close

Вход

Забыли?

вход по аккаунту

?

Proton-conducting beta"-alumina via microwave-assisted synthesis and mechanism of enhanced corrosion prevention of a zinc rich coating with electronic control

код для вставкиСкачать
PROTON CONDUCTING β’’-ALUMINA VIA MICROWAVE ASSISTED SYNTHESIS
AND
MECHANISM OF ENHANCED CORROSION PREVENTION OF
A ZINC RICH COATING WITH ELECTRONIC CONTROL
Brent William Kirby
A DISSERTATION
PRESENTED TO THE FACULTY
OF PRINCETON UNIVERSITY
IN CANDIDACY FOR THE DEGREE
OF DOCTOR OF PHILOSOPHY
RECOMMENDED FOR ACCEPTANCE
BY THE DEPARTMENT OF CHEMISTRY
Adviser: Andrew Bocarsly
JUNE 2008
UMI Number: 3308035
Copyright 2008 by
Kirby, Brent William
All rights reserved.
UMI Microform 3308035
Copyright 2008 by ProQuest Information and Learning Company.
All rights reserved. This microform edition is protected against
unauthorized copying under Title 17, United States Code.
ProQuest Information and Learning Company
300 North Zeeb Road
P.O. Box 1346
Ann Arbor, MI 48106-1346
© Copyright by Brent William Kirby, 2008.
All rights reserved.
Abstract
Proton Conducting β’’-alumina via Microwave Assisted Synthesis: The microwave
assisted synthesis of proton conducting Mg- and Li-stabilized NH4+/H3O+ β’’-alumina
from a solution based gel precursor is reported. β’’-alumina is a ceramic fast ion
conductor containing two-dimensional sheets of mobile cations. Na+-β’’-alumina is the
most stable at the sintering temperatures (1740°C) reached in a modified microwave
oven, and can be ion exchanged to the K+ form and then to the NH4+/H3O+ form. β-phase
impurity is found to be 20% for Mg-stabilized material and 30-40% for Li-stabilized
material. The composition of the proton conducting form produced here is deficient in
NH4+ as compared to the target composition (NH4)1.00(H3O)0.67Mg0.67Al10.33O17.
Average grain conductivity for Li-stabilized material at 150°C is 6.6x10-3 ±
1.6x10-3 S/cm with 0.29 ± 0.05 eV activation energy, in agreement with single crystal
studies in the literature. Grain boundary conductivity is found to be higher in the Listabilized material. A hydrogen bond energy hypothesis is presented to explain these
differences. Li-stabilized NH4+/H3O+ β’’-alumina is demonstrated as a fuel cell
electrolyte, producing 28 μA/cm2 of electrical current at 0.5 V.
Mechanism of Enhanced Corrosion Prevention of a Zinc Rich Coating with
Electronic Control: A corrosion inhibition system consisting of high weight-loading
zinc rich coating applied to steel panels is examined. An electronic control unit (ECU)
consisting of a battery and a large capacitor in series with the panel is shown to improve
corrosion protection upon immersion in 3% NaCl solution. Weekly solution changes to
avoid zinc saturation in solution system were necessary to see well differentiated results.
iii
The corrosion product, hydrozincite [Zn5(CO3)2(OH)6] is observed to deposit
within the pores of the coating and on the surface as a barrier layer. Simonkolleite
[Zn5(OH)8Cl2·H2O] is found to form in place of the original zinc particles. The barrier
layer is denser and more adherent with the ECU in place. A mechanism is proposed in
which the characteristic time constant of the ECU is roughly matched to the time scale of
ionic motion within the coating. The capacitive nature of the ECU retards the motion of
ions, and affects the formation of denser corrosion products.
iv
Acknowledgements
My adviser, Prof. Andy Bocarsly, has skillfully shepherded me through graduate
school. He has been patient in the face of (occasionally successful) attempts to prove him
wrong, exhortations to throw something, anything, away, and even musical satire. Thank
you, Andy, for a great experience.
Prof. Bob Cava was instrumental in my success through lending actual
instruments (LCR meter and optical pyrometer), insightful comments and enthusiastic
discussions, and friendship formed on the softball field. He also found the time to read
this dissertation despite duties as Department Chair. His students are busy, but always
friendly; a reflection of the Boss. Thank you.
Prof. Jay Benziger has made me think more carefully about fuel cells and has
been generous with his time. Prof. Steve Bernasek shows a keen interest in all things
scientific and has made me want to learn to fly airplanes. Thanks to you both for your
service on my committee.
I would like to acknowledge those who directly contributed to this work: Paul
Majsztrik designed and built the EIS test stand; Martina Vondrova took endless EIS
spectra with me; Tyrel McQueen took long-scan XRD data, and taught me to use the
LCR meter; Ellazar Niangar produced electrodes for the β’’-alumina fuel cell; Gloria
Gussie, Barbara Fornalska, and David Dowling provided corrosion data from New York.
Thank you to the New Jersey Commission on Science and Technology for funding, and
to everyone at Applied Semiconductor, Inc. for involving me in such an interesting
project.
v
Martina Vondrova let me experience the excitement of a new rock climber, and
was my science buddy. Paul “The Engineer” Majsztrik and Jonathan “Kludgemaster”
Mann were excellent companions in the lab, and graduated just in time to leave me alone
to write my dissertation. Carolyn Mordas, Lakshmi Krishnan, Tao Zhang, Christine
Burgess, Emily Barton, Kate Keets, Amanda Tricarico, Pearl Dickerson, Ely Niangar
made this whole ride enjoyable.
My parents, Barbara and Alan, always encouraged me to ask questions. Well, it’s
taken me this far, and I haven’t stopped yet. I appreciate all that you’ve done for me, now
more than ever.
My wife, Mary, has sustained me: emotionally, spiritually, and nutritionally. If I
have accomplished anything, it is because I have a partner who is committed with me to
my dreams. Ryan, my son, you have given me permission to be a kid again. You are a joy
beyond any other. And we’ll have a new playmate for you real soon.
vi
Table of Contents
ABSTRACT
.............................................................................................................................................. III ACKNOWLEDGEMENTS ........................................................................................................................... V LIST OF FIGURES ...................................................................................................................................... IX LIST OF TABLES ...................................................................................................................................... XV CHAPTER 1: INTRODUCTION TO SOLID STATE PROTON CONDUCTORS ................................... 16 1.1. Motivation ................................................................................................................................ 16 1.2. Fuel Cells .................................................................................................................................. 16 1.2.1. 1.2.2. 1.2.3. 1.2.4. 1.2.5. 1.3. Overview ............................................................................................................................... 16 PEMFC’s .............................................................................................................................. 19 High temperature PEM-type electrolytes ............................................................................. 20 SOFC’s ................................................................................................................................. 21 Solid-state proton conductors ............................................................................................... 21 β’’-Alumina .............................................................................................................................. 22 Na+ β’’-alumina .................................................................................................................... 22 Ion exchange ......................................................................................................................... 26 Proton conductivity ............................................................................................................... 26 Microwave assisted synthesis ............................................................................................... 28 Hydrogen bond energy hypothesis ........................................................................................31 1.3.1. 1.3.2. 1.3.3. 1.3.4. 1.3.5. CHAPTER 2: NH4+/H3O+ β’’-ALUMINA: MICROWAVE ASSISTED SYNTHESIS.............................. 33 2.1. Introduction .............................................................................................................................. 33 2.2. Experimental ............................................................................................................................. 34 2.2.1. 2.2.2. 2.2.3. 2.2.4. 2.2.5. 2.2.6. 2.2.7. 2.2.8. 2.3. Pechini process ..................................................................................................................... 34 Scale up to 50 gram quantities..............................................................................................37 Ball milling and uniaxial pressing ........................................................................................ 38 Reaction vessel...................................................................................................................... 38 Temperature measurement by optical pyrometry .................................................................40 Microwave processing ..........................................................................................................41 Ion exchange ......................................................................................................................... 42 Characterization ................................................................................................................... 43 Results ...................................................................................................................................... 44 2.3.1. 2.3.2. 2.3.3. 2.3.4. 2.3.5. 2.3.6. 2.3.7. 2.3.8. 2.3.9. 2.3.10. Mullite-like alumina precursor ............................................................................................. 44 Grinding and ball milling ..................................................................................................... 45 Optical pyrometry ................................................................................................................. 48 Microwave heating ............................................................................................................... 49 XRD analysis ........................................................................................................................ 52 Direct K+ synthesis ............................................................................................................... 53 High temperature potassium ion exchange ...........................................................................58 Molten ammonium nitrate exchange ..................................................................................... 60 Density measurements ..........................................................................................................60 TGA analysis......................................................................................................................... 63 2.4. Discussion................................................................................................................................. 65 2.5. Conclusions .............................................................................................................................. 68 vii
CHAPTER 3: NH4+/H3O+ β’’-ALUMINA: ELECTRONIC PROPERTIES ............................................... 70 3.1. Introduction .............................................................................................................................. 70 3.1.1. 3.1.2. 3.2. Model circuits ....................................................................................................................... 70 Conductivity .......................................................................................................................... 73 Experimental ............................................................................................................................. 74 3.2.1. 3.2.2. 3.3. Electrochemical impedance spectroscopy ............................................................................ 74 Fuel cell configuration.......................................................................................................... 76 Results and Discussion ............................................................................................................. 78 3.3.1. 3.3.2. 3.4. Conductivity measurements by EIS ....................................................................................... 78 Fuel cell configuration.......................................................................................................... 86 Conclusions .............................................................................................................................. 88 CHAPTER 4: INTRODUCTION TO ZINC RICH COATINGS ................................................................ 89 4.1. Motivation ................................................................................................................................ 89 4.2. Zinc Rich Coatings ................................................................................................................... 89 4.3. Electrochemical Impedance Spectroscopy ............................................................................... 90 4.4. Corrosion Products ................................................................................................................... 92 4.5. Electrochemical Noise .............................................................................................................. 95 CHAPTER 5: MECHANISM OF ENHANCED CORROSION PROTECTION OF A ZINC RICH
COATING WITH ELECTRONIC CONTROL .................................................................... 97 5.1. Introduction .............................................................................................................................. 97 5.2. Experimental ............................................................................................................................. 99 5.2.1. 5.2.2. 5.2.3. 5.2.4. 5.2.5. 5.3. Sample preparation............................................................................................................... 99 Testing protocol .................................................................................................................. 100 Characterization ................................................................................................................. 101 Electrochemical impedance spectroscopy .......................................................................... 103 Large volume study ............................................................................................................. 105 Results .................................................................................................................................... 106 5.3.1. 5.3.2. 5.3.3. 5.3.4. 5.3.5. 5.3.6. 5.3.7. 5.3.8. 5.3.9. 5.3.10. 5.4. 5.4.1. 5.4.2. 5.4.3. 5.5. Visual evidence ................................................................................................................... 106 Corrosion potential ............................................................................................................. 107 Aqueous zinc analysis ......................................................................................................... 107 Large volume study ............................................................................................................. 110 Microscopy and structural analysis .................................................................................... 113 Identification of corrosion products ................................................................................... 115 Porosity ............................................................................................................................... 118 Density and thickness of the barrier layer, Group F .......................................................... 119 Adhesion of the barrier layer .............................................................................................. 122 Impedance spectroscopy ..................................................................................................... 125 Discussion............................................................................................................................... 149 Zinc surface area ................................................................................................................ 149 Complex equilibrium........................................................................................................... 150 Porosity of the barrier layer and ion mobility .................................................................... 153 Conclusions ............................................................................................................................ 158 REFERENCES ............................................................................................................................................ 160 viii
List of Figures
Figure 1.1. PEM fuel cell schematic. ................................................................................ 18 Figure 1.2. Unit cell of (a) Na+ β’’-alumina and (b) Na+ β-alumina. ............................... 24 Figure 1.3. Na+ β’’ (a) and Na+ β (b) alumina conduction planes. ................................... 25 Figure 1.4. Proposed ordered structure of the conduction plane of the 60/40 composition
of ammonium/hydronium β’’-alumina, (NH4)1.0(H3O)0.67Mg0.67Al10.33O17. ..................... 28 Figure 1.5. Projection of four unit cells of m-alumina showing tethrahedral and
octahedral aluminum ions. ................................................................................................ 30 Figure 1.6. Schematic comparison showing similarities between the β-, β’’-, and
m-alumina structures. ........................................................................................................ 31 Figure 2.1. Final insulated reaction vessel design for microwave heating. ...................... 39 Figure 2.2. Optical pyrometry set-up for measuring high temperatures during microwave
heating. .............................................................................................................................. 41 Figure 2.3. XRD pattern of mullite-like alumina phase after calcining precursor to 900°C.
Expected pattern for m-Al2O3 is shown (PDF card #12-0539)......................................... 44 Figure 2.4. SEM micrograph of a dried gel after grinding with mortar and pestle. Many
5-10 μm particles are visible. ............................................................................................ 45 Figure 2.5. SEM micrograph of a dried gel after milling for four hours. Nearly all
particles are less than 2 μm in diameter. ........................................................................... 46 ix
Figure 2.6. The same sample from Figure 2.5 at reduced magnification showing
agglomeration of 1-2 μm primary particles into 20-50 μm agglomerates after four hours
of milling........................................................................................................................... 47 Figure 2.7. Calibration between emissivity, ε, values in microwave pyrometry setup. ... 49 Figure 2.8. Typical heating profile for microwave processing of a precursor pellet. ....... 51 Figure 2.9. Powder XRD spectrum of sintered Na+ β’’-alumina pellet............................ 54 Figure 2.10. Evolution of XRD pattern from m-alumina to Na+ β’’-alumina during
microwave heating. ........................................................................................................... 55 Figure 2.11. Powder XRD spectrum of microwave heated but unsintered direct-synthesis
K+ β’’-alumina. ................................................................................................................. 56 Figure 2.12. Powder XRD spectrum of microwave sintered direct-synthesis
K+ β’’-alumina. ................................................................................................................. 57 Figure 2.13. EDS spectrum of microwave-processed Na+ β’’-alumina showing peaks for
O, Na, Mg (stabilizing dopant), Al, and Ir (conductive sputter coated layer). Also shown
is the same material after vapor phase KCl exchange and sonication rinse, showing
replacement of Na by K. ................................................................................................... 59 Figure 2.14. Comparison of microwave processed sample after KCl vapor exchange to
commercial reference samples. ......................................................................................... 60 Figure 2.15. Density (as percentage of the theoretical Na+ β’’-alumina density: 2.86
g/mL) vs. shrinkage during sintering as measured by the decrease in diameter of the
pellet. ................................................................................................................................. 63 x
Figure 2.16. TGA scan of Mg-stabilized gel precursor with 20% excess Na. .................. 64 Figure 2.17. TGA scan of NH4+/H3O+ exchanged β’’-alumina sample (Mg stabilized,
20% excess Na in precursor) at 0.5°C/min. ...................................................................... 65 Figure 3.1. Model circuits and their corresponding Nyquist plots. .................................. 71 Figure 3.2. Schematic of H2/air NH4/H3O+ β’’-alumina fuel cell setup. .......................... 77 Figure 3.3. Nyquist plot (10 kHz to 10 MHz) of Mg-stabilized and Li-stabilized proton
conducting β’’-alumina, both at 100°C. ........................................................................... 79 Figure 3.4. Arrhenius plot for a Li-stabilized sample showing grain and grain boundary
conductivities. ................................................................................................................... 81 Figure 3.5. Evolution of Nyquist plots with increasing temperature for Li-stabilized
NH4+/H3O+ β’’-alumina from 22-150°C. .......................................................................... 83 Figure 3.6. Comparison of Li-stabilized β’’-alumina as measured by LCR meter and
lock-in amplifier at room temperature. ............................................................................. 84 Figure 3.7. Current-voltage response for NH4+/H3O+ β’’-alumina fuel cell operating with
H2 and ambient air at room temperature. .......................................................................... 87 Figure 4.1. Model circuit for ZRC corrosion proposed by Xie, et al.61,62. ........................ 90 Figure 4.2. Model circuit for polymer coatings proposed by Mansfeld63. ........................ 91 Figure 4.3. Transmission line model circuit proposed by Feliu et al.55. ........................... 91 Figure 4.4. Two views of simonkolleite from Hawthorne68. ............................................ 93 xi
Figure 4.5. Oblique view of the simonkolleite structure from Hawthorne68 showing the
conformation of adjacent sheets and the arrangement of the interstitial H2O groups. ..... 94 Figure 4.6. A view from Ghose69 of a single oxy-hydroxy-zinc sheet of hydrozincite,
viewed along [100]. .......................................................................................................... 95 Figure 5.1. Schematic of the ECU connected to a ZRC coated steel panel. ..................... 97 Figure 5.2. Schematic of setup and wiring of control and ECU panels. ......................... 100 Figure 5.3. Schematic of electrochemical contact cell for weekly EIS measurements on
Groups G and H. ............................................................................................................. 104 Figure 5.4. Photographs of F Group after 161 days of immersion. Visual appearance
remained similar until the end of the study. .................................................................... 106 Figure 5.5. Average corrosion potentials for Groups A, B, and F. ................................. 107 Figure 5.6. Zinc concentration averages for the first 140 days of immersion, measured by
ICP. ................................................................................................................................. 109 Figure 5.7. Detail of later zinc concentrations for Group F. ........................................... 110 Figure 5.8. Large volume (100 L) experiment to monitor zinc concentration. Solution was
buffered to pH 8.2 with 44 mM borate/boric acid. ......................................................... 111 Figure 5.9. Photographs of the panels in the large volume experiment. ......................... 112 Figure 5.10. Solubility of zinc species with respect to pH. 81. ........................................ 112 Figure 5.11. Cross-sectional SEM image of a virgin panel showing spherical zinc
particles within the aluminosilicate matrix on the underlying steel substrate. ............... 114 Figure 5.12. ECU panel (A5) after four months of corrosion. ........................................ 115 xii
Figure 5.13. Cross-sectional SEM image of an ECU panel showing simonkolleite
corrosion product in place of completely and partially corroded zinc particles. ............ 116 Figure 5.14. XRD of panels F4 (Control) and F5 (ECU). .............................................. 117 Figure 5.15. Control panel, Group A, after one month of corrosion. ............................. 118 Figure 5.16. Control plate after one year of corrosion. Line scan showing x-ray counts for
each element, terminating in a pure zinc particle. .......................................................... 120 Figure 5.17. ECU plate after one year of corrosion. Line scan showing x-ray counts for
each element, terminating in a pure zinc particle. .......................................................... 121 Figure 5.18. Cross-sectional SEM images of control panels from Group F at 1600x
magnification. ................................................................................................................. 123 Figure 5.19. Cross-sectional SEM images of ECU panels from Group F at 1600x
magnification. ................................................................................................................. 124 Figure 5.20. Typical Nyquist plot (circles) for early immersion times (13 days). ......... 125 Figure 5.21. Typical Nyquist plot (circles) for mid-range immersion times (83 days). . 126 Figure 5.22. Typical Nyquist plot (circles) for long immersion times (159 days). ........ 126 Figure 5.23. Model circuit used for EIS analysis of zinc rich coating on steel. ............. 127 Figure 5.24. Nyquist plots for Group F after one year of immersion. ............................ 129 Figure 5.25. Photographs of a panel after six months of weekly EIS testing. ................ 130 Figure 5.26. Inductance (L) results, Group G. ................................................................ 131 Figure 5.27. Resistance (R0) results, Group G. .............................................................. 132 xiii
Figure 5.28. Resistance (R1) results, Group G. .............................................................. 133 Figure 5.29. Capacitance (Y1) results, Group G. ........................................................... 134 Figure 5.30. Exponent (n1) results, Group G.................................................................. 135 Figure 5.31. Resistance (R2) results, Group G. .............................................................. 136 Figure 5.32. Capacitance (Y2) results, Group G. ........................................................... 137 Figure 5.33. Exponent (n2) results, Group G.................................................................. 138 Figure 5.34. Warburg coefficient (W) results, Group G. ................................................ 139 Figure 5.35. Inductance (L) results, Group H. ................................................................ 140 Figure 5.36. Resistance (R0) results, Group H. .............................................................. 141 Figure 5.37. Resistance (R1) results, Group H. .............................................................. 142 Figure 5.38. Capacitance (Y1) results, Group H. ........................................................... 143 Figure 5.39. Exponent (n1) results, Group H.................................................................. 144 Figure 5.40. Resistance (R2) results, Group H. .............................................................. 145 Figure 5.41. Capacitance (Y2) results, Group H. ........................................................... 146 Figure 5.42. Exponent (n2) results, Group H.................................................................. 147 Figure 5.43. Warburg coefficient (W) results, Group H. ................................................ 148 Figure 5.44. Two-dimensional preponderance diagram for (Zn2+ + H2O + CO2) at 25°C
and Zn2+ activity of 10-4 with partial pressure of CO2 plotted against pH.84 .................. 152 Figure 5.45. Model circuit for the corrosion process within a ZRC protected by an ECU.
......................................................................................................................................... 157 xiv
List of Tables
Table 1.1. Comparison of β- and β’’-alumina structural properties. ................................ 25 Table 2.1. Mass percentages of water and ammonia observed from several β’’-alumina
samples. Expected mass percentages from the 60/40 composition are given for
comparison, for both Li and Mg stabilized material. See above for discussion of
composition calculations. .................................................................................................. 68 Table 3.1. Conductivity and activation energies for grains and grain boundaries in proton
conducting β’’-aluminas. Grain conductivity could not be determined in Mg-stabilized
samples due to the overlap in frequency response of grains and grain boundaries. ......... 85 Table 5.1. Barrier layer thickness (with standard deviations) measured by SEM. ......... 122 xv
Chapter 1: Introduction to Solid State Proton Conductors
1.1.
Motivation
The energy landscape around the world continues to shift with changing resource
availability and increasing populations. While estimates vary as to the future of
petroleum, coal, and other fossil fuel resources, we can be certain that we will eventually
have to seek new sources of energy as well as become more efficient in how we use the
energy available on this planet. Fuel cells offer an efficient way to store and produce
electrical energy from many sources. Any shift in energy usage will ultimately be driven
by economic factors. The sooner fuel cells are cost-competitive with more traditional
power sources, the sooner the transition can take place.
This work focuses on improving the performance and decreasing the cost of one
class of fuel cells, proton exchange membrane fuel cells (PEMFC’s), by seeking a solidstate alternative to current polymer-based technology for the proton exchange membrane.
Performance and efficiency can be greatly enhanced by increasing the operating
temperature of PEMFC’s to 150-200°C. A highly proton conductive, solid-state material
at these elevated temperatures could go a long way toward putting PEMFC’s on the
public energy landscape.
1.2.
Fuel Cells
1.2.1. Overview
A fuel cell is an electrochemical device that produces an electric current through
an external circuit by the reaction of a fuel at the anode with an oxidant at the cathode,
separated by an ion-conducting electrolyte. They are differentiated from batteries by their
16
supply of reactants. In a battery, all chemical components are self-contained; once the
chemical reactions reach completion, the battery must either be discarded or recharged. A
fuel cell is externally supplied with fuel and oxidant; as long as reactants are supplied,
electricity is generated. A fuel cell can quickly be refueled rather than recharged to
continue producing electricity.
A schematic of an H2/O2 PEMFC is shown in Figure 1.1. Hydrogen gas is passed
into the anode chamber where it reacts at the catalyst surface, typically nanoparticulate
platinum supported on larger carbon particles. Hydrogen gas is split into protons which
pass through the electrolyte membrane, and electrons which do useful work in the
external circuit. Protons, electrons, and oxygen gas combine at the cathode catalyst to
form water, which provides the chemical driving force for the process. Thus, the
chemical energy in the fuel is directly converted to electrical energy, rather than heat in
the case of direct combustion. This also means that fuel cells are not subject to the
fundamental Carnot efficiency limit for heat engines.
17
Electric Current
C
Pt
4H+ + 4e- + O2
Anode
2H2O
Cathode
H2
O2
Catalyst Layers
Pt
H2
2H+
C
+
Electrolyte Membrane
2e-
Figure 1.1. PEM fuel cell schematic. Hydrogen gas reacts at the anode catalyst to form
protons which pass through the electrolyte membrane, and electrons which do useful work in the
external circuit. Protons, electrons, and oxygen gas combine at the cathode catalyst to form water,
which provides the chemical driving force for the process.
Hydrogen is the fuel most commonly used in fuel cells today and oxygen, either
pure or in air, the most common oxidant. Many other fuels can be used for various
applications such as methanol1, ethanol2, formic acid3, glucose4, hydrocarbons5, and
others.
Different common types of fuel cells are referred to by their electrolyte. They
conduct different ions at widely varying temperatures: polymer electrolyte membrane
fuel cells (PEMFC’s) transport H+ or H3O+ at 50-100°C, though newer materials offer
higher operating temperature; alkaline fuel cells (AFC’s) transport OH- at 90-100°C;
18
phosphoric acid fuel cells (PAFC’s) transport H+ or H3O+ at 150-200°C; molten
carbonate fuel cells (MCFC’s) transport CO32- at 600-700°C; solid oxide fuel cells
(SOFC’s) transport O2- at 650-1000°C. Each type of fuel cell is suited to separate
applications based on scale and duty cycle considerations.
There are many critically important aspects of fuel cell design such as the catalyst
and support, fuel and oxidant flow patterns, water management, and thermal
management. This work will focus on the electrolyte material for PEM-type fuel cells.
1.2.2. PEMFC’s
The most common electrolyte for PEMFC’s is DuPont’s Nafion, a perfluorinated
ionomer consisting of a polytetrafluoroethylene backbone with perfluorinated side chains
terminated by a sulfonic acid group. Proton conduction occurs through connected clusters
of hydrated sulfonic acid groups, though the exact morphology is still debated6. Since
Nafion only conducts protons in its hydrated state, water management is an important
issue for PEMFC operation. Humidification is required, especially at temperatures above
80°C, adding complexity to the system. Also, Nafion undergoes a glass transition at
temperatures slightly higher than 100°C6, reducing its performance dramatically.
Start up and shut down of PEMFC’s is fast due to the relatively low operating
temperature (50-100°C), making them well suited for transportation and portable
applications (especially with methanol fuel). However, low operating temperature also
has several drawbacks. Cooling the cell is difficult due to the small thermal gradient with
the environment. This also results in low quality waste heat for combined heat and power
(CHP) applications. CO tolerance is low, an important factor since most hydrogen
produced today is converted from fossil fuel sources and contains some CO that must be
19
removed at additional cost. CO competes with H2 for adsorption sites on the platinum
catalyst surface. Product water in liquid form can also cause flooding problems at the
cathode. All these issues can be mitigated by elevating the operating temperature of
PEMFC’s to 150-200°C.
1.2.3. High temperature PEM-type electrolytes
The advantages of improved cooling, high quality waste heat, CO tolerance,
simplified water management and improved electrode kinetics have driven efforts to
elevate the operating temperature of PEMFC’s. Nafion itself is not suitable for operation
above 100°C due to dehydration and mechanical instability. Other electrolyte materials
must be found.
Composite Nafion and metal oxide membranes have been shown by the Bocarsly
group7 and others8,9 to improve the mechanical and conductive properties of Nafion at
elevated temperature and reduced relative humidity. Polybenzimidazole (PBI)
membranes impregnated with phosphoric acid have shown good results at elevated
temperature10,11 and are commercially available (Celtec-P, BASF Fuel Cells, formerly
PEMEAS). Celtec-P based cells operating at 160°C with H2 and O2 produce a current
density of 800 mA/cm2 at 0.5 V.
Degredation and failure modes for polymer membranes include pinhole
formation, membrane rupture, and catalyst dissolution into the polymer at high operating
potential12. A robust solid-state material would avoid these complications.
20
1.2.4. SOFC’s
SOFC’s are the highest temperature fuel cells, operating between 650 and
1000°C. Yttria-stabilized zirconia is used as the oxide ion conductor, and hydration is
clearly not required. Noble catalysts are not required at these high temperatures, and a
variety of fuels can be easily oxidized, including hydrocarbons like propane, or even
gasoline5. SOFC’s offer high quality waste heat that can be utilized for highly efficient
CHP operation.
High temperatures also lead to the most serious drawbacks for SOFC’s. The
thermal expansion coefficients for the anode, electrolyte, and cathode must be closely
matched to avoid cracking during thermal cycling, limiting the materials that can be used.
Long startup times to reach operating temperature limit SOFC’s to applications where
continuous operation is expected. This work seeks to develop a solid-state proton
conducting electrolyte that can combine the operating flexibility of PEMFC’s together
with the anhydrous nature and physical robustness of SOFC’s.
1.2.5. Solid-state proton conductors
There are a wide variety of proton conductors that can function at various
temperatures13, but very few have the conductivity necessary to be utilized in an elevated
temperature fuel cell. Nafion has a conductivity of 0.1 S/cm (where S stands for
Siemen ≡ Ω-1) at 100°C, fully humidified. This is an approximate target value for any
protonic electrolyte.
Superprotonic solid acids are one material that can compete with Nafion.
CsHSO4, for example, has a superprotonic phase transition at 141°C above which
temperature the conductivity can reach 10-2 S/cm 14. A fuel cell produced using this
21
material produced a current density at 160°C of 20 mA/cm2 at 0.5 V. Challenges with this
material include solubility in liquid water, ductility at high temperatures, and reduction to
sulfur after prolonged exposure to the fuel cell environment. Progress has been made in a
commercial effort to engineer solid acid fuel cells by Superprotonic, Inc. who now report
600 mA/cm2 at 0.5 V.
Another promising material is protonic β’’-alumina, a ceramic material with twodimensional conduction planes that can accommodate a variety of cations. The highest
conductivities have been observed for NH4+/H3O+ β’’-alumina. Single crystal studies
have shown in-plane conductivity as high at 3x10-2 S/cm 15. NH4+/H3O+ β’’-alumina
forms the basis of this work and is elaborated upon below.
1.3.
β’’-Alumina
1.3.1. Na+ β’’-alumina
The family of β- and β’’-aluminas has been extensively studied, primarily
Na+ β’’-alumina used in high energy density Na/S batteries16,17. The structures consist of
layers of spinel blocks with intervening conducting planes of cations, usually singly
charged (Figure 1.2) 20. These cations are highly mobile and can be ion-exchanged to
produce various conductive materials. The β- and β’’ forms are closely related, though
the β’’-form exhibits higher conductivity by approximately one order of magnitude for
various intercalated cations18,19.
Figure 1.3 shows the different arrangement of cations in the β and β’’ structures21.
In the β structure, the conduction plane is a mirror plane and the oxygen layers are
directly on top of one another. A cation in a trigonal hole in both the upper and lower
22
spinel blocks in the β structure is said to occupy a Beevers-Ross site (BR, labeled 1). The
anti-Beevers-Ross site (aBR, labeled 3) creates the smallest constriction in the conduction
plane at 2Å. The site between two bridging oxygens is called the mid-oxygen site (mO,
labeled 2). Conduction in β-alumina occurs by cation motion around the bridging oxygen
atoms from sites 1-2-3-2-1.
The conduction plane is not a mirror plane in the β’’ structure; spinel oxygens
layers are offset creating a more open structure. There are two equivalent BR-type sites in
β’’-alumina (labeled 1 and 3), increasing the narrowest part of the conduction plane to
3Å. Conduction again proceeds from sites 1-2-3-2-1.
There are more Na+ ions than BR sites in β-alumina, forcing some Na+ to pair up
and straddle a BR site. These extra Na+ ions are charge balanced by interstitial oxygen
defects in the conduction plane. In β’’-alumina, there are more BR-type sites than
Na+ ions, and the conduction proceeds via a vacancy mechanism. Excess Na+ is charge
balanced by substitution of some Al3+ by Mg2+. This allows for the non-stoichiometric
excess of Na+ to be maintained in the β’’ form without blocking the conduction plane
with interstitial oxygen. It is this fact, combined with the removal of the narrow aBR site
that accounts for the higher conductivity of β’’-aluminas. Substitutional stabilization by
Mg2+ results in a formula of (NH4)1.67-y(H3O)yMg0.67Al10.33O17 20. Li+ stabilization is also
possible resulting in a composition of (NH4)1.67-y(H3O)yLi0.33Al10.67O17. Table 1.1
summarizes the properties of β- vs. β’’-alumina.
23
FIGURE 1.2. Unit cell of (a) Na+ β’’-alumina and (b) Na+ β-alumina. Alternating with closely
packed spinel blocks are the more loosely packed conduction planes20. Reprinted with permission of
the Materials Research Society.
24
Oxygen, upper
close-packed plane
Oxygen, lower
close-packed plane
Oxygen, conduction
plane
Na+ ion
Figure 1.3. Na+ β’’ (a) and Na+ β (b) alumina conduction planes. Sodium ions diffuse along a
zigzag 1-2-3-2-1 pathway21. Reprinted with permission from AAAS.
Table 1.1. Comparison of β- and β’’-alumina structural properties.
β-alumina
β’’-alumina
Na1+xMgxAl11-xO17 or
Composition
Na1+xAl11O17+x/2
Na1+xLix/2Al11-x/2O17
Distinct Na+ Sites
Three: BR, aBR, mO
Two: BR-type, mO
Na+ Description
Interstitial Pairs
Na+ vacancies
Charge Balance
Interstitial O2-
Mg2+ or Li+ at Al3+ site
Narrowest Gap in Plane
2Å
3Å
25
1.3.2. Ion exchange
Na+ β’’-alumina can be ion exchanged with H3O+ and/or NH4+ to produce a
proton conducting material. Of these NH4+/H3O+ β’’-alumina has the highest observed
proton conductivity15 and was therefore chosen as the basis of this work.
Ion exchange from polycrystalline Na+ to NH4+/H3O+ β’’-alumina is
accomplished in two steps. First, a high temperature exchange in KCl vapor is required to
insert the larger K+ ions without catastrophic ion exchange stresses along the crystal’s
c-axis22. Proton conducting material is produced by ion exchange of K+ β’’-alumina in
molten ammonium nitrate. K+ and NH4+ ions have similar radii allowing the exchange to
occur. It was found that it was not possible to produce a sample of pure
NH4+ β’’-alumina, although NH4+ β-alumina (no primes) is known23. The hydronium ions
in the β’’-form are most likely produced by the decomposition of ammonium nitrate in
the melt15.
NH4NO3 → N2O + 2H2O
EQUATION 1.1
NH4+ + H2O → NH3 + H3O+
EQUATION 1.2
1.3.3. Proton conductivity
The room temperature proton conductivity of single crystal NH4+/H3O+
β’’-alumina stated in the literature varies from 3x10-5 to 1x10-3 S/cm 15,21,24-26. This
spread in values is primarily due to compositional variation. Conduction follows
Arrhenius behavior with quoted activation energies ranging from 0.24 to 0.31 eV. This
corresponds to conductivities at 150°C from 5x10-4 to 3x10-2 S/cm, making this material
a viable option for fuel cell applications.
26
The mechanism of proton transport has been studied, but has not been
conclusively determined. The proton conductivity of NH4+/H3O+ β’’-alumina is in large
part due to a Grotthus mechanism via the hydrogen bonded network of NH4+ and H3O+,
as shown by Thomas and Farrington27. NH4+ groups have one hydrogen bond along the
c-axis to the adjacent spinel block around which they can rotate. Some proton sites are
available due to a small amount of neutral NH3 and H2O in the structure. Rotations allow
hydrogen bonds to form between NH4+ and H2O or between NH3 and H3O+ and can result
in a proton hop from the donor to the acceptor. The extent that NH4+ or H3O+ diffuse
through the solid (a vehicle mechanism for proton transport), and over what ranges of
composition, is unclear. Dennuzio and Farrington15 suggest a maximal hydrogen bonded
network for the specific composition (NH4)1.0(H3O)0.67Mg0.67Al10.33O17 that is depicted in
Figure 1.4. This is referred to as the 60/40 composition and has the highest measured
proton conductivity of all the protonated β’’-aluminas.
27
Column Oxygen
Hydronium Ion
Ammonium Ion
Vacancy
Hydrogen Bond
Figure 1.4. Proposed ordered structure of the conduction plane of the 60/40 composition of
ammonium/hydronium β’’-alumina, (NH4)1.0(H3O)0.67Mg0.67Al10.33O17. The hydronium ions form
hydrogen bonds (solid lines) with the column oxygens, and the vacancies are stabilized by the
ammonium ions15.
1.3.4. Microwave assisted synthesis
The original Ford synthesis28 of β’’-alumina involves grinding the raw materials
α-alumina, magnesium oxide, and sodium oxide together before calcining them in two
stages. The zeta-process28 produced Li-stabilized material with the added refinement of
prereacting the lithium nitrate or lithium oxalate with alumina at 1250°C. These
processes are time and energy intensive. Several researchers have developed alternative
methods to mix precursors in solution form and then process them either by spray
drying29 or gel formation30-33. In particular Subasri, Mathews, et. al32 converted a
precursor gel to sintered Mg-stabilized β’’-alumina in one step using microwave
radiation.
28
Subasri reports using the Pechini process to produce a precursor gel from sodium
nitrate, magnesium acetate, and aluminum nitrate complexed with citric acid. Upon heat
treatment to 750°C, a crystalline phase forms that is characterized as mullite-like
alumina, or simply m-alumina. Further heat treatment to 1100-1200°C results in
β’’-alumina formation.
Crystallization of m-alumina was observed to occur over the range of
750-950°C34. Both Elliot and Huggins35 as well as Mazza and Vallino36 report that the
m-alumina phase is only observed when the components were intimately mixed on the
atomic scale, for example quenching from a melt,37 using the Pechini method outlined
here, or using a similar solution based method.
Mullite itself is an aluminosilicate compound with the formula 3Al2O3·2SiO2.
Replacing all the Si4+ in the structure with Al3+ results in a charge imbalance that can be
compensated either by oxygen vacancies or by the addition of Na+ or K+, which avoids
disturbing the oxygen distribution. These two cases are shown below in Kröger-Vink
notation:
′
••
EQUATION 1.3
or ′
•
EQUATION 1.4 The m-alumina phase (Figure 1.5) is very similar in structure to the desired
β’’-alumina phase. Where β’’-alumina contains blocks of four close packed oxygen
layers separated by a conduction plane of bridging oxygen and mobile sodium ions,
m-alumina alternates single close packed planes of oxygen ions with planes of bridging
29
oxygen and sodium ions. A schematic comparison of β-, β’’-, and m-alumina is shown in
Figure 1.6. Conversion from m-alumina to β’’-alumina only requires migration of the
requisite sodium ions and bridging oxygens across one or two close packed planes. This
accounts for the observed rapid phase conversion at relatively modest temperatures
(1100-1200°C) in this system.
Figure 1.5. Projection of four unit cells of m-alumina showing tethrahedral and octahedral
aluminum ions. Large dark circles, O2-; small dark circles, Na+; small light circles, Al3+.33 Reprinted
with permission from Wiley-Blackwell.
30
Figure 1.6. Schematic comparison showing similarities between the β-, β’’-, and m-alumina
structures. Na+ ions need only to migrate across one or two close packed planes of oxygen to convert
m-alumina to β’’-alumina.34 Reprinted with permission from Elsevier.
1.3.5. Hydrogen bond energy hypothesis
Given the proton conduction mechanism of NH4+/H3O+ β’’-alumina relies on the
making and breaking of hydrogen bonds in the conduction plane, it was hypothesized that
by altering the strength of these hydrogen bonds, the proton conductivity would be
affected. Subtle changes in the energetics of rotation and proton hopping could have a
large effect on conductivity, and it is unlikely that the pure Mg-stabilized material
possesses the ideal energetics for maximum conductivity.
31
The current work seeks to demonstrate a difference in proton conductivity
between Li-stabilized and Mg-stabilized NH4+/H3O+ β’’-alumina. This is the largest
change that can be made to the electronic structure of the spinel block without disturbing
the β’’ structure. Lithium is more electronegative than magnesium which would tend to
draw electron density away from the axial hydrogen bonds to the NH4+ groups in the
conduction plane. This in turn may affect the in plane hydrogen bonding. While the exact
mechanism of conduction remains unclear, the altering of hydrogen bond energies must
certainly impact conduction in some way. If an effect can be proven, more subtle or
mixed doping could be explored to fine tune conductivity including with other
appropriately sized ions (critical radius: 0.97 nm)38 to remain in the spinel block such as:
Ni2+, Co2+, Cu2+, Zn2+, Mn2+, and Cd2+.19
The solution based microwave synthesis offers a relatively facile method to
pursue doped structures. Composition can be accurately and easily controlled by varying
solution concentrations, and the heating period is short, reducing the possibility of loss of
volatile components at sintering temperatures above 1600°C.
32
Chapter 2: NH4+/H3O+ β’’-Alumina: Microwave Assisted Synthesis
2.1.
Introduction
The family of β-aluminas is a well-studied class of materials with structures
consisting of layers of spinel blocks with intervening conducting planes of cations,
usually singly charged. These cations can vary widely, including proton conducting
forms. NH4+/H3O+ β’’-alumina is the most highly conductive of the proton conducting
β’’-aluminas15. The room temperature proton conductivity of single crystal NH4+/H3O+
β’’-alumina stated in the literature varies from 3x10-5 to 1x10-3 S/cm 15,21,24-26, with
activation energies ranging from 0.24 to 0.31 eV. This corresponds to conductivities at
150°C from 5x10-4 to 3x10-2 S/cm, making this material a viable option for elevated
temperature PEM-type fuel cell applications.
The β’’ form is more conductive then the β form due to a more open conduction
plane. β’’ is preferentially stabilized by replacing some Al3+ in the conduction plane with
Mg2+ or Li+ to account for the non-stoichiometric excess of Na+. This results in the
following formulas: (NH4)1.67-y(H3O)yMg0.67Al10.33O17 or
(NH4)1.67-y(H3O)yLi0.33Al10.67O17.
To avoid the high temperatures and long times associated with traditional
synthesis of β’’-alumina from powder precursors28, several solution based methods have
been developed29-33. In particular Subasri, Mathews, et. al32 converted a precursor gel to
sintered Mg-stabilized β’’-alumina in one step using microwave radiation.
33
This approach was developed and modified to test the hypothesis that Mgstabilized and Li-stabilized NH4+/H3O+ β’’-alumina would show different proton
conductivites. Lithium is more electronegative than magnesium which would tend to
draw electron density away from the axial hydrogen bonds to the NH4+ groups in the
conduction plane. This in turn may affect the in plane hydrogen bonding, and therefore
the proton conductivity. Small sintered pellets of both the Li- and Mg-stabilized material
were synthesized. Electronic properties are reported in Chapter 3.
2.2.
Experimental
2.2.1. Pechini process
The Pechini process39 involves coordinating metal salts with citric acid to achieve
intimate mixing on the molecular level. This method was employed by Subasri, et al.32
during a microwave assisted synthesis of Mg-stabilized Na+ β’’-alumina. Aqueous
solutions of metal salts were prepared to target a final composition of
Na1.67Mg0.67Al10.33O17.
Reagents used were as follows: sodium nitrate, NaNO3 (Sigma-Aldrich);
magnesium acetate tetrahydrate, (C2H3O2)Mg·4H2O (Sigma-Aldrich, 98+%); aluminum
nitrate nonahydrate, Al(NO3)3·9H2O (Sigma-Aldrich, 98+%); citric acid (Sigma-Aldrich,
99.5+%); ethylene glycol, HOCH2CH2OH (J.T. Baker, 100.0%). Later experiments with
lithium stabilization also utilized lithium nitrate, LiNO3 (Fluka).
To produce 5 g of the final ceramic product, Na1.67Mg0.67Al10.33O17, the following
procedure was initially used. 1.172 g sodium nitrate, 1.187 g magnesium acetate
34
tetrahydrate, and 32.01 g aluminum nitrate nonahydrate were dissolved in 40 mL
deionized water. The solution was stirred in a 250 mL beaker and heated on a hot plate
until approximately 15 mL of water were evaporated off. Boiling was avoided. The
excess water served to allow easy dissolution of the large quantity of salts and insure
uniform mixing. The goal was to have a minimum amount of water present to facilitate
drying of the resultant gel. 20.10 g of citric acid were used for this composition.
In other preparations, quantities were adjusted to produce solutions with a 20%
stoichiometric excess of sodium based on a private communication from Mathews40. He
stated that excess sodium allowed for easier microwave heating of the precursor and that
the stoichiometric excess would be volatilized away during sintering. To produce 5 g of
the final product, with 20% excess sodium, the following quantities were used: 1.407 g
sodium nitrate, 1.187 g magnesium acetate tetrahydrate, 32.01 g aluminum nitrate
nonahydrate, and 20.64 g citric acid.
Lithium stabilized precursors were also produced in a similar fashion using
lithium nitrate in place of magnesium acetate tetrahydrate. Half the moles of Li+ are
required than for Mg2+ since the substitution of one Li+ for an Al3+ results in the
incorporation of two Na+ ions in the structure to maintain charge neutrality. The resulting
quantities for 5 g of lithium-stabilized material with 20% stoichiometric excess sodium
were: 1.371 g sodium nitrate, 0.196 g lithium nitrate, 33.38 g aluminum nitrate
nonahydrate, and 20.74 g citric acid.
Meanwhile, a quantity of citric acid was warmed on a heating mantle with 10 mL
of DI water to near boiling in a 250 mL round bottom flask equipped with a condenser.
The amount of citric acid was chosen so that there was a 1:1 mole ratio between citric
35
acid and the sum of the three metal ions (Na+, Mg2+, Al3+). This small amount of water
was added to allow for easier heating and faster mixing once the salt solution was added.
The citric acid did not all dissolve and was effectively a slurry.
After the salt solution had been sufficiently concentrated, it was poured into the
round bottom flask containing the citric acid. The combined solution, now containing a
total of 35 mL of water, was heated to boiling over approximately 20 minutes and
allowed to reflux for 10 minutes. During heating the solution became yellowish. After 10
minutes of refluxing the solution was removed from the heat and 5 mL of ethylene glycol
were added to thicken the solution. Stirring was continued for 5 minutes.
If the salt solution was allowed to become more concentrated (for example
evaporating away 20 mL of water rather than the usual 15 mL), the reaction would
proceed more quickly once the salt solution was added to the citric acid, generating large
quantities of NO2 gas during the refluxing period. Vigorous bubbling was observed as
brown gas spewed out of the condenser. The precursor material behaved in all other ways
identically to that prepared by the normal procedure, resulting in the same β’’-alumina
final product. Warning: NO2 is a toxic gas.
Half the solution was poured into each of two glass boats constructed of a bottle
cut in half lengthwise (so as to fit well in a tube furnace for later heating), and kept in a
drying oven at 60-80°C. Over the course of two to three days, the solution released a
brown gas, assumed to be nitrogen dioxide, NO2. The release of gas typically started a
few hours after the solution was placed in the oven. The solution appeared to be boiling
with brown bubbles. As the reaction continued and water was removed the solution, a gel
36
was formed. Further drying resulted in a sticky white gel with large trapped bubbles of
NO2 inside.
Once the foaming process slowed down at 60-80°C (two to three days), the gel
was placed in a tube furnace and heated from 80 to 250°C at 1°C per minute. During
heating, the gel expanded to about ten times the volume of the original solution and
became pure white, releasing trapped NO2 gas. At 250°C the white gel became brown; a
varying toasted shade due to the initiation of pyrolysis of the organic components. This
brown gel was very dry, light, and crispy, looking something like a baked chocolate cake.
The brown dried gel was crushed using a mortar and pestle to reduce its volume
and then placed in a large alumina boat for heat treatment in a tube furnace. Temperature
was initially set to 250°C and ramped to 900°C at 1°C per minute, held for two hours,
and then cooled at 5°C per minute. During the heating process dark smoke was observed
exiting the tube furnace as combustion progressed. The resulting calcined powder was
pure white, loosely packed, and significantly reduced in volume.
2.2.2. Scale up to 50 gram quantities
Larger quantities were produced so that batch-to-batch precursor variability could
be removed. Quantities given earlier were multiplied by ten to give a final yield of 50 g
β’’-alumina. Salts were stirred and heated in a 2 L beaker, and refluxing was performed
in a 1 L, three-neck round bottom flask with a condenser on the central neck. The other
two necks were stoppered except during addition of salt solution or ethylene glycol. After
addition of ethylene glycol, the solution was divided between two, 2 L beakers and
placed in an oven at 60-80°C.
37
The reaction proceeded in similar fashion as with the 5 g quantities. Large
quantities of NO2 were released and the viscous solution became a sticky gel with
incorporated NO2 bubbles over the course of two to three days. The temperature of the
oven was then increased in 20° increments up to 220°C. This dried gel was white to light
brown, with the more brown regions being closer to the heating elements at the bottom of
the oven.
2.2.3. Ball milling and uniaxial pressing
To achieve smaller particle size for improved packing density in the pressed
pellets, a vibrating mill was utilized (Micronizing Mill, Chemplex). Hard cylindrical
Al2O3 milling media (approximately 30 pieces) were placed with the dried gel into a
tightly closed plastic milling container. The container was placed into the mill where it
was rapidly vibrated for four hours. The impact of the milling media against each other
effectively crushed the powder. Dried gel was milled for four hours, and then heat treated
up to 500°C (5°C/min ramp, 1 hour soak).
A standard 13 mm steel die (Pike Technologies) was used for uniaxial pressing.
0.60 g of powder were loaded in a steel cylinder between two polished cylindrical anvils.
A hydraulic press was used to apply pressure of 300 MPa (9000 lbs. force over a 13 mm
diameter) for 5 minutes. The resulting 3 mm-thick precursor pellets were fragile, but held
together well enough to be transferred and readied for microwave processing.
2.2.4. Reaction vessel
The reaction vessel utilized a powder insulation design (Figure 2.1), with zirconia
powder (ZrO2, <5 μm, 99%, Aldrich) inside a large high-form alumina crucible (50 mL,
38
99.8% alumina, Mattech). Microwave treatment occasionally led to runaway heating due
to partial melting of the sample, so a reaction vessel was required which could tolerate
the melting temperature of alumina (2050°C). Zirconia is a highly refractory material
with a melting point of 2700°C. The alumina crucible was partially filled with zirconia
powder. The powder was gently packed by tapping the crucible on the lab bench. Then
the precursor pellet was placed in the center of the packed zirconia. A wood cone was
held in place on top of the pellet to maintain a hole while more zirconia was added to fill
the crucible. Again, the powder was gently packed. This allowed for careful removal of
the wood cone while the powder held its shape. Care was taken to remove larger pieces
of zirconia from near the edge of the hole before removal of the wood cone.
Figure 2.1. Final insulated reaction vessel design for microwave heating. The precursor
pellet is surrounded by packed zirconia powder, with a hole in the powder to allow the pellet surface
to be viewed from above.
39
2.2.5. Temperature measurement by optical pyrometry
Sintering temperatures for β’’-alumina are >1600°C41, well beyond the operating
range of most thermocouples. Thus, optical pyrometry, which determines temperature
based on the black body emission spectrum of by an object, was employed. An optical
pyrometer was used with the reaction vessel shown in Figure 2.1.
A Vanzetti Systems series 3009 thermal monitor was used with a model OH2S
silicon detector head and an optical probe with a 0.10 mm spot size at 2.5 cm distance.
Since the optical probe had to be mounted outside of the microwave cavity, the 2.5 cm
working distance was not adequate. As shown in Figure 2.2, a quartz lens was mounted
on the top of the microwave to focus the light from the reaction vessel at a point 2.5 cm
from the optical probe. Proper focus and alignment were regularly confirmed by shining a
laser pointer in reverse down the fiber optic cable of the optical probe and observing the
focused spot in the microwave chamber.
40
Detector Head
Fiber Optic Cable
Focusing Lens
Detector
Microwave Cavity
Figure 2.2. Optical pyrometry set-up for measuring high temperatures during microwave
heating (thermal monitor unit not shown).
2.2.6. Microwave processing
The microwave oven used for heating precursor pellets was a 950 watt GE Model
JES738W02 commercial microwave oven. The GE models’ design, including a rightangle antenna out of the magnetron emitter, proved to be superior at avoiding feedback to
the emitter when operating with a very small absorber (effectively running the microwave
empty). Other models were tried that resulted in sparking inside the oven and, on one
occasion, melting of the end of the emitter.
In order to convert this system into a high-temperature solid-state reactor, several
modifications were made. The first was to remove the turntable device. Original
41
experiments relied on incandescent color to estimate temperature, and so a switch was
spliced into the light bulb circuit of the microwave so that the optical illumination of the
microwave cavity could be controlled. A hole was drilled in the top of the microwave
through both layers of sheet metal to allow addition of an optical pyrometer (see Section
2.2.5). Since overheating of the magnetron became an issue during pulsed operation, the
fan circuit was jumpered such that the fan would remain on to cool the magnetron at all
times when the door was closed.
The sample within the reaction vessel was heated in a furnace to 900°C (5°C/min,
1 hr. soak), then transferred to the microwave. The position of the sample was chosen to
be at the point of maximum intensity in the oven, which was centered front to back, 10
cm from the left wall, and 4 cm above the floor. The microwave was controlled manually,
heating for anywhere from 3 to 30 seconds per pulse and then allowing the pellet to cool
approximately halfway back to the previous temperature to allow heat to distribute evenly
within the pellet. The target temperature was 1740°C, reached on three pulses before
ending microwave irradiation and allowing the pellet to cool in the closed microwave
chamber.
2.2.7. Ion exchange
Once Na+ β’’-alumina pellets were produced, they were ion exchanged in two
steps to get to the final NH4+/H3O+ composition. KCl exchange followed the method of
Crosbie and Tennenhouse42. 3 g of potassium chloride, KCl, were placed in the bottom of
a cylindrical alumina crucible with close fitting lid (10 mL, 99.8% alumina, McDanel).
Inside the crucible, a wire stand was fashioned from a length of Pt-10% Rh wire. This
wire was chosen for its extreme chemical inertness even at high temperatures. The wire
42
stand was 2 cm above the surface of the potassium chloride. A Na+ β’’-alumina pellet
was placed on the wire stand and the crucible was covered with the lid. The crucible was
heated to 1000°C (1°C/min ramp, 24 hour soak), and cooled at 5°C/min.
After exchange for potassium, the β’’-alumina pellets were exchanged in a molten
bath of ammonium nitrate, NH4NO3. An elongated ceramic boat was partially filled with
ammonium nitrate powder. Four K+ β’’-alumina pellets were placed on top, and then
covered with more ammonium nitrate. Temperature was ramped at 1°C/min up to 180°C
at which point the ammonium nitrate was molten. The samples were held at this
temperature for ten days to allow for complete ion exchange43. Upon cooling,
recrystallization of the ammonium nitrate into a bulk solid occured. To avoid stresses that
might crack the embedded pellets they were removed at temperature and placed in an
adjacent boat in the furnace. Cooling was done at 1°C/min. CAUTION: Care must be
taken when working with ammonium nitrate, a possible explosion hazard. Use
minimum quantities, clean ceramic ware, and utilize a hood or blast shield. No
problems were encountered in this work using quantities up to six grams.
2.2.8. Characterization
A group of sintered Na+ β’’-alumina pellets were tested using Archimedes method
for density measurement. Pellets were first weighed in air. Then pellets were gripped
with forceps, submerged (without contacting the sides or bottom) in a small beaker filled
with mixed hexanes on a digital balance, and the mass recorded. These masses were used
to calculate the density of the pellet.
Reaction progress was followed by powder x-ray diffraction (XRD) on a Rigaku
MiniFlex diffractometer with a Cu Kα source. Thermogravimetric analysis (TGA) was
43
employed using a Perkin Elmer TGA 7 with TAC 7/DX Thermal Analysis Controller.
Scanning electron microscopy (SEM) and energy dispersive x-ray spectroscopy (EDS)
were performed on an FEI XL30 FEG-SEM with a PGT-IMIX PTS EDS system.
2.3.
Results
2.3.1. Mullite-like alumina precursor
The structure of the gel precursor was confirmed by XRD analysis as the expected
mullite-like alumina phase (m-alumina, PDF card #12-0539)33. Figure 2.3 shows the
XRD pattern of a sample after calcination to 900°C. This pattern is in good agreement
with the crystalline phases observed by several other groups34,36,44. Extra peaks near 20°,
36°, and 45° were not assigned.
Intensity (counts)
1000
800
600
400
200
0
20
30
40
50
2 θ (degrees)
60
70
80
Figure 2.3. XRD pattern of mullite-like alumina phase after calcining precursor to 900°C.
Expected pattern for m-Al2O3 is shown (PDF card #12-0539).
44
2.3.2. Grinding and ball milling
Powder processing is very important for obtaining a well-sintered and uniform
final product. It is especially important when the material is exposed to the high heating
rates of the microwave process.
In early experiments, dried gel (treated at 220-250°C) was ground by hand with a
mortar and pestle before being calcined at 900°C for two hours with a 1°C/min heating
ramp rate, and a 5°C/min cooling ramp rate. The resulting white powder was then ground
by hand again before being pressed into a pellet for further thermal processing to form
β’’-alumina.
Figure 2.4 shows the relatively large particle sizes of 5-10 μm that result from this
process. These larger particles were more difficult to densely pack in a pressed pellet and
led to variable results during microwave sintering including failure to heat, cracking, and
uneven heating resulting in melting.
Figure 2.4. SEM micrograph of a dried gel after grinding with mortar and pestle. Many
5-10 μm particles are visible.
45
The milling time was chosen after SEM analysis showed that the particles were
finer after four hours of milling than after only one hour. Resulting particle size can be
seen in Figure 2.5. Nearly all primary particles are less than 2 μm in diameter. These
particles form larger agglomerates as shown in Figure 2.6.
Technical discussions45 led to the conclusion that even though the primary
particles were ground down to a micron in diameter, calcining to 900°C would effectively
sinter the large agglomerates in place, and that milling would have insufficient energy to
break apart the sintered agglomerates. This would negate the packing advantage of the
smaller particles.
Figure 2.5. SEM micrograph of a dried gel after milling for four hours. Nearly all particles
are less than 2 μm in diameter.
46
Figure 2.6. The same sample from Figure 2.5 at reduced magnification showing
agglomeration of 1-2 μm primary particles into 20-50 μm agglomerates after four hours of milling.
Therefore in later experiments the processing procedure was changed. Dried gel
was ground and heated to 500°C, a temperature high enough for nearly all of the
combustion mass loss to occur from the precursor, but low enough to avoid sintering the
particle agglomerates (see TGA analysis, Section 2.3.10). The resulting powder was a
medium gray color, rather than pure white, confirming that there was some incomplete
burnout from the sample. Even with the additional mass loss that occurred after pressing
a pellet of the gray ash, more consistent and stronger pellets were produced after
microwave heating than by using the fully calcined 900°C material. Of the two powders,
the 900°C treated powder was more difficult to remove from the die after pressing into a
pellet, and the die was more difficult to clean. Based on TGA data the mass used for each
47
pellet was increased to 0.60 g for the 500°C treated samples to approximate the same
final mass as 0.50 g of the 900°C treated samples used originally.
2.3.3. Optical pyrometry
The optical pyrometer included an adjustment for emissivity, ε, of the sample in
the range 0-100. Emissivity adjustment was used to correct for the specific setup with the
focusing lens in place. The reaction vessel was heated in the furnace and then quickly
transferred to the microwave. The pyrometer agreed closely with the furnace temperature
at ε=5.
The display range of the pyrometer was 660-1625°C. Full sintering was not
achieved even at the maximum range using ε=5. Therefore, ε=95 was used to extend the
upper range of the measurement. Various temperatures were compared at emissivities of
5 and 95 yielding the linear relationship in Figure 2.7. The target display temperature of
1300°C at ε=95 corresponded to a real temperature of 1740°C at ε=5. This gives a rough
idea of the maximum temperatures reached inside the reaction vessel. Temperatures
reported here are the extrapolated values assuming a linear sensor response factor.
48
Real Temperature (C), ε = 5
1800
1600
1400
1200
1000
800
800
1000
1200
1400
Indicated Temperature (C), ε = 95
Figure 2.7. Calibration between emissivity, ε, values in microwave pyrometry setup. An
emissivity of 5 was found to accurately reflect reference furnace temperatures, while an emissivity of
95 was used to extend the upper range of the measurement (display range up to 1625°C).
2.3.4. Microwave heating
Commercial microwave ovens with turntables are not designed to have a uniform
field within the cooking chamber. There are high-intensity and low-intensity areas and by
rotating the contents to be heated, the overall heating is somewhat averaged out (different
radiuses from the center might experience different microwave intensities). The field
inside the microwave used in these experiments was roughly mapped by heating paper
soaked in concentrated cobalt chloride solution, CoCl2, at various heights in the
microwave. Hydrated cobalt chloride is pink, while the anhydrous form is blue. By
observing the appearance of blue color during heating (and hence drying), the local
49
intensity inside the oven was ascertained. The position of the sample was chosen to be at
the point of maximum intensity in the oven, which was centered front to back, 10 cm
from the left wall, and 4 cm above the floor.
In order to deposit microwave energy in the pellet, it was first heated to 900°C.
Subasri, et al.32 report heating from room temperature in the microwave, but this was not
found to be possible. At 900°C there is enough ionic mobility in the precursor to
effectively absorb microwave energy. Successful microwave heating was also observed
by pre-heating as low as 600°C, but 900°C was chosen as the standard method (5°C/min
ramp, 1 hour soak) to ensure easy coupling to the microwave and therefore decreased
strain on the magnetron that would arise from longer heating times.
The microwave was controlled manually, heating for anywhere from 3 to 30
seconds per pulse and then allowing the pellet to cool to allow heat to distribute evenly
within the pellet. This was an important consideration due to the very uneven field in the
microwave chamber. Several experiments showed that there was a focused, oblong hot
region roughly 2 mm wide by 5 mm long. If samples were heated too quickly, especially
at the beginning of the heating process, that small region would melt and cause runaway
heating. Once in liquid form, the ions are extremely mobile and the material becomes an
excellent absorber of microwave energy. Cracking of the sample due to thermal stress
was also observed with rapid heating, even if melting was avoided. By pulsing the oven
manually, the overall heating rate could be controlled and the temperature of the pellet
could be kept more uniform.
A typical heating profile is presented in Figure 2.8. Time is shown on the x-axis
as the time the microwave was actually turned on. Vertical lines indicate cooling between
50
microwave pulses. The target temperature was 1740°C, chosen to give good sintering
while avoiding bulk melting and plasma generation seen at higher temperatures. The
microwave used here required two seconds to start up the magnetron at the beginning of
each pulse. The early pulses were often as short as three seconds, meaning only one
second worth of microwave energy was absorbed by the sample. The heating rate was
very fast at the lower temperatures, approximately 100°C per second of microwave
energy. Note the transition between 1100 and 1200°C to much slower heating rates,
roughly 5°C per second of microwave energy. This is due to the transition from
m-alumina to β’’-alumina, and is discussed further in Section 2.3.5.
Temperature (C)
1800
1600
1400
1200
1000
800
0
50
100
150
200
250
Total Microwave Heating Time (seconds)
300
350
Figure 2.8. Typical heating profile for microwave processing of a precursor pellet. All
temperatures are reported at ε=5. A transition occurs near 1200°C, due to the formation of
β’’-alumina from the precursor m-alumina.
Each microwave heating experiment required about ten minutes of total elapsed
time. The average overall heating rate to go from 900°C to 1740°C was therefore
80°C/min. After heating, the pellet was allowed to cool completely in the microwave.
While the cooling rate was certainly not faster than the heating rate while exposed to
microwave energy, it was faster (at least at the highest temperatures) than the average
51
overall heating rate of 80°C/min. Some samples had cracks after cooling, but it is unclear
if these cracks occurred during the heating process or during passive cooling.
2.3.5. XRD analysis
Powder X-ray Diffraction (XRD) analysis of the microwave treated samples
revealed the expected Na+ β’’-alumina structure (PDF card #19-1173) shown in Figure
2.9, indicating successful conversion of the m-alumina precursor by microwave heating.
The competing phase, Na+ β-alumina (PDF card #25-775), is shown for comparison.
Some β-phase impurity is indeed present, most clearly seen in the peak at 33°, and the
shoulder at 42°. Based on quantitative analysis of the peaks at 21° (β’’), 33° (β), and 46°
(β’’) with respect to their expected intensities, it is estimated that the β-phase accounts
for 20% of the material. Two samples of Li-stabilized material were analyzed similarly,
resulting in estimated β-phase impurity of 30% and 40%, respectively.
Evolution of the crystal structure with temperature was monitored by XRD to
follow the transition from m-alumina to Na+ β’’-alumina. While sintering requires
temperatures >1600°C, the phase conversion temperature is important to understanding
the heating behavior in the microwave and the role of m-alumina phase in the synthesis
process. Separate precursor pellets were pressed and heated to 900°C in the furnace. A
sample was reserved with no microwave heat treatment, and subsequent samples were
heated to 1000°C, 1050°C, 1100°C, and 1150°C as measured by optical pyrometry with
ε=5. Heating followed the procedure detailed in Section 2.3.4, with five cycles reaching
the target temperature. Results are presented in Figure 2.10, focusing on diffraction peaks
near 16° and 26°. Conversion began even by 1000°C where both phases were present.
52
Further conversion occurred through 1050°C, with full conversion to β’’-alumina
complete by 1100°C.
2.3.6. Direct K+ synthesis
Figure 2.11 shows the powder XRD spectrum of directly synthesized
K+ β’’-alumina after microwave treatment to ~1200°C, hot enough to convert the
m-alumina, but not hot enough to sinter the pellet. Analysis of the diffraction peaks at
33° (β) and 34° (β’’) yields a β-phase impurity of 29%. Recall that even microwave
sintered Na+ β’’-alumina showed a β-phase impurity of only 20%. Upon full sintering of
a similar K+ sample at ~1700°C, the β-phase impurity was found to be 54% (see Figure
2.12).
53
1400
1200
Intensity (counts)
1000
800
600
400
200
0
+
Na -β
+
Na -β''
20
30
40
2 θ (degrees)
50
60
70
Figure 2.9. Powder XRD spectrum of sintered Na+ β’’-alumina pellet. Expected peaks from both the Na+ β’’ (black) and Na+ β (gray) are
shown. A small amount of β-phase impurity is visible, especially in the peak at 33°, and the shoulder at 42°.
54
Intensity (arbitrary units)
1150°C
1100°C
1050°C
1000°C
900°C
furnace
+
Na β''-alumina
m-alumina
14
16
18
20
22
24
26
28
2 θ (degrees)
Figure 2.10. Evolution of XRD pattern from m-alumina to Na+ β’’-alumina during microwave heating. Expected peaks for m-alumina (black)
and Na+ β’’-alumina (gray) are shown. Temperatures were measured by optical pyrometry (ε=5) except the 900°C sample which was measured in the
furnace and was not heat treated in the microwave.
55
1200
1000
Intensity (counts)
800
600
400
200
0
+
K -β
+
K -β''
20
30
40
2 θ (degrees)
50
60
70
Figure 2.11. Powder XRD spectrum of microwave heated but unsintered direct-synthesis K+ β’’-alumina. Expected peaks from both the K+ β’’
(black) and K+ β (gray) are shown. A small amount of β-phase impurity is visible. At lower temperatures (~1200°C) it is possible to stabilize the
K+ β’’-phase.
56
1200
1000
Intensity (counts)
800
600
400
200
0
+
K -β
+
K -β''
20
30
40
2 θ (degrees)
50
60
70
Figure 2.12. Powder XRD spectrum of microwave sintered direct-synthesis K+ β’’-alumina. Expected peaks from both the K+ β’’ (black) and
K+ β (gray) are shown. A mixture of both forms is clearly present. Reaching sintering temperatures (~1700°C) unavoidably resulted in a β/β’’ mix.
57
2.3.7. High temperature potassium ion exchange
Once Na+ β’’-alumina pellets were produced, they were ion exchanged in two
steps to get to the final NH4+/H3O+ composition. The first step is to exchange Na+ to K+
in order to expand the lattice along the c-axis. This must be done in the vapor phase at
high temperatures to avoid stress cracking in polycrystalline material42,43. 1000°C was
sufficient to melt the potassium chloride and provide sufficient vapor pressure to displace
Na+ with K+ in the crystal structure of the β’’-alumina.
Figure 2.13 presents EDS spectra for the microwave-processed Na+ β’’-alumina
and the same material after the vapor-phase KCl exchange. The exchanged pellets were
rinsed and sonicated in DI water to remove surface salts. Signals for all the components
of the material are seen: oxygen, sodium or potassium, magnesium (the stabilizing
dopant), and aluminum. The peak assigned to iridium is due to the coating process
necessary to make the sample conductive for analysis in the SEM. The disappearance of
sodium and the appearance of potassium can be clearly seen in the spectra. Quantification
of the amount of residual sodium was not done, nor was any further potassium exchange
(such as in molten potassium nitrate). Since the potassium served only to expand the
lattice to allow exchange for H3O+/NH4+, the nearly complete exchange shown by EDS
was sufficient.
Further evidence of potassium ion exchange is shown in Figure 2.14. The XRD
spectrum of a microwave processed pellet after KCl vapor exchange is compared to the
XRD spectra of commercial Na+ and K+ β’’-alumina samples (Ionotec, Cheshire,
England). While most peaks overlap directly, a few x-ray reflections shift with the
58
changing spacing of the conduction planes. Of the four peaks shown, all overlap except
for the peak near 34.5°, which has shifted from the Na+ to the K+ position.
+
K β''-alumina
+
Na β''-alumina
Intensity (arbitrary units)
Al
K
Mg
O
Ir
Na
0
1
2
3
X-ray Energy (kV)
4
5
Figure 2.13. EDS spectrum of microwave-processed Na+ β’’-alumina showing peaks for O,
Na, Mg (stabilizing dopant), Al, and Ir (conductive sputter coated layer). Also shown is the same
material after vapor phase KCl exchange and sonication rinse, showing replacement of Na by K.
59
Intensity (arbitrary units)
Microwave processed,
+
K exchanged
+
Ionotec K β''-alumina
+
Ionotec Na β''-alumina
33
34
35
36
2 θ (degrees)
37
38
39
Figure 2.14. Comparison of microwave processed sample after KCl vapor exchange to
commercial reference samples (Ionotec). While most peaks in Na+ and K+ β’’-alumina overlap, some
peaks, such as 34.5°, show that the potassium has indeed altered the structure.
2.3.8. Molten ammonium nitrate exchange
All samples experienced some cracking during the molten salt exchange ranging
from minor edge cracking to breakage into several pieces. This could have been due to
exacerbation of stresses carried over from the KCl exchange. It may also have been due
to incomplete removal of Na+ in the KCl exchange resulting in cracking due to c-axis
lengthening in the lower temperature molten ammonium nitrate where stress relaxation is
not possible. Lithium stabilized pellets were generally more structurally robust than
magnesium-stabilized pellets after molten ammonium nitrate exchange.
2.3.9. Density measurements
A group of sintered Na+ β’’-alumina pellets were tested using Archimedes method
for density measurement. Pellets were first weighed in air. Then pellets were gripped
60
with forceps and submerged in a small beaker (without contacting the sides or bottom)
filled with mixed hexanes (density 0.6637 ± 0.0003 g/mL) on the digital balance. This
allows the volume of the sample to be calculated. The final density is given by the
following:
EQUATION 2.1
where ρ is the density of the pellet, ρhex is the density of the hexanes, mair is the mass in
air, and mhex is the mass in hexanes. The density of air was neglected. An additional
correction was made to the mass in hexanes to account for wicking of liquid up the
forceps. A reproducible mass loss of 13 mg was recorded when the forceps contacted the
liquid surface, so 13 mg was added to each mass measured in hexanes.
The theoretical density was calculated using structural data from Boilot, et al.46.
The rhombohedral lattice parameters of their sample were determined to be a=5.31 Å,
c=33.54 Å. The stoichiometry was Al(11-y)O17MgyNa(1+y) with y=0.71 (very close to the
target composition of y=0.67). Using these data the theoretical density of
Na+ β’’-alumina is 2.86 g/mL.
With no sintering, the density of pressed pellets in this work was measured at
65-68% of theoretical. Geometric green density was calculated from the mass of the
green pellet and its cylindrical dimensions after uniaxial pressing, giving a value of 49%
of theoretical. The geometric green density is more accurate than that measured by the
procedure above due to the infiltration of hexanes into the porous green body. If large
enough interconnected pores exist in the solid to allow the weighing fluid to enter, the
results will be skewed toward erroneously high density. Escaping air bubbles were
observed upon submersion of the unsintered samples, indicating this infiltration was
61
occurring. Though care was taken to minimize the effect and record the mass at the
instant of immersion, the results are inaccurate for porous samples.
Visibly sintered samples ranged from 88% to 106% of theoretical density. The
diameter of each sample was measured after microwave treatment as a quick check of the
extent of sintering. The original green diameter was 13.2 mm (slightly larger than the
nominal press diameter of 13 mm). Results are shown in Figure 2.15 with density plotted
against percent shrinkage. The expected trend is apparent that as shrinkage of the
diameter increases, so does measured density. The two high outliers are due to excessive
heating which resulted in partial melting and thickening at the edges of the pellet,
exaggerating the shrinkage along the diameter. Once the procedure was optimized, good
shrinkage was observed for most heating cycles, resulting in Na+ β’’-alumina pellets near
theoretical density.
62
Density (% of Theoretical)
110
100
90
80
Measured Density
Geometric Green Density
70
60
50
0
5
10
Sintering Shrinkage, Diameter (%)
15
Figure 2.15. Density (as percentage of the theoretical Na+ β’’-alumina density: 2.86 g/mL) vs.
shrinkage during sintering as measured by the decrease in diameter of the pellet. High outliers are
due to excessive shrinkage from edge melting. Calculated geometric density of the green body is also
given since measurements were complicated by infiltration of the measuring fluid into the porous
green samples.
2.3.10. TGA analysis
Thermal gravimetric analysis (TGA) was used to monitor the decomposition
progress from the dried gel precursor to the mullite-like alumina during heating. Several
mass loss events occur as water and organic components of the gel are burned away as
shown in Figure 2.16. This progression was key to developing the proper powder
processing procedure as discussed in Section 2.3.2. After 450°C no mass loss occurs until
the event at 700°C, at which point 25% of the remaining mass is lost. Achieving full
mass loss by heating to 800 or 900°C resulted in irreversible sintering of the powder
agglomerates and complicated further pellet pressing and microwave heating operations.
63
0.00
80
-0.05
Percent Mass
Rate of Mass Change
60
-0.10
-0.15
40
-0.20
20
-0.25
0
0
100
200
300 400 500
Temperature (C)
600
700
Rate of Mass Change (mg/degree)
Percent Mass (mg)
100
800
Figure 2.16. TGA scan of Mg-stabilized gel precursor with 20% excess Na. Conducted under
flowing argon at 5°C/min. Several mass loss events can be seen as water loss and pyrolysis of organic
components occur.
TGA was also used to characterize the final product after ion exchange to the
NH4+/H3O+ form. Previous work analyzing the effluent stream has shown that a mass loss
from NH4+/H3O+ β’’-alumina at 250-300°C is attributed to loss of NH3 from the
conduction plane15. A mass loss event from 130-250°C was attributed to water loss from
the conduction plane. Results from the current work agree with those results, as shown in
Figure 2.17.
64
0
98
-5
NH3
1.5%
96
-10
94
-15
H2O
4.9%
92
-20
Percent Mass
Rate of Mass Change
90
-25
50
100
150
200
250
300
Temperature (C)
350
Rate of Mass Change (μg/degree)
Percent Mass (mg)
100
400
Figure 2.17. TGA scan of NH4+/H3O+ exchanged β’’-alumina sample (Mg stabilized, 20%
excess Na in precursor) at 0.5°C/min. More water and less ammonia was observed than would be
expected from the target composition of (NH4)1.00(H3O)0.67Mg0.67Al10.33O17.
2.4.
Discussion
The relatively high β-phase impurity (see Section 2.3.5) of 20% for Mg-stabilized
material and 30-40% for Li-stabilized material is not ideal since the β-phase is at least an
order of magnitude less conductive than the β’’-phase. More careful control of the
heating cycle could reduce this impurity. Earlier work on Na+ β’’-alumina28 found
optimum conductivity (likely due to less β-phase) to occur with a two peak firing
schedule (1510°C and 1625°C), and an increase in strength by reaching 1625°C rather
than 1610°C. In this work the microwave was not controlled to this level of accuracy, and
the outcome of heating was only evaluated by the amount of sintering shrinkage
observed.
65
Conversion from m-alumina to β’’-alumina using solution based precursors
occurs at relatively low temperatures compared to the formation of β’’-alumina from
powdered precursors. This transition was observed to occur in the microwave from 1000
to 1100°C as monitored by optical pyrometry (Section 2.3.5). In a similar study using
conventional heating, Subasri33 reports no β’’-alumina present after heating to 1000°C,
mixed phases at 1100°C, and full conversion of the m-alumina to β’’-alumina at 1200°C.
The current work had better resolution, but less accurate temperature readings. If the
optical pyrometry measurements were erroneously low by 50°C the two results would be
well matched. It is, however, unlikely that the pyrometry measured low by that amount.
Taken together these two studies would indicate that the conversion from m-alumina to
β’’-alumina begins at 1000-1050°C and is complete by 1100-1150°C. This confirms that
the change in microwave susceptibility shown in Figure 2.8 is indicative of a phase
change, and that m-alumina is more susceptible to microwave radiation than β’’-alumina.
Several researchers have attempted to directly synthesize K+ β’’-alumina and
avoid the high-temperature ion exchange necessary to convert the Na+-form into the
K+-form22,30,47. It has been found to be difficult to maintain phase purity in the potassium
system, as the thermodynamics seem to tilt toward K+ β-alumina at high (i.e. sintering)
temperatures. Attempts at the direct K+ synthesis were made here, confirming the
previous reports. The larger K+ cation affects the relative stability of the two structures
enough that a dramatic difference is seen in the phase behavior at high temperatures.
Many samples were observed to have fine cracks along a diameter of the pellet
after K+ exchange, either on one or both sides, although the pellets remained in one piece.
Park and Hellstrom48 undertook a careful study of the parameters such as temperature,
66
distance from the melt, sample geometry, and crucible size that affect the Na+ to K+
exchange rate and structural integrity of β’’-alumina samples. They concluded that rather
than diffusion of K+ into the structure, the limiting step was transport of NaCl back to the
melt. They found evidence of NaCl on the walls of their crucibles which would lead to
higher vapor pressures of NaCl than if all the NaCl were to be diluted in the KCl melt.
They also report that exchanges that occur too quickly can lead to cracking at 900°C, and
that higher temperatures could be used to mitigate cracking. The current work utilized
1000°C with a 1°C/min ramp. So it is possible that if exchange was too rapid given the
geometry of the setup, cracking could have occurred during the ramp. Perhaps a faster
ramp should be used to more quickly reach temperatures at which stress relaxation is
possible.
Quantities of H2O and NH3 released from the NH4+/H3O+ β’’-alumina structure do
not agree with the target composition of (NH4)1.00(H3O)0.67Mg0.67Al10.33O17 (see Section
1.3.3). More water and less ammonia than expected were routinely observed in the TGA
analysis (see
67
Table 2.1). Projected compositions were calculated using the following
assumptions: (1) Nominal values of stabilizing element were maintained when enough
protonic ions were present; (2) moles of NH4+ were taken as measured; (3) moles of H3O+
were assigned to make up the charge balance, with the rest of the water peak assigned as
excess water.
68
Table 2.1. Mass percentages of water and ammonia observed from several β’’-alumina
samples. Expected mass percentages from the 60/40 composition are given for comparison, for both
Li and Mg stabilized material. See above for discussion of composition calculations.
Sample
Stabilizer
Mass,
Mass,
Mass,
H2O
NH3
Total
(%)
(%)
(%)
Projected Composition
60/40
Mg
2.02
2.85
4.87
(NH4)1.00(H3O)0.67Mg0.67Al10.33O17
1 (a)*
Mg
3.53
1.52
5.06
(NH4)0.53(H3O)1.14Mg0.67Al10.33O17·0.03H2O
1 (b)
Mg
3.32
1.76
5.08
(NH4)0.62(H3O)1.05Mg0.67Al10.33O17·0.06H2O
2
Mg
4.85
1.50
6.35
(NH4)0.53(H3O)1.14Mg0.67Al10.33O17·0.50H2O
60/40
Li
2.04
2.87
4.91
(NH4)1.00(H3O)0.67Li0.33Al10.67O17
3 (a)
Li
4.50
1.12
5.62
(NH4)0.39(H3O)1.28Li0.33Al10.67O17·0.21H2O
3 (b)
Li
3.19
1.26
4.45
(NH4)0.44(H3O)1.04Li0.24Al10.76O17
*Letters in parentheses indicates separate trials
It is not clear why the compositions are significantly different than those
previously reported, but it is likely due to differences between the single crystals in
previous studies and the polycrystals in the current work. As evidenced in Chapter 3,
grain boundary effects are clearly apparent in the microwave processed material. Phase
impurities and grain-to-grain interfaces may have hindered the diffusion of NH4+ ions.
2.5.
Conclusions
Proton conducting Mg- and Li-stabilized NH4+/H3O+ β’’-alumina were
successfully synthesized via microwave assisted synthesis from a solution based gel
precursor. Polycrystalline Na+-β’’-alumina was sintered in a modified microwave oven
69
from pressed pellets of ground, heat-treated gel. Preheating the sample was necessary to
couple to the microwave radiation. Ion exchange to the K+ form and finally to the
NH4+/H3O+ form was completed. Utilizing the same method to synthesize the K+ form
directly resulted in mixed β/β’’ material. This is the first known instance of Li-stabilized
material being produced with this method, as well as the first conversion of microwave
processed β’’-alumina to the proton conducting form.
The transition from the crystalline mullite-like alumina to β’’-alumina was
observed to occur between 1000 and 1100°C. β-phase impurity was found to be 20% for
Mg-stabilized material and 30-40% for Li-stabilized material. The composition of the
proton conducting form produced by this method was deficient in NH4+ as compared to
the target 60/40 composition (NH4)1.00(H3O)0.67Mg0.67Al10.33O17.
70
Chapter 3: NH4+/H3O+ β’’-Alumina: Electronic Properties
3.1.
Introduction
3.1.1. Model circuits
Electrochemical impedance spectroscopy (EIS) was used to determine the ionic
conductivity of the proton conducting β’’-alumina pellets described in Chapter 2. EIS involves
applying a small AC voltage of varying frequency across a sample and monitoring the resulting
current along with the phase difference between the current and the voltage. As frequency
changes, different processes within the material become evident based on their inherent resistive
and capacitive components. A typical conductivity process is modeled by a resistor and a
capacitor in parallel. The resistor represents the electronic or ionic resistance to the flow of
charge in the material. In an ionic conductor it is simply the resistance due to moving ions within
the material. The capacitor represents capacitance due to interfaces: between an ionic conductor
and the electrode (electronic conductor); or between two distinct ion-conducting phases within a
given material. Data are typically analyzed using a Nyquist plot; total impedance and phase shift
are transformed into the real and imaginary components of impedance. A resistor and capacitor
in parallel give rise to a semicircle in this representation. Some illustrative example plots are
shown in Figure 3.1. These plots indicate various possible results from a material with two
distinct charge transfer processes, and are elaborated upon below.
71
Zimaginary (Ω)
500
1MHz
(a)
400
200
10MHz
100
10 nF
C2
100kHz
500
Zimaginary (Ω)
R2
1000 Ω
300
0
400
800
Zreal (Ω)
1200
(b)
400
R1
200 Ω
R2
1000 Ω
100 pF
C1
10 nF
C2
10kHz
300
200
100kHz
10MHz
100
1kHz
1MHz
0
400
2000
Zimaginary (Ω)
R1
200 Ω
800
Zreal (Ω)
1200
R1
200 Ω
(c)
R2
1000 Ω
W
1500
1000
10kHz
1kHz
500
100 pF
C1
10 nF
C2
1MHz
0
0
1000
2000
3000
Zreal (Ω)
Figure 3.1. Model circuits and their corresponding Nyquist plots. (a) Basic RC circuit (known as a
Randles circuit) showing a semicircular trace. The high frequency intercept with the real axis corresponds to
the first resistance, while the diameter of the semicircle corresponds to the resistance in the parallel RC
component. (b) Two RC components give rise to two semicircles when they have significantly different
frequency responses. (c) Adding a Warburg component, which models diffusion, causes a linear rising tail at
low frequency.
72
Figure 3.1(a) shows what is known as the Randles circuit, consisting of a resistor (R1) in
series with a parallel resistor (R2) and capacitor. When viewed as a Nyquist plot, this gives rise
to a semicircle which intercepts the real axis at R1 (high frequency) and R1+R2 (low frequency).
Thus, R2 is the diameter of the semicircle. R1 corresponds to the resistance of the leads or the
total impedance due to another charge transfer process that would be evident with higher
frequency measurements. At which frequency the semicircle appears depends upon the specific
values of resistance and capacitance in the parallel circuit component. Specifically,
EQUATION 3.1
where f0 is the frequency in hertz at the top of the semicircle, R is the resistance in ohms, and C is the capacitance in farads. Figure 3.1(b) demonstrates that two parallel RC components in series can result in two
visible semicircles if the frequency responses are well separated (in this case by more than two
orders of magnitude). In this case the high frequency semicircle intercepts the real axis at 0 Ω
and 200 Ω (R1), while the second semicircle intercepts at 200 Ω (R1) and 1200 Ω (R1+R2).
The Warburg component, indicated by “W”, is used to model semi-infinite linear
diffusion. Physically, this is due to the motion of ions within a material. At low frequencies,
diffusive motion becomes the limiting factor in current flow. The effect of Warburg impedance
is to impart a rising tail at a 45 degree slope. Figure 3.1(c) depicts the same model circuit from
Figure 3.1(b) with Warburg impedance included.
In general, Nyquist plots can be fit to simple model circuits. The difficulty lies in
assigning physical meaning to the various components in the fit. In addition, frequency
constraints can limit which processes are measurable. Equipment considerations typically limit
73
the high end of frequency, while available time limits the very low end since several cycles are
necessary for accurate measurement.
3.1.2. Conductivity
The proton conduction mechanism of NH4+/H3O+ β’’-alumina relies on the making and
breaking of hydrogen bonds in the conduction plane. Li-stabilized and Mg-stabilized material
was produced to explore if the stabilizing dopant would have an effect on the proton conductivity
by affecting the energetics of conduction plane hydrogen bonding. A significant challenge in this
endeavor is producing reproducible results considering the variability in single crystal
conductivity values reported in the literature. Reproducible results were achieved for
polycrystalline Li-stabilized material. Grain-boundary effects dominated, however, and it was
not possible to determine the grain conductivity of Mg-stabilized material. Li-stabilized material
also showed significantly higher grain-boundary conductivity than Mg-stabilized material. If the
grain boundary effects are due to the detected β-phase impurity, that would imply a conductivity
change due to hydrogen-bonding energy in the closely related β-phase material. With more
careful control of heating conditions to improve phase purity, a similar effect might be resolved
in the β’’-phase material.
One primary goal of this research was to test proton conducting β’’-alumina in a fuel cell
configuration. These results are presented in Section 3.3.2. The only known demonstration of
proton conducting β’’-alumina as a fuel cell electrolyte was by Munshi and Nicholson with H3O+
β/β’’-alumina49. Their cell with a 3mm thick electrolyte produced 23 μA/cm2 at 0.5 V and
150°C. Preliminary demonstration of NH4+ β’’-alumina as a hydrogen sensor has also been
carried out50, although this material was almost certainly the NH4+/H3O+ form.
74
3.2.
Experimental
3.2.1. Electrochemical impedance spectroscopy
Pellets of NH4+/H3O+ β’’-alumina were first gently sanded smooth and flat using
diamond polishing cloth (Precision Surfaces International, Houston, TX), first with 220 mesh
(70 μm), then with 600 mesh (30 μm). This process removed any sintered ZrO2 that occasionally
remained from the microwave heating step (Section 2.3.4), allowed for accurate thickness
measurement by removing any extra thickness around the edges, and created a smooth and
reproducible surface for applying electrodes.
Some cracking occurred in all ion-exchanged samples. After the high temperature KCl
exchange (Section 2.3.7) cracks running roughly along the midline of the pellet were often
observed. After molten ammonium nitrate exchange (Section 2.3.8), samples were often cracked
around the edges and sometimes broken into several pieces. Thus, sample preparation had to be
undertaken very carefully to avoid further damage to the samples. Samples were ground by hand
with light pressure and the diamond cloth was cleared regularly of β’’-alumina powder and any
pieces that broke off from the edges of the pellets.
Circular gold blocking electrodes were sputter coated (VCR IBS/TM250 Ion Beam
Sputterer) onto both sides of the β’’-alumina pellets (or pieces of a pellet). Electrodes were 4 mm
in diameter and 5 nm thick. A home built stainless steel combination mask and clamp was used
to hold the pellets during sputter coating. Two #5 stainless steel machine screws (shaft OD
0.125”, 3.18 mm) and nuts were used to hold two pieces of 0.010” (0.25 mm) stainless steel
sheets together. A 4 mm hole was drilled through both sheets creating an aligned mask. The
protruding shaft of the machine screws matched the diameter of common SEM stubs used in the
75
sputter coater. One screw protruded from each side, allowing the sample to be flipped over
without removal from the clamp. In this way, aligned discs of gold with very well-defined edges
were deposited on the β’’-alumina surfaces.
Thin wire leads were added to the gold pads using a silver adhesive that dries at room
temperature (Electron Microscopy Sciences, Hatfield, PA, Cat. #12686-15). A strand of wire
(0.20 mm diameter) was curled at the end and taped down to a clean surface. The pellet was
placed under the curled end of the wire so that wire was held flat on the surface by spring
tension. Then two coats of silver adhesive were applied to hold the wire to the gold pad and
make electrical contact. Wires were left ~10 cm long so they could be manipulated without
placing excessive stress on the adhesive joint.
The pellets were then placed in an existing cylindrical stainless steel bomb with
dimensions: ID 2.5” (6.4 cm); inside height 5.5” (14.0 cm); wall thickness 0.75” (1.9 cm). The
lid of the bomb was attached with four bolts and had a pass through for four wire leads and a Ktype thermocouple. The leads were soldered together in two pairs, and one pair soldered to each
lead wire on the pellet. The four pass-through wires were externally soldered to the four co-axial
leads (two for current, two for voltage) of an HP 4275A Multi-Frequency LCR Meter. The
ground sheaths of all four co-axial leads near the bomb were bundled and connected to the bomb
via one of the bolts for the lid, creating a grounded shield around the pellet for the high
frequency measurements.
The LCR meter allowed for accurate measurement of total impedance and phase angle
for the following ten frequencies: 10 kHz, 20 kHz, 40 kHz, 100 kHz, 200 kHz, 400 kHz, 1 MHz,
2 MHz, 4 MHz, 10 MHz. A perturbation of 100 mV was used, and data were recorded by hand.
The bomb was wrapped in heating tape and controlled by a PID temperature controller to
76
facilitate elevated temperature measurements at 50, 75, 100, 125, and 150°C. Impedance at
10 kHz was monitored as the set point temperature was reached to ensure the pellet had reached
thermal equilibrium before the full impedance spectrum was taken.
Room temperature EIS from 100 kHz to 10 Hz was also performed for comparison
utilizing a potentiostat (PAR model 273A) and lock-in amplifier (PAR model 5210) controlled
by Princeton Applied Research PowerSINE software. Forty-eight points were collected,
distributed logarithmically. The same lead wires on the pellets were used, and the connection
was made to the potentiostat with standard alligator clips. No shielding of the sample was used.
Impedance data from both the LCR meter and the lock-in amplifier were analyzed by
least-squares fitting using ZSimpWin version 3.21 using one of several model circuits, discussed
below.
3.2.2. Fuel cell configuration
The most structurally robust pellet, sample BK22a-3, was used to construct a working
H2/O2 fuel cell. An 8 mm diameter circle was masked off on both sides of the pellet using
adhesive tape. An ink was prepared by mixing 112.5 mg of 20% by weight platinum dispersed
on XC-72 high surface area carbon (De Nora North America, ETEK Division, Lot #3145302)
with 16 mL isopropyl alcohol. The ink was sonicated overnight to achieve good dispersion.
Electrodes applied to the β’’-alumina pellet for conductivity testing were removed by sanding on
diamond cloth. The catalyst ink was applied to both sides of the pellet using an airbrush under a
heat lamp to speed evaporation of the isopropyl alcohol solvent. Significant overspray around the
edges was expected and unavoidable. The final loading was determined gravimetrically as
0.8 mg Pt/cm2. A thin overcoat of solubilized Nafion (Liquion, Ion Power Inc., New Castle, DE)
77
was sprayed on to help adhere the catalyst to the β’’-alumina. Even so, some small pieces of
catalyst flaked off upon removal of the mask.
Figure 3.2 presents a schematic of the fuel cell setup. Current collectors were applied to
the sprayed carbon electrodes by dabbing three small spots of silver adhesive around the edges of
each electrode. Some resistance would be encountered passing current from the electrode out to
the edges, but this was considered negligible compared to the much larger electrolyte resistance
of the β’’-alumina. The anode side was mounted directly to one side of a 1/4" stainless steel
Swagelok tee fitting. The pellet was placed over the opening of the tee fitting such that the silver
adhesive pads contacted the edge of the opening. The parts were joined together using more
silver adhesive around the edge resulting in a sealed system. The cathode lead was prepared by
using silver adhesive to attach a thin wire across the three current collector spots.
Wire
Platinum on Carbon
Ambient Air
Cathode
Anode
Silver Adhesive
Silver Adhesive
H2 Out
Swagelok Tee
H2 In
Figure 3.2. Schematic of H2/air NH4/H3O+ β’’-alumina fuel cell setup. The Swagelok tee served as the
hydrogen conduit as well as the anode contact by touching the silver adhesive current collector pads on the
anode surface.
78
Hydrogen gas was flowed through the other two ports on the tee fitting so that hydrogen
could reach the anode. A small fan directed ambient air to the cathode. A potentiostat (PAR
model 173 with model 179 digital coulometer) was used to measure the current-voltage response
at room temperature. The reference electrode and counter electrode leads were joined together
and connected to the anode via an alligator clip on the body of the Swagelok tee fitting. The
working electrode was attached via an alligator clip to the wire on the cathode current collector
spots. The potentiostat was operated under voltage control.
3.3.
Results and Discussion
3.3.1. Conductivity measurements by EIS
The original goal of this research was to explore the effects of stabilizing dopants on the
proton conductivity of β’’-alumina utilizing microwave assisted synthesis. In light of the results
presented below, the polycrystalline nature of the microwave processed materials makes this
comparison very difficult. Grain boundary resistance dominated the overall resistance of the
pellets. The grain resistance was, in fact, only apparent in the Li-stabilized material within the
limits of the measurement. The stabilizing dopant did have a measureable effect, however, just
not the one that was hypothesized: the Li-stabilized material showed reduced grain boundary
resistance as compared to the Mg-stabilized material.
79
Mg-stabilized
Li-stabilized
40
Zimaginary (kΩ)
10
30
8
6
20
4
Increasing
Frequency
2
0
10
0
4
8
12
0
0
20
40
60
Zreal (kΩ)
80
100
Figure 3.3. Nyquist plot (10 kHz to 10 MHz) of Mg-stabilized and Li-stabilized proton conducting
β’’-alumina, both at 100°C. Large arc for Mg-stabilized material intercepts real axis at zero. Li-stabilized
material exhibits a reproducible non-zero intercept at high frequency (inset).
Examples of Li-stabilized and Mg-stabilized Nyquist plots are shown in Figure 3.3. Data
are presented from the LCR meter at 100°C. The significantly larger resistance of the Mgstabilized material is immediately apparent. This resistance is due to grain boundaries within the
material51. The expected grain conductivity from single crystal studies at 100°C is
10-3 S/cm 21,25,26. The large arcs indicate conductivites of 1.4x10-5 S/cm for Mg-stablized and
6.1x10-5 S/cm for Li-stabilized, both nearly two orders of magnitude lower than the expected
value. The observed intercept in the Li-stabilized material gives a conductivity of 3.9x10-3, in
line with the prediction. Capacitance derived from the large arcs was 0.62 nF for Mg-stabilized
and 5.9 nF for Li-stabilized. Higher capacitance and lower resistance for the Li-stabilized
material resulted in similar frequency response (see Equation 3.1). The higher capacitance seen
with Li-stabilization could be due to more grain boundaries, given the higher β-phase impurity
80
found with Li-stabilization. Both grain and grain boundary conductivity exhibit an Arrhenius
relationship to temperature. Li-stabilization leads to higher grain boundary conductivity than
Mg-stabilization with 93% by a statistical t-test on the data at 150°C presented in Table 3.1.
The source of the grain boundary resistance is not explicitly known. Geometrical
constraints involved in the random packing of the two-dimensional conducting grains accounts
for some of this high resistance. However, β-alumina phase impurity will also impact the
conductivity. Both Mg- and Li-stabilized material had significant amounts of the β-phase present
(see Section 2.3.5); Li-stabilized having more than Mg-stabilized. It is entirely possible that what
is referred to here as grain boundary conductivity is, in fact, due to conduction through grains of
the less conductive β-phase. The effect will continue to be referred to as “grain boundary”
henceforth, but the reader should keep in mind the existence of the phase impurity. The fact that
Li-stabilized shows a higher overall conductivity, even with more phase impurity, is
encouragement for future work to produce single crystals of Li-stabilized NH4+/H3O+ β’’alumina; it may be that the Li-stabilization is raising the conductivity of the β-phase, and could
have the same effect on the β’’-phase.
The inset of Figure 3.3 shows the high frequency region of the two Nyquist plots. The Listabilized material makes a small positive intercept with the real axis, while the Mg-stabilized
material intercepts at zero or negative values. This was consistently observed, and quantified
using the ZSimpWin fitting software.
At low temperatures, the resistance of the grains and that of the grain boundaries are the
most different, as shown in the Arrhenius plots in Figure 3.4. That is, the conductivity of the
grain boundary rises more quickly than that of the grains. This separation in resistance allows the
best chance for resolution in frequency of two RC semicircles, one corresponding to grain
81
conduction, and one corresponding to grain boundary conduction52. All model circuits discussed
here utilize constant phase elements (CPE) in place of capacitors. Random crystallite orientation
and surface roughness give rise to a depressed semi-circular trace in the Nyquist plot that can be
fit using a CPE52.
T (ºC)
150
125
100
75
50
25
-4
-1
ln(σ) (Ω cm )
-6
-1
-8
Grain, Ea=0.22eV
Grain Boundary, Ea=0.56eV
-10
-12
-14
2.2
2.4
2.6
2.8
1000/T
3.0
3.2
3.4
Figure 3.4. Arrhenius plot for a Li-stabilized sample showing grain and grain boundary
conductivities.
The evolution of the Nyquist plots with increasing temperature for the Li-stabilized
material is presented in Figure 3.5. At low temperature, the effects of the high frequency arc due
to the grains can be seen (Figure 3.5a). At room temperature a full arc is apparent, while at 50°C
the curve simply tails off as the low and high frequency arcs begin to overlap. These types of
curves were fit with the model circuit shown to account for the two distinct arcs.
Higher temperatures saw the disappearance of the high-frequency arc, but maintained the
positive intercept with the real axis (Figure 3.5b). In terms of model circuits this corresponds to a
simple resistor for the grains in front of the RC component for the grain boundaries.
82
As grain boundary resistance further decreased at increasing temperature (note the decreasing
scales of the graphs) the characteristic frequency response was shifted to higher frequencies. As
the test frequencies remain the same, this has the effect of revealing more detail on the low
frequency (right) side of the arc. Figure 3.5c shows the appearance of a Warburg impedance tail
at 125°C and 150°C. This component was added to the model circuit as shown when appropriate.
The Warburg tail is further confirmed from wider frequency sweeps taken on the lock-in
amplifier. Figure 3.6 shows both LCR and lock-in data for Li-stabilized material at room
temperature. By going to low enough frequencies, the Warburg tail can be seen even at this
lower temperature.
83
50
20
22ºC
(a)
50ºC
30
Grain
Grain Boundary
20
10
0
Zimaginary (kΩ)
Zimaginary (kΩ)
40
(b)
75ºC
15
10
Grain
5
Grain
Boundary
100ºC
0
0
10
20
30
Zreal (kΩ)
40
50
1400
0
10
20
Zreal (kΩ)
30
(c)
Zimaginary (Ω)
1200
1000
800
125ºC
600
400
200
150ºC
0
0 1000 2000 3000 4000 5000
Zreal (Ω)
Figure 3.5. Evolution of Nyquist plots with increasing temperature for Li-stabilized NH4+/H3O+
β’’-alumina from 22-150°C. (a) A high frequency semi-circle (toward the left of the plots) is clearly evident in
the 22°C curve, while a tail is visible at 50°C. This is the response from the grain conductivity. (b) The high
frequency tail disappears at 75°C and 100°C, but a non-zero intercept for the large grain-boundary arc is
evident. (c) The non-zero intercept remains at high frequency and the beginning of a low-frequency Warburg
tail appears at 125°C and 150°C. Appropriate model circuits are shown with each graph. Capacitors were
replaced with constant phase elements in all cases to account for the polycrystalline samples (i.e. depressed
semi-circles).
84
Zimaginary (kΩ)
400
Lock-In (10Hz-100kHz)
LCR Meter (10kHz-10MHz)
300
200
100
0
0
200
400
600
800
1000
Zreal (kΩ)
Figure 3.6. Comparison of Li-stabilized β’’-alumina as measured by LCR meter and lock-in
amplifier at room temperature.
All conductivity data are summarized below in Table 3.1, including activation energies
from the Arrhenius plots. Arrhenius behavior was very linear for all samples. Conductivity is
reported at 150°C. Literature references for single crystal NH4+/H3O+ β’’-alumina are included in
the lower half of the table for comparison, with reported activation energies and calculated
conductivities at 150°C. All literature values are from Mg-stabilized material.
Grain conductivity is not reported for microwave processed Mg-stabilized material since
no high frequency arcs were visible in EIS analysis. That is not to imply that grain conductivity
is higher or lower in Mg-stabilized material, but only that the frequency responses of the grains
and grain boundaries overlapped. The average grain boundary conductivity for Mg-stabilized
material at 150°C was 1.7x10-4 ± 1.4x10-4 S/cm. For Li-stabilized material, grain boundary
conductivity at 150°C was 4.6x10-4 ± 3.3x10-4 S/cm. These two means are statistically different
to 93% confidence as determined by a t-test.
85
Table 3.1. Conductivity and activation energies for grains and grain boundaries in proton conducting
β’’-aluminas. Grain conductivity could not be determined in Mg-stabilized samples due to the overlap in
frequency response of grains and grain boundaries.
Grain
Sample
Precursor Stabilizer
Conductivity
at 150°C
(S/cm)
BK20i-2
BK21a-1
BK21a-3
BK21a-4
BK22a-1
BK22a-3
BK22a-4
BK22a-6
Excess
Na
Stoich.
Na
Stoich.
Na
Stoich.
Na
Excess
Na
Excess
Na
Excess
Na
Excess
Na
Grain
E a,
Boundary
Grain
Conductivity
(eV)
at 150°C
(S/cm)
E a,
Grain
Boundary
(eV)
Mg
1.5E-05a
0.69
Mg
9.3E-05
0.60
Mg
2.2E-04
0.53
Mg
3.3E-04
0.50
Li
4.8E-03
0.31
2.8E-04
0.29b
Li
6.0E-03
0.22
1.5E-04
0.56c
Li
8.6E-03
0.34c
5.2E-04
0.46
Li
7.1E-03
0.29
8.9E-04
0.49
Crystal24
Na
Mg
4.8E-04
0.24
Crystal
21
Na
Mg
3.2E-03
0.30
Crystal
26
Na
Mg
6.4E-03
0.25
Crystal
25
Na
Mg
8.2E-03
0.31
Crystal15
Na
Mg
2.9E-02
0.27
a
Reported at 125°C, the highest temperature measured for this sample.
b
Sample failed due to a crack causing a short between contacts.
c
Cooling cycle used to calculate Ea.
86
Grain conductivities were measured for the Li-stabilized material resulting in an average
conductivity at 150°C of 6.6x10-3 ± 1.6x10-3 S/cm and an average activation energy of
0.29 ± 0.05 eV. The good reproducibility is due to proper powder processing, polished sample
surfaces, and very reproducible sputtered gold electrodes. These average values put the grain
conductivity squarely in the middle of the pack of literature values for single crystal NH4+/H3O+
β’’-alumina conductivity.
3.3.2. Fuel cell configuration
The motivation for this work was to produce a highly conductive, solid state, anhydrous
proton conductor for use in elevated temperature PEM-type fuel cells. Thus, a demonstration of
the material produced by microwave assisted synthesis as a fuel cell electrolyte was undertaken.
The highest open circuit potential observed in the fuel cell configuration was 0.975 V.
The current-voltage response is displayed in Figure 3.7 in terms of current density (electrode
diameter was 8 mm, an area of 0.5 cm2). Resistance from the current-voltage curve agrees well
with the conductivity measured by EIS. The slope of the line corresponds to the resistance of the
β’’-alumina electrolyte, and was calculated to be 15.2 kΩ·cm2. Incorporating the thickness of
2.0 mm, results in electrolyte conductivity of 1.32x10-5 S/cm. Conductivity as measured by EIS
at 22°C was 1.12x10-5 S/cm, in close agreement with the fuel cell derived value.
87
0.93 V
1.0
Cell Voltage (V)
0.8
0.6
0.4
0.2
0.0
0
20
40
2
Current Density (µA/cm )
60
80
Figure 3.7. Current-voltage response for NH4+/H3O+ β’’-alumina fuel cell operating with H2 and
ambient air at room temperature. Measured response agrees with EIS conductivity data and yields an open
circuit potential of 0.93 V. Conductivity data were used to calculate the expected response at 100°C and
150°C.
The pellet cracked during preparation for elevated temperature measurements so no data
are available. However, the close agreement in conductivity values gives confidence that the
conductivity of the fuel cell electrolyte would follow the Arrhenius behavior shown in the EIS
results. Further improvement in conductivity would be achieved by using a thinner electrolyte.
With optimization of the sintering and ion exchange processes, 1 mm thick material could have
sufficient strength to be used as a fuel cell electrolyte. This would double the conductivity and
therefore roughly double the current density.
The only known demonstration of proton conducting β’’-alumina as a fuel cell electrolyte
was by Munshi and Nicholson using H3O+ β’’-alumina, 3 mm thick49. Their current-voltage
response was also very linear with an open circuit potential near 1.0 V. Reported current
88
densities at 0.5 V cell voltage were 19 μA/cm2 at 100°C and 23 μA/cm2 at 150°C. In this work,
the measured current at 0.5 V cell voltage was 28 μA/cm2 at room temperature, an improvement
of 47% over the previously reported response at 100°C.
3.4.
Conclusions
Both Li-stabilized and Mg-stabilized NH4+/H3O+ β’’-alumina were tested for proton
conductivity by EIS. Average grain conductivity for Li-stabilized material at 150°C was
6.6x10-3 ± 1.6x10-3 S/cm with 0.29 ± 0.05 eV activation energy. This is in the middle of the
range of literature values for conductivity of single crystal NH4+/H3O+ β’’-alumina. Grain
conductivity was not resolved for Mg-stabilized material due to grain boundary effects.
The average grain boundary conductivity results at 150°C were: Mg-stabilized,
1.7x10-4 ± 1.4x10-4 S/cm; Li-stabilized, 4.6x10-4 ± 3.3x10-4 S/cm. These two means are
statistically different to 93% confidence. Grain boundary conductivity was at least in part due to
β-phase impurity in the samples. It is possible that the proton conductivity in the two β-phase
materials is being affected by the hydrogen bonding energetics in the conduction plane. More
careful control of phase purity and sintering conditions, or single crystal studies would be
necessary to confirm this conclusion.
Li-stabilized NH4+/H3O+ β’’-alumina was demonstrated as a fuel cell electrolyte,
producing 28 μA/cm2 at 0.5 V.
89
Chapter 4: Introduction to Zinc Rich Coatings
4.1.
Motivation
Corrosion is a costly problem worldwide. Studies of the costs of corrosion have been
undertaken in various countries and estimates range from 2-5% of gross national product53.
Corrosion of steel is chief among these issues, affecting buidings, roads, bridges, vehicles,
ships, etc. Prevention of steel corrosion is in itself a huge industry. Any advances in corrosion
protection have the potential for significant impact on global economies.
Zinc is commonly used as a protective coating on steel in the galvanization process. The
more reactive zinc preferentially corrodes leaving the underlying steel intact. Hot dip
galvanization leaves a thin layer of zinc over the entire surface. Other coating systems are more
complex utilizing a zinc rich coating primer, often an adhesive mid-coat, and a barrier topcoat.
The discussion here will focus on the properties of the zinc rich coating.
4.2.
Zinc Rich Coatings
Zinc rich coatings (ZRC’s) have long been employed to prevent corrosion on steel
structures54. ZRC’s consist of zinc dust (typically >80% by weight) bound in an inorganic
(e.g. ethyl silicate) or an organic (e.g. epoxy) binder. It is widely accepted that protection occurs
initially by sacrificial galvanic protection offered by the zinc particles which are electrically
connected to each other and to the steel substrate55. After a period of weeks or months, zinc
corrosion products build up within and on top of the ZRC, resulting in a barrier layer56. This
physical prevention of access by corrosive species to the underlying steel becomes the primary
means of corrosion inhibition.
90
4.3.
Electrochemical Impedance Spectroscopy
The simplest way to montitor the electrochemical behavior of a ZRC is by monitoring the
potential against a reference electrode. It has been generally observed that corrosion potential in
a ZRC on steel system rises from the potential of zinc toward the potential of steel over time57-60.
While this is useful information, more detailed analysis of the system is often required. One
technique that allows such analysis is electrochemical impedance spectroscopy (EIS).
EIS was discussed in Chapter 3. Briefly, EIS involves applying a small AC voltage of
varying frequency across a sample and monitoring the resulting current along with the phase
difference between the current and the voltage. As frequency changes, different processes within
the material become evident based on their inherent resistive and capacitive components. A
model circuit is developed to give physical meaning to the observed components.
With respect to ZRC’s, several model circuits have been proposed in the literature. Xie,
et al.61,62 propose the following:
Figure 4.1. Model circuit for ZRC corrosion proposed by Xie, et al.61,62.
R0 is the uncompensated resistance due to wire leads and solution resistance; R1 is the charge
transfer resistance associated with zinc dissolution; C1 is the double layer capacitance for zinc
dissolution; R2 is the resistance through the binder; C2 is the capacitance between zinc particles
separated by the binder.
91
Mansfeld63 has proposed the following model for protection by polymer coatings:
Cc
RΩ
Cdl
Rpore
Rp
Figure 4.2. Model circuit for polymer coatings proposed by Mansfeld63.
RΩ is the uncompensated resistance due to wire leads and solution resistance; Cc is the
capacitance of the polymer coating; Rpore is the resistance due to ionically conducting paths
across the coating; Rp is the polarization or charge transfer resistance; Cdl is the double layer
capacitance.
The two models give similar fits mathematically, defined by two time constants visible as
semi-circles on the Nyquist plot. A tangibly different model circuit, called a transmission line
model, has been proposed by Feliu et al.55 to take into account the uneven current distribution
across the surface of the coating.
Figure 4.3. Transmission line model circuit proposed by Feliu et al.55.
92
In practice the unevenness of the coating is most often modeled by replacing the
capacitors in the first two model circuits with constant phase elements (CPE’s). The expressions
for impedance as a function of frequency are given below for a resistor, capacitor, and a CPE.
EQUATION 4.1
EQUATION 4.2
EQUATION 4.3
Where Z is impedance, j is the imaginary unit, ω is the angular frequency, R is resistance, and C
is capacitance. Thus, for the CPE the Y term corresponds to a capacitance, while the exponent, n,
alters the phase response. At n=1 a CPE behaves exactly like a capacitor. At n=0.5 the center of
the semicircle response on the Nyquist plot would be depressed by 45°.
4.4.
Corrosion Products
The physical structure of the barrier layer is important to its function in corrosion
inhibition. Desirable properties of the barrier layer include good adhesion to the ZRC substrate
and low permeability to corrosive species in solution. For a given compound, adhesion and
permeability will be dependent on the specific morphology of the barrier layer: crystallinity,
crystallize size, density, etc.
Compounds previously observed to form on ZRC’s include zinc hydroxide64 [Zn(OH)2],
zinc oxide [ZnO], zinc carbonate65 [ZnCO3], zinc hydroxycarbonates65-67, including hydrozincite
[Zn5(CO3)2(OH)6], and zinc hydroxychlorides67, including simonkolleiete [Zn5(OH)8Cl2·H2O].
The chloride containing simonkolleite and the carbonate containing hydrozincite have
similar structures. Simonkolleite, Zn5(OH)8Cl2·H2O, is rhombohedral, space group R3m,
a=6.3412, c=23.646 Å, three formula units per cell68. It consists of sheets of edge-sharing
Zn(OH)6 octahedra, with a vacant octahedron in every other site of every other row, offset to
93
form a hexagonal superlattice of vacant sites. Each of these vacant sites is capped both above and
below by a zinc ion tetrahedrally coordinated to the three available hydroxide groups plus one
chloride ion capping the tetrahedron; tetrahedra therefore point away from the sheet. Sheets stack
to intermesh arrays of tetrahedral and to contain interstitial H2O groups. Hydrogen bonding from
OH- to Cl- and H2O hold the layers together.
Figure 4.4. Two views of simonkolleite from Hawthorne68. Zn octahedra are yellow; Zn tetrahedra
are green; H atoms are small blue circles; H2O groups are red circles; thin black lines indicate hydrogen
bonds. Reprinted with permission from the authors.
94
Figure 4.5. Oblique view of the simonkolleite structure from Hawthorne68 showing the conformation
of adjacent sheets and the arrangement of the interstitial H2O groups. Reprinted with permission from the
authors.
Hydrozincite, Zn5(CO3)2(OH)6, is monoclinic with a=13.62, b=6.30, c=5.42 Å, β=95.83°,
space group C2/m, two formula units per cell69. Hydrozincite also consists of edge-sharing
Zn(OH)6 octahedra with a nearly identical pattern of vacancies as in simonkolleite. Again, the
vacancies appear at every other site along every other row of octahedra, but they are aligned so
as to form a square superlattice of vacant sites. Vacancies are similarly capped on both sides by
tetrahedral zinc, with the apical ligand in this case being one of the carbonate oxygens. A second
carbonate oxygen is coordinated to an octahedral zinc in the adjoining layer, while the third is
involved in hydrogen bonding to three hydroxide groups. In this way, the carbonate ions link the
sheets together.
95
Figure 4.6. A view from Ghose69 of a single oxy-hydroxy-zinc sheet of hydrozincite, viewed along
[100]. Reprinted with permission of the International Union of Crystallography.
4.5.
Electrochemical Noise
Corrosion processes cause random fluctuations in electrode potential, known as
electrochemical noise. The nature of the electrochemical noise can be used as a tool to
understand corrosion rate70 and has been studied specifically in the ZRC on steel system71-73. As
discreet corrosion events occur, local current flow develops between anodic and cathodic regions
of the electrode. A ZRC will degrade by these bursts of corrosion rather than by a smooth,
uniform process.
Corrosion noise measurement is generally non-invasive. The use of matched pairs of
electrodes as employed by Skerry and Eden73, as well as Recarey et al.72 allows for the
measurement of current fluctuations between the two panels. Voltage fluctuations can be
measured against a reference electrode with a suitably sensitive detector. Recarey used this
96
technique to delineate various periods in the development of the corrosion process of a ZRC on
steel.
97
Chapter 5: Mechanism of Enhanced Corrosion Protection of a Zinc Rich
Coating with Electronic Control
5.1.
Introduction
Zinc rich coatings (ZRC’s) are a common way of protecting steel from corrosion. They
operate first by galvanic protection, and then as a barrier coating as corrosion products build up
on the surface and in the pores of the coating55,56. This work examines a commercially available
high weight-loading ZRC in an aluminosilicate binder (Zetan, Applied Semiconductor Inc., New
York). The Zetan system also includes an electronic control unit (ECU), the laboratory version of
which consisted of a large 12V battery and a 0.56 F capacitor wired in series to the steel sample
(Figure 5.1). The purpose of this study was to determine if the application of the ECU enhanced
the corrosion protection of the ZRC, and if so, by what mechanism.
+ 12 V -
ECU
ZRP on
Steel
Figure 5.1. Schematic of the ECU connected to a ZRC coated steel panel.
Galvanic protection of steel by zinc in electrical contact is one form of cathodic
protection. That is, the more active zinc preferentially corrodes, becoming the anode in the
galvanic couple, and protects the steel by maintaining it as the cathode. The impressed current
98
technique is another form of cathodic protection in which an external power source is used to
constantly supply electrons to the steel, again maintaining it as a cathode and preventing iron
dissolution. While the ZRC itself offers cathodic protection to the steel, the ECU is a unique
system and should not be mistaken for an impressed current technique. All the potential from the
battery is dropped across the capacitor. Thus, there is not a continuous flow of current from the
battery to the steel.
The ECU concept was originally developed by William Riffe of Marine Environmental
Research, Inc., Morehead City, NC74. The original hypothesis was that the ECU maintained a net
negative charge on the ZRC which kept a higher concentration of Zn2+ ions close to the surface,
thereby slowing the corrosion process. In later disclosures75,76, the mechanism of enhanced
corrosion inhibition is described as due to the semiconductor properties at the interface of zinc
metal with a zinc oxide coating. A p-n junction at the Zn/ZnO interface is proposed which
behaves like a diode, and under reverse bias prevents electron flow across the interface. Random
voltage fluctuations can cause the bias to shift to the forward direction allowing more rapid
corrosion. The role of the ECU is described as limiting these fluctuations and thereby reducing
the corrosion of the zinc particles.
More recently, the technology has been developed by Applied Semiconductor, Inc.
Dowling77 considers the electronic filtering provided by the capacitor in the circuit. It is argued
that by suppressing the random voltage fluctuations (electrochemical noise) associated with the
corrosion process, corrosion proceeds more slowly and the life of the coating is extended.
Dowling and Khorrami78 report an active tunable device to match the frequency response of the
ECU to that of the corrosion noise experienced by each object to be protected. The structure of
zinc surrounded by zinc oxide, in turn surrounded by a conductive silicate binder is described as
99
a varistor (variable resistor) which behaves like back-to-back diodes. Within a range of voltage,
resistance is high and current flow is significantly reduced. Outside of this range, resistance of
the varistor breaks down and corrosion proceeds rapidly. Again, the ECU operates by filtering
out these voltage fluctuations. Previous experiments have indicated that the size and/or
construction of the capacitor in the ECU can significantly affect performance79, supporting the
role of the ECU as a filter.
In this work, the chemistry occurring at ECU protected and non-ECU protected
steel/ZRC interfaces is investigated. This interface is explored immersed in 3% by weight NaCl
solution for up to a year. Corrosion potential measurements, cross-sectional SEM and elemental
analysis, x-ray diffraction studies, and electrochemical impedance spectroscopy (EIS) were used
to explore the mechanism of enhanced protection due to the ECU.
5.2.
Experimental
5.2.1. Sample preparation
Sandblasted steel test panels (100 mm x 150 mm) were coated on both sides and all edges
with a solvent-based inorganic ZRC containing a dry zinc content of greater than 90% with a
nominal dry film thickness of 90 μm. Measured film thickness by SEM over 22 samples was
105 ± 20 μm. Two fine multi-stranded wires were soldered to the underlying steel at the upper
corners of each panel for potential measurements. The wires were connected one to the
conductor and one to the shield of a female co-axial connector. Each panel was suspended in a
separate two-liter beaker containing 3 wt.% sodium chloride solution at pH 7 and room
temperature such that the panel was immersed to 75% of its height under an acrylic cover. All
solutions were slowly stirred with a magnetic stir. Observations were made over six to twelve
100
months to determine any influence of the ECU on the corrosion process. Concurrent experiments
were conducted in New York City, NY and Princeton, NJ.
An electronic control unit (ECU) was included on half of the panels. The ECU consisted
of a 0.56 F capacitor connected to the positive terminal of a large 12 V Zn/MnO2 battery
(Eveready #732). The capacitor and the negative battery terminal were then connected to the
female co-axial connector mentioned above, making contact with the two corners of a steel test
panel. This produced a circuit where the capacitor was charged to 12 V, and no current flowed
across the panel, barring electrochemical activity (Figure 5.2).
ECU
+ 12 V -
Figure 5.2. Schematic of setup and wiring of control and ECU panels.
5.2.2. Testing protocol
Group A (New York) consisted of twelve panels: six with an ECU and six control.
Solutions were replaced four times throughout the study period (on days 35, 64, 105, and 141). A
pair (one ECU, one control) of Group A panels was removed at one month intervals for analysis,
resulting in six months of observation, or 183 days.
101
Group B (New York) was identical to Group A except that each panel contained a 50 mm
scribe mark across the middle of one side exposing bare steel beneath the ZRC. Again, a pair of
panels was removed at one month intervals for analysis.
In Group F (New York), solutions were replaced weekly to prevent buildup of aqueous
corrosion products. Six replicates were tested: three panels with an ECU; and three control
panels. These panels were observed for one year.
Group G (Princeton) consisted of 6 panels, three with an ECU and three control.
Solutions were replaced in accordance with Groups A and B (on days 35, 64, 105, and 141).
These plates were monitored weekly by electrochemical impedance spectroscopy (EIS) for five
months.
Group H (Princeton) was identical to Group G except that each panel contained a 50 mm
scribe mark across the middle of one side exposing bare steel beneath the ZRC.
The pH of each solution was monitored daily and adjusted to 7.0±0.1 as necessary by the
dropwise addition of 1 M NaHCO3 or 1 M HCl. Typically the pH drifted up during immersion
and HCl was required to maintain pH 7.
5.2.3. Characterization
Digital photographs of the panel surface were taken to monitor visual evidence of
corrosion (Nikon Coolpix 880).
Corrosion potential was measured against a saturated calomel electrode (SCE), and
recorded weekly. Potential measurements were made by immersing an SCE into the solution
approximately 5 cm from the panel. All measurements were taken in New York using the same
SCE.
102
Samples of salt solution were reserved at each solution change for aqueous zinc analysis.
Samples from Day 7-140 were analyzed by inductively coupled plasma (ICP) emission
spectroscopy (Perkin-Elmer model 4300 DV). Samples from Day 147-345 were analyzed by
stripping voltammetry using a static mercury drop electrode (Princeton Applied Research, Model
303A Static Mercury Drop Electrode). An integrated reference electrode within the mercury drop
electrode apparatus consisted of a glass frit separating a chamber containing saturated AgCl and
an Ag wire electrode (Ag/AgCl: -1.19 vs. SCE) and was located 1 cm from the working
electrode (mercury drop). For stripping voltammetry the potential was held -1.15 V vs. the
internal Ag/AgCl reference electrode for 10 seconds, followed by a 0.75 V sweep at 100 mV/s to
-0.40 V vs. Ag/AgCl. The hold potential was sufficiently negative to reduce zinc into the
mercury droplet forming an amalgam. Sweeping the potential positive gave a reproducible peak
in current. Peak heights were used for calibration with the baseline taken at -1.09 V vs. Ag/AgCl
and the peak current falling at -0.93 V vs. Ag/AgCl. Calibration standards were prepared by
serial dilution from ZnCl2 in concentrations of 10, 5, 2, and 1 ppm (by weight of Zn). Calibration
points were taken as a set of five replicates including a blank while three replicates were used for
analysis runs. Calibration curves were linear, with R2 values ranging from 0.998 to 0.9999.
Samples were prepared for scanning electron microscopy (SEM) (FEI XL30 FEG-SEM)
analysis by cutting 1 cm2 pieces from a panel using a band saw. Pieces were then mounted in
cross section in low viscosity epoxy (EpoFix, Struers); pieces were held upright by a sample clip
and epoxy was slowly poured over the sample in a reusable sample cup (MultiForm, 25 mm
diameter, Struers). Curing was allowed to proceed overnight. Mounted samples were removed
from the cups and the top edge ground down with 120 grit SiC sandpaper on an 8” diameter
rotating wheel under flowing deionized water. Once material undamaged by the band saw was
103
exposed, the surface was ground with 240, 400, and 600 grit SiC sandpaper. Two polishing steps
followed utilizing alumina slurries (60 g/L), first with 1 μm particles, and finally with 0.3 μm
particles. Samples were rotated 90° between each step. After polishing, samples were rinsed first
in copious amounts of deionized water and then in ethanol to dry the samples and avoid further
oxidation. The back surface of each sample was ground just enough to expose bare metal so that
when mounted on a 1” diameter aluminum SEM stub an electrical contact was made from the
stub to the polished surface. Samples were sputter coated with 3 nm of iridium and imaged at
15 kV accelerating voltage. Energy dispersive x-ray spectroscopy (EDS) (PGT-IMIX PTS EDS
system) was performed in the SEM to obtain elemental composition data.
Long scan x-ray diffraction (XRD) (Bruker D8 Focus Flat-Plate Diffractometer with a Cu
Kα source, a graphite diffracted beam monochromator, and a scintillation counter) was used to
determine the crystal structure of corrosion products on the panels. 1 cm2 pieces were cut from a
Group A control panel after five months of immersion and a control panel from Group A after
four months of immersion. Panels had been rinsed, and had seen no further preparation, such as
scraping the surface. Samples were placed flat in the diffractometer on a recessed stage.
Scanning parameters were: 2θ = 10-70°, 0.020° step, 5.5 s step time.
5.2.4. Electrochemical impedance spectroscopy
Each panel in Groups G and H was withdrawn from solution weekly for testing by
electrochemical impedance spectroscopy (EIS). Panels were withdrawn, and rinsed with
deionized water from a squirt bottle on both sides. The ECU was disconnected if present. Panels
were then mounted in the contact cell shown schematically in Figure 5.3. A polycarbonate
cylinder (2.5” (64 mm) inner diameter) was fitted with an o-ring and clamped to the plate using a
104
modified C-clamp with a Teflon pad for the back of the panel. The counter electrode was a thin
disk of stainless steel with a hole cut in the center. The reference electrode assembly consisted of
R.E.
Capillary
Contact Cell
C.E.
O-ring
Test Panel
W.E.
Figure 5.3. Schematic of electrochemical contact cell for weekly EIS measurements on Groups
G and H. The working electrode (W.E.) was the panel; the counter electrode (C.E.) was a thin disk of stainless
steel with a hole cut out in the center; and the reference electrode (R.E.) was an SCE mounted within a glass
capillary which passed through a the counter electrode. An o-ring sealed the cell to the panel.
an SCE within a tapered glass capillary which passed through the center of the counter electrode.
1 M NaCl was used as the electrolyte.
EIS spectra were taken from 100 kHz to 1 Hz with 10 mV RMS excitation voltage using
a potentiostat (PAR model 273A) and lock-in amplifier (PAR model 5210) controlled by
Princeton Applied Research PowerSINE software. Two replicate spectra of each panel were
taken. Panels were then rinsed again with dionized water, placed back in their two-liter beakers,
and the ECU reattached if necessary.
105
Similar EIS analysis was also done on panels withdrawn for examination by microscopy,
including Groups A, B, and F. Panels had been shipped dry and were soaked in 3% NaCl
overnight before EIS testing.
5.2.5. Large volume study
A large volume study was undertaken to follow the evolution of zinc concentration
without the limitation of saturation. For these studies, the pH was buffered to 8.2 using
borate/boric acid to mimic ocean pH. Two new, clean polyethylene garbage cans were each filled
with 100 L of deionized water. 3.03 kg of NaCl (technical grade, McMaster Carr) was added to
each container resulting in a 3% by weight solution. 272 g of boric acid was added to each
container and the pH adjusted to 8.2 (ocean pH) with approximately 46 g of NaOH, making a
buffer with a total borate concentration of 44 mM. This concentration was shown to have
sufficient buffer capacity based on preliminary experiments. The solutions were stirred using a
250 gallon/hour magnetic drive impeller pump (Mag Drive 250, Danner) placed at the bottom of
the container and powered by a standard 110 V outlet. Pumps were oriented to draw solution in
from the top and expel solution tangentially near the bottom of the container, resulting in a
steady circular flow throughout the container. Test panels were suspended in solution to 75% of
their height, and one was connected to an ECU. Containers were kept covered and deionized
water added as needed to maintain the volume. 15 mL samples were withdrawn at intervals (1 h,
4 h, 16 h, 24 h, then daily, then less often) from halfway between the center of the panel and the
container wall. Aqueous zinc analysis was performed by stripping voltammetry.
106
5.3.
Results
5.3.1. Visual evidence
Group F showed the clearest visual evidence of corrosion inhibition due to the ECU (see
Figure 5.4). White corrosion products were evident on the surface of the control panels as soon
as 7 days after immersion. The unique patterns of corrosion products were evident by Day 14.
The same general appearance persisted until the end of the study. The ECU plates showed no
visible signs of corrosion for the duration of the study.
Figure 5.4. Photographs of F Group after 161 days of immersion. Visual appearance remained
similar until the end of the study.
107
5.3.2. Corrosion potential
Upon immersion in salt water, the panels immediately begin to corrode. This process was
followed by monitoring the corrosion potential. Results are presented in Figure 5.5 for the
average of the ECU and control plates in Groups A, B, and F. The potential of the bare steel used
in this study is shown for reference along with the empirically determined limit below which
cathodic protection is active for steel (-780 mV)80.
-600
Corrosion Potential (mV vs SCE)
Bare Steel Potential
-700
Cathodic Protection Limit
-800
Group A, Control
Group A, ECU
-900
Group B, Control
Group B, ECU
Group F, Control
Group F, ECU
-1000
0
50
100
150
Immersion Time (Days)
200
250
Figure 5.5. Average corrosion potentials for Groups A, B, and F. Potential of the bare steel and the
accepted potential below which cathodic protection is operative80 are shown for reference. Weekly solution
changes led to a large difference in performance for Group F.
5.3.3. Aqueous zinc analysis
Aqueous zinc concentrations for samples reserved at every solution change were
followed by either ICP (Day 7-140) or stripping voltammetry (Day 147-345). Data are presented
108
in Figure 5.6 for the first 140 days after immersion as averages for each group shown. Error bars
indicate plus and minus one standard deviation. Group F had solutions changed weekly, while
Groups A and B had solutions changed approximately monthly. Groups A and B show roughly
the same zinc concentration after 35 days as Group F after 7 days. This suggests saturation
behavior at ~50 ppm Zn2+. At intermediate times the F Group shows the ECU panels adding
more aqueous zinc to the solution than the control panels, especially on days 28, 35, and 42.
Interestingly, these days are correlated with the most rapid deviations in the F Group corrosion
potential shown in Figure 5.5. After extended immersion the level of zinc in solution from all
109
70
Group A, Control
Group A, ECU
50
Group B, Control
Group B, ECU
40
Aqueous Zn
2+
(mg/kg) (ppm)
60
Group F, Control
Group F, ECU
30
20
10
0
7
13
21
28
35
42
49
56
63
70
77
84
91
98
105 112 120 126 133 140
Immersion Time (Days)
Figure 5.6. Zinc concentration averages for the first 140 days of immersion, measured by ICP. Error bars show plus and minus one standard
deviation for the average of each group. Later times showed similar concentrations near 5 ppm (see Figure 5.7).
110
F-Group panels approaches ~5 ppm after a week in a given solution. These values were found to
fluctuate together over time, as seen in Figure 5.7. This is not an artifact of stripping
voltammetry as many days worth of samples were tested on a single day. An environmental
factor such as lab temperature may have affected zinc concentration.
10
F1, ECU
F2, Control
F3, ECU
F4, Control
F5, ECU
F6, Control
Aqueous Zn
2+
(mg/kg) (ppm)
9
8
7
6
5
4
3
2
1
0
147 175 183 190 196 203 210 217 224 231 240 247 253 260 268 275 282 289 295 301
Immersion Time (Days)
Figure 5.7. Detail of later zinc concentrations for Group F. Concentrations vary together suggesting
an environmental factor such as lab temperature affecting zinc concentration.
5.3.4. Large volume study
Results of the borate-buffered, pH 8.2, large volume study are shown in Figure 5.8. Both
panels output nearly identical zinc to solution during the first five days. Then deviation begins
with the control solution containing more aqueous zinc. Equilibrium is reached where the control
solution contains 4-5 ppm Zn, and the ECU solution contains 3-4 ppm Zn, with an average
111
difference (from Day 14 onwards) of 1.0 ppm. Concentrations vary up and down together,
suggesting an environmental factor affecting zinc solubility.
After one day, white deposits began to appear on both plates. After two days a somewhat
gelatinous blob was observed on both plates, most likely Zn(OH)2 (see Figure 5.9). Solubility of
zinc varies strongly with pH in this pH region; by as much as two orders of magnitude between
5
4
3
Aqueous Zn
2+
(mg/kg) (ppm)
pH 7 and pH 881 (Figure 5.10).
pH 8.2, Borate Buffer
2
100 L, Control
100 L, ECU
1
0
0
5
10
15
20
25
30
35
40
Immersion Time (Days)
45
50
55
Figure 5.8. Large volume (100 L) experiment to monitor zinc concentration. Solution was buffered to
pH 8.2 with 44 mM borate/boric acid.
112
Figure 5.9. Photographs of the panels in the large volume experiment. Both sides are shown for each
panel on Day 2 and Day 26. The appearance was similar from Day 26 to the end of the experiment.
Figure 5.10. Solubility of zinc species with respect to pH. Reproduced from Stumm and Morgan81.
113
5.3.5. Microscopy and structural analysis
Cross-sectional SEM analysis is a useful tool for understanding the structure of the
coating and relating that structure to the observed corrosion chemistry. Figure 5.11 shows a
cross-sectional SEM image of a virgin panel, having never been exposed to salt water. Densely
packed spherical zinc particles are clearly visible. Grey intervening space is indicative of the
aluminosilicate matrix which binds the zinc particles together. Dark black regions are due to the
mounting epoxy, including pores in the coating where epoxy could infiltrate. The underlying
steel substrate is visible at the bottom of the image. All these structures were verified by EDS
elemental analysis of various spots on the sample.
114
Zinc
Pore
Epoxy
Matrix
Steel
Figure 5.11. Cross-sectional SEM image of a virgin panel showing spherical zinc particles within the
aluminosilicate matrix on the underlying steel substrate. Dark black areas correspond to the mounting epoxy
including pores where epoxy could infiltrate the coating.
Upon exposure to salt water, corrosion begins and corrosion products begin to form a
barrier layer on the surface. Figure 5.12 shows the SEM cross-section of a panel after four
months of corrosion (panel A5). Corrosion products can be seen filling the pores of the coating
as well as covering the entire surface in a barrier layer. It is this sealing of the surface by
corrosion products that give ZRC’s in general, and Zetan in particular, their excellent long-term
corrosion inhibition properties56. See Section 5.3.6 for identification of these products.
115
Barrier layer
Filled pores
Figure 5.12. ECU panel (A5) after four months of corrosion. Pores have been filled in by corrosion
products and a barrier layer is present over the entire surface.
5.3.6. Identification of corrosion products
Corrosion products could be found in three distinct locations: in place of zinc particles
(see Figure 5.13), filling the original pores, and in the barrier layer (see Figure 5.12). Spherical
zinc particles corroded in place, leaving behind a zinc chloride containing species, as detected by
EDS. Slow scan XRD analysis clearly identified two species of corrosion products (Figure 5.14):
simonkolleite68, Zn5(OH)8Cl2·H2O; and hydrozincite69, Zn5(CO3)2(OH)6. There was also a minor
signal observed for zincite, ZnO. Zn(OH)2 was not observed, though it may be present in an
amorphous form, and therefore not detected by XRD. Since the corroded zinc particle sites were
116
the only areas on the sample where a strong chloride signals were observed, it is concluded that
simonkolleite is what replaces the metallic zinc. It is inferred, then, that hydrozincite is the
species found both filling the original pores and in the barrier layer. This is supported by EDS
analysis showing zinc and oxygen but no chloride signal in these regions. In addition, images
like Figure 5.12 indicate a smooth transition between the material in the barrier layer and that in
the pores, suggesting they are the same material.
Partially corroded
zinc particle
Corroded
zinc particle
Barrier layer
Figure 5.13. Cross-sectional SEM image of an ECU panel showing simonkolleite corrosion product in
place of completely and partially corroded zinc particles.
117
Figure 5.14. XRD of panels F4 (Control) (red, offset) and F5 (ECU) (black). Major signals are from simonkolleite (purple, PDF 7-0155),
hydrozincite (green, PDF 19-1458), and zinc metal (blue, PDF 4-0831). A minor signal is apparent for ZnO (orange, 65-3411). Crystalline Zn(OH)2 was
not detected. Zinc metal peaks were allowed to go off-scale to show detail.
118
5.3.7. Porosity
A virgin panel was shown in Figure 5.11, indicating pores in the coating into which
mounting epoxy was able to infiltrate. These pores fill with corrosion products over time as
shown in Figure 5.12. Pores were observed to be completely filled after four months of
corrosion. After only one month of corrosion, the barrier layer was already in place to the extent
that it blocked infiltration of the mounting epoxy into the pores (see Figure 5.15). After two and
three months, pores became successively filled in with corrosion products. The slow pace of this
process is indicative of the mitigated transport of oxygen, chloride ions, and carbonate ions
across the barrier layer.
Figure 5.15. Control panel, Group A, after one month of corrosion. Void spaces are evident in the
pores. Pores were not penetrated by the mounting epoxy due to the blocking effect of the barrier layer.
119
5.3.8. Density and thickness of the barrier layer, Group F
Analysis of Group F by SEM and EDS indicates a higher density in the barrier layer for
panels undergoing corrosion with an ECU in place. The barrier layer is somewhat porous for
both ECU and control panels; by observing the carbon signal due to the mounting epoxy as a
function of position via EDS, it is clear that epoxy penetrates partway into the barrier layer,
though the weak carbon signal makes this difficult to quantify. An examination by EDS of the
relative amount of zinc in the portion of the barrier layer adjacent to the original ZRC was
undertaken to quantify any differences due to the ECU. The mean density of zinc (as a
percentage of x-ray counts vs. pure zinc) in the barrier layer was 40% ± 5% for control panels
(example scan shown in Figure 5.16) and 48% ± 4% for ECU panels (example scan shown in
Figure 5.17). A statistical t-test on these distributions yielded a 99.99% probability that the two
means were statistically different.
Sixteen EDS line scans consisting of 100 points each were taken across three control
panels. Similarly, sixteen line scans were taken across three panels with an ECU. Lines were
selected to end within a particle known to be pure zinc, providing an in-situ standard for x-ray
counts from pure zinc. For the barrier layer, an average of zinc x-ray counts was taken for the ten
points in the barrier layer adjacent to the interface with the original ZRC. This allows analysis of
the earliest-formed portion of the barrier layer. The location of the interface was indicated by a
strong rise in the Al, Si, or Cl signals; Al and Si are present in the ZRC, and Cl is evident in the
simonkolleite, Zn5(OH)8Cl2·H2O, left behind when a zinc particle corrodes in place. The ratio of
the average zinc x-ray counts from the barrier layer to that from the pure zinc particles yielded a
zinc density in the barrier layer, expressed as a percentage vs. pure zinc.
120
100 points were taken in each line scan regardless of the thickness of the barrier layer.
Points were evenly spaced across the line scan. Therefore, the average densities for the thicker
layers associated with the control plates were calculated over larger physical region than for the
thinner barrier layers on the control plates. This was done under the assumption that the thicker
layers grew faster than the thinner ones, and so taking an average of the ten points nearest the
ZRC/barrier layer interface would give a consistent measure of relatively consistent time period
of growth. A second analysis was done avoiding this assumption by calculating the density based
only on the one point in the measurement nearest the interface. This gave similar results with
600
500
X-ray Counts
400
O
Al
Si
Cl
Zn
300
200
100
0
0.00
5.00
10.00
15.00
20.00
Distance (um, from top to bottom)
25.00
30.00
Figure 5.16. Control plate after one year of corrosion. Line scan showing x-ray counts for each
element, terminating in a pure zinc particle. Barrier layer near the interface produces 37.7% of the x-ray
counts produced by a pure zinc particle.
121
600
X-ray Counts
500
400
O
Al
Si
Cl
Zn
300
200
100
0
0
5
10
15
20
Distance (um, from top to bottom)
25
30
Figure 5.17. ECU plate after one year of corrosion. Line scan showing x-ray counts for each element,
terminating in a pure zinc particle. Barrier layer near the interface produces 49.1% of the x-ray counts
produced by a pure zinc particle.
122
larger standard deviations. With single point analysis, control panels showed an average density
of 41% ± 7% of a pure zinc particle, and the ECU showed 48% ± 6%. These means are
statistically different to 99.7% probability.
Barrier layer thickness was spot checked at several positions on all samples during
microscopy analysis, and data are presented in Table 5.1. The Group F control panels showed the
greatest barrier layer thickness, as well as the greatest variability.
Table 5.1. Barrier layer thickness (with standard deviations) measured by SEM.
Group A
Group B
Group F
Control
24 ± 9 μm, n=25
18 ± 6 μm, n=22
34 ± 23 μm, n=24
ECU
18 ± 5 μm, n=24
18 ± 9 μm, n=22
12 ± 4 μm, n=24
5.3.9. Adhesion of the barrier layer
ECU samples also showed superior adhesion of the barrier layer to the ZRC substrate.
Significant cracks were often observed at the barrier layer/ZRC interface for the control samples
(Figure 5.18), while the ECU samples showed very few such cracks (Figure 5.19).
123
Figure 5.18. Cross-sectional SEM images of control panels from Group F at 1600x magnification.
Cracks are visible where the barrier layer meets the ZRC, as indicated by the arrows.
124
Figure 5.19. Cross-sectional SEM images of ECU panels from Group F at 1600x magnification. Note
the lack of cracks where the barrier layer meets the ZRC.
125
5.3.10. Impedance spectroscopy
Impedance spectroscopy results changed over time for both Group G and Group H. The
Nyquist plots typically displayed two main time constants. Inductive behavior at high frequency
was observed for earlier immersion times (Figure 5.20). A Warburg diffusion tail was observed
at longer immersion times (Figure 5.22).
12
8
15
6
Zimaginary (Ω)
Zimaginary (Ω)
10
4
10
5
0
2
0
0
0
5
10
Zreal (Ω)
15
5
10
Zreal (Ω)
15
20
Figure 5.20. Typical Nyquist plot (circles) for early immersion times (13 days). Line is the fit from the
model circuit shown below. Inset shows equal axes, while main figure is shown full scale to show details.
126
50
30
100
Zimaginary (Ω)
Zimaginary (Ω)
40
20
80
60
40
20
10
0
0
0
0
20
40
60
Zreal (Ω)
80
20 40 60 80 100
Zreal (Ω)
100
Figure 5.21. Typical Nyquist plot (circles) for mid-range immersion times (83 days). Line is the fit
from the model circuit shown below. Inset shows equal axes, while main figure is shown full scale to show
details.
100
10
Zimaginary (Ω)
Zimaginary (Ω)
15
80
60
40
20
5
0
0
20 40 60 80 100
Zreal (Ω)
0
0
20
40
60
Zreal (Ω)
80
100
Figure 5.22. Typical Nyquist plot (circles) for long immersion times (159 days). Line is the fit from
the model circuit shown below. Inset shows equal axes, while main figure is shown full scale to show details.
127
The model circuit employed to fit the weekly data is shown in Figure 5.23. This circuit
was chosen as a simple model and was based on that used by Xie, et al.62 in the analysis of ZRC
on steel, with either epoxy-polyamide or ethyl silicate binder. The inductor component and
Warburg impedance were added for the current study based on clear features in the Nyquist
plots. The more complex transmission line model55 may be a more realistic representation of the
system. However, with the variability from panel to panel, it was desired to follow the general
trends in the data, which was most expediently done with the model shown here. The same
model was applied across all the data. Any parameter that did not converge to an error less than
100% was discarded. Clear outliers were also discarded.
R1
L
R2
R0
W
CPE1
(Y1, n1)
CPE2
(Y2, n2)
Figure 5.23. Model circuit used for EIS analysis of zinc rich coating on steel.
L: The inductance associated with the many current paths through the material.
R0: The external resistance due to wire leads and solution resistance.
R1: Charge transfer resistance associated with zinc dissolution.
CPE1: Double layer capacitance for zinc dissolution. Contains capacitive term, Y1, and
exponent, n1.
R2: Resistance through the binder.
128
CPE2: Capacitance between zinc particles separated by the binder. Contains capacitive term,
Y2, and exponent, n1.
W: Warburg coefficient, given as calculated in the ZSimpWin software as the magnitude of the
admittance (1/Z) at a frequency of 1 rad/s. Therefore, a lower value of W indicates greater
Warburg impedance.
Full results of least-squares fitting are presented below for all eight parameters from
Group G analysis (Figure 5.26-Figure 5.33) and Group H analysis (Figure 5.35-Figure 5.42).
Parameters that did not converge in the fit were removed from the data. Key results are
summarized here:
1) Inductive behavior uniformly decreases over time.
2) Charge transfer resistance, R1, increases over time.
3) Double layer capacitance, Y1, decreases over time, reaching a stable low value after
70 days for Group G and after 100 days for Group H.
4) Binder resistance, R2, increases over time, reaching a peak near 120 days for several
plates in Group G and Group H before decreasing slightly.
5) Particle-to-particle capacitance, Y2, decreases over time, reaching a stable value
higher than that of the double layer capacitance, Y1.
6) The values of n1 and n2 are close to 0.5, indicative of a porous electrode60.
7) Warbug impedance increases over time (i.e. W, the Warburg coefficient decreases).
Group F panels tested after a year of immersion showed the largest difference between
the control panels and the ECU panels. Nyquist plots for Group F are presented in Figure 5.24.
The overall resistances are similar (barring the F-4 outlier), but the ECU plates consistently show
a Warburg impedance tail.
129
50
Zimaginary (Ω)
40
Control:
F-2
F-4
F-6
30
20
ECU:
F-1
F-3
F-5
10
50
100
150
Zreal (Ω)
200
250
Figure 5.24. Nyquist plots for Group F after one year of immersion. Resistances are approximately
equal, barring the F-4 outlier. ECU panels show a Warburg tail beginning at higher frequency than the
control panels.
Photos after completing the study of Groups G and H (five months) show discolored
circles due to EIS measurement, either due to clamping or due to the applied voltage during the
measurement.
130
Figure 5.25. Photographs of a panel after six months of weekly EIS testing. The left image is of the
front of the panel which was subjected to clamping by an o-ring and was the working electrode in EIS. The
right image is the back of the panel which was contacted by a circular Teflon pad during clamping.
131
Control:
G-2
G-4
G-6
5
R1
L
R2
R0
ECU:
4
Inductance, L (uH)
Group G, L
W
CPE1
(Y1, n1)
G-1
G-3
G-5
CPE2
(Y2, n2)
3
2
1
0
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.26. Inductance (L) results, Group G. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus and
minus one standard deviation from least-squares fitting to the model circuit shown.
132
4
Control:
G-2
G-4
G-6
Group G, R0
R1
L
R2
R0
W
ECU:
Resistance, R0 (Ω)
3
G-1
G-3
G-5
CPE1
(Y1, n1)
CPE2
(Y2, n2)
2
1
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.27. Resistance (R0) results, Group G. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus and
minus one standard deviation from least-squares fitting to the model circuit shown.
133
40
Control:
G-2
G-4
G-6
Group G, R1
R1
L
ECU:
Resistance, R1 (Ω)
30
R2
R0
G-1
G-3
G-5
W
CPE1
(Y1, n1)
CPE2
(Y2, n2)
20
10
0
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.28. Resistance (R1) results, Group G. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus and
minus one standard deviation from least-squares fitting to the model circuit shown.
134
6
5
Control:
G-2
G-4
G-6
Group G, Y1
Capacitance, Y1 (mF)
ECU:
G-1
G-3
G-5
4
R1
L
R2
R0
W
3
CPE1
(Y1, n1)
CPE2
(Y2, n2)
2
1
0
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.29. Capacitance (Y1) results, Group G. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus and
minus one standard deviation from least-squares fitting to the model circuit shown.
135
1.4
1.2
Control:
G-2
G-4
G-6
Group G, n1
R1
L
R0
W
ECU:
G-1
G-3
G-5
CPE1
(Y1, n1)
1.0
Exponent, n1
R2
CPE2
(Y2, n2)
0.8
0.6
0.4
0.2
0.0
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.30. Exponent (n1) results, Group G. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus and minus
one standard deviation from least-squares fitting to the model circuit shown.
136
R1
L
Group G, R2
R2
R0
W
400
Resistance, R2 (Ω)
CPE1
(Y1, n1)
CPE2
(Y2, n2)
300
200
Control:
G-2
G-4
G-6
ECU:
100
G-1
G-3
G-5
0
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.31. Resistance (R2) results, Group G. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus and
minus one standard deviation from least-squares fitting to the model circuit shown.
137
30
Control:
G-2
G-4
G-6
25
Group G, Y2
R1
L
R2
R0
W
Capacitance, Y2 (mF)
ECU:
G-1
G-3
G-5
20
CPE1
(Y1, n1)
CPE2
(Y2, n2)
15
10
5
0
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.32. Capacitance (Y2) results, Group G. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus and
minus one standard deviation from least-squares fitting to the model circuit shown.
138
1.4
Control:
G-2
G-4
G-6
1.2
Group G, n2
R1
L
R2
R0
W
ECU:
G-1
G-3
G-5
Exponent, n2
1.0
CPE1
(Y1, n1)
CPE2
(Y2, n2)
0.8
0.6
0.4
0.2
0.0
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.33. Exponent (n2) results, Group G. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus and minus
one standard deviation from least-squares fitting to the model circuit shown.
139
Control:
H-2
H-4
H-6
-1 1/2
Warburg Admittance (Ω s ) at ω=1 rad/s
0.12
Group G, W
R1
ECU:
L
R2
R0
W
H-1
H-3
H-5
0.10
CPE1
(Y1, n1)
0.08
CPE2
(Y2, n2)
0.06
0.04
0.02
0.00
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.34. Warburg coefficient (W) results, Group G. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus
and minus one standard deviation from least-squares fitting to the model circuit shown.
140
Control:
H-2
H-4
H-6
4
Group H, L
R1
L
ECU:
H-1
H-3
H-5
Inductance, L (uH)
R2
R0
W
CPE1
(Y1, n1)
3
CPE2
(Y2, n2)
2
1
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.35. Inductance (L) results, Group H. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus and
minus one standard deviation from least-squares fitting to the model circuit shown.
141
Group H, R0
R1
L
W
CPE1
(Y1, n1)
1.5
Resistance, R0 (Ω)
R2
R0
CPE2
(Y2, n2)
1.0
0.5
Control:
H-2
H-4
H-6
ECU:
H-1
H-3
H-5
0.0
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.36. Resistance (R0) results, Group H. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus and
minus one standard deviation from least-squares fitting to the model circuit shown.
142
80
Control:
H-2
H-4
H-6
Group H, R1
ECU:
Resistance, R1 (Ω)
60
H-1
H-3
H-5
R1
L
R0
R2
W
CPE1
(Y1, n1)
40
CPE2
(Y2, n2)
20
0
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.37. Resistance (R1) results, Group H. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus and
minus one standard deviation from least-squares fitting to the model circuit shown.
143
Group H, Y1
15
Capacitance, Y1 (mF)
Control:
H-2
H-4
H-6
R1
L
ECU:
10
R2
R0
W
CPE1
(Y1, n1)
H-1
H-3
H-5
CPE2
(Y2, n2)
5
0
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.38. Capacitance (Y1) results, Group H. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus and
minus one standard deviation from least-squares fitting to the model circuit shown.
144
1.6
Group H, n1
1.4
Control:
H-2
H-4
H-6
1.2
R1
L
R2
R0
W
CPE1
(Y1, n1)
CPE2
(Y2, n2)
Exponent, n1
ECU:
H-1
H-3
H-5
1.0
0.8
0.6
0.4
0.2
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.39. Exponent (n1) results, Group H. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus and minus
one standard deviation from least-squares fitting to the model circuit shown.
145
2000
Control:
H-2
H-4
H-6
Group H, R2
R1
L
R2
R0
W
ECU:
Resistance, R2 (Ω)
1500
H-1
H-3
H-5
CPE1
(Y1, n1)
CPE2
(Y2, n2)
1000
500
0
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.40. Resistance (R2) results, Group H. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus and
minus one standard deviation from least-squares fitting to the model circuit shown.
146
60
Group H, Y2
Capacitance, Y2 (mF)
50
40
Control:
H-2
H-4
H-6
R1
L
R2
R0
W
CPE1
(Y1, n1)
ECU:
H-1
H-3
H-5
CPE2
(Y2, n2)
30
20
10
0
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.41. Capacitance (Y2) results, Group H. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus and
minus one standard deviation from least-squares fitting to the model circuit shown.
147
Control:
H-2
H-4
H-6
1.0
Group H, n2
R1
L
ECU:
Exponent, n2
W
CPE1
(Y1, n1)
H-1
H-3
H-5
0.8
R2
R0
CPE2
(Y2, n2)
0.6
0.4
0.2
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.42. Exponent (n2) results, Group H. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus and minus
one standard deviation from least-squares fitting to the model circuit shown.
148
Control:
H-2
H-4
H-6
-1 1/2
Warburg Admittance (Ω s ) at ω=1 rad/s
0.10
Group H, W
R1
ECU:
L
R2
R0
W
H-1
H-3
H-5
0.08
CPE1
(Y1, n1)
CPE2
(Y2, n2)
0.06
0.04
0.02
0.00
20
40
60
80
100
Immersion Time (Days)
120
140
Figure 5.43. Warburg coefficient (W) results, Group H. Control panels are open symbols, ECU panels are filled symbols. Error bars show plus
and minus one standard deviation from least-squares fitting to the model circuit shown.
149
5.4.
Discussion
5.4.1. Zinc surface area
The corrosion potential starts near that of pure zinc and slowly rises to that of steel as
corrosion progresses. This can be roughly understood as following the ratio of available surface
area of the two metals: initially, the high surface area of the particulate zinc dominates; later the
surface area of zinc decreases as particles become smaller and corrosion products block the
surface; finally, nearly all zinc surface area becomes inaccessible, and the potential settles near
that of bare steel.
Capacitance data from EIS experiments give some insight into zinc surface area behavior.
Double layer capacitance, Y1, decreased steadily over time, reaching a stable low value after 70
days for Group G (see Figure 5.29), and 100 days for Group H (see Figure 5.38). This behavior
is likely due to the reduced surface area of zinc as spherical particles corrode, as well as the
increased passivation by simonkolleite deposits. The fact that the charge transfer resistance, R1
(Figure 5.28 and Figure 5.37), also increases over time supports the interpretation of an
increasing conversion of metallic zinc to simonkolleite around the zinc particles. Particle-toparticle capacitance, Y2, also falls over time as the surface area of zinc falls.
The inductive behavior, especially at early immersion times, is an unexpected
observation (Figure 5.26 and Figure 5.35). A possible explanation is that current flow to actively
corroding areas of the sample induces currents in parallel pathways of electrically connected zinc
particles. As these particles corrode, they lose electrical continuity, and reduce the variety of
electrically connected pathways within the ZRC. Hence, inductive behavior drops over time.
150
5.4.2. Complex equilibrium
The patterns of corrosion products are quite different over the three control panels. This
may be due to the complex specific flow patterns within each beaker, with effects due to small
differences in the placement of the panel within the beaker and the centering of the beaker on the
stir plate. It is clear from microscopy (see Section 0) that a relatively dense barrier layer of
corrosion products is present on both control and ECU panels. The white deposits seen on the
surface of the control panels may be indicative of corrosion that was too rapid to be completely
precipitated in the barrier layer and formed blooms of corrosion products on the surface.
While each group exhibited a slower rise in corrosion potential (indicating slower
corrosion) with the ECU in place, the largest difference was seen in Group F with weekly
solution changes. The Group F ECU results are approximately in line with Groups A and B,
while the Group F control shows significantly increased corrosion potential. Groups A and B
were effectively inhibited against corrosion by the buildup of aqueous corrosion products,
primarily Zn2+. Confinement to the two-liter beaker resulted in high concentrations of Zn2+
which inhibited the further dissolution of Zn (see Section 5.3.3). This is unrealistic in open
systems such as the ocean or soil, and by changing the solution regularly in Group F this effect
was minimized. It was then the ECU which inhibited corrosion in the ECU panels while the
control panels showed a more rapid rise in potential when allowed to corrode freely.
After extended immersion the level of zinc in solution from all F Group panels
approaches ~5 ppm after a week in a given solution. These values were found to fluctuate
together over time, as seen in Figure 5.7. This is not an artifact of stripping voltammetry as many
days worth of samples were tested on a single day. It seems some environmental factor caused
the concentrations to move in unison, possibly the temperature in the lab or the ambient CO2
151
concentration affecting the equilibrium. It is possible that the ~5 ppm zinc concentrations come
from contamination of the solution used, however it is unlikely that zinc was present in such
concentrations. Commercial distilled water was used to prepare the solutions and while blank
zinc concentrations were unfortunately not measured, but the US Environmental Protection
Agency secondary standard for zinc in public drinking water is 5 ppm82. With such a low value
expected in public drinking water, it is unlikely that residue from the distillation process would
be up to the ppm range.
Groups A and B show roughly the same zinc concentration after 35 days as Group F after
7 days (Figure 5.6). This suggests saturation behavior at ~50 ppm Zn2+. There is a complex
equilibrium established between aqueous zinc, chloride, carbonate, and hydroxide ions,
molecular oxygen, and solid zinc and zinc corrosion products (zincite, hydrozincite, and
simonkolleite, see Section 5.3.6)83. At long immersion times the zinc concentration for all plates
settles into the 5 ppm range. Certainly corrosion is less rapid at this point with the barrier layer
firmly in place, but the fact that all panels, regardless of duration since the previous solution
change, show the same concentration argues that it is again an equilibrium effect. This may be
due to a simpler equilibrium that is primarily based on hydrozincite since the solution only has
access to the hydrozincite barrier layer which seals off access to the interior of the coating (see
Section 5.3.6).
Preis and Gamsjager84 report a detailed thermodynamic analysis of the (Zn2+ + H2O +
CO2) system including hydrozincite [Zn5(OH)6(CO3)2], smithsonite [ZnCO3], and zincite [ZnO].
A preponderance diagram of these phases with respect to partial pressure of CO2 and pH is
presented in Figure 5.44. These calculations were made at a temperature of 25°C and assuming a
Zn2+ activity of 10-4. log(p(CO2)) in ambient air is -3.5. The dependence on pH is significant. At
152
pH 7 this diagram would indicate that Zn2+ would be the stable phase, but the pH locally within
the ZRC and barrier layer may be significantly higher than the pH of the bulk solution. Any OHproduced at cathodic regions would need to diffuse out into solution (or be precipitated in a solid
deposit). Especially under the slow diffusion conditions present with the buildup of corrosion
products, a concentration gradient will exist for OH- from high values (high pH) within the
coating to the lower values (pH 7) in solution. At ambient CO2 levels, pH above 7.2 would favor
the deposition of hydrozincite.
Figure 5.44. Two-dimensional preponderance diagram for (Zn2+ + H2O + CO2) at 25°C and Zn2+
activity of 10-4 with partial pressure of CO2 plotted against pH.84 Reprinted with permission from Elsevier.
At intermediate times for the F Group the ECU panels add more aqueous zinc to the
solution than the control panels, especially on days 28, 35, and 42 (Figure 5.6). Interestingly,
these days are correlated with the most rapid deviations in the F Group corrosion potential shown
in Figure 5.5. According to the above hypothesis, the control panels reach the pure hydrozincite
equilibrium more quickly, though it appears corrosion at these times continues to happen rapidly
153
as evidenced by the rapid rise in corrosion potential (i.e. loss of active Zn surface area). This
could be explained by a more rapidly grown but less dense barrier layer in the control panels. A
more porous barrier layer would allow corrosive species (Cl-, O2) access to the coating to
continue rapid corrosion, but a large surface area of hydrozincite would facilitate precipitation of
zinc ions as they diffused out. The ECU panels, on the other hand, may have densely formed
hydrozincite layers in some areas at these intermediate times, while still maintaining a more
complex equilibrium with simonkolleite and perhaps metallic zinc directly exposed to solution.
In the pH 7 study, more aqueous zinc was seen from the ECU panels. In the large volume
study buffered at pH 8.2 the reverse was true (Figure 5.8). Perhaps at pH 8.2 the equilibrium
remains complex and there is a different mix of compounds on the surface between the control
and ECU panels. These panels showed much more visible corrosion products than those
immersed at pH 7. Up to Day 5, the control and ECU panels agree closely and exhibit a steady
rate of concentration increase, suggesting that any barrier layer is not at all active during this
period.
5.4.3. Porosity of the barrier layer and ion mobility
Corrosion proceeds in the ZRC with the anodic dissolution of zinc balanced by oxygen
reduction at cathodic regions, either on zinc particles or the steel substrate.
Anode:
Cathode:
Zn → Zn2+(aq) + 2e½O2(aq) + H2O + 2e- → 2OH-(aq)
EQUATION 5.1
EQUATION 5.2
The concomitant diffusion of ions to maintain charge balance cannot be ignored. There is
migration of chloride ions to the anodic regions while sodium ions migrate to the cathodic
regions. The already high Cl- concentration in the electrolyte (3% NaCl by weight, or 0.5M)
would therefore be increased locally at the site of zinc dissolution. This favors the formation of
154
the insoluble simonkolleite [Zn5(OH)8Cl2·H2O] precipitate, and explains the spatial observation
of simonkolleite formation replacing the original metallic zinc particles.
The hydrozincite barrier layer was present after only one month of immersion (Figure
5.15). At this point it was already sufficiently dense so as to block the infiltration of the
mounting epoxy into the still-empty pores of the ZRC. The formation of this layer is believed to
be the key to the corrosion inhibition process. The specific morphology of the barrier layer is
affected by the ECU making it more effective for limiting access of the electrolyte to the ZRC
and underlying steel substrate.
It was observed that the barrier layer on the ECU panels is indeed more dense and thinner
than on the control panels. The Group F control panels showed the greatest barrier layer
thickness, as well as the greatest variability (Table 5.1). This further supports the hypothesis of a
thick but relatively porous barrier layer for the control panels. ECU samples showed a denser
barrier layer (Figure 5.16 and Figure 5.17) and superior adhesion of the barrier layer to the ZRC
substrate as evidenced by the lack of cracks at the barrier layer/ZRC interface.
XRD results presented in Figure 5.14 of an ECU and control panel after one year of
corrosion indicate some subtle differences that support the conclusion of a denser barrier layer
due to the ECU. Several angles, most notably at 13° and 24°, show a stronger diffraction signal
for hydrozincite from the ECU plate than the control. This would indicate a higher degree of
crystallinity. The low angle peak at 13° corresponds to a d-spacing of 6.8 Å and is due to the
stacking of the layers within the hydrozincite structure. The a-axis in the unit cell is 13.6 Å (2 x
6.8) and encompasses two stacked planes69. The presence of more stacking faults within the
structure on the control panel would reduce the signal at 13° and weaken the crystallites of
hydrozincite.
155
Resistance due to corrosion products is indicated by R2 in the EIS results (Figure 5.31
and Figure 5.40). Resistance rises over time, and peaks for several samples at around Day 120.
Differences between the control and ECU panels are not clearly evident, though this is likely due
to the limited solution changes done with Groups G and H. The drop in resistance after the peak
may be indicative of breaches in the barrier layer.
One possibility for the difference in density is that overall corrosion rate is reduced by the
effect of the ECU on ionic diffusion within the porous binder and corrosion product matrix. If
the time scale for ion transport between anodic and cathodic regions is similar to the
characteristic time scale of the ECU, diffusive charge transport could be hindered. Slower
diffusion of corrosive species to the zinc particles, as well as slower diffusion of Zn2+ and OHproducts after corrosion occurs would both act to slow the corrosion process. The resultant
slower growth of the barrier layer may allow for denser crystallite packing and reduced stacking
faults in the hydrozincite structure.
Further evidence of the effect of the ECU on ion mobility is provided by the EIS Nyquist
plots for Group F after one year of immersion. This is the clearest evidence of a difference
between control and ECU panels seen by EIS. Figure 5.24 clearly demonstrates a reproducible
Warburg impedance tail for the three ECU panels. While the F-4 control panel also shows a
Warburg tail, the resistance is much lower than all the others, making direct comparison difficult.
Warburg impedance appears when ions cannot diffuse quickly enough in the material to support
the charge transfer processes driven by the perturbation frequency. The Warburg impedance
becomes visible by the low frequency limit of this study, 1 Hz, for the ECU panels. This
indicates that diffusions limited charge transfer at these perturbation frequencies, while diffusion
did not yet impact the control panels. The built-up corrosion products more strongly resist the
156
diffusion of ions in the ECU panels, resulting in diffusion limited behavior at higher frequencies
than in the control panels.
These low frequencies of a few hertz are on the order of the time constant of the full
circuit with the ECU capacitor in place. A characteristic frequency of 1 Hz from a capacitor of
0.56 F would require a resistance in series of 0.3 Ω. Typical resistances measured for Group F
were 0.4 Ω.
The following model circuit is proposed as a representation of the corrosion process
within the ZRC during the galvanic phase of protection (Figure 5.45). Anodic and cathodic sites
are each represented by a charge transfer resistance and double layer capacitance in parallel. The
electrolytic coupling between them is represented as a Warburg impedance. This represents the
actual corrosion process and is distinct from the EIS process and the model circuit employed
there. The three-electrode setup of the EIS experiments effectively probes only one half-cell of
the corrosion system, while during active corrosion anodic and cathodic sites are both active.
EECU: Potential of the ECU (12 V)
CECU: Capaciance of the ECU (0.56 F)
Enoise: Random potential fluctuations due to discrete corrosion events
Ecorr: Average corrosion potential of the ZRC/steel system
RCT: Charge transfer resistance at the anode (a) or cathode (c)
CDL: Double layer capacitance at the anode (a) or cathode (c)
W: Warburg impedance of the electrolytic coupling between anodic and cathodic sites
157
ECU
EECU
CECU
RCT,a
RCT,c
W
Enoise Ecorr
CDL,a
CDL,c
Figure 5.45. Model circuit for the corrosion process within a ZRC protected by an ECU.
The ECU is comprised of a charged capacitor. There is also capacitance within the
corrosion system. As the corrosion potential fluctuates over time a flow of charge is required,
either via corrosion reactions or from the ECU. If charge can flow from the ECU to the corrosion
system before more corrosion occurs, the fluctuation in the corrosion potential can be brought
back to the average value. A key parameter appears to be the Warburg impedance associated
with the electrolyte path between anodic and cathodic sites. This affects the time constant of
corrosion and therefore the required time constant of the ECU.
The above circuit was simulated using National Instruments Multisim 10 software
(formerly Electronics Workbench), with the following parameters: EECU=12 V; CECU=0.56 F;
Enoise=1 mV at various frequencies; Ecorr=1 V; RCT,a=200Ω;CDL,a=1.8 mF; W was simulated as a
resistor of 1kΩ; RCT,c=20Ω; CDL,c=1 mF. The current was monitored over time across EECU (the
158
ECU battery) as simulated sinusoidal noise of varying frequency was applied at Enoise. The
current response was sinusoidal and ranged in magnitude from 1.2 μA at 1 mHz to 1.7 μA at
10 kHz. These results indicate that charge does indeed flow from the ECU battery in response to
corrosion noise in the circuit.
This, then, indicates the best hypothesis for the mechanism of enhanced corrosion
protection by the ECU. A circuit is formed with a characteristic frequency on the order of the
diffusion time of ions within the coating. The capacitive action reduces diffusive charge transport
within the pores and the barrier layer resulting in slower precipitate formation and denser
precipitates. The improved barrier layer more effectively inhibits infiltration of the electrolyte,
and therefore extends the life of the coating.
5.5.
Conclusions
An electronic control unit (ECU) consisting of a 12 V battery and a 0.56 F capacitor in
series provided enhanced corrosion protection to a high weight-loading zinc rich coating on steel
upon immersion in 3% NaCl solution as seen visually and by corrosion potential measurements.
Weekly solution changes to avoid zinc saturation in solution were necessary to see well
differentiated results.
Simonkolleite [Zn5(OH)8Cl2·H2O] was found to form as a corrosion product in place of
the original zinc particles. Hydrozincite [Zn5(CO3)2(OH)6] was deposited within the pores of the
coating and on the surface as a barrier layer. The barrier layer was in place at the first
observation after 30 days of immersion. The barrier layer was denser and more adherent with the
ECU in place.
The ECU is proposed to act by a mechanism in which the characteristic time constant of
the ECU is roughly matched to the time scale of ionic motion within the coating. The capacitive
159
nature of the ECU mitigates random fluctuations in corrosion potential before extensive
corrosion can occur.
160
References
(1)
Deluca, N. W.; Elabd, Y. A. Journal of Polymer Science Part B-Polymer Physics
2006, 44, 2201-2225.
(2)
Mann, J.; Yao, N.; Bocarsly, A. B. Langmuir 2006, 22, 10432-10436.
(3)
Zhu, Y. M.; Ha, S. Y.; Masel, R. I. Journal of Power Sources 2004, 130, 8-14.
(4)
Bullen, R. A.; Arnot, T. C.; Lakeman, J. B.; Walsh, F. C. Biosensors &
Bioelectronics 2006, 21, 2015-2045.
(5)
Gorte, R. J.; Kim, H.; Vohs, J. M. Journal of Power Sources 2002, 106, 10-15.
(6)
Mauritz, K. A.; Moore, R. B. Chemical Reviews 2004, 104, 4535-4585.
(7)
Adjemian, K. T.; Dominey, R.; Krishnan, L.; Ota, H.; Majsztrik, P.; Zhang, T.;
Mann, J.; Kirby, B.; Gatto, L.; Velo-Simpson, M.; Leahy, J.; Srinivasant, S.; Benziger, J. B.;
Bocarsly, A. B. Chemistry of Materials 2006, 18, 2238-2248.
(8)
Antonucci, P. L.; Arico, A. S.; Creti, P.; Ramunni, E.; Antonucci, V. Solid State
Ionics 1999, 125, 431-437.
(9)
Watanabe, M.; Uchida, H.; Seki, Y.; Emori, M.; Stonehart, P. Journal of the
Electrochemical Society 1996, 143, 3847-3852.
(10)
Schechter, A.; Savinell, R. F. Solid State Ionics 2002, 147, 181-187.
(11)
Seland, F.; Berning, T.; Borresen, B.; Tunold, R. Journal of Power Sources 2006,
160, 27-36.
(12)
Paik, C. H.; Saloka, G. S.; Graham, G. W. Electrochemical and Solid State Letters
2007, 10, B39-B42.
161
(13)
Colomban, P. Proton conductors: Solids, membranes and gels - materials and
devices; University Press: Cambridge, 1992.
(14)
Haile, S. M.; Boysen, D. A.; Chisholm, C. R. I.; Merle, R. B. Nature 2001, 410,
910-913.
(15)
Denuzzio, J. D.; Farrington, G. C. Journal of Solid State Chemistry 1989, 79, 65-
(16)
Dell, R. M.; Moseley, P. T. Journal of Power Sources 1981, 7, 45-63.
(17)
Dell, R. M.; Moseley, P. T. Journal of Power Sources 1981, 6, 143-160.
(18)
Briant, J. L.; Farrington, G. C. Journal of Solid State Chemistry 1980, 33, 385-
(19)
Stevens, R.; Binner, J. G. P. Journal of Materials Science 1984, 19, 695-715.
(20)
Sudworth, J. L.; Barrow, P.; Dong, W.; Dunn, B.; Farrington, G. C.; Thomas, J.
74.
390.
O. MRS Bulletin 2000, 25, 22-26.
(21)
Farrington, G. C.; Briant, J. L. Science 1979, 204, 1371-1379.
(22)
Schafer, G. W.; Kim, H. J.; Aldinger, F. Solid State Ionics 1995, 77, 234-239.
(23)
Ochadlick, A. R.; Story, H. S.; Farrington, G. C. Solid State Ionics 1981, 3-4, 79-
(24)
Baffier, N.; Badot, J. C.; Colomban, P. Solid State Ionics 1984, 13, 233-236.
(25)
Smoot, S. W.; Whitmore, D. H.; Halperin, W. P. Solid State Ionics 1986, 18-9,
84.
687-693.
(26)
Tsai, Y. T.; Smoot, S.; Whitmore, D. H.; Tarczon, J. C.; Halperin, W. P. Solid
State Ionics 1983, 9-10, 1033-1040.
162
(27)
Thomas, J. O.; Farrington, G. C. Acta Crystallographica Section B-Structural
Science 1983, 39, 227-235.
(28)
Sudworth, J. L.; Tilley, A. R. The Sodium Sulfur Battery; Chapman and Hall:
London, 1985.
(29)
Sutorik, A. C.; Neo, S. S.; Treadwell, D. R.; Laine, R. M. Journal of the American
Ceramic Society 1998, 81, 1477-1486.
(30)
Kim, W. S.; Kim, W. J.; Lim, S. K. Journal of Industrial and Engineering
Chemistry 2004, 10, 78-84.
(31)
Kutty, T. R. N.; Jayaraman, V.; Periaswami, G. Materials Research Bulletin 1996,
31, 1159-1168.
(32)
Subasri, R.; Mathews, T.; Sreedharan, O. M.; Raghunathan, V. S. Solid State
Ionics 2003, 158, 199-204.
(33)
Subasri, R.; Roy, S.; Matusch, D.; Nafe, H.; Aldinger, F. Journal of the American
Ceramic Society 2005, 88, 1740-1746.
(34)
Subasri, R. Materials Science and Engineering B-Solid State Materials for
Advanced Technology 2004, 112, 73-78.
(35)
Elliot, A. G.; Huggins, R. A. Journal of the American Ceramic Society 1975, 58,
497-500.
(36)
Mazza, D.; Vallino, M.; Busca, G. Journal of the American Ceramic Society
1992, 75, 1929-1934.
(37)
Foster, P. A. Journal of the Electrochemical Society 1959, 106, 971-975.
(38)
Boilot, J. P.; Thery, J. Materials Research Bulletin 1976, 11, 407-414.
(39)
Pechini, M. P. US Patent 3,330,697, 1967.
163
(40)
Mathews, T., 20% excess sodium.
(41)
Nicholson, P. S.; Nagai, M.; Yamashita, K.; Sayer, M.; Bell, M. F. Solid State
Ionics 1985, 15, 317-326.
(42)
Crosbie, G. M.; Tennenhouse, G. J. Journal of the American Ceramic Society
1982, 65, 187-191.
(43)
Iyi, N.; Grzymek, A.; Nicholson, P. S. Solid State Ionics 1989, 37, 11-16.
(44)
Angel, R. J.; Prewitt, C. T. American Mineralogist 1986, 71, 1476-1482.
(45)
Milius, D., Princeton University, Personal Communication, 2007.
(46)
Boilot, J. P.; Collin, G.; Colomban, P.; Comes, R. Physical Review B 1980, 22,
5912-5923.
(47)
de Kroon, A. P.; Schafer, G. W.; Aldinger, F. Chemistry of Materials 1995, 7,
878-887.
(48)
Park, S. M.; Hellstrom, E. E. Solid State Ionics 1991, 46, 221-231.
(49)
Munshi, M. Z.; Nicholson, P. S. Solid State Ionics 1987, 23, 203-209.
(50)
Kuo, C. K.; Tan, A. C.; Nicholson, P. S. Solid State Ionics 1992, 53-6, 58-62.
(51)
Tan, A. C.; Kuo, C. K.; Nicholson, P. S. Solid State Ionics 1991, 45, 137-142.
(52)
de Kroon, A. P.; Gstrein, F.; Schafer, G. W.; Aldinger, F. Solid State Ionics 2000,
133, 107-120.
(53)
Biezma, M. V.; Cristobal, J. R. S. Corrosion 2006, 62, 1051-1055.
(54)
Munger, C. G.; Vincent, L. D. Corrosion Prevention by Protective Coatings; 2nd
ed.; NACE: Houston, 1999.
(55)
Feliu, S.; Barajas, R.; Bastidas, J. M.; Morcillo, M. Journal of Coatings
Technology 1989, 61, 63-69.
164
(56)
Feliu, S.; Barajas, R.; Bastidas, J. M.; Morcillo, M. Journal of Coatings
Technology 1989, 61, 71-76.
(57)
Abreu, C. M.; Izquierdo, M.; Merino, P.; Novoa, X. R.; Perez, C. Corrosion 1999,
55, 1173-1181.
(58)
Gervasi, C. A.; Disarli, A. R.; Cavalcanti, E.; Ferraz, O.; Bucharsky, E. C.; Real,
S. G.; Vilche, J. R. Corrosion Science 1994, 36, 1963-1972.
(59)
Romagnoli, R.; Vetere, V. F. Jocca-Surface Coatings International 1993, 76,
(60)
Vilche, J. R.; Bucharsky, E. C.; Giudice, C. A. Corrosion Science 2002, 44, 1287-
(61)
Xie, D. M.; Wang, J. M.; Hu, J. M.; Zhang, J. Q. Transactions of Nonferrous
208-&.
1309.
Metals Society of China 2002, 12, 1036-1039.
(62)
Xie, D. M.; Wang, J. M.; Hu, J. M.; Zhang, J. Q. Transactions of Nonferrous
Metals Society of China 2003, 13, 421-425.
(63)
Mansfeld, F. Journal of Applied Electrochemistry 1995, 25, 187-202.
(64)
Feliu, S.; Morcillo, M.; Bastidas, J. M. Journal of Coatings Technology 1991, 63,
(65)
Kalewicz, Z. JOCCA-Surface Coatings International 1985, 68, 299-305.
(66)
Lindqvist, S. A.; Meszaros, L.; Svenson, L. JOCCA-Surface Coatings
31-34.
International 1985, 68, 10-14.
(67)
Newton, D. S.; Sampson, F. G. Jocca-Surface Coatings International 1965, 48,
(68)
Hawthorne, F. C.; Sokolova, E. Canadian Mineralogist 2002, 40, 939-946.
382.
165
(69)
Ghose, S. Acta Crystallographica 1964, 17, 1051-1057.
(70)
Gusmano, G.; Montesperelli, G.; Pacetti, S.; Petitti, A.; Damico, A. Corrosion
1997, 53, 860-868.
(71)
Bastidas, J. M.; Feliu, S.; Morcillo, M. Progress in Organic Coatings 1990, 18,
265-273.
(72)
Recarey, L. E.; Bermudez, A. S.; Madrinan, S. U.; Alvela, F. B. Revista De
Metalurgia 2001, 37, 24-33.
(73)
Skerry, B. S.; Eden, D. A. Progress in Organic Coatings 1991, 19, 379-396.
(74)
Riffe, W. J. US Patent 5,055,165, 1991.
(75)
Riffe, W. J. US Patent 5,352,342, 1994.
(76)
Riffe, W. J. US Patent 5,478,451, 1995.
(77)
Dowling, D. B. US Patent 6,562,201, 2003.
(78)
Dowling, D. B.; Khorrami, F. US Patent 6,811,681, 2004.
(79)
Dowling, D. B., Applied Semiconductor, Inc., Personal communication, 2006.
(80)
Uhlig, H. H.; Revie, R. W. Corrosion and Corrosion Control; 3rd ed.; John Wiley
& Sons, Inc.: New York, 1985.
(81)
Stumm, W.; Morgan, J. J. Aquatic chemistry: chemical equilibria and rates in
natural waters; 3rd ed.; John Wiley & Sons: New York, 1995.
(82)
U.S. EPA, Drinking Water Contaminants,
http://www.epa.gov/safewater/contaminants/index.html, 2008.
(83)
Alwan, A. K.; Williams, P. A. Transition Metal Chemistry 1979, 4, 128-132.
(84)
Preis, W.; Gamsjager, H. Journal of Chemical Thermodynamics 2001, 33, 803-
819.
166
Документ
Категория
Без категории
Просмотров
0
Размер файла
6 111 Кб
Теги
sdewsdweddes
1/--страниц
Пожаловаться на содержимое документа