n t WindowSize = 1000 ; //Length o f c y c l e f o r power on/ o f f
16 unsigned long p r e vp r i n tM i l l i s = 0 ;
17 const long p r in tde l ay = 5000 ; // de lay amount be f o r e p r i n t i n g v ia s e r i a l
18 unsigned long windowStartTime ;
19 unsigned long prevRampTimer = 0 ;
20 unsigned long MaxOP = . 9 5 ; // Max Operating Power
21
118
22 RunningAverage myRA(10) ; // Set up running average with # of
↪→ measurements
23
24 PID myPID(&Input , &Output , &workingSet , 250 , 50 , 8 , DIRECT) ; // Spec i f y the
↪→ l i n k s and i n i t i a l tuning parameters
25
26 void setup ( ) {
27 pinMode ( RelayPin , OUTPUT) ; // d e f i n e RelayPin as an output
28
29 // s t a r i n g s e r i a l communications
30 S e r i a l . begin (9600) ;
31 S e r i a l . p r i n t l n ( ”PID Cont r o l l e r Output” ) ;
32 S e r i a l . p r i n t l n ( ” Setpo int \ t Working SP\ t TC Temp\ t Output” ) ; // tab
↪→ de l iminated f o r l ogg ing
33
34 // setup f o r thermocouple reader
35 max . begin ( ) ;
36 max . setThermocoupleType (MAX31856 TCTYPE K) ; // change type i f needed
37
38 windowStartTime = m i l l i s ( ) ; // s t a r t t imer f o r duty cy c l e window
39
40 myRA. c l e a r ( ) ; // e x p l i c i t l y s t a r t c l ean
41
42 // i n i t i a l i z e the v a r i a b l e s we ’ re l i nked to
43 workingSet = 20 ;
44 Setpo int = 20 ;
45
119
46 myPID. SetOutputLimits (0 , MaxOP ∗ WindowSize ) ; // t e l l the PID to range
↪→ between 0 and the f u l l window s i z e
47 myPID. SetMode (AUTOMATIC) ; // turn the PID on
48 }
49
50 // f u c t i on to r e c i e v e data v ia s e r i a l and proce s s when endmarker i s
↪→ r e c i e v ed
51 void recvWithEndMarker ( ) {
52 s t a t i c byte ndx = 0 ;
53 char endMarker = ’ \n ’ ; // d e f i n e l i n e endmarker f o r c l i e n t
54 char rc ;
55
56 whi le ( S e r i a l . a v a i l a b l e ( ) > 0 && newData == f a l s e ) {
57 rc = S e r i a l . read ( ) ;
58 i f ( r c != endMarker ) {
59 rece ivedChars [ ndx ] = rc ;
60 ndx++;
61 i f ( ndx >= numChars ) {
62 ndx = numChars − 1 ;
63 }
64 }
65 e l s e {
66 rece ivedChars [ ndx ] = ’ \0 ’ ; // terminate the s t r i n g
67 ndx = 0 ;
68 newData = true ;
69 }
70 }
120
71 }
72
73 //Function to repeat new s e tpo i n t f o r con f i rmat ion
74 void showNewData ( ) {
75 i f ( newData == true ) {
76 St r ing r e c i e v edS t r i n g = St r ing ( rece ivedChars ) ;
77 S e r i a l . p r i n t ( ” Setpo int changed to . . . ” ) ;
78 S e r i a l . p r i n t l n ( r e c i e v edS t r i n g . toFloat ( ) ) ;
79 Setpo int = r e c i e v edS t r i n g . toF loat ( ) ;
80 newData = f a l s e ;
81 }
82 }
83
84 void loop ( ) {
85 recvWithEndMarker ( ) ; // check f o r new s e tpo i n t
86 showNewData ( ) ; // s e t s new s e tpo i n t and d i s p l a y s i t
87 myRA. addValue (max . readThermocoupleTemperature ( ) ) ; //add themocouple
↪→ read to r o l l i n g average
88 Input = myRA. getAverage ( ) ; // s e t s PID input to r o l l i n g average value
89 myPID. Compute ( ) ; // c a l c u l a t e PID output
90
91 // p r in t de lay f o r s e r i a l output so not to f l o od l o g s
92 unsigned long p r i n tM i l l i s = m i l l i s ( ) ;
93 i f ( p r i n tM i l l i s − p r e vp r i n tM i l l i s >= pr in tde l ay ) {
94 p r e vp r i n tM i l l i s = p r i n tM i l l i s ;
95 S e r i a l . p r i n t ( Setpo int ) ; S e r i a l . p r i n t ( ”\ t ” ) ; S e r i a l . p r i n t ( workingSet ) ;
↪→ S e r i a l . p r i n t ( ”\ t ” ) ; S e r i a l . p r i n t (max . readThermocoupleTemperature
121
↪→ ( ) ) ; S e r i a l . p r i n t ( ”\ t ” ) ; S e r i a l . p r i n t l n (Output/WindowSize ∗100) ;
96 }
97
98 // turn the output pin on/ o f f based on pid output
99 i f ( m i l l i s ( ) − windowStartTime>=WindowSize )
100 { // time to s h i f t the Relay Window
101 windowStartTime += WindowSize ;
102 }
103 i f (Output < m i l l i s ( ) − windowStartTime ) d i g i t a lWr i t e ( RelayPin ,LOW) ; //
↪→ HIGH and LOW se t to d e f au l t o f f
104 e l s e d i g i t a lWr i t e ( RelayPin ,HIGH) ;
105
106 // s e t s ramp ra t e f u n c t i o n a l i t y
107 unsigned long ramptimer = m i l l i s ( ) ;
108 unsigned long ramprate = 6000 ; // m i l l i s e c ond s f o r 1C change
109 i f ( workingSet != Setpo int )
110 {
111 i f ( ramptimer − prevRampTimer >= ramprate && workingSet < Setpo int )
112 {
113 workingSet++;
114 prevRampTimer = ramptimer ;
115 }
116 i f ( ramptimer − prevRampTimer >= ramprate && workingSet > Setpo int )
117 {
118 workingSet−−;
119 prevRampTimer = ramptimer ;
120 }
122
121 }
122 }
C.2 MKS Gas Controller and Scheduler
The purpose of these LabVIEW programs is to communicate with and control
MFCs for remote and automated operation of the TGA. Individual MFCs can be
turned on or off and have their set point changed. After basic controls are developed,
a new VI, a scheduling program, is provided which reads a text file automatically
changing flow rates at predetermined times. The VI gas controller and gas scheduler
VIs, along with a sample gas schedule, are available under the GitHub repository
“mks-mfc.” The VIs are saved for LabVIEW version 2013 and later. This work is the
continuation and development of programs previously written by others.
C.2.1 Hardware and Other Libraries
These programs are written for the use of a MKS 647C Multi Channel Flow
Ratio/Pressure Controller with Type M100B MFCs. It can accommodate up to
8 channels for MFCs and communicates via serial. Drivers for LabVIEW commu-
nications with the MKS 647C are available online from the National Instruments
(NI) website. Serial communications occur over the VISA standard and require the
installation of the VISA drivers from NI.
An Arduino Uno R3 is utilized with a servo motor and 3D printed parts to
manually turn a shut-off valve, compensating for a leaking MFC. Details on the
installation of the Arduino controlled shut-off including STL files for 3D printing are
123
available at https://hackaday.io/project/52519-automated-gas-flow-shut-off. To con-
trol the Arduino from LabVIEW, the LINX library is used and must be downloaded
from the VI package manager. Additionally, the Arduino will need to be flashed
with the LINX control sketch before use. This component can be easily removed if
not required for the application.
C.2.2 Program Overview
The block diagram of the Gas Controller VI, which manually controls the
MFCs, is provided in Figure C.1 and is the fundamental starting block for the gas
scheduling program. Starting on the far left of the program, initial communications
variables such as baud rate and port information are passed to the sub-VIs which
open communications. The program then enters a while loop until the stop button
is pressed or an error occurs, at which point serial communications are closed by the
appropriate sub-VI.
Within the while loop, the program sequentially goes to each channel, checking
the desired status, setting the control variables and reading the flow. If the program
is set to not use an MFC, the internal valve is shut and the flow is read to ensure
no leakage. If an MFC is set to be used, the percent of full flow is calculated based
on the size of the MFC, the gas correction factor, and the desired flow rate. The
percent of full flow is then passed to the MFC and the flow is read. For the case of
channel 3, which possesses a leaky MFC, the addition of an external shutoff valve
controlled by Arduino was implemented. If channel 3 is to be used, the Arduino
124
sends a pulse width modulation (PWM) signal to the servo motor which corresponds
to the valve being in the open position. In the case where channel 3 is to be closed,
a PWM signal is sent, which closes the valve.
Finally, after each channel has been set, the program checks for the dangerous
combination of hydrogen and oxygen gasses. If the channels for hydrogen and oxygen
(in this case channels 3 and 4) are both open, the program immediately shuts the
valves and gives an error warning the user and ends the program. As long as this is
not the case, the program proceeds to loop back to setting the values for channel 1.
125
126
Figure C.1: MFC Controller program
127
Figure C.1: MFC Controller program (cont.)
128
Figure C.1: MFC Controller program (cont.)
129
Figure C.1: MFC Controller program (cont.)
130
Figure C.1: MFC Controller program (cont.)
131
Figure C.1: MFC Controller program (cont.)
The Gas Scheduler program uses the Gas Controller to automate the switching
of gasses based on the instructions provided in a text file. The text file, referred
to as the gas schedule, is a nine column tab delimited file where the first column
is the amount in seconds for which that gas flow is to be run, followed by the
flows for each channel. The program sequentially goes though the rows until it
encounters a negative time, at which point it ends, leaving that row’s gas flows
running. The program also additionally records the measured flows to a csv file for
manual inspection later.
Figure C.2 gives the block diagram of the program. Before the start of the
program, the path to the gas schedule must be specified. At the start of the program,
the location to record measured gas flows is requested and setup as a csv file type with
a header, variables to count the row of the program and time elapsed are initialized
and communications variables are passed to sub-VIs. After this, the program enters
a while loop. This while loop is ended if an error occurs, if the stop button is pressed,
or if the amount of time in the stage (given by the first column in the gas schedule)
is less than zero.
Once the start button is pushed, a timer is started to measure the amount of
time elapsed in the stage. If the time has elapsed the amount prescribed (starting
with zero seconds), the program reads the gas schedule, increments the stage number,
and passes the flow rates to a gas controller sub-VI, after which the VI loops. If time
has not elapsed, the program records the flows based on a second elapsed time timer
to limit the total number of recordings. After a program end condition is met, the
communications channels are closed.
132
133
Figure C.2: Gas Scheduler program
C.3 Temperature and pO Measurements2
The purpose of this LabVIEW program is to periodically measure the tempera-
ture from a Eurotherm process controller and calculate pO2 based on voltage from a
Keithley multimeter for the TGA system. The VI is available on GitHub under the
repository “eurotherm-keithley-measure” and is saved for LabVIEW 2013 and later.
This work is the continuation and development of programs previously written by
others.
C.3.1 Hardware and Other Libraries
Temperature measurements are taken directly from a Eurotherm 2408 controller
with optional serial communications module. A driver package for all 2400 series
Eurotherm controllers is available from the NI website. Serial communications again
occur by VISA and require the NI VISA drivers.
pO2 was calculated from voltage measurements taken of a YSZ sensor by a
Keithley 2000 digital multimeter. Communications were established with the Keithley
via GPIB (IEEE-488) and thus require the NI-488 driver from NI. Additionally,
drivers for the Keithley 2000 are available online from NI and are required for the
VI.
C.3.2 Program Overview
A copy of the block diagram is given in Figure C.3. When started, the program
asks for the location to save the measurements. The program then creates the csv
134
file time in seconds, the voltage read, the calculated log pO2 , and the temperature
from the controller. Communications variables, such as baud rate and channels, are
sent the appropriate sub-VIs to begin communications to the device and a timer is
started. The program backs up the Keithley’s current settings, resets the device and
prepares it to measure voltage on one channel, while displaying “Running” on the
front panel display. After the program is complete, the Keithley is reset once again
and its original settings are restored before communications are closed to it and the
Eurotherm.
The rest of the program occurs in a while loop. The loop is delayed a fixed
amount of time, set by the Measurement Period. If the timer functionality is turned
on, the loop repeats for a fixed amount of time based on the Length of Run variable,
after which a stop signal is sent. During each loop, the process value (temperature)
of the Eurotherm is measured and any alarms or flags are checked for. If any occur,
they are passed to the error handler and the program stops. The voltage of the
specified channel of the Keithley is measured and the log(pO2) is calculated from it.
These values are then displayed, plotted, saved to the csv file.
135
136
Figure C.3: Temperature and pO2 measurement program
Bibliography
[1] U.S. Energy Information Administration. World energy demand and economic
outlook. Technical Report May 2016, U.S. Energy Information Administration,
2016.
[2] Stephen G Osborn, Avner Vengosh, Nathaniel R Warner, and Robert B Jackson.
Methane contamination of drinking water accompanying gas-well drilling and
hydraulic fracturing. Proceedings of the National Academy of Sciences of the
United States of America, 108(20):8172–6, 2011.
[3] Avner Vengosh, Robert B Jackson, Nathaniel Warner, Thomas H Darrah,
and Andrew Kondash. A Critical Review of the Risks to Water Resources
from Shale Gas Development and Hydraulic Fracturing in the United States.
Environmental Science & Technology, 16(4):6838, 2014.
[4] Susan Solomon, Gian-Kasper Plattner, Reto Knutti, and Pierre Friedlingstein.
Irreversible climate change due to carbon dioxide emissions. Proceedings of the
National Academy of Sciences of the United States of America, 106(6):1704–9,
2009.
[5] J Hansen, D Johnson, A Lacis, S Lebedeff, P Lee, D Rind, and G Russell. Cli-
mate impact of increasing atmospheric carbon dioxide. Science, 213(4511):957–
966, 1981.
[6] U.S. Environmental Protection Agency. Inventory of U.S. Greenhouse Gas
Emissions and Sinks: 1990 - 2015. Technical report, U.S. Environmental
Protection Agency, 2017.
[7] E. D. Wachsman and K. T. Lee. Lowering the Temperature of Solid Oxide
Fuel Cells. Science, 334(6058):935–939, 2011.
[8] Robert H. Lasseter and Paolo Paigi. Microgrid: A conceptual solution. In
IEEE Power Elecrronics Speacialists Conference, volume 6, pages 4285–4290,
2004.
137
[9] Seungdoo Park, John M. Vohs, and Raymond J. Gorte. Direct oxidation of
hydrocarbons in a solid-oxide fuel cell. Nature, 404(6775):265–267, 2000.
[10] Wikimedia Commons. Solid Oxide Fuel Cell Diagram.
[11] James Larminie and Andrew Dicks. Fuel Cell Systems Explained, volume 93.
2001.
[12] I. Pilatowsky, R.J. Romero, C.A. Isaza, S.A. Gamboa, P.J. Sebastian, and
W. Rivera. Thermodynamics of Fuel Cells. Cogeneration Fuel Cell-Sorption
Air Conditioning Systems, pages 25–36, 2011.
[13] Mogens Mogensen, Nigel M. Sammes, and Geoff A. Tompsett. Physical,
chemical and electrochemical properties of pure and doped ceria. Solid State
Ionics, 129(1):63–94, 2000.
[14] Mikko Pihlatie, Andreas Kaiser, and Mogens Mogensen. Mechanical properties
of NiO/Ni-YSZ composites depending on temperature, porosity and redox
cycling. Journal of the European Ceramic Society, 29(9):1657–1664, 2009.
[15] J. Laurencin, G. Delette, O. Sicardy, S. Rosini, and F. Lefebvre-Joud. Impact
of ‘redox’ cycles on performances of solid oxide fuel cells: Case of the electrolyte
supported cells. Journal of Power Sources, 195(9):2747–2753, 2010.
[16] Jiang Liu, Brian D Madsen, Zhiqiang Ji, and Scott A Barnett. A Fuel-Flexible
Ceramic-Based Anode for Solid Oxide Fuel Cells. Electrochemical and Solid-
State Letters, 5(May 2012):122–124, 2002.
[17] Sossina M. Haile. Fuel cell materials and components. Acta Materialia,
51(19):5981–6000, 2003.
[18] Harumi Yokokawa, Hengyong Tu, Boris Iwanschitz, and Andreas Mai. Funda-
mental mechanisms limiting solid oxide fuel cell durability. Journal of Power
Sources, 182(2):400–412, 2008.
[19] D. A. Andersson, S. I. Simak, N. V. Skorodumova, I. A. Abrikosov, and
B. Johansson. Optimization of ionic conductivity in doped ceria. Proceedings
of the National Academy of Sciences, 103(10):3518–3521, 2006.
[20] G. Bryan Balazs and Robert Glass. ac impedance studies of rare earth oxide
doped ceria. Solid State Ionics, 76(1-2):155–162, feb 1995.
[21] Shaowu Zha, Changrong Xia, and Guangyao Meng. Effect of Gd (Sm) doping
on properties of ceria electrolyte for solid oxide fuel cells. Journal of Power
Sources, 115(1):44–48, 2003.
[22] Hideaki Inaba and Hiroaki Tagawa. Ceria-based solid electrolytes. Solid State
Ionics, 83(1-2):1–16, jan 1996.
138
[23] F. Gutierrez-Mora, J. M. Ralph, and J. L. Routbort. High-temperature
mechanical properties of anode-supported bilayers. Solid State Ionics, 149(3-
4):177–184, 2002.
[24] Ji Haeng Yu, Gun Woo Park, Shiwoo Lee, and Sang Kuk Woo. Microstructural
effects on the electrical and mechanical properties of Ni-YSZ cermet for SOFC
anode. Journal of Power Sources, 163(2):926–932, 2007.
[25] William D. Callister and David G. Rethwisch. Influence of Porosity. Materials
Science and Engineering: An Introduction, pages 373–381, 2014.
[26] Michel Barsoum. Processing and Surface Flaws. In Fundamentals of Ceramics,
volume 2174, pages 374–381. 2003.
[27] M. Radovic and E. Lara-Curzio. Mechanical properties of tape cast nickel-based
anode materials for solid oxide fuel cells before and after reduction in hydrogen.
Acta Materialia, 52(20):5747–5756, 2004.
[28] Miladin Radovic and Edgar Lara-Curzio. Elastic Properties of Nickel-Based
Anodes for Solid Oxide Fuel Cells as a Function of the Fraction of Reduced
NiO. J. Am. Ceram. Soc., 87(12):2242–2247, 2004.
[29] J. Laurencin, G. Delette, B. Morel, F. Lefebvre-Joud, and M. Dupeux. Solid
Oxide Fuel Cells damage mechanisms due to Ni-YSZ re-oxidation: Case of the
Anode Supported Cell. Journal of Power Sources, 192(2):344–352, 2009.
[30] D. Sarantaridis and A. Atkinson. Redox cycling of Ni-based solid oxide fuel
cell anodes: A review. Fuel Cells, 7(3):246–258, 2007.
[31] Florian P. Nagel, Tilman J. Schildhauer, Josef Sfeir, Alexander Schuler, and
Serge M A Biollaz. The impact of sulfur on the performance of a solid oxide
fuel cell (SOFC) system operated with hydrocarboneous fuel gas. Journal of
Power Sources, 189(2):1127–1131, 2009.
[32] K. Haga, S. Adachi, Y. Shiratori, K. Itoh, and K. Sasaki. Poisoning of SOFC
anodes by various fuel impurities. Solid State Ionics, 179(27-32):1427–1431,
2008.
[33] Davide Papurello, Andrea Lanzini, Sonia Fiorilli, Federico Smeacetto, Rahul
Singh, and Massimo Santarelli. Sulfur poisoning in Ni-anode solid oxide fuel
cells (SOFCs): Deactivation in single cells and a stack. Chemical Engineering
Journal, 283:1224–1233, 2016.
[34] A. Lussier, S. Sofie, J. Dvorak, and Y. U. Idzerda. Mechanism for SOFC anode
degradation from hydrogen sulfide exposure. International Journal of Hydrogen
Energy, 33(14):3945–3951, 2008.
139
[35] Shung-Ik Lee, John M. Vohs, and Raymond J. Gorte. A Study of SOFC
Anodes Based on Cu-Ni and Cu-Co Bimetallics in CeO2-YSZ. Journal of The
Electrochemical Society, 151(9):A1319, 2004.
[36] X. C. Lu and J. H. Zhu. Cu(Pd)-impregnated La0.75Sr0.25Cr0.5Mn0.5O3-δanodes
for direct utilization of methane in SOFC. Solid State Ionics, 178(25-26):1467–
1475, 2007.
[37] Hongpeng He, Raymond J. Gorte, and John M. Vohs. Highly Sulfur Tolerant
Cu-Ceria Anodes for SOFCs. Electrochemical and Solid-State Letters, 8(6):A279,
2005.
[38] John B. Goodenough and Yun Hui Huang. Alternative anode materials for
solid oxide fuel cells. Journal of Power Sources, 173(1):1–10, 2007.
[39] Shaowu Zha, Philip Tsang, Zhe Cheng, and Meilin Liu. Electrical properties
and sulfur tolerance of La0.75Sr0.25Cr1-xMnxO3 under anodic conditions. Journal
of Solid State Chemistry, 178(6):1844–1850, 2005.
[40] S. Primdahl, J. R. Hansen, L. Grahl-Madsen, and P. H. Larsen. Sr-Doped
LaCrO3 Anode for Solid Oxide Fuel Cells. Journal of The Electrochemical
Society, 148(1):A74, 2001.
[41] A. Mohammed Hussain, Jens V.T. Høgh, Wei Zhang, Peter Blennow, Nikolaos
Bonanos, and Bernard A. Boukamp. Effective improvement of interface modified
strontium titanate based solid oxide fuel cell anodes by infiltration with nano-
sized palladium and gadolinium-doped cerium oxide. Electrochimica Acta,
113:635–643, 2013.
[42] A. Mohammed Hussain, Jens V.T. Høgh, Torben Jacobsen, and Nikolaos
Bonanos. Nickel-ceria infiltrated Nb-doped SrTiO3 for low temperature SOFC
anodes and analysis on gas diffusion impedance. International Journal of
Hydrogen Energy, 37(5):4309–4318, 2012.
[43] A. Mohammed Hussain, Ke-Ji Pan, Ian A. Robinson, Thomas Hays, and
Eric D. Wachsman. Stannate-Based Ceramic Oxide as Anode Materials for
Oxide-Ion Conducting Low-Temperature Solid Oxide Fuel Cells. Journal of
The Electrochemical Society, 163(10):F1198–F1205, 2016.
[44] Yun-Hui Huang, Ronald I Dass, Zheng-Liang Xing, and John B Goodenough.
Double Perovskites as Anode Materials for Solid-Oxide Fuel Cells. Science,
2(April):254–258, 2006.
[45] O Yamamoto. Perovskite-type oxides as oxygen electrodes for high temperature
oxide fuel cells. Solid State Ionics, 22(2-3):241–246, 1987.
[46] Harlan U. Anderson. Review of p-type doped perovskite materials for SOFC
and other applications. Solid State Ionics, 52(1-3):33–41, 1992.
140
[47] Tatsumi Ishihara. Perovskite Oxide for Solid Oxide Fuel Cells. Springer, 2009.
[48] Y.-H. Huang, G Liang, M Croft, M Lehtimäki, M Karppinen, and J B Goode-
nough. Double-perovskite anode materials Sr2MMo06 (M = Co, Ni) for solid
oxide fuel cells. Chemistry of Materials, 21(11):2319–2326, 2009.
[49] Yun-Hui Huang, Ronald I. Dass, Jonathan C. Denyszyn, and John B. Good-
enough. Synthesis and Characterization of Sr2MgMoO6-δ. Journal of The
Electrochemical Society, 153(7):A1266, 2006.
[50] Carlos Bernuy-Lopez, Mathieu Allix, Craig A. Bridges, John B. Claridge, and
Matthew J. Rosseinsky. Sr2MgMoO6-δ: Structure, phase stability, and cation
site order control of reduction. Chemistry of Materials, 19(5):1035–1043, 2007.
[51] Ke-Ji Pan, A. Mohammed Hussain, and Eric D. Wachsman. High Performance
Low-Temperature Solid Oxide Fuel Cells with SrFe0.2Co0.4Mo0.4O3-δ based
Ceramic Anodes.
[52] A. Mohammed Hussain, Yi-Lin Huang, Ke-Ji Pan, Ian Robinson, and Eric D.
Wachsman. A Redox-Robust Ceramic Anode for Low-Temperature SOFC.
[53] A. Mohammed Hussain, Ian Robinson, Yi-Lin Huang, Ke-Ji Pan, and Eric D.
Wachsman. Stable Molybdate-Based Ceramic Anodes for Low-Temperature
Solid Oxide Fuel Cells.
[54] Qin Zhang, Tao Wei, and Yun-hui Huang. Journal of Power Sources, pages
59–65.
[55] Nguyen Q. Minh. Solid oxide fuel cell technology - Features and applications.
Solid State Ionics, 174(1-4):271–277, 2004.
[56] Vanessa Cascos, José Antonio Alonso, and Maŕıa Teresa Fernández-Dı́az. Nb5+-
Doped SrCoO3-δ Perovskites as Potential Cathodes for Solid-Oxide Fuel Cells.
Materials, 9(7):579, 2016.
[57] Zhangwei Chen, Xin Wang, Finn Giuliani, and Alan Atkinson. Fracture
Toughness of Porous Material of LSCF in Bulk and Film Forms. Journal of
the American Ceramic Society, 98(7):2183–2190, 2015.
[58] Melanie Kuhn, Jae Jin Kim, Sean R Bishop, and Harry L Tuller. Oxy-
gen Nonstoichiometry and Defect Chemistry of Perovskite- Structured
BaxSr1-xTi1-yFeyO3-y/2+δ Solid Solutions. Chemistry of Materials, 25(1):2970–
2975, 2013.
[59] Jeongpill Ki and Daejong Kim. Computational model to predict thermal
dynamics of planar solid oxide fuel cell stack during start-up process. Journal
of Power Sources, 195(10):3186–3200, may 2010.
141
[60] Natsuko Sakai, Harumi Yokokawa, Teruhisa Horita, and Katsuhiko Yamaji.
Lanthanum Chromite-Based Interconnects as Key Materials for SOFC Stack
Development. International Journal of Applied Ceramic Technology, 1(1):23–30,
2005.
[61] Zhenguo Yang, Guanguang Xia, Prabhakar Singh, and Jeffry W. Stevenson.
Electrical contacts between cathodes and metallic interconnects in solid oxide
fuel cells. Journal of Power Sources, 155(2):246–252, 2006.
[62] S. Taniguchi, M. Kadowaki, T. Yasuo, Y. Akiyama, Y. Miyake, and K. Nishio.
Improvement of thermal cycle characteristics of a planar-type solid oxide fuel
cell by using ceramic fiber as sealing material. Journal of Power Sources,
90(2):163–169, 2000.
[63] Steven P. Simner and Jeffry W. Stevenson. Compressive mica seals for SOFC
applications. Journal of Power Sources, 102(1-2):310–316, 2001.
[64] S.H Chan, K.a Khor, and Z.T Xia. A complete polarization model of a solid
oxide fuel cell and its sensitivity to the change of cell component thickness.
Journal of Power Sources, 93(1-2):130–140, 2001.
[65] Felix Fleischhauer, Andreas Tiefenauer, Thomas Graule, Robert Danzer, An-
dreas Mai, and Jakob Kuebler. Failure analysis of electrolyte-supported solid
oxide fuel cells. Journal of Power Sources, 258:382–390, 2014.
[66] J. Laurencin, G. Delette, F. Lefebvre-Joud, and M. Dupeux. A numerical tool
to estimate SOFC mechanical degradation: Case of the planar cell configuration.
Journal of the European Ceramic Society, 28(9):1857–1869, 2008.
[67] S.R. Bishop, D. Marrocchelli, C. Chatzichristodoulou, N.H. Perry, M.B. Mo-
gensen, H.L. Tuller, and E.D. Wachsman. Chemical Expansion: Implications
for Electrochemical Energy Storage and Conversion Devices. Annual Review
of Materials Research, 44(1):205–239, 2014.
[68] Keith L. Duncan, Yanli Wang, Sean R. Bishop, Fereshteh Ebrahimi, and Eric D.
Wachsman. Role of point defects in the physical properties of fluorite oxides.
Journal of the American Ceramic Society, 89(10):3162–3166, 2006.
[69] Yanli Wang, Keith Duncan, Eric D. Wachsman, and Fereshteh Ebrahimi.
The effect of oxygen vacancy concentration on the elastic modulus of fluorite-
structured oxides. Solid State Ionics, 178(1-2):53–58, 2007.
[70] David J. Green. An Introduction to the Mechanical Properties of Ceramics.
Cambridge University Press, 1998.
[71] Zhen-Yan Deng, Jihong She, Yoshiaki Inagaki, Jian-Feng Yang, Tatsuki Ohji,
and Yoshihisa Tanaka. Reinforcement by crack-tip blunting in porous ceramics.
Journal of the European Ceramic Society, 24(7):2055–2059, jun 2004.
142
[72] Milan Ambrožič and Tomaž Kosmač. Optimization of the bend strength of
flat-layered alumina-zirconia composites. Journal of the American Ceramic
Society, 90(5):1545–1550, may 2007.
[73] Akio Yonezu and Xi Chen. Micro-scale damage characterization in porous
ceramics by an acoustic emission technique. Ceramics International, 40(7
PART A):9859–9866, 2014.
[74] S. Meille, M. Lombardi, J. Chevalier, and L. Montanaro. Mechanical properties
of porous ceramics in compression: On the transition between elastic, brittle,
and cellular behavior. Journal of the European Ceramic Society, 32(15):3959–
3967, nov 2012.
[75] Pulin Nie, Heping Lv, Tao Zhou, Xun Cai, and Paul K. Chu. Interfacial
adhesion measurement of a ceramic coating on metal substrate. Journal of
Coatings Technology and Research, 7(3):391–398, may 2010.
[76] F. Vales, R. Rezakhanlou, and C. Olagnon. Determination of the fracture
mechanical parameters of porous ceramics from microstructure parameters
measured by quantitative image analysis. Journal of Materials Science, 4:4081–
4088, 1999.
[77] Bent F. Sørensen and Andy Horsewell. Crack Growth along Interfaces in Porous
Ceramic Layers. Journal of the American Ceramic Society, 84(9):2051–2059,
dec 2004.
[78] S. M. Barinov, V. F. Ponomarev, and V. Ya. Shevchenko. Effect of Hot
Isotatic Pressing on the Mechanical Properties of Aluminum Oxide Ceramics.
Refractories and Industrial Ceramics, 38(1-2):9–12, 1997.
[79] J. J. Roa, M. A. Laguna-Bercero, A. Larrea, V. M. Orera, and M. Segarra.
Mechanical properties of highly textured porous Ni-YSZ and Co-YSZ cer-
mets produced from directionally solidified eutectics. Ceramics International,
37(8):3123–3131, 2011.
[80] Arata Nakajo, Jakob Kuebler, Antonin Faes, Ulrich F. Vogt, Hans Jürgen
Schindler, Lieh Kwang Chiang, Stefano Modena, Jan Van Herle, and Thomas
Hocker. Compilation of mechanical properties for the structural analysis of
solid oxide fuel cell stacks. Constitutive materials of anode-supported cells.
Ceramics International, 38(5):3907–3927, 2012.
[81] K Yasuda, K Uemura, and T Shiota. Sintering and mechanical properties
of gadolinium-doped ceria ceramics. Journal of Physics: Conference Series,
339(1):012006, jan 2012.
[82] Sophie Giraud and Jérôme Canel. Young’s modulus of some SOFCs materials as
a function of temperature. Journal of the European Ceramic Society, 28(1):77–
83, 2008.
143
[83] Takuto Kushi, Kazuhisa Sato, Atsushi Unemoto, Shinichi Hashimoto, Koji
Amezawa, and Tatsuya Kawada. Elastic modulus and internal friction of SOFC
electrolytes at high temperatures under controlled atmospheres. Journal of
Power Sources, 196(19):7989–7993, 2011.
[84] Koji Amezawa, Takuto Kushi, Kazuhisa Sato, Atsushi Unemoto, Shin Ichi
Hashimoto, and Tatsuya Kawada. Elastic moduli of Ce0.9Gd0.1O2-δ at high
temperatures under controlled atmospheres. Solid State Ionics, 198(1):32–38,
2011.
[85] M. Jenkins, T. Chang, and A. Okura. A Simple Maching Jig for Chevron-
notched Specimens. Experimental Techniques, 12(8):20–22, 1988.
[86] ASTM Standard C1421. Standard Test Methods for Determination of Fracture
Toughness of Advanced Ceramics at Ambient Temperature. ASTM Interna-
tional, pages 1–33, 2016.
[87] ASTM Standard C1221. Standard Test Method for Flexural Strength of
Advanced Ceramics at Elevated Temperatures. ASTM International, 11(Reap-
proved 2008):1–16, 2013.
[88] Brian H Toby and Robert B. Von Dreele. GSAS-II: The genesis of a modern
open-source all purpose crystallography software package. Journal of Applied
Crystallography, 46(2):544–549, 2013.
[89] Koichi Momma and Fujio Izumi. VESTA 3 for three-dimensional visualization
of crystal, volumetric and morphology data. Journal of Applied Crystallography,
44(6):1272–1276, 2011.
[90] Masatsugu Oishi, Keiji Yashiro, Kazuhisa Sato, Junichiro Mizusaki, and Tat-
suya Kawada. Oxygen nonstoichiometry and defect structure analysis of B-site
mixed perovskite-type oxide (LaSr)(CrM)O3-δ (M=Ti, Mn and Fe). Journal
of Solid State Chemistry, 181(11):3177–3184, 2008.
[91] Junichiro Mizusaki, Naoya Mori, Hiroshi Takai, Yuki Yonemura, Hidehiro
Minamiue, Hiroaki Tagawa, Masayuki Dokiya, Hideaki Inaba, Kazunori Naraya,
Tadashi Sasamoto, and Takuya Hashimoto. Oxygen nonstoichiometry and
defect equilibrium in the perovskite-type oxides La1-xSrxMnO3+d. Solid State
Ionics, 129(1):163–177, 2000.
[92] Finn Willy Poulsen. Defect chemistry modelling of oxygen-stoichiometry,
vacancy concentrations, and conductivity of (La1-xSrx)yMnO3±δ. Solid State
Ionics, 129(1):145–162, 2000.
[93] Shang Xian Wu. Fracture toughness determination of bearing steel using
chevron-notch three point bend specimen. Engineering Fracture Mechanics,
19(2):221–232, 1984.
144
[94] Roy W. Rice. Porosity of Ceramics: Properties and Applications. Marcel
Dekker. Inc., New York, 1998.
[95] Rodger W. Davidge and Anthony G. Evans. The Strength of Ceramics. Mate-
rials Science and Engineering, (6):281–298, 1970.
[96] Barry Carter and Grant Norton. Ceramic Materials Science and Engineering.
2007.
[97] Oldřich Ševeček, Michal Kotoul, Dominique Leguillon, Eric Martin, and Raul
Bermejo. Assessment of crack-related problems in layered ceramics using
the finite fracture mechanics and coupled stress-energy criterion. Procedia
Structural Integrity, 2:2014–2021, 2016.
[98] James T Richardson, Robert Scates, and Martyn V Twigg. X-ray diffraction
study of nickel oxide reduction by hydrogen. Applied Catalysis A: General,
246:137–150, 2003.
[99] Clemens Ritter, Javier Blasco, Jose-Maria De Teresa, David Serrate, Luis
Morellon, Joaquin Garcia, and Manuel Ricardo Ibarra. Structural and magnetic
details of 3d-element doped Sr2Fe0.75T0.25MoO6, volume = 6, year = 2004. Solid
State Sciences, (5):419–431.
[100] D. S. Tsvetkov, I. L. Ivanov, D. A. Malyshkin, A. S. Steparuk, and A. Yu.
Zuev. The defect structure and chemical lattice strain of the double perovskites
Sr2BMoO6-δ (B = Mg, Fe). Dalton Transactions, 45(32):12906–12913, 2016.
[101] B. Levasseur and S. Kaliaguine. Effects of iron and cerium in
La1-yCeyCo1-xFexO3 perovskites as catalysts for VOC oxidation. Applied Catal-
ysis B: Environmental, 88(3-4):305–314, 2009.
[102] Qiang Liu, Daniel E. Bugaris, Guoliang Xiao, Maxwell Chmara, Shuguo
Ma, Hans Conrad Zur Loye, Michael D. Amiridis, and Fanglin Chen.
Sr2Fe1.5Mo0.5O6-δ as a regenerative anode for solid oxide fuel cells. Journal of
Power Sources, 196(22):9148–9153, 2011.
[103] S. Vasala, M. Lehtimäki, Y. H. Huang, H. Yamauchi, J. B. Goodenough, and
M. Karppinen. Degree of order and redox balance in B-site ordered double-
perovskite oxides, Sr2MMoO6-δ (M=Mg, Mn, Fe, Co, Ni, Zn). Journal of Solid
State Chemistry, 183(5):1007–1012, 2010.
[104] D. Marrero-López, J. Peña-Mart́ınez, J. C. Ruiz-Morales, M. Gabás, P. Núñez,
M. A G Aranda, and J. R. Ramos-Barrado. Redox behaviour, chemical
compatibility and electrochemical performance of Sr2MgMoO6-δ as SOFC
anode. Solid State Ionics, 180(40):1672–1682, 2010.
[105] R Kircheisen and J. Töpfer. Nonstoichiometry, point defects and magnetic
properties in Sr2FeMoO6-δ double perovskites. Journal of Solid State Chemistry,
185:76–81, 2012.
145
[106] Keith L. Duncan and Eric D. Wachsman. General model for the functional
dependence of defect concentration on oxygen potential in mixed conducting
oxides. Ionics, 13(3):127–140, 2007.
[107] A. Holt, T. Norby, and R. Glenne. Defects And Transport In SrFe1-xCoxO3.5.
Ionics, 5(5-6):434–443, 1999.
[108] S R Bishop, K L Duncan, and E D Wachsman. Surface and bulk oxygen
non-stoichiometry and bulk chemical expansion in gadolinium-doped cerium
oxide. Acta Materialia, 57(12):3596–3605, 2009.
[109] Patrick Stanley, A Mohammed Hussain, Yi-lin Huang, J Evans Gritton,
and Eric D Wachsman. Defect chemistry and oxygen non-stoichiometry in
SrFe0.2Co0.4Mo0.4O3-delta for solid-oxide fuel cells. pages 1–19.
[110] David Waldbillig, Anthony Wood, and Douglas G. Ivey. Thermal analysis of
the cyclic reduction and oxidation behaviour of SOFC anodes. Solid State
Ionics, 176(9-10):847–859, 2005.
[111] Robert Danzer, Tanja Lube, Peter Supancic, and Rajiv Damani. Fracture of
ceramics. Advanced Engineering Materials, 10(4):275–298, 2008.
[112] Carol A. Markowski and Edward P Markowski. Conditions for the Effectiveness
of a Preliminary Test of Variance. The American Statistician, 44(4):322, nov
1990.
[113] B. L. Welch. The Generalization of ‘Student’s’ Problem when Several Different
Population Variances are Involved. Biometrika, 34(1/2):28, 1947.
[114] David S. Moore, George P. McCable, and Bruce A. Craig. Introduction to the
practice of statistics. 2009.
[115] David Tanner. t for One, or Two. In Using Statistics to Make Educational
Decisions, pages 139–171. SAGE Publications, 2013.
[116] Robert L. Anderson. Practical Statistics for Analytical Chemists. Van Nostrand
Reinhold, 1987.
[117] W. Weibull. A statistical distribution function of wide applicability, 1951.
[118] George D. Quinn and Janet B. Quinn. A practical and systematic review of
Weibull statistics for reporting strengths of dental materials. Dental Materials,
26(2):135–147, feb 2010.
[119] ASTM Standard C1161. Standard Test Method for Flexural Strength of
Advanced Ceramics at Ambient Temperature. ASTM International, pages
1–19, 2018.
146
[120] George D Quinn. NIST Recommended Practice Guide Fractography of Ce-
ramics and Glasses. Technical report, National Institute of Standards and
Technology, Gaithersburg, MD, 2016.
147