Cathode Design for High Energy Molten Salt Lithium-Oxygen Batteries - CaltechTHESIS

Cathode Design for High Energy Molten Salt Lithium-Oxygen Batteries - CaltechTHESIS
CaltechTHESIS
A Caltech Library Service
About
Browse
Deposit an Item
Instructions for Students
Cathode Design for High Energy Molten Salt Lithium-Oxygen Batteries
Citation
Tozier, Dylan Douglas
(2018)
Cathode Design for High Energy Molten Salt Lithium-Oxygen Batteries.
Dissertation (Ph.D.), California Institute of Technology.
doi:10.7907/TG0K-8776.
Abstract
State of the art commercial lithium ion batteries use cathodes such as lithium cobalt oxide which rely on insertion and removal of lithium ions from a host material. However, insertion cathode materials are limited in their capacity, and replacing them with a cathode that employs growth and dissolution of new phases could significantly increase a battery’s energy density. For example, oxygen and sulfur cathodes have been widely researched to this end, with both cases involving the growth of a lithium-rich compound on a current collector/catalyst support.
We begin by describing the effect of using a molten salt electrolyte in a lithium-oxygen battery. In particular, we focus on how the electrochemical performance and discharge product, lithium peroxide, differ from that of a traditional organic electrolyte. In addition, we discuss the enhanced peroxide solubility in a molten salt and its implications for lithium peroxide growth and coulombic efficiency. Finally, we address the cell death of a galvanostatically cycled battery.
We then introduce a similar phase-forming conversion chemistry, whereby a molten nitrate salt serves as both an active material and the electrolyte. Molten nitrate salts were previously studied as an active material in a primary lithium battery where lithium oxide irreversibly forms as nitrate reduces to nitrite. We will describe how the use of a nanoparticle heterogeneous catalyst allows the reversible growth and dissolution of micron-scale lithium oxide crystals in this system.
After introducing these molten salt lithium batteries, we address the effect of cathode geometry on electrochemical performance. In particular, we note that the growth of such large, solid phase species on the surface of the catalyst support imposes new design restrictions when optimizing a cathode for energy density. As a proof of concept, we design and implement an architected electrode with large pore volume and relatively small surface area, comparing it with the more typical geometries of thin films and nanoparticles.
Item Type:
Thesis (Dissertation (Ph.D.))
Subject Keywords:
Battery; Lithium-Oxygen; Molten Salt; Structured Electrode
Degree Grantor:
California Institute of Technology
Division:
Engineering and Applied Science
Major Option:
Materials Science
Thesis Availability:
Public (worldwide access)
Research Advisor(s):
Greer, Julia R.
Thesis Committee:
Fultz, Brent T. (chair)
Greer, Julia R.
Johnson, William Lewis
See, Kimberly
Defense Date:
31 May 2018
Record Number:
CaltechTHESIS:06072018-140534228
Persistent URL:
DOI:
10.7907/TG0K-8776
Related URLs:
URL
URL Type
Description
DOI
Article adapted for Ch. 2
ORCID:
Author
ORCID
Tozier, Dylan Douglas
0000-0001-9489-8824
Default Usage Policy:
No commercial reproduction, distribution, display or performance rights in this work are provided.
ID Code:
11037
Collection:
CaltechTHESIS
Deposited By:
Dylan Tozier
Deposited On:
08 Jun 2018 23:59
Last Modified:
08 Nov 2023 00:27
Thesis Files
Preview
PDF
- Final Version
See Usage Policy.
10MB
Repository Staff Only:
item control page
CaltechTHESIS is powered by
EPrints 3.3
which is developed by the
School of Electronics and Computer Science
at the University of Southampton.
More information and software credits
Cathode Design for High Energy Molten Salt
Lithium-Oxygen Batteries

Thesis by

Dylan D. Tozier

In Partial Fulfillment of the Requirements for the
Degree of
Doctor of Philosophy

CALIFORNIA INSTITUTE OF TECHNOLOGY
Pasadena, California

2018
Defended May 31, 2018

ii

Dylan D. Tozier
ORCID: 0000-0001-9489-8824

iii

ACKNOWLEDGEMENTS

Just over seven years ago I somehow stumbled into Prof Ajayan’s lab at Rice
University. I had never even considered going to grad school, but my time in his
lab working with Akshay Mathkar sent me down the path to Caltech. Throughout
this time, I’ve met some extraordinary people, my advisor Julia being one of them.
I admire your infectious enthusiasm, and you’ve made doing research at Caltech a
wonderful experience. To the rest of the Greer Group, I am always amazed by how
you can be both so good at science and so much fun. I couldn’t have asked for better
group members. To everyone over at Liox, you have been fantastic collaborators
and like a second group to me. I’ve learned so much from you all, and I can’t
believe how lucky I was to get to work as closely with you as I did. To the many
other amazing people I have had the pleasure to work with, Betar, Carol, Nathan,
and others I’m sure I am leaving out, I appreciate how generous you were with your
time. It’s people like you who make Caltech the great place that it is. To all the
current and former members of Steuben House, thank you for the good times. It’s
been like coming home to family, and I would be hard pressed to find a better group
of people to live with. To Patricia, thanks for putting up with me all these years. I
guess this means it’s your turn to go to grad school. And finally, to my parents who
have encouraged and supported me every step of the way, it’s hard to believe that the
last time I wasn’t a student, you were making homework for me to do with Garrett
so I wouldn’t feel left out. I guess I’m finally finished. Mom and Dad, thanks for
everything.

iv

ABSTRACT

State of the art commercial lithium ion batteries use cathodes such as lithium cobalt
oxide which rely on insertion and removal of lithium ions from a host material.
However, insertion cathode materials are limited in their capacity, and replacing
them with a cathode that employs growth and dissolution of new phases could
significantly increase a battery’s energy density. For example, oxygen and sulfur
cathodes have been widely researched to this end, with both cases involving the
growth of a lithium-rich compound on a current collector/catalyst support.
We begin by describing the effect of using a molten salt electrolyte in a lithiumoxygen battery. In particular, we focus on how the electrochemical performance
and discharge product, lithium peroxide, differ from that of a traditional organic
electrolyte. In addition, we discuss the enhanced peroxide solubility in a molten salt
and its implications for lithium peroxide growth and coulombic efficiency. Finally,
we address the cell death of a galvanostatically cycled battery.
We then introduce a similar phase-forming conversion chemistry, whereby a molten
nitrate salt serves as both an active material and the electrolyte. Molten nitrate salts
were previously studied as an active material in a primary lithium battery where
lithium oxide irreversibly forms as nitrate reduces to nitrite. We will describe how
the use of a nanoparticle heterogeneous catalyst allows the reversible growth and
dissolution of micron-scale lithium oxide crystals in this system.
After introducing these molten salt lithium batteries, we address the effect of cathode
geometry on electrochemical performance. In particular, we note that the growth of
such large, solid phase species on the surface of the catalyst support imposes new
design restrictions when optimizing a cathode for energy density. As a proof of
concept, we design and implement an architected electrode with large pore volume
and relatively small surface area, comparing it with the more typical geometries of
thin films and nanoparticles.

PUBLISHED CONTENT AND CONTRIBUTIONS

1. Giordani, V., Tozier, D., Tan, H., Burke, C. M., Gallant, B. M., Uddin, J., Greer,
J. R., McCloskey, B. D., Chase, G. V. & Addison, D. A Molten Salt LithiumOxygen Battery. Journal of the American Chemical Society 138, 2656–2663.
doi: 10.1021/jacs.5b11744 (2016).
D. Tozier assisted with battery assembly and electrochemical analysis, performed the ex situ characterization, and participated in the writing of the
manuscript.
2. Addison, D., Tozier, D., Tan, H., Uddin, J., Gallant, B. M., McCloskey, B. D.,
Greer, J. R., Chase, G. V. & Giordani, V. A New Rechargeable Battery: Lithium
Oxide Growth through Molten Salt Nitrate Reduction. (In Preparation).
D. Tozier assisted with battery assembly and electrochemical analysis, performed the ex situ characterization, and wrote much of the manuscript.
3. Tozier, D., Giordani, V., Chase, G. V., Addison, D. & Greer, J. R. The Role of
Battery Cathode Architecture in Conversion Reaction Chemistries. (In Preparation).
D. Tozier fabricated and characterized the samples, assembled the batteries,
performed the electrochemical analysis, and wrote the manuscript.

vi

TABLE OF CONTENTS

Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iii
Abstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iv
Published Content and Contributions . . . . . . . . . . . . . . . . . . . . . . v
Table of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi
List of Illustrations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
List of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix
Chapter I: Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 A Brief History of Batteries . . . . . . . . . . . . . . . . . . . . . . 1
1.2 The Lithium-Oxygen Battery . . . . . . . . . . . . . . . . . . . . . 4
1.3 Thesis Overview . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
Chapter II: A Molten Salt Lithium-Oxygen Battery . . . . . . . . . . . . . . 11
2.1 Molten Salt Electrolytes . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2 Oxygen Electrochemistry in a Molten Salt . . . . . . . . . . . . . . 13
2.3 Lithium Peroxide and Wulff Constructions . . . . . . . . . . . . . . 18
2.4 Battery Cycling Performance and Cell Death . . . . . . . . . . . . . 26
2.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Chapter III: A Nitrate Reduction Battery . . . . . . . . . . . . . . . . . . . . 30
3.1 Nitrate Reduction Chemistry . . . . . . . . . . . . . . . . . . . . . . 31
3.2 Catalysis of Nitrite Oxidation . . . . . . . . . . . . . . . . . . . . . 33
3.3 Lithium Oxide as a Discharge Product . . . . . . . . . . . . . . . . . 37
3.4 Nitrate to Nitrite and Back Again . . . . . . . . . . . . . . . . . . . 42
3.5 Battery Cycling Performance . . . . . . . . . . . . . . . . . . . . . 45
3.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
Chapter IV: Role of Cathode Architecture in Conversion Reaction Chemistries 48
4.1 Capacity Limitations . . . . . . . . . . . . . . . . . . . . . . . . . . 49
4.2 Fabrication of Architected Electrodes . . . . . . . . . . . . . . . . . 53
4.3 Electrochemical Surface Area of Nickel Electrodes . . . . . . . . . . 57
4.4 Performance of Architected Electrodes . . . . . . . . . . . . . . . . 62
4.5 Commercial Foam Electrode Comparison . . . . . . . . . . . . . . . 66
4.6 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71
Chapter V: Summary and Outlook . . . . . . . . . . . . . . . . . . . . . . . 72
Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
Appendix A: Detailed Cell Construction . . . . . . . . . . . . . . . . . . . . 82
Appendix B: Ion Exchange Chromatography Calibration and Calculation . . . 85
Appendix C: Nanoscribe Speed Optimization . . . . . . . . . . . . . . . . . 88
Appendix D: Wetting of Architected Electrodes . . . . . . . . . . . . . . . . 91

vii

LIST OF ILLUSTRATIONS

Number
Page
1.1 Lithium-ion battery schematic . . . . . . . . . . . . . . . . . . . . . 3
1.2 Lithium-oxygen battery schematic . . . . . . . . . . . . . . . . . . . 5
1.3 Typical lithium-oxygen first cycle . . . . . . . . . . . . . . . . . . . 6
1.4 Typical morphologies of Li2 O2 . . . . . . . . . . . . . . . . . . . . 7
1.5 Growth mechanisms of Li2 O2 . . . . . . . . . . . . . . . . . . . . . 8
1.6 Gutmann acceptor and donor number of solvents . . . . . . . . . . . 9
2.1 Cyclic voltammetry of oxygen reduction and evolution . . . . . . . . 14
2.2 Galvanostatic cycling of oxygen reduction and evolution . . . . . . . 16
2.3 Mass spectrometry of lithium-oxygen charge process . . . . . . . . . 18
2.4 X-ray diffraction of lithium-oxygen cathode across one cycle . . . . . 19
2.5 Wulff construction schematic . . . . . . . . . . . . . . . . . . . . . 20
2.6 Wulff construction of Li2 O2 . . . . . . . . . . . . . . . . . . . . . . 21
2.7 Morphology of oxygen cathode discharge product . . . . . . . . . . . 22
2.8 Elemental map of oxygen cathode discharge product . . . . . . . . . 23
2.9 Elemental map of oxygen cathode after full cycle . . . . . . . . . . . 23
2.10 Linear sweep voltammograms of Li2 O2 oxidation . . . . . . . . . . . 25
2.11 Cycling behavior of lithium-oxygen battery . . . . . . . . . . . . . . 26
2.12 X-ray diffraction of cycled lithium-oxygen battery . . . . . . . . . . . 27
2.13 Scanning electron microscopy of cycled lithium-oxygen battery . . . 27
3.1 Nitrate reduction first cycle on a Super P carbon cathode . . . . . . . 32
3.2 Nitrate reduction schematic . . . . . . . . . . . . . . . . . . . . . . 32
3.3 Catalysis comparison for nitrate reduction reversibility . . . . . . . . 33
3.4 Galvanostatic intermittent titration technique of nitrate reduction . . . 34
3.5 X-ray photoelectron spectroscopy of nickel nanoparticles . . . . . . . 35
3.6 Transmission electron microscopy of nickel nanoparticles . . . . . . 36
3.7 Lithium-doped nickel oxide as a catalyst . . . . . . . . . . . . . . . . 38
3.8 X-ray diffraction of nitrate reduction cathode across one cycle . . . . 39
3.9 Transmission electron microscopy of nitrate reduction cathode . . . . 39
3.10 Morphology of nitrate reduction cathode discharge product . . . . . . 40
3.11 Elemental map of nitrate reduction cathode discharge product . . . . 41
3.12 Raman spectroscopy of nitrate reduction electrolyte across one cycle . 42

viii
3.13
3.14
3.15
4.1
4.2
4.3
4.4
4.5
4.6
4.7
4.8
4.9
4.10
4.11
4.12
4.13
4.14
4.15
A.1
B.1
C.1
D.1
D.2

Charging profile of newly constructed cells preloaded with Li2 O . . .
Ion exchange chromatography of nitrate/nitrite solution . . . . . . . .
Cycling behavior of nitrate reduction battery . . . . . . . . . . . . .
Surface of fully discharge cathodes . . . . . . . . . . . . . . . . . .
Example structured nickel electrodes . . . . . . . . . . . . . . . . .
Model of lattice electrode unit . . . . . . . . . . . . . . . . . . . . .
Fabrication of polymer lattice . . . . . . . . . . . . . . . . . . . . .
Sputtered nickel film coating a polymer lattice . . . . . . . . . . . .
Nickel lattice electrode . . . . . . . . . . . . . . . . . . . . . . . . .
Surface electrochemistry of nickel . . . . . . . . . . . . . . . . . . .
Electrochemical surface area of sputtered nickel . . . . . . . . . . . .
Electrochemical surface area of lattice electrode . . . . . . . . . . .
Cyclic voltammetry of nitrate reduction on different cathodes . . . .
Galvanostatic discharge of nitrate reduction on different cathodes . .
Nitrate reduction discharge product on lattice electrode . . . . . . . .
Foam electrode morphologies . . . . . . . . . . . . . . . . . . . . .
Cyclic voltammetry of nitrate reduction on foam cathodes . . . . . .
Galvanostatic discharge of nitrate reduction on foam cathodes . . . .
Picture of Swagelok cell . . . . . . . . . . . . . . . . . . . . . . . .
Ion exchange chromatography nitrate calibration scans . . . . . . . .
Nanoscribe parameter sweep . . . . . . . . . . . . . . . . . . . . . .
Wettability of lattice electrode . . . . . . . . . . . . . . . . . . . . .
Wetting of a lattice electrode . . . . . . . . . . . . . . . . . . . . . .

43
45
46
51
52
54
55
56
57
58
60
61
62
64
64
67
68
69
83
85
89
91
92

ix

LIST OF TABLES
Number
Page
1.1 Theoretical energy density of batteries . . . . . . . . . . . . . . . . . 4
2.1 Molten salt electrolyte compositions . . . . . . . . . . . . . . . . . . 13
4.1 Acid bath for nickel surface cleaning . . . . . . . . . . . . . . . . . 59
4.2 Normalized discharge capacity of nitrate reduction on different cathodes 65
4.3 Structure properties of discharged foam cathodes . . . . . . . . . . . 69

Chapter 1

INTRODUCTION

The ability to store energy and release it as needed plays a surprisingly important
role in our lives, so much so that we often forget how integral energy storage has
become to a modern lifestyle. For instance, this thesis is written on a laptop, a
device that is only possible due to significant advancements in battery technology
over the course of decades. And as the performance of batteries has grown over
the years, so too has the range of applications where they can play a part. While
today it may seem inevitable that electric vehicles are set to replace their gasoline
counterparts as automakers announce electrified lineups and countries announce
electrification goals[1], it is easy to forget that just 10 years ago Tesla released their
first electric vehicle to a skeptical public. Such a dramatic change in perspective
is largely thanks to the precipitous drop in battery prices over that time period[2].
Along a similar vein, renewable energy sources such as wind and solar have begun to
provide a significant portion of the power in our electric grid, but their intermittency
has created a need for large scale, cheap energy storage. We are already seeing
batteries begin to play a role in this space, and the improvement of existing battery
chemistries and development of new ones will only see their part grow. Between
consumer electronics, electric vehicles, grid storage, and many more, the range of
applications for batteries, in addition to the demands on them, is only growing.
The scope of this thesis is dedicated to understanding some of these new battery
chemistries, but to fully do so, it is instructive to look back at how we got to where
we are today.
1.1

A Brief History of Batteries

For the purposes of this thesis, it is best to divide the history of batteries into
three simple categories: before, during, and after lithium-ion batteries. As this is
a brief history, many chemistries will be overlooked. A commonality across all of
these battery chemistries is that the underlying physics is the same: an electrolyte
separates two electrodes where redox reactions occur, one an oxidation and the
other a reduction, and the difference in reduction potential of the two reactions is
the voltage across the battery.

Before Lithium-Ion
The first true battery was invented in 1800 by Alessandro Volta, after whom the
unit Volt is named[3]. While electrochemistry was not understood at the time, his
voltaic stack was a true electrochemical cell consisting of zinc and copper electrodes
separated by a sulfuric acid electrolyte, giving the half cell reactions of Equation
1.1.
Zn + SO42− → ZnSO4 + 2e−

(1.1a)

2H + + 2e− → H2

(1.1b)

Similar batteries were used through much of the 1800s until the development of the
lead-acid battery, the first rechargeable battery, in 1859[4]. The ability to reuse these
batteries led to an expanded range of applications for batteries, and the lead-acid
chemistry is still widely used today.
After the development of the lead-acid battery, many other primary and reversible
chemistries came into use, including zinc-carbon, nickel-cadmium, alkaline, and
nickel-metal hydride to name a few. However, a battery with larger energy density
was desired, and lithium became a heavily researched battery constituent due to its
low atomic mass and reduction potential. While lithium batteries were commercialized as primary batteries, developing a reversible chemistry proved challenging.
The Lithium-Ion Battery
Beginning in the 1970s, several groups performed the work necessary to develop
the lithium-ion battery, with Stanley Whittingham demonstrating the reversible
intercalation of lithium[5, 6], John Goodenough the LiCoO2 cathode[7], and Rachid
Yazami the reversible intercalation of lithium into a graphite anode[8]. Building
off of these works, Sony was able to commercialize the lithium-ion battery in 1991,
first with a hard carbon anode, and later a graphite one.
The operation of the lithium-ion battery is significantly different than the rechargeable batteries that preceded it. Both electrodes employ a lithium intercalation
mechanism, a process by which lithium ions reversibly enter and exit a host crystal
structure. Typical positive electrodes are transition metal oxides such as LiCoO2
and LiFePO4 [9], and the typical negative electrode is a carbon such as graphite or
amorphous carbon. Because the reduction potential of lithium insertion into these
carbons is close to that of lithium (ca. 0.2 V vs Li/Li+ ) and the reduction potential
of lithium insertion into one of these transition metal oxides is around 3 - 4.5 V vs

Li/Li+ , the cell potential is far too large to use an aqueous electrolyte. As a result,
an organic electrolyte such as a mixture of lithium hexafluorophosphate salt in an
ethylene carbonate and diethyl carbonate solution (LiPF6 in EC/DEC) is used[10].
A schematic of the discharge process in such a lithium-ion battery is shown in Figure
1.1, with the lithium ions leaving the host structure of the negative electrode and
entering the host structure of the positive electrode, and the charge process sees the
reverse happen.

Figure 1.1: Lithium-ion battery schematic. Adapted from [10].
The lithium-ion chemistry makes a compelling battery for several reasons, but chief
among them is its achievable energy density. Since its implementation in 1991,
many new transition metal oxides cathodes (LMO, NMC, and NCA to name a few)
have been developed, allowing this energy density to continue to grow. The capacity
of a typical graphite anode is ca. 300 mAh/g, and that of a typical transition metal
oxide is approaching 200 mAh/g, allowing a real full cell energy density over 250
Wh/kg[11]. This is an impressive value, considering that the first Sony lithium-ion
battery in 1991 achieved an 80 Wh/kg energy density[12]. There is still plenty of
rich work being done to continue improving the lithium-ion battery, such as the
implementation of silicon anodes[13], 5 V cathodes[14], and lithium-rich transition
metal oxide cathodes[15].
Beyond Lithium-Ion
While the progress in lithium-ion batteries has been quite impressive over the last
couple decades, there is still a desire for larger energy density, cheaper, and safer
batteries. As a result, research groups have looked beyond the typical intercalation
chemistries of the lithium-ion battery. This can take many forms: replacing the
graphite anode with silicon or lithium metal, replacing the liquid inorganic elec-

trolyte with an aqueous or solid one, replacing the transition metal oxide cathode
with a conversion reaction chemistry, or moving away from lithium altogether, replacing it with a cheaper or multivalent ion. Each of these options is a field in
its own right. This thesis will focus on chemistries that replace the intercalation
based transition metal oxide cathode with one that does not require a host structure
for lithium, and Table 1.1 compares the theoretical energy density of current and
future intercalation cathodes with two of the more prominent "beyond lithium-ion"
cathode chemistries. By removing the mass of the host intercalation structure, far
higher theoretical energy densities can be achieved.

Cathode
LiCoO2
(LCO)
LiMn1.5 Ni0.5 O4
(high-voltage spinel)
LiNi0.8 Co0.15 Al0.05 O2
(NCA)
LiNi0.33 Mn0.33 Co0.33 O2
(NMC)
Li1.2 Ni0.15 Mn0.55 Co0.10 O2
(LMR-NMC)
Li2 S
(lithium-sulfur)
Li2 O2
(lithium-oxygen)

Theoretical
Avg. discharge Theoretical
specific capacity potential
specific energy
(mAh g-1 )
(V vs Li/Li+)
(Wh kg-1 )
164
3.8
625
147

4.7

690

195

3.8

740

166

3.8

620

267

3.65

970

1168

2.15

2510

1168

2.8

3330

Table 1.1: Theoretical energy density of batteries, as paired with a lithium metal
anode. Adapted from [16].

1.2

The Lithium-Oxygen Battery

The large theoretical energy density of lithium-oxygen batteries makes them an
attractive candidate for research, because if even a small fraction of that theoretical
value were practically realized, it would compete well with state of the art lithiumion chemistries. The first demonstration of the lithium-oxygen battery chemistry
was in 1996[17], and it has been well studied since, leading to a good understanding
of the underlying mechanisms and appreciation for what still needs to be solved.

Oxygen as a Cathode
When using oxygen as a cathode, the discharged state of the battery is less clear than
it is for a well structured intercalation one. For instance, one might imagine that
lithium oxide (Li2 O) is the discharged state. And this would be a good thing, as each
mole of oxygen gas could host 4 moles of lithium. However, as suggested by Table
1.1, the discharged state is in fact lithium peroxide (Li2 O2 ), giving 2 moles lithium
for each mole oxygen. Figure 1.2 is a simple schematic describing this discharge
process when paired with a lithium metal anode (as is typical in lithium-oxygen
research).

Figure 1.2: Lithium-oxygen battery schematic.

Note that a porous cathodic current collector, often a high surface area carbon, must
support the growth of the discharge product while also allowing diffusion of oxygen
from the gas phase to the electrode surface. In addition, this simple schematic
obfuscates the complex process of bringing together two lithium ions and an oxygen
molecule to the surface of the electrode where a two electron redox process must
take place. Due to this complexity, the possibility of a one step reaction is untenable,
and Equation 1.2 describes the generally accepted series of reactions[18].
O2 + e− → O2−

(1.2a)

O2− + Li + → LiO2

(1.2b)

2LiO2 → Li 2 O2 + O2

(1.2c)

Unlike the three step oxygen reduction reaction (ORR) on discharge, the oxygen
evolution reaction (OER) on charge is a single step evolution of oxygen. Regardless
of the reaction steps, the thermodynamic standard potential for the overall reaction
of Equation 1.2 is 2.96 V[19]. However, when performing this chemistry in an
actual battery, there is typically a significant overpotential on both discharge and
charge. Figure 1.3 shows the first cycle of lithium-oxygen batteries with different
carbons serving as the cathodic current collector.

Figure 1.3: Typical lithium-oxygen first cycle. 0.1 mA cm-2 . Adapted from [20].
This large overpotential suggests a few things: the mechanism of Li2 O2 growth on
discharge and dissolution on charge is probably complicated; the wide operating
potential window both reduces the round-trip energy efficiency of the battery and
lends the system to unwanted side reactions; and an effective catalyst for the reaction,
or some other mechanism of reducing the overpotential, is necessary for the lithiumoxygen battery to be a successful secondary battery.
Lithium Peroxide Growth Mechanisms
While Equation 1.2 describes the chemistry involved in forming Li2 O2 , it does not
address how it is supported by the cathodic current collector. In particular, it gives
no understanding of the morphology of Li2 O2 as it grows, the mechanism of Li2 O2
growth, or the relationship between this morphology and the performance of the
battery. There are a wide range of morphologies reported throughout literature[21–
23], and Figure 1.4 gives a few typical examples of them.
It is worth noting that Li2 O2 is insulating, so one would expect the thin film morphology in Figure 1.4 due to the inability of electrons to leave the surface of the

Figure 1.4: Typical morphologies of Li2 O2 . (A) Thin film (scale bar: 300 nm), (B)
platelet (scale bar: 1 µm), and (C) toroid (scale bar: 300 nm). Adapted from [24],
[25], and [26], respectively.

electrode once an insulating layer has formed. However, other morphologies are
also observed, including "toroid" structures on the scale of a micron, and the growth
of these larger particles should significantly increase the amount of Li2 O2 hosted by
the cathode, in turn increasing the battery’s capacity. If all the chemical reactions in
Equation 1.2 happened at the surface of the electrode, this morphology would not be
possible. This curious observation of large, electrochemically grown, yet insulating
crystals was difficult to explain until the recent development of more detailed growth
mechanisms[25, 27].
A schematic detailing the two potential growth mechanisms of Li2 O2 is shown in
Figure 1.5. In the surface growth mechanism, the first two steps of Equation 1.2
occur on the surface of the electrode, with a second reduction reaction then taking
place at the electrode. Since the entire ORR takes place on the electrode surface,
electrons must tunnel through any insulating Li2 O2 already present, resulting in a
film thickness no larger than about 5-10 nm[28]. In the solution mediated growth
mechanism, the first two steps of Equation 1.2 also take place on the surface of
the electrode. However, here the intermediate species, lithium superoxide (LiO2 ),
is at least sparingly soluble in the electrolyte, and as a result, the disproportionation reaction can take place away from the electrode surface. This allows the final
step in Equation 1.2 to occur in a location which does not have access to an electron, therefore enabling the growth of micron scale insulating particles through an
electrochemical process.
Then the Li2 O2 growth mechanism is dictated by the electrolyte, namely how well
the electrolyte can solvate the LiO2 intermediate species. In other words, the
extent of Lewis acidity and basicity of the electrolyte will determine which growth

Figure 1.5: Growth mechanisms of Li2 O2 . Adapted from [25].

mechanism occurs. A quantitative measure of this can be given by the Gutmann
acceptor and donor numbers, which essentially dictate how well a solvent can
donate or accept electrons[29]. Figure 1.6 gives the acceptor and donor numbers for
some commonly used lithium-oxygen electrolyte solvents such as dimethoxyethane
(DME) and dimethyl sulfoxide (DMSO) as well as impurities such as water. When
using DME as an electrolyte solvent, the surface growth mechanism takes place
and the capacity is poor, but when using DMSO or having a water impurity in
DME, the solution mediated growth mechanism predominates and the capacity is
far greater[27]. This suggests that finding an electrolyte with large acceptor and
donor numbers would increase the solvation of LiO2 , and thus give an even larger
discharge capacity.
An interesting result of the importance of LiO2 solvation is the role that a trace
amount of water can play in the electrolyte[30]. Increasing the water content from
an anhydrous rated <4ppm concentration to 1000ppm can increase the discharge
capacity by a factor of 10[31]. While this water impurity can act as an enabler of
the solution mediated growth process, it also can serve as an example of the role
that side reactions can play in the lithium-oxygen chemistry. In a similar approach,
researchers have attempted using soluble catalysts, known as redox mediators, to
encourage solution mediated growth without requiring solvation of LiO2 [32, 33].
Instability and Side Reactions
As with any secondary battery chemistry, the ability to reversibly cycle a lithiumoxygen battery is predicated on avoiding side reactions that consume active parts of
the battery. For example, the development of carbonate electrolytes in lithium-ion

Figure 1.6: Gutmann acceptor and donor number of solvents. Free energy of LiO2
dissolution, relative to DME. Adapted from [25].

batteries allowed the formation of a stable solid electrolyte interphase (SEI) on a
graphite anode, preventing further decomposition through exfoliation and allowing
graphite to intercalate reversibly[34]. Along a similar vein, it is important to select
the appropriate electrolyte and cathode material for a lithium-oxygen battery to cycle
reversibly.
As described above, water impurities can play a significant role in the lithiumoxygen battery. While it can enhance the solubility of LiO2 , it can also engage in a
side reaction, forming lithium hydroxide (LiOH), which is an irreversible reaction
in this system[30]. In addition to water impurities, degradation of the organic
electrolyte can lead to a wide range of irreversible decomposition products[35]. An
expansive search for organic electrolytes has yet to find one that does not suffer
from degradation[36–41]. Interestingly, it is thought that LiO2 is a primary culprit,
considering the reactivity of the superoxide radical to organic species. This leads
to the unfortunate situation where increasing the capacity through the solution
mediated growth mechanism also leads to irreversibility due to superoxide attack of
the electrolyte.
In addition to electrolyte degradation, the high surface area carbons often employed
as the cathode catalyst or the support for other catalysts also suffers from side
reactions. In particular, lithium carbonate (Li2 CO3 ) is a common side product
which is irreversible on charge[42, 43]. Searching for electrodes which are both
catalytic to ORR and OER while also being immune to these side reactions is an
active area of research[44–46].

10
The accumulation of these side products which are not removed during OER lead
to a passivation of the cathode as insulating species prevent electrons from reaching
the electrolyte. When enough of these side products have accumulated, the cell can
no longer function and the discharge capacity rapidly approaches zero in an event
called "sudden cell death"[42].
The instabilities of both the cathode and electrolyte mentioned above, in addition to a
host of other practical complications not mentioned here, demonstrate the difficulties
associated with successfully implementing a lithium-oxygen battery. This quick
review has only covered the aspects of the lithium-oxygen battery most relevant for
this thesis, and there are several excellent, thorough reviews for further reading[16,
19, 47, 48]. The complexity of the lithium-oxygen chemistry is a long way from
that first primary battery of Volta described in Equation 1.1. However, this lends it
to being a rich area of research with many problems to solve.
1.3

Thesis Overview

The aim of this thesis is to give a deeper understanding of battery cathodes that
employ a phase forming conversion reaction chemistry, and in doing so, provide
design guidelines so that one could optimize such a cathode. The first chapter of
this thesis has provided a brief introduction to the current state of battery research and
then given a detailed look into the lithium-oxygen battery. The second chapter will
probe the effects of operating a lithium-oxygen battery with a molten salt electrolyte,
noting the implications this gives for the chemistry more broadly. The third chapter
will introduce nitrate reduction chemistry, which is similar to the lithium-oxygen
battery, and demonstrate its ability to serve as a secondary battery. The fourth
chapter will use this nitrate reduction chemistry to explore the effects of cathode
structure on battery performance, drawing conclusions which can be more broadly
applied to other battery chemistries. Finally, the fifth chapter will reflect on the
implications of this thesis and the direction of the field moving forward.

11
Chapter 2

A MOLTEN SALT LITHIUM-OXYGEN BATTERY
Adapted with permission from
1. Giordani, V., Tozier, D., Tan, H., Burke, C. M., Gallant, B. M., Uddin, J., Greer,
J. R., McCloskey, B. D., Chase, G. V. & Addison, D. A Molten Salt LithiumOxygen Battery. Journal of the American Chemical Society 138, 2656–2663.
doi: 10.1021/jacs.5b11744 (2016).

Chapter Abstract
Despite the promise of extremely high theoretical capacity (2Li + O2 → Li 2 O2 ,
1675 mAh per gram of oxygen), many challenges currently impede development
of lithium-oxygen battery technology. Finding suitable electrode and electrolyte
materials remains the most elusive challenge to date. A radical new approach is
to replace volatile, unstable and air-intolerant organic electrolytes common to prior
research in the field with alkali metal nitrate molten salt electrolytes and operate
the battery above the liquidus temperature (>80 °C). In this chapter we demonstrate an intermediate temperature lithium-oxygen battery using a lithium anode, a
molten nitrate-based electrolyte (e.g., LiNO3 –KNO3 eutectic) and a porous carbon
oxygen cathode with high energy efficiency (~95%) and improved rate capability
because the discharge product, lithium peroxide, is stable and moderately soluble
in the molten salt electrolyte. The results, supported by essential state-of-the-art
electrochemical and analytical techniques such as in situ pressure and gas analyses,
scanning electron microscopy, rotating disk electrode voltammetry, demonstrate that
Li2 O2 electrochemically forms and decomposes upon cycling with discharge/charge
overpotentials as low as 50 mV. We show that the cycle life of such batteries is
limited only by carbon reactivity and by the uncontrolled precipitation of Li2 O2 ,
which eventually becomes electrically disconnected from the oxygen electrode.

12
2.1

Molten Salt Electrolytes

The Introduction Chapter demonstrated both the potential of lithium-oxygen batteries as a replacement for lithium-ion ones and the many challenges that must still be
overcome for them to be practically realized. In particular, finding sufficiently stable
electrolytes in the presence of the active lithium-oxygen cathode electrochemistry
has been the most elusive challenge to date. In addition, the insulating and insoluble nature of Li2 O2 passivates the cathode surface, limiting discharge capacity to
a small fraction of the large theoretical capacity provided by the lithium-oxygen
electrochemistry.
The approach described in this chapter is to replace unstable aqueous or organicbased electrolytes common in prior lithium-oxygen battery research with an organicfree molten alkali metal nitrate electrolyte and operate the battery above the eutectic
melting point, typically between 80 and 200 °C[49]. Research on molten nitrate
electrolytes for lithium batteries and thermal batteries traces back to the late 1970s
when it was observed that the reaction between metallic lithium and the nitrate anion
generated a SEI composed of Li2 O that was sufficiently stable for primary cells and
rechargeable cells with limited cycle life[50, 51]. However, reports on oxygen
electrode behavior in these electrolyte mixtures are somewhat lacking. Previous
works reported that equilibria between superoxide, peroxide, and oxide forms of
reduced oxygen could coexist in the molten salt electrolyte with high reversibility
at a platinum rotating disk electrode[52, 53]. Stability toward lithium, low melting
point relative to other inorganic salts, high thermal stability above 500 °C, nonvolatility, high ionic conductivity and acceptable electrochemical stability window
(typically between 2.2 and 3.8 V vs Li/Li+ on amorphous carbon black electrodes)
are attractive features that make the molten nitrate and nitrite class of electrolytes
interesting for lithium-oxygen batteries.
Furthermore, we hypothesize that LiO2 and Li2 O2 discharge products will have
enhanced solubility in this electrolyte compared to room temperature organic electrolytes, because previous work on molten salts as solvents have found them to have
acceptor numbers >100, a value well off the chart in Figure 1.6[54]. Together with
improved electrode kinetics at elevated temperature, we anticipate high reversibility
and higher rate capability for the oxygen electrode.
The composition and relevant material properties of molten salts used in this work
are described in Table 2.1. All salts were purchased from Sigma-Aldrich and vacuum dried for at least one week before use. The chosen compositions correspond to

13
the eutectic composition of the described mixture. Melting point was measured by
differential scanning calorimetry (DSC), lithium cation transference number by the
Bruce and Vincent method[55], and ionic conductivity at 150 °C by using a conductivity cell of known cell constant suitable for elevated temperature measurements.

Electrolyte
LiNO3 –
KNO3
LiNO3 –
KNO2 –
CsNO3

Chemical
Melting Lithium
Ionic
composition point
transference conductivity
(mol %)
(°C)
number
(mS cm-1 )
42–58
125
0.68
88
37–39–24

90

0.28

115

Table 2.1: Molten salt electrolyte compositions, with measurements of conductivity
properties at 150 °C.

2.2

Oxygen Electrochemistry in a Molten Salt

A combination of cyclic voltammetry (CV) and galvanostatic cycling are used to
compare the electrochemical behavior of a traditional organic electrolyte with the
molten salt electrolytes described in Table 2.1. These experiments were performed
in Swagelok cells with known volume that are paired with a pressure sensor and
valve to allow filling the head-space with gas and connecting the cell to a mass
spectrometer. Cathodes consisted of about 5 mg of Super P carbon and PTFE binder
95:5 mass ratio pressed onto a stainless steel mesh. A glass microfiber separator
was impregnated with about 300 mg of molten salt electrolyte and vacuum dried
overnight at 200 °C. For organic electrolytes, the separator was simply wet with
purchased premixed electrolyte. Anodes were a simple lithium foil. All assembly
took place in a glovebox, then the cells were filled with oxygen after a leak test,
and finally placed in an oven to be held at the operating temperature. A detailed
description of the cell design and assembly is given in the appendix.
Cyclic Voltammetry
Figure 2.1 compares CVs of oxygen reduction at a Super P carbon cathode in
a conventional lithium-oxygen battery electrolyte, 0.1 M LiClO4 –DMSO, and a
LiNO3 –KNO3 eutectic, at 30 and 150 °C, respectively. As expected, results indicate
that in the common organic electrolyte, oxygen reduction to form Li2 O2 is a highly
irreversible electrochemical process, as only a strong cathodic current and no anodic

14

Figure 2.1: Cyclic voltammetry of oxygen reduction and evolution. (a,c) 0.1 M
LiClO4 –DMSO at 30 °C and (b,d) LiNO3 –KNO3 eutectic at 150 °C. Working
electrode: Super P porous carbon (A = 0.785 cm2 ), carbon loading ~5 mg/cm2 .
Counter and reference electrodes: lithium metal. Scan rate: 0.05 mV/s. Voltage
window: 2.5–3.2 V vs Li/Li+ . Current density expressed in mA per g of carbon.

current is observed in the operating potential window. Previous results have shown
that high anodic potentials (>3.8 V) are necessary to oxidize Li2 O2 in DMSObased electrolytes, although at these potentials, a substantial fraction of the Li2 O2
oxidation current results in parasitic decomposition of the electrolyte. On the other
hand, CVs performed in the oxygen-saturated LiNO3 –KNO3 eutectic at 150 °C
shows high reversibility with Qa/Qc ratio of ~0.9 (anodic charge Qa and cathodic
charge Qc were obtained by integrating the “current vs time” plot derived from the
CV; Qa/Qc would ideally equal 1). To verify nitrate electrolyte electrochemical
stability in the operating voltage window, we performed the same CV experiment
under argon gas (without any oxygen present) and did not observe any faradaic
currents. Furthermore, pressure monitoring clearly demonstrates gas is consumed
upon reduction (cathodic scan) and generated upon oxidation (reverse, anodic scan).
Mass spectrometry on the head-space following the CV indicates that only oxygen,
and no other gas, is evolved during the anodic scan. As expected, no pressure rise is

15
observed during the anodic CV scan of a DMSO-based battery, as no anodic current
is observed, indicating that no product is being oxidized. The standard electrode
potential (E0 ) for the electrochemical formation of Li2 O2 (2Li + +O2 +2e−
Li 2 O2 )
can be calculated from the free energy of formation, and is generally considered to
be ~2.96 V vs Li/Li+ at 30 °C. However, operating the battery as 150 °C gives a
calculated potential of ~2.83 V vs Li/Li+ , which agrees with the median potential,
measured from the CV.
It is important to note that quasi-reversible CVs are observed at a carbon oxygen
electrode in the molten salt eutectic, with cathodic peak to anodic peak separation
of about 200 mV. Typical CVs for conventional organic electrolytes at room temperature usually display anodic peaks in the 3.4-4.5 V window, which are attributed
to side reactions (e.g., solvent decomposition) alongside direct Li2 O2 oxidation.
Greater currents are obtained when using the molten salt electrolyte, which likely is
a result of less cathode passivation due to the higher solubility of intermediate LiO2
and discharge product Li2 O2 . The high cathodic peak potential for oxygen reduction
in the melt (~2.73 V vs Li/Li+ ) is likely a result of a combination of the following
effects: (i) improved electrode kinetics from elevated temperature operation, (ii)
lower internal cell resistance as a result of high ionic conductivity of the molten
salt, (iii) improved intermediate solubility (as will be conclusively shown later),
such as LiO2 , which in turn reduces Li2 O2 induced electrode passivation[56, 57].
Consistent with a Qa/Qc ratio of ~0.9, a slight inefficiency in the pressure profile
for the nitrate electrolyte is attributed to Li2 O2 precipitating in the separator, away
from the current collector carbon, and becoming electrically disconnected. Some
dissolved Li2 O2 may also have traveled across the electrolyte and been reduced to
Li2 O on the lithium metal. No particular effort had been made to optimize the
porous volume and electrode architecture to date. These results constitute a significant step forward in the search for stable electrolyte materials. We anticipate that
the LiO2 or Li2 O2 could react with carbon at the operating temperature of these
batteries to generate Li2 CO3 at the surface of carbon. This is consistent with oxygen
consumption/evolution inefficiencies observed during the CV.
Galvanostatic Cycling
In a similar vein as the CV scans of Figure 2.1, Figure 2.2 compares galvanostatic
cycles of oxygen reduction at a Super P carbon cathode in a conventional lithiumoxygen battery electrolyte , 0.1 M LiClO4 –DMSO, and a molten salt electrolyte.
However, in this case the molten salt is a LiNO3 –KNO2 –CsNO3 eutectic mixture, and

16

Figure 2.2: Galvanostatic cycling of oxygen reduction and evolution with pressure
profiles measured in (a,c) 0.1 M LiClO4 –DMSO at 30 °C between 2.5 and 4.2 V
and (b,d) LiNO3 –KNO2 –CsNO3 molten salt electrolyte at 120 °C between 2.65 and
3.0 V. Positive electrode: Super P carbon:PTFE 95:5 wt %, current: 0.25 mA (~80
mA/g carbon). Electrolyte loading: 150 µL. Carbon loading: ~4 mg/cm2 . Pressure
profiles depicted in (c) and (d) are for the first cycle of the battery. Battery capacity
expressed in mAh per g of carbon. E0 represents the thermodynamic potential for
the reaction 2Li + + 2e− + O2
Li 2 O2 at 30 °C (c) and 120 °C (d)

they operate at 30 and 120 °C, respectively. The ternary mixture of nitrate/nitrite
salts has a lower melting point compared to the lithium/potassium nitrate binary
mixture, and therefore allowed the operation of the battery at a lower temperature.
The battery that uses the organic electrolyte has poor capacity retention and high
voltage hysteresis, whereas the battery using the molten nitrate/nitrite salts can be
cycled between 2.65 and 3.0 V with remarkably low voltage hysteresis (~0.1 V) and
good capacity retention. Consistent with the CV data, a ~10% irreversible capacity
loss is observed in the molten nitrate/nitrite battery. The voltage profile comparison
for the first three cycles of the batteries clearly demonstrates the superior behavior
of the molten salt lithium-oxygen battery compared to prior state of the art in terms
of (i) dramatically reduced voltage gap, (ii) excellent capacity retention, and (iii)
clear end of half cycle based on voltage turn-down on discharge and turn-up on

17
charge. The challenge will be to couple the molten nitrate electrolytes with stable
cathode materials to achieve long cycle life. Pressure monitoring in the organic
electrolyte battery during cycling clearly shows that only a small fraction of oxygen
consumed during discharge is evolved during charge. Side reactions govern the
electrochemistry and account for rapid capacity fade in the DMSO-based battery.
Many authors have reported high chemical instability of DMSO solvent in the
lithium-oxygen battery[36, 58].
Figure 2.2d demonstrates that oxygen can be consumed and subsequently evolved
with relatively high efficiency (∆Pcharge/ ∆Pdischarge ≈ 90%) when recharging the
molten salt battery, with identical slopes of pressure variation within the battery
during discharge and charge. Enhanced LiO2 /Li2 O2 solubility in the melt leads to
electrical disconnection and carbon decomposition, and are thought to be responsible
for the ~10% irreversible pressure loss during cycling. We found the stoichiometry
for the battery reaction, often reported as e- /O2 molar ratio in the literature, to be 2.0
by combining coulometry and gas analysis, for both discharge (ORR) and charge
(OER) half-cycles. Oxygen partial pressure, internal battery volume, and operating
temperature are all carefully controlled parameters. Molar ratios that deviate from
2.0 are indicative of parasitic processes occurring during cycling, often related to
a reaction between the electrolyte (or the electrode) and reduced forms of oxygen.
Gas analysis by mass spectrometry of the battery head-space is routinely performed
to ensure O2 (mass 32) is the only gaseous species formed during cycling. Typically,
gases such as CO (28), CO2 (44), NO (30) and NO2 (46) are monitored by mass
spectrometry. These would arise from oxidation reactions of the carbon electrode
material and the electrolyte materials, namely nitrite and nitrate.
Gas Analysis of Charging Process
Figure 2.3 shows in situ gas analysis performed during a lithium-oxygen battery
charge half-cycle in LiNO3 –KNO3 electrolyte at 150 °C. The battery was first discharged under oxygen to roughly 30% depth of discharge (~400 mAh/g). Then,
during open circuit voltage (OCV) between discharge and charge, the valve connecting the battery to the mass spectrometer was opened and the gases were allowed
to flow from the battery head-space directly into the mass spectrometer chamber.
The charge half-cycle was therefore carried out under a vacuum which would not be
possible with common organic electrolytes due to solvent volatility. As the electrode
begins to recharge, oxygen mass 32 signal increases and remains steady throughout
the entire charge, which is consistent with the pressure data observed in Figure 2.2d.

18
Other gases, such as carbon dioxide (mass 44) and nitric oxide (mass 30) remain at
background levels.

Figure 2.3: Mass spectrometry of lithium-oxygen charge process. Battery employed
a LiNO3 –KNO3 molten salt electrolyte, a Super P carbon:PTFE cathode, and was
charged at 150 °C, ~80 mA/g, to a 3.0 V cutoff. Capacity expressed in mAh per g
of carbon.

2.3

Lithium Peroxide and Wulff Constructions

Confident in the reversibility in both the electrochemistry and gas analysis afforded
by a molten salt electrolyte, a combination of X-ray diffraction (XRD), scanning
electron microscopy (SEM), and energy-dispersive X-ray spectroscopy (EDS) are
used to characterize the discharge product grown on the cathode. Typically, the
cathode was extracted from heated batteries inside an argon-filled glovebox and
rinsed with N-Methylacetamide (NMA) solvent to remove residual nitrate and nitrite
salts.
X-ray Diffraction
After extraction from the battery following the procedure described above, carbon
electrodes were then sealed in Kapton tape for protection against air contamination.
Diffraction was performed on a PANalytical X’Pert Pro X-ray powder diffractometer
with Cu Kα1 radiation.
XRD patterns of the carbon electrode before discharge, following a ~1400 mAh/g
discharge under O2 to 2.6 V cathodic cutoff, and after a completed cycle (electrode
fully recharged to 3.0 V) are depicted in Figure 2.4. The XRD results clearly
demonstrate that crystalline Li2 O2 forms on the carbon surface during discharge
and can subsequently be oxidized upon battery charge. XRD also reveals the

19

Figure 2.4: X-ray diffraction of lithium-oxygen cathode across one cycle. Scans
taken following an OCV period (before discharge), a single discharge to 2.6 V,
and a discharge/charge cycle between 2.6 and 3.0 V, in LiNO3 –KNO3 molten salt
electrolyte at 150 °C with Super P carbon cathode.

presence of Li2 CO3 alongside Li2 O2 , confirming sustained reaction between carbon
and the oxygen reduction products. In addition, this Li2 CO3 side product remains
after charge.
Wulff Constructions
Before analyzing the discharge product morphologies observed in this molten salt
electrolyte, it is worthwhile to give a brief review on Wulff constructions. In
particular, this section will describe the underlying theory, geometrical construction,
and necessary conditions of Wulff constructions.
A simple summary of the theory and proof for a Wulff construction is provided here.
For the detailed version of the proof on which this section is based, see [59]. The
surface energy of a crystal with i atoms (the difference between the free energy of i
atoms in an infinite crystal and the free energy of the i atoms in this crystal) is given
by Gibbs[60] to be
γjOj
(2.1)
∆Gi =

where γ j is the surface energy of jth face, and O j its area. Minimizing this energy

20
for a constant volume gives

γ j δO j = 0

(2.2)

Now choose a point inside the crystal. Drawing perpendiculars h j from each face to
this point, the volume of this crystal (assuming the crystal is convex) can be given
as
1Õ
V=
hjOj
(2.3)
3 j
Taking a small change in crystal shape at constant volume, after some algebra which
is omitted here, yields the expression
h j δO j = 0
(2.4)

Combining Equations 2.2 and 2.4 with an arbitrary constant λ gives
(h j − λγ j )δO j = 0

(2.5)

Because the small change in crystal shape was arbitrary, so is δO j , requiring h j −
λγ j = 0 for all j. Then h j = λγ j , and the length of this perpendicular line is
proportional to the surface energy of the face it corresponds to, proving the idea
proposed by Wulff[61].

Figure 2.5: Wulff construction schematic. Adapted from [62].
Turning the idea of this derivation backwards, one can construct the shape of a
crystal with minimized surface energy. Begin with a plot of surface energy for every

21
crystal orientation, γ(i j k). Draw a line (representing the h j of above) from the origin
to each point on this plot. At this intersection, construct the plane perpendicular to
this hi j k (representing the O j of above). The crystal with minimum surface energy
is then the one bounded by all of these planes. A schematic of this process in two
dimensions is shown in Figure 2.5.
While this construction represents a minimization of the surface energy, and thus
the thermodynamic equilibrium crystal shape, like many topics in materials science,
a competition between thermodynamics and kinetics determines the physically realized outcome. For instance, several groups have performed Wulff constructions
of Li2 O2 [63, 64], and Figure 2.6 gives a representative example. By comparison
to the previously observed morphologies of Li2 O2 shown in Figure 1.4, it appears
that the equilibrium hexagonal structure is not achieved in a lithium-oxygen battery
for either the surface or solution mediated growth mechanisms. This suggests that
the kinetics of the system are playing a significant role in determining the crystal
morphology.

Figure 2.6: Wulff construction of Li2 O2 . Adapted from [63].

Microscopy and Elemental Analysis
After extraction from the battery following the procedure described above, carbon
electrodes were briefly exposed to air (~30 seconds) as they were loaded into an
SEM vacuum chamber. SEM analysis was performed in a Versa 3D DualBeam
focused ion beam (FIB) microscope (FEI), and the elemental analysis with a ZEISS
1550VP FESEM microscope equipped with an Oxford X-Max SDD X-ray energy
dispersive spectrometer.
Figure 2.7 shows SEM images of the carbon electrode surface before and after
a discharge under oxygen to 2.6 V at 80 mA/g in a LiNO3 –KNO3 molten salt.
Typically, large particles ranging from 500 nm to several microns were observed at
the carbon surface of the discharged cathode. These particles are much larger than

22

Figure 2.7: Morphology of oxygen cathode discharge product, taken with SEM. (A)
Super P carbon nanoparticles before discharge (scale bar: 250 nm). (B,C) Cathode
following ~1400 mAh/g discharge under oxygen to 2.6 V cutoff (scale bar: 1 µm
and 500 nm, respectively).

the ones grown in past work using an organic electrolyte, such as those in Figure
1.3. In addition, the observed morphology of the discharge product, believed to
be Li2 O2 , consists of stacks of hexagonal layers. This morphology agrees with
theoretical predictions of the Wulff construction of Li2 O2 being a hexagonal prism
as shown in Figure 2.6.
The observation of these particularly large particles suggests an enhanced solubility
of the discharge products, leading to the solution mediated precipitation of Li2 O2 .
The increased solubility agrees with the large acceptor and donor number of previously studied molten salts described earlier. This enhanced solubility, in addition
to the elevated temperature of the battery, would reduce the kinetic restrictions on
crystal growth, allowing these Li2 O2 crystals to grow with equilibrium shape.
Figure 2.8 displays the elemental analysis performed on a discharged carbon cathode
and centered on a grown Li2 O2 crystal. While EDS is typically unable to observe
lithium directly, we do observe only the presence of oxygen, and notably the absence
of carbon or potassium. The lack of carbon demonstrates that this crystal is not
Li2 CO3 or some other side product due to decomposition of the Super P carbon on
the cathode, and the lack of potassium demonstrates that this crystal is not frozen salt
or some side product due to decomposition of the electrolyte. The observed particle
is clearly crystalline as evidenced by the well defined facets, so the diffraction of
Figure 2.4 corroborates that this is in fact Li2 O2 .
SEM analysis of a Super P carbon electrode fully charged to 3.0 V (1st cycle)
in Figure 2.9 showed no evidence of Li2 O2 , and many areas of the cathode were
covered by needle-like particles of several hundred nanometers. Consistent with

23

Figure 2.8: Elemental map of oxygen cathode discharge product, taken with EDS.
(grayscale) Electron micrograph (scale bar: 2.5 µm) with (red) carbon, (green) oxygen, and (purple) potassium elemental maps of a Super P carbon cathode following
~1400 mAh/g discharge under oxygen to 2.6 V cutoff.

Figure 2.9: Elemental map of oxygen cathode after full cycle, taken with EDS.
(grayscale) Electron micrograph (scale bar: 2 µm) with (red) carbon and (green)
oxygen of a Super P carbon cathode following ~1400 mAh/g discharge and charge
to 3.0 V cutoff under oxygen.

XRD data showing Li2 CO3 at the end of charge, elemental analysis supports the
coprecipitation of Li2 CO3 at the cathode surface.

24
Lithium Peroxide Solubility
While the morphology of Li2 O2 shown in the previous section suggests an enhanced
solubility in the molten salt electrolyte, a more quantitative measure would help
confirm this. Li2 O2 solubility and diffusivity in the molten salt electrolyte were
investigated by rotating disk electrode (RDE) measurements. Typically a PTFEbased electrochemical cell was used inside an argon-filled glovebox. The working
electrode consisted of a platinum RDE (5.0 mm OD, 99.99% pure, mirror polished,
Pine Research), the counter electrode a platinum wire (1.0 mm OD, Sigma-Aldrich)
and the reference electrode a lithium metal rod (Sigma-Aldrich). The 3-electrode
cell was loaded with roughly 35 mL of carefully dried LiNO3 –KNO3 eutectic and
maintained at 150 °C by means of thermocouple and suitable beaker heating mantle.
Once reproducible CVs were obtained in the 2.8–3.3 V range, Li2 O2 powder (Alfa
Aesar, 95% pure) was added to the molten salt electrolyte, and using the RDE tip,
the solution was vigorously stirred for several minutes until saturation. Linear sweep
voltammograms (anodic scans from OCV to 3.3 V vs Li/Li+ ) were then performed
at 1 mV/s, varying the electrode rotation rate from 100 to 2500 rpm. Levich/Cottrell
analysis was used to determine bulk concentration and diffusivity of Li2 O2 in the
molten salt electrolyte. The Levich equation contains the diffusivity as well as the
bulk concentration of the reacting species Li2 O2 , hence one of these quantities must
be known to estimate the other. For that we used the Cottrell equation derived from
transient current measurements under quiescent conditions.
Figure 2.10 shows this linear sweep voltammetry data obtained for Li2 O2 bulk
oxidation at 150 °C at a platinum RDE. Upon addition of Li2 O2 , the OCV of
the battery shifted to ~2.8 V vs Li/Li+ which is consistent with thermodynamic
data. Slightly sloped anodic limiting currents (i.e., mass-transport limited currents)
are due to interference with oxidation by impurities in the melt such as nitrite
and hydroxide ions (nitrite ions form when nitrate ions get chemically reduced by
the lithium reference electrode, whereas LiOH is a known impurity in LiNO3 ).
Consistent with fast electrode kinetics, the current observed at any given potential
along the voltammogram varies linearly with the square root of the rotation rate
and the line intercepts the vertical axis at zero. We measured the peroxide anion
concentration to be 5.2e-4 mol/L, which is an order of magnitude greater than values
calculated in organic solvents such as DMSO or DMF at 30 °C (in the 10-5 M
range). For those organic solvents, lithium atomic absorption spectroscopy (AAS)
was typically performed on Li2 O2 -saturated samples after dilution in water[65]. It
appears that elevated temperature has an influence on Li2 O2 solubility. Solvation

25
effects between the molten salt electrolyte and the peroxide anions require further
investigation.

Figure 2.10: Linear sweep voltammograms of Li2 O2 oxidation, recorded at a platinum RDE (A = 0.196 cm2 ) in a molten salt electrolyte containing Li2 O2 . v = 1
mV/s, from OCV to 3.3 V vs Li/Li+ . RDE rotation rate increased from 0 to 2500
rpm. For comparison, a LSV scan of the molten salt electrolyte without Li2 O2 was
performed at each rotation rate to measure the background current (plotted is the
900 rpm scan). Performed in a LiNO3 –KNO3 electrolyte at 150 °C with platinum
wire counter electrode and lithium metal rod reference electrode.
In the LiNO3 –KNO3 eutectic molten electrolyte, the rate of the electrochemical
half-reaction O22− → O2 + 2e− at 150 °C is governed only by mass transport to the
electrode surface. A diffusion coefficient of 3.1e-8 cm2 /s was found for the O2 2anion, several orders of magnitude lower than typical values for O2 (10-5 , 10-6 cm2 /s
range), but consistent with the fact that peroxide has two negative charges and could
exist as an ion pair with Li+ cations. Appropriate methodology to determine the
solubility and diffusivity of oxygen in molten nitrates is being developed.
The data shown throughout this section support the reversible, solution mediated
growth of Li2 O2 in a molten salt electrolyte. In particular, the solvating ability of
the molten salt, paired with the elevated temperature of the battery, enables fast
electrode kinetics, resulting in the extremely low discharge/charge overpotentials
(~50 mV).

26
2.4

Battery Cycling Performance and Cell Death

In order to understand the failure mechanisms of this system, the lithium-oxygen
battery depicted in Figure 2.11 was galvanostatically cycled in LiNO3 –KNO3 eutectic at 150 °C with limited depth of discharge (2 mAh discharge, ~45% DOD). The
battery was stopped after 50 cycles (~340 h of cycling) when the discharge capacity
had significantly dropped.

Figure 2.11: Cycling behavior of lithium-oxygen battery, including voltage profile
for cycle 1, 10, 20, 30, 40, and 50. Battery cycled in LiNO3 –KNO3 molten salt
electrolyte at 150 °C using a Super P carbon cathode, current density of ~0.64
mA/cm2 or ~160 mA/g of carbon, and carbon loading of ~4 mg/cm2 . Arrows
indicate increasing voltage hysteresis with cycle number.
Interestingly, if we assume that all of the coulombic inefficiency accumulated over
the life of the battery is attributed to the side reaction resulting in the formation of
Li2 CO3 , it would require more carbon than is present in the cathode. To explain this
apparent discrepancy, the cathode was analyzed by XRD and SEM similarly to the
previous sections.
Despite ending the cycling after a complete charging half-cycle to 3.0 V, the diffraction of Figure 2.12 reveals the presence of Li2 O2 at the carbon surface, alongside
the expected Li2 CO3 . We believe that electrically disconnected Li2 O2 accumulates
away from the electrode during cycling, accounting for the rest of the coulombic
inefficiency, while Li2 CO3 passivates carbon and is responsible for the continuous
increase in battery polarization with cycle number.
SEM analysis of the cycled carbon cathode in Figure 2.13 further supports this
argument. First, the surface of the cathode is clearly electrically insulating, as
indicated by the charging of Figure 2.13a, with what appear to be clusters of Li2 O2

27

Figure 2.12: X-ray diffraction of cycled lithium-oxygen battery, from a cell similar
to the one in Figure 2.11.

Figure 2.13: Scanning electron microscopy of cycled lithium-oxygen battery from
a cell similar to the one in Figure 2.11. (A) Image charging of cathode due to
electrically insulating surface (scale bar: 15 µm). (B) Li2 CO3 particles covering
carbon surface (scale bar: 5 µm). (C) Li2 O2 particles agglomerating on a separator
microfiber away from electrode surface (scale bar: 10 µm).

and Li2 CO3 covering the carbon. A closer observation in Figure 2.13b shows large
crystals of Li2 CO3 (fuzzy “needle-like” particles) which are seen throughout the
carbon surface. In addition, Figure 2.13c shows large, >10 µm, clusters of Li2 O2
hexagonal prisms growing on the residual separator glass microfibers leftover on
the harvested cathode. It is noteworthy that electrochemically grown Li2 O2 is
growing on the battery separator, which is insulating by design. This serves as a
direct confirmation of the solution-mediated growth process and is consistent with
uncontrolled precipitation of Li2 O2 during battery cycling.
A sustained reaction between oxygen reduction products and carbon is responsible

28
for the cycle life limitation of the present batteries. Electrode passivation by Li2 CO3 ,
often reported as “carbon corrosion”, is a known phenomenon in lithium-oxygen
batteries[42, 43]. As the number of cycles increases, the voltage hysteresis of the
battery increases, consistent with the growth in electrode impedance. The origin of
capacity fade in these systems is therefore 2-fold. First, one needs to better control
Li2 O2 dissolution and precipitation during discharge, and second, cathode materials
stable to oxygen reduction products need to be developed in order to achieve high
coulombic efficiency and maintain high energy efficiency.
2.5

Summary

We developed chemically stable electrolytes for the lithium-oxygen battery by replacing volatile, unstable, and air-intolerant aqueous or organic-based electrolytes
with an inorganic molten salt. We used eutectic binary or ternary mixtures of alkali
metal nitrate/nitrite salts and studied the oxygen electrochemistry at 120 and 150 °C.
Oxygen reduction was found to proceed based on a 2e- /O2 reaction leading to the formation of Li2 O2 . We observed the shape of deposited Li2 O2 to be hexagonal, which
agrees with equilibrium surface energy calculations of its Wulff construction. XRD
confirmed the reversible formation of crystalline Li2 O2 , and in situ gas and pressure
analyses demonstrated that oxygen could be efficiently evolved during charge at a
very low overpotential. A discharge/charge voltage gap of only ~0.1 V constitutes
the lowest value reported for a lithium-oxygen battery. Improved reversibility and
rate capability are thought to originate from enhanced solubility of the discharge
products, which alleviate the intrinsic electronic transport limitations of Li2 O2 deposited on the lithium-oxygen battery cathode. More fundamental work needs to be
done to understand solvation effects and the role of nitrate/nitrite anions in enhancing solubility of the discharge products. In the present study we used Super P carbon
as the oxygen electrode material and showed that it tends to react with oxygen reduction products to form Li2 CO3 . We also demonstrated that decomposition of the
amorphous carbon electrode causes battery failure, and therefore the identification
of suitable non-carbonaceous oxygen electrode materials will be key to success for
this chemistry. Although we found that Li2 CO3 is sparingly soluble in the melt and
can be bulk oxidized at about 3.5 V vs Li/Li+ at a platinum disk electrode, the side
product passivates the carbon electrode when operating the battery in the typical
voltage regime (2.6–3.0 V) and is responsible for increased polarization and capacity
loss. XRD analysis of a cycled carbon cathode revealed substantial accumulation
of electronically disconnected Li2 O2 . Large clusters, >10 µm in diameter, of Li2 O2

29
crystals deposited on the glass fiber separator were observed by SEM. Solubility of
Li2 O2 allows growth of large particles and clusters via a solution phase mechanism
which enables high areal capacity and low overpotential, but creates challenges in
achieving high coulombic efficiency. Uncontrolled diffusion and precipitation of
soluble Li2 O2 is a major cause of capacity loss, and the low oxygen solubility is
a challenge for achieving high rate capability. Methods to address these issues are
important areas of future work.

30
Chapter 3

A NITRATE REDUCTION BATTERY
Adapted from
1. Addison, D., Tozier, D., Tan, H., Uddin, J., Gallant, B. M., McCloskey, B. D.,
Greer, J. R., Chase, G. V. & Giordani, V. A New Rechargeable Battery: Lithium
Oxide Growth through Molten Salt Nitrate Reduction. (In Preparation).

Chapter Abstract
Next generation lithium batteries, such as the lithium-oxygen battery of the previous
chapter, often employ the growth of lithium-rich phases, allowing specific capacities
much higher than typical insertion cathode materials. In this chapter we introduce
a battery chemistry along a similar vein. We demonstrate that a nanoparticle
heterogeneous catalyst allows the reversible growth and dissolution of micron scale
Li2 O crystals through the effective catalysis of nitrate reduction and nitrite oxidation.
This enables a new rechargeable battery system with a full cell theoretical specific
energy of 1289 Whkg-1 where a molten nitrate salt serves as both active material
and electrolyte.

31
3.1

Nitrate Reduction Chemistry

The Introduction Chapter discussed the potential benefits of moving beyond the typical intercalation cathode materials of lithium-ion batteries, looking at chemistries
that employ the growth and dissolution of new phases such as oxygen and sulfur cathodes, both widely studied systems. The Second Chapter proposed a novel approach
to the lithium-oxygen battery, replacing the traditional organic electrolyte with a
molten salt one in hopes of overcoming some of practical challenges involved in
that system. This molten salt electrolyte also led to interesting growth mechanisms
of that systems lithium-rich phase, Li2 O2 .
This chapter considers a similar phase-forming conversion chemistry, whereby a
molten nitrate salt serves as both an active material and the electrolyte. Molten
nitrate salts have been previously studied as an active material in a primary lithium
battery where Li2 O irreversibly forms as nitrate reduces to nitrite[50, 51, 66, 67].
The reaction and thermodynamic potential at 150 °C are given in Equation 3.1.
2Li + LiNO3 → Li 2 O + LiNO2

E 0 = 2.44 V

(3.1)

When using a eutectic composition LiNO3 –KNO3 electrolyte (the same one as in
Chapter Two), Equation 3.1 has a theoretical capacity of 612 mAhg-1 , and paired
with a lithium metal anode at the thermodynamic potential of 2.44 V, has a full
cell theoretical specific energy of 1289 Whkg-1 . This is an encouraging value when
looking at Table 1.1, especially when considering the electrolyte serves as the active
material. If only a fraction of this theoretical specific energy were available, it would
still be competitive with state of the art battery chemistries.
Previous work on this chemistry as a primary battery used a high surface area carbon current collector, resulting in a large overpotential on discharge[67]. When
attempting to reverse the chemistry after discharge as in Figure 3.1, there is significant overpotential accompanied by gas evolution, suggesting the irreversibility
of the reaction. However, by using a cathodic current collector which catalyzes
Equation 3.1, in particular the reverse reaction of nitrite oxidation, this chemistry
could be implemented as a secondary battery. Such a catalyst for nitrite oxidation is
important because Li2 O will directly oxidize to oxygen above 2.8 V in a similar way
to Li2 O2 in a lithium-oxygen battery, and this reaction contributes to the pressure
rise of Figure 3.1.
If such a catalyst were found, a secondary nitrate reduction battery could be operated
as schematically described in Figure 3.2. Just as in the lithium-oxygen system, a

32

Figure 3.1: Nitrate reduction first cycle on a Super P carbon cathode, with pressure
monitoring of the cell. LiNO3 –KNO3 electrolyte at 150 °C with carbon loading of
~4 mg/cm2 .

lithium-rich phase would grow on the surface of the catalytic porous cathode support
on discharge, and it would be removed on charge. In addition, note the presence of
Li2 O on the surface of the lithium anode. Interestingly, this natural SEI, which was
previously reported for lithium in molten nitrate salts[67, 68], is the same reaction
as the one occurring at the cathode, Equation 3.1, except that it occurs chemically.
While a lithium anode directly exposed to the electrolyte is not practical for extended
cycling, it is stable enough here to enable the study of this cathode chemistry.

Figure 3.2: Nitrate reduction schematic, with depictions of the beginning of the
discharge and charge processes. Molten nitrate electrolyte in blue and Li2 O in
yellow on both the lithium metal anode and conductive catalyst cathode support.

33
3.2

Catalysis of Nitrite Oxidation

A combination of galvanostatic cycling and galvanostatic intermittent titration technique (GITT) are used to compare the electrochemical behavior of various catalyst
candidates and electrochemically confirm the reversibility of any promising material. These experiments were performed in the same Swagelok cells used in Chapter
Two. Cathodes consisted of about 10-20 mg/cm2 of metal nanoparticles (or microparticles) pressed onto a stainless steel or aluminum mesh without the use of
binder. A glass microfiber separator was impregnated with about 200 mg of molten
salt electrolyte, LiNO3 –KNO3 unless otherwise noted, and vacuum dried overnight
at 200 °C. Anodes were a simple lithium foil. All assembly took place in a glovebox,
then the cells were filled with argon after a leak test, and finally placed in an oven
to be held at the 150 °C. A detailed description of the cell design and assembly is
given in the appendix.
Searching for a Catalyst
A single galvanostatic cycle for various metal nanoparticles (all around 30-50 nm),
including many typically used as catalysts for OER and ORR in other systems, is
shown in Figure 3.3.

Figure 3.3: Catalysis comparison for nitrate reduction reversibility. First cycle of
nitrate reduction using various nanoparticle (except where indicated) cathodes, at
0.1 mA cm-2 .

34
A comparison of discharge and charge overpotentials between the different metals
serves as a good metric for catalytic activity toward either nitrate reduction or
nitrite oxidation. The overpotential for nickel is clearly the smallest (~100 mV),
demonstrating that nickel nanoparticles can effectively catalyze both the reduction
of nitrate and the oxidation of nitrite. In addition, micron-sized nickel particles
exhibit a large overpotential and low capacity, demonstrating the importance of
using a high surface area catalyst. While these nickel nanoparticles do have a large
surface area (~10 m2 /g), Super P carbon has a surface area an order of magnitude
larger, suggesting that the nickel surface does in fact serve a catalytic role.
To confirm electrochemical reversibility and determine the OCV when using nickel
nanoparticles as a cathode catalyst, we employ GITT in Figure 3.4with two different
electrolyte compositions: one all nitrate, the other a nitrate/nitrite mixture. In the
case of a LiNO3 –KNO2 electrolyte, the measured OCV is ~2.47 V, in agreement
with the calculated potential of 2.44 V in Equation 3.1. In addition, the OCV is
the same on discharge and charge which suggests that the forward and backward
reactions have the same standard potential, a good indication that the chemistry of
Equation 3.1 is happening reversibly.

Figure 3.4: Galvanostatic intermittent titration technique of nitrate reduction, with
a nickel nanoparticle cathode in both an all nitrate and nitrate/nitrite electrolyte.

35
In the case of a LiNO3 –KNO3 electrolyte, the measured OCV is ~2.55 V initially,
while gradually decreasing on discharge and then increasing on charge. This process
agrees with what we would expect from the Nernst equation[69], reproduced in
Equation 3.2.
RT
αRed
E=E −
ln
(3.2)
zF
αOx
The concentration of nitrite, the reduced species, is initially negligible, while that
of nitrate, the oxidized species, is well above standard conditions. Then the reaction
quotient is less than one and the Nernstian contribution is positive, causing the
observed OCV to be above the standard potential. As the concentration of nitrite
increases throughout discharge, its activity approaches that of nitrate and thus the
OCV approaches the calculated standard potential.
Catalytic Properties of Nickel
With a good understanding of the role that nickel nanoparticles can play in the
catalysis of nitrate reduction and nitrite oxidation, a combination of X-ray photoelectron spectroscopy (XPS) and transmission electron microscopy (TEM) are used
to identify the surface state of the nickel and a potential explanation for its catalytic
activity, and samples were cleaned as described previously. XPS was performed
with a Surface Science Instruments M-Probe XPS system under monochromatic Al
Kα X-ray (1486.6 eV) illumination[70], and samples were briefly exposed to air as
they were put into the transfer chamber.

Figure 3.5: X-ray photoelectron spectroscopy of nickel nanoparticles, with depth
profiling and performed after exposure to molten nitrate. (A) Ni 2p and (B) O 1s
binding energies.

36
Figure 3.5 contains an XPS depth profile of nickel nanoparticles after exposure to
molten nitrate. The surface state of the nickel is a nickel oxide with both Ni2+ and
Ni3+ states. As the depth profile proceeds, the spectrum progresses from a nickel
oxide state to a metallic nickel one. This suggests that there is a nickel oxide surface
layer, but that the metallic nickel is retained in the center of the nanoparticles. In
addition, the existence of a mixed nickel oxidation state indicates that the surface
nickel oxide is nickel deficient. While a native oxide layer with a small nickel
vacancy concentration does form on nickel surfaces exposed to air[71], we believe
that the exposure to a molten nitrate salt enhances the nickel deficiency by an in situ
introduction of lithium substitutions at nickel sites.
TEM is used to observe and better understand these nickel nanoparticles and their
nickel oxide surface. To prepare samples, cathode material was scraped from the
mesh current collector, ground between glass slides, and put onto a copper TEM
grid with holey carbon film. TEM was performed using a FEI Tecnai F30ST (300
kV), and the TEM grid was briefly exposed to air upon inserting the TEM holder.

Figure 3.6: Transmission electron microscopy of nickel nanoparticles, performed
after exposure to molten nitrate. (A) High resolution of two nickel particles (scale
bar: 10 nm). (B) Bright field and (C) dark field of a collection of particles (scale
bar: 100 nm). (D) Diffraction pattern of area in panels B,C. Dark field taken at
location of red arrow on faint diffraction ring.

37
TEM analysis of nickel nanoparticles is shown in Figure 3.6. The high resolution
micrograph of Figure 3.6a demonstrates the surface layer present on these nickel
nanoparticles. In order to understand what this surface layer is, diffraction is
performed on the collection of particles in Figure 3.6b. The resulting pattern of
Figure 3.6d has all the expected rings of nickel, in addition to two faint extra ones.
The most clear of these is pointed at by a red arrow, and a bright field image at this
spot is shown in Figure 3.6c. Interestingly, these extra rings correspond to those
of nickel oxide with a rock salt structure. Note that the bright field consists of a
diffuse contrast across all the particles, consistent with the fact that the surface layer
is nickel oxide.
We propose that it is this cation deficient nickel oxide surface, and not metallic nickel,
which serves as the catalyst for reversible nitrate reduction. In particular, these point
defect sites, whether in the form of a nickel vacancy or lithium substitution, serve
as an oxide ion transfer catalyst enabled by the neighboring nickel atom’s ability to
cycle between 2+ and 3+ oxidation states.
In order to confirm this, we synthesize lithium-doped nickel oxide (L-NiO) particles
for use as a cathode catalyst. 2.0 mL of water:iso-propanol (2:1) was added to
a mixture of Ni(OH)2 (1.0 g, 10.79 mmol) and LiOH (12.93 mg, 0.54 mmol) in
a vial. The suspension was sonicated at room temperature for 2 hours to make
it a homogeneous mixture. Water and iso-propanol were then removed using a
rotary evaporator under vacuum. The resulting wet slurry was dried in an oven at
50 °C overnight. The Ni(OH)2 –LiOH solid mix was then heated to 300 °C in a
furnace for 3 hours to give nanoparticles of L-NiO. BET surface area measurements
were performed using a Micromeritics TriStar II Plus surface area analyzer. A
BET surface area of 120 m2 /g was measured, following a 17 hour degassing at
120 °C under nitrogen flow. Note that such a doped nickel oxide displays p-type
conductivity, allowing its use without a conductive additive[72–74].
Figure 3.7 compares the first cycle of nitrate reduction and nitrite oxidation on this
L-NiO to the nickel nanoparticles used previously. The performance on both charge
and discharge is nearly identical to that of the nickel nanoparticles, confirming that
it is in fact this doped nickel oxide which serves as the catalyst for this reaction.
3.3

Lithium Oxide as a Discharge Product

With a catalyst allowing the reversible reduction of nitrate based on the electrochemical data, a combination of XRD, SEM, EDS, and TEM are used to characterize

38

Figure 3.7: Lithium-doped nickel oxide as a catalyst.

the discharge product grown on the cathode. As before, the cathode was extracted
from heated batteries inside an argon-filled glovebox and rinsed with NMA solvent
to remove residual nitrate and nitrite salts.
X-ray and Electron Diffraction
After extraction from the battery following the procedure used previously, nickel
electrodes were then sealed in Kapton tape for protection against air contamination. Diffraction was again performed on a PANalytical X’Pert Pro X-ray powder
diffractometer with Cu Kα1 radiation.
XRD patterns of the nickel electrode before discharge, following a discharge to 2.3
V cathodic cutoff, and after a completed cycle (electrode fully recharged to 2.75
V) are depicted in Figure 3.8. The XRD results clearly demonstrate that crystalline
Li2 O forms on the nickel surface during discharge and can subsequently be oxidized
upon battery charge. Unlike the case with the lithium-oxygen battery from before,
there is no carbon present in this system (lithium metal anode, ceramic separator,
molten salt electrolyte, nickel cathode, no binder). As a result, there is no means
of forming carbonaceous side products, and the only crystalline discharge product
observed is the desired Li2 O.
Because structural confirmation of electrochemically grown Li2 O is not (to our
knowledge) reported in literature, TEM was used to further confirm its existence.

39

Figure 3.8: X-ray diffraction of nitrate reduction cathode across one cycle. Scans
taken following an OCV period (before discharge), a single discharge to 2.3 V,
and a discharge/charge cycle between 2.3 and 2.75 V, in LiNO3 –KNO3 molten salt
electrolyte at 150 °C with nickel nanoparticle cathode. Peaks not attributed to Li2 O
are from the nickel nanoparticles and aluminum grid they are pressed onto.

As before, cathode material was cleaned as described above, scraped from the mesh
current collector, ground between glass slides, and put onto a copper TEM grid with
holey carbon film. TEM was performed using a FEI Tecnai F30ST (300 kV), and
the TEM grid was briefly exposed to air upon inserting the TEM holder.

Figure 3.9: Transmission electron microscopy of nitrate reduction cathode. (A)
Standard and (B) bright field images of discharge product (scale bar: 200 nm), with
selected area indicated for (C) diffraction pattern, indexed to [011] orientation of
Antifluorite structure Li2 O. Sample taken from cathode discharged to 2.3 V at 150
°C in LiNO3 –KNO3 electrolyte.

40
While Li2 O2 is extremely beam sensitive, requiring the use of cryo-TEM to effectively image it[75], Li2 O appears to be only relatively beam sensitive, allowing one
to obtain a single diffraction pattern at ambient temperature by working quickly.
Interestingly, Li2 O2 first decomposes to Li2 O under the electron beam, painstakingly shown with detailed cryo-TEM experiments[21]. Figure 3.9 shows both the
micrograph and diffraction pattern of discharge product from a cathode discharged
to 2.3 V. Indexing the diffraction confirms that this is the Antifluorite crystal structure of Li2 O. While a full analysis would involve tilting of the sample and obtaining
diffraction patterns along different zone axes, the instability of Li2 O under the beam
allowed only the single pattern (without tilting, purely luck) at ambient temperature
before the crystal decomposed under the electron beam.
Microscopy and Elemental Analysis
As done previously, after extraction from the battery following the procedure described above, nickel electrodes were briefly exposed to air (~30 seconds) as they
were loaded into an SEM vacuum chamber. SEM analysis was performed in a Versa
3D DualBeam focused ion beam (FIB) microscope (FEI), and the elemental analysis
with a ZEISS 1550VP FESEM microscope equipped with an Oxford X-Max SDD
X-ray energy dispersive spectrometer.

Figure 3.10: Morphology of nitrate reduction cathode discharge product. (A)
Nickel nanoparticles before discharge (scale bar: 500 nm). (B,C) Cathode following
discharge to 2.3 V cutoff (scale bar: 20 µm and 5 µm, respectively.)
Figure 3.10 shows SEM images of the nickel cathode surface before and after discharge to 2.3 V in a LiNO3 –KNO3 molten salt. Typically, large particles several
microns in size and a thin film are observed growing over the bed of nickel nanoparticles, and the cyrstallinity of this discharge product is evident. In particular, note
the octahedral morphology of large crystals. In a similar vein to the Li2 O2 of the

41
previous chapter, this morphology agrees with theoretical predictions of the Wulff
construction of Li2 O being an octahedron[76].
While there is no formal proposal of growth mechanisms of electrochemically grown
Li2 O as there is for Li2 O2 , the growth of these large particles which should be
insulating suggests that a similar thing is happening here as the case of the solutionmediated Li2 O2 growth mechanisms described previously. There is evidence of
oxide solubility in molten nitrates[52, 77], and based on the size of these crystals,
such a solution growth mechanism makes sense. As before with Li2 O2 , the high
solubility of discharge products, in addition to the elevated temperature of the battery,
would reduce the kinetic restrictions on crystal growth and encourage the growth of
the equilibrium Wulff shape.

Figure 3.11: Elemental map of nitrate reduction cathode discharge product, taken
with EDS. (grayscale) Electron micrograph (scale bar: 40 µm) with (purple) nickel,
(green) oxygen, and (pink, bottom right) potassium elemental maps of a nickel
nanoparticle cathode following discharge to 2.3 V cutoff.
Figure 3.11 displays the elemental analysis performed on a discharged nickel cathode, performed with a wide field of view capturing many Li2 O crystals. While EDS
is typically unable to observe lithium directly, we do observe only the presence of
oxygen, and notably the absence of nickel or potassium. As discussed previously,
there is no carbon in this system. The observed particles are clearly crystalline, so
the diffraction of Figure 3.8 corroborates that this is in fact Li2 O.

42
3.4

Nitrate to Nitrite and Back Again

By contrast to the lithium-oxygen system, this nitrate reduction chemistry has two
reaction products which must be verified. Having confirmed the growth and dissolution of Li2 O on discharge and charge, it is time to address the other reaction
product, nitrite. A combination of Raman spectroscopy, clever electrochemistry,
and ion exchange chromatography are used to characterize the electrolyte constituents. Electrolytes were extracted from batteries inside an argon-filled glovebox
by disassembling the entire Swagelok fixture without remelting the frozen salt.
Raman Spectroscopy
After extraction from the battery, frozen electrolytes were taken for Raman spectroscopy in air as the species of interest are not air sensitive. Raman spectroscopy
was performed on a Renishaw M1000 Mirco Raman Spectrometer System with an
argon ion laser at 514.5 nm.

Figure 3.12: Raman spectroscopy of nitrate reduction electrolyte across one cycle.
Spectra taken following an OCV period (before discharge), a single discharge to 2.3
V, and a discharge/charge cycle between 2.3 and 2.75 V, in LiNO3 –KNO3 molten
salt electrolyte at 150 °C with nickel nanoparticle cathode.
Raman spectra of the initially LiNO3 –KNO3 frozen electrolyte before discharge,
following a discharge to 2.3 V cathodic cutoff, and after a completed cycle (electrode fully recharged to 2.75 V) are depicted in Figure 3.12. The spectra clearly
demonstrate the presence of a nitrite peak upon discharge and most of the peak going
away upon charge. The peak does not fully go away for two reasons. The coulombic

43
efficiency of the first cycle is not 100%, and the lithium anode SEI growth forms
nitrite as described previously.
Ion Exchange Chromatography
Batteries of varying electrolyte compositions were assembled as described previously and discharged to a 2.3 V cutoff. The cells were taken to the glovebox and
disassembled, separating the electrolyte and cathode. For some cells, the cathode
was cleaned using the procedure above and then put into another cell with new
lithium anode and electrolyte. Several electrolyte compositions were tested, and
these cells with discharged cathodes (preloaded with Li2 O) were then charged a
fixed amount, 2 mAh.

Figure 3.13: Charging profile of newly constructed cells preloaded with Li2 O,
with various electrolyte compositions. Clean cathode contains no pre-grown Li2 O,
and the other three cells contain a nickel nanoparticle cathode harvested from a
discharged cell and put into a new cell containing the labeled electrolyte composition.
Figure 3.13 demonstrates the effect of attempting this charging process in newly
constructed cells. When using an all nitrite electrolyte and clean cathode, the cell
polarizes as expected due to the absence of Li2 O. Now consider the other three cells
which contain a cathode harvested from another cell that was discharged to 2.3 V,
preloading the cells with Li2 O. Note that the cell with a purely nitrate electrolyte
also polarizes, because the absence of nitrite disallows the charging reaction. By
contrast, the cells with partial or entire nitrite electrolyte and a discharged cathode
are able to charge, showing that both Li2 O and nitrite must be present to allow the

44
charging reaction. In addition, note that the charging potential of the all nitrite
electrolyte is lower than that of the part nitrate, part nitrite one, in agreement with
the discussion of the Nernst potential in this system.
While we have used several techniques to qualitatively confirm the formation of
both nitrite and Li2 O on discharge, a quantitative technique that can be corroborated
with electrochemical data would make an even stronger argument that the chemistry
described herein is in fact happening. In a similar vein, differential electrochemical
mass spectrometry (DEMS) found use in the lithium-oxygen battery field to directly
quantify the amount of oxygen consumed and produced, whereas before a simple
qualitative confirmation of Li2 O2 was seen as sufficient[39, 43]. To this end, we use
ion exchange chromatography to measure the ratio of nitrate to nitrite anions in the
electrolyte. This allows a direct comparison between the electrochemical data and
the amount of nitrate and nitrite present in the electrolyte.
Once the process of Figure 3.13 was complete, the battery was again disassembled
using the same procedure. All electrolytes of interest were collected and dissolved
in mQ water (very pure water) of known volume. Following a dilution procedure,
the ion concentration was in the 10–500 µM range which is appropriate for ion
exchange chromatography. A Dionex ICS-2000 Ion Chromatography System was
used for all experiments, and a calibration procedure for both nitrite and nitrate was
performed before each set of samples. See the appendix for a detailed calibration
procedure.
Some typical ion exchange chromatography data is shown in Figure 3.14, demonstrating the peak separation for nitrate and nitrite, in addition to the inevitable
carbonate impurity due to carbon dioxide in the air. This peak separation is important as the concentration of species in solution is proportional to the area under the
curve.
All three of the electrolyte compositions in Figure 3.13 were put through ion exchange chromatography, but the purely nitrite electrolyte is the most revealing. A
detailed reasoning for focusing on the all nitrite system is given in the appendix. A
ratio of nitrate to nitrite anions was calculated from the measurements, and using
the initial mass of electrolyte in the cell, the number of moles of each could be
calculated. After accounting for the small impurity amounts found in nitrite, all
of the nitrate can be attributed to the charging reaction. Considering this reaction
transfers 2e- /mol nitrite, we can convert the number of moles of produced nitrate
to the equivalent amount of charge passed due to this reaction. From the electro-

45

Figure 3.14: Ion exchange chromatography of nitrate/nitrite solution, prepared as a
standard for calibration procedures.

chemical data in Figure 3.13, we expect 2 mAh of charge to have been passed, and
the ion exchange chromatography data suggest that 2.09 mAh equivalent of nitrate
is present in the electrolyte. This suggests that the charge process is in fact the
electrochemical reaction described in Equation 3.1.
3.5

Battery Cycling Performance

With a good understanding the the chemistry of Equation 3.1 can serve as a secondary
battery chemistry, both in the reversibility of nitrite and Li2 O formation and the role
of catalysis afforded by the surface L-NiO, we can now probe the cycling behavior
of the system. In order to study the reversibility of the chemistry and not the ability
of the cathode structure to accommodate Li2 O growth (a subject for Chapter Four),
we cycle with a capacity limitation of 1.27 mAh cm-2 on discharge, and to avoid
oxygen evolution at higher potentials, we cycle with a potential limitation of 2.75 V
on charge.
Note that the pressure in the head-space of the cell in Figure 3.15 remains constant
over the course of 1000 hours of cycling, suggesting that we are not engaging any side
reactions with gaseous evolution. In addition, the cycling profile remains unchanged
over 100 cycles except for the gradual drop in potential due to accumulating nitrite
and the increase in coulombic efficiency as the electrolyte and cathode saturate
with nitrite and Li2 O. This is again in agreement with what we would expect

46

Figure 3.15: Cycling behavior of nitrate reduction battery. Pressure of cell headspace with evolution of cycling profile and discharge/charge capacity with coulombic
efficiency for a cell using a nickel nanoparticle cathode (0.26 mA cm-2 , 1.27 mAh
cm-2 capacity limitation). Nickel loading of 15 mg cm-2 and nitrate loading of 200
mg cm-2 .

from the Nernstian contribution to the potential. The initially poor coulombic
efficiency can be attributed to the diffusion of nitrite away from the cathode and
the sparing solubility of Li2 O in molten nitrate, allowing the discharge product
to become electrically disconnected[68]. Although the current design allows only
~10% nitrate utilization (over 6 mAh cm-2 ), this is a cell engineering problem related
to the cathode structure which is addressed in Chapter Four, and not any limitation
of the chemistry. While prolonged cycling of this chemistry is hindered by the long
term stability of a lithium anode (an issue in most “beyond lithium-ion” chemistries),
the cycling stability of the cathodic reaction stands in contrast to similar systems
such as the lithium-oxygen battery which is plagued by degrading side reactions[42].
3.6

Summary

Through careful characterization of the nitrate reduction reaction described in Equation 3.1, we have demonstrated the ability of a nickel nanoparticle catalyst to enable
the reversibility of a previously irreversible system. In doing so, we have created a
new, high capacity secondary battery chemistry in which the electrolyte also serves
as the active material. In addition, we have shown that it is the surface oxide layer,
L-NiO, which is responsible for the catalytic activity. Having described the catalytic
behavior of nickel nanoparticles, we explicitly confirmed the reversible growth and
dissolution of crystalline Li2 O, and we observed that this Li2 O grows according to

47
its equilibrium Wulff construction. We also quantitatively confirmed the reversible
formation of nitrite in the electrolyte. Finally, we validated this reversibility by
cycling the battery over the course of 1000 hours, achieving a steady coulombic
efficiency fluctuating around 100%. The lack of an obvious degradation mechanism
on the cathode side suggests that this chemistry should be able to cycle for far
longer, but the limitation of uncontrolled SEI growth on the lithium anode makes
such an experiment untenable. The development of a protected lithium anode is key
to enabling this and many other "beyond lithium-ion" systems. While the utilization
of active material was quite low, Chapter Four will address the capacity limitations
of this system and possible means of alleviating them.

48
Chapter 4

ROLE OF CATHODE ARCHITECTURE IN CONVERSION
REACTION CHEMISTRIES
Adapted from
1. Tozier, D., Giordani, V., Chase, G. V., Addison, D. & Greer, J. R. The Role of
Battery Cathode Architecture in Conversion Reaction Chemistries. (In Preparation).

Chapter Abstract
The previous two chapters have investigated the use of cathodes which employ the
growth of lithium-rich phases. Using the growth of Li2 O through nitrate reduction
as a model system, we address the effect of cathode geometry on the electrochemical
performance of such a battery chemistry. In particular, we note that the growth of
such large, solid phase species on the surface of the catalyst support imposes new
design restrictions when optimizing a cathode for energy density. For instance, it is
not just the surface area of the catalyst support that determines the discharge capacity,
but also the amount of usable pore volume which can accommodate this solid phase
discharge product. As a proof of concept, we design and implement an architected
electrode with large pore volume and relatively small surface area, comparing it
with the more typical electrode geometries of thin films and nanoparticles. Finally,
we directly vary the pore volume of an electrode, noting the implications these
structural changes will have on a battery’s energy density.

49
4.1

Capacity Limitations

Chapters Two and Three discussed the reversibility of battery chemistries which
employ growth and dissolution of a new phase that serves as the lithium host in
a battery’s cathode. In addition, the growth mechanism of these phases, Li2 O2
and Li2 O, were explored, and both of these molten salt systems saw the growth of
large, several micron crystals which must be formed each battery cycle. These types
of chemistries attracted the attention of researchers due to their potentially large
discharge capacities.
Defining Capacity in Phase Forming Chemistries
In the case of an intercalation cathode such as the ones described in the Introduction
Chapter, it is easy to define the capacity of the cathode in terms of the mass of active
material. For example, the stoichiometric ratio of lithium in LiCO2 along with the
fraction of lithium which can be reversibly removed from the crystal structure gives
a well defined theoretical capacity of 164 mAh g-1 as shown in Table 1.1. Then for
a given mass loading of active material, a well controlled value when assembling
a battery, the expected capacity is easily calculated, and the actual capacity can
be normalized to this known weight of active material, giving a clear comparison
between ideal and real capacity.
This procedure is far less straightforward in a system which does not use an intercalation mechanism for lithium storage. For instance, consider the lithium-oxygen
system. As is often imagined for such a battery (quite impractically, but it makes
for a good thought experiment), the active material, oxygen, is taken from the atmosphere. This means that a lithium-oxygen battery has an infinite reservoir of
active material as its cathode. Clearly one cannot normalize the capacity of such a
battery by an infinite amount of oxygen. This raises the question of what value is a
reasonable one to normalize by so that battery capacities can be readily compared
and still have meaning.
As a result of this issue, the typical approach for normalizing battery capacity in
lithium-oxygen cells is to use the mass of the cathode catalyst support material. In
the case of the molten salt lithium-oxygen battery of Chapter Two, this corresponded
to the mass of Super P carbon that was then pressed onto the mesh current collector.
However, such an approach can lead to confusion and misleading reported capacities.
For instance, if a very small amount of cathode catalyst material, say on the order of
µg, is used in conjunction with a relatively large surface area current collector, one

50
can achieve outrageous normalized capacity values which begin to lose meaning[78].
When replacing this mass normalization with a geometric footprint normalization,
the true cell level capacity is revealed and often insignificant compared to realistic
values found in traditional lithium-ion cells. But normalizing by the footprint area
gives little information about the material performance of the system, a metric which
we are after.
The real question then is what is it that determines the capacity of such a battery. Why
were the lithium-oxygen batteries of Chapter Two generally limited to ~1400 mAh
gcarbon -1 for ~4 mg cm-2 carbon loading? In addition, why were the nitrate reduction
batteries of Chapter Three limited to 5–10% utilization of active material (nitrate
in the electrolyte)? When normalizing similarly to the lithium-oxygen system, this
utilization corresponded to ~800 mAh gnickel -1 for ~5 mg cm-2 nickel loading. To
further complicate things, these normalized capacities are not constant for varying
mass loadings of the cathode catalyst material, suggesting some property other than
mass is responsible for determining the observed capacity.
Consider the mechanism of charge storage in these types of batteries. A solid phase
which is electronically insulating is growing on the surface of this cathode catalyst
support material. Then having more surface to accommodate this growth should
result in a larger capacity, and one would expect the capacity to scale with the "true"
or electrochemical surface area of the electrode. The electrochemical surface area
corresponds to the total surface area of the electrode able to make contact with the
electrolyte and participate electrochemically. This surface area has been used as a
normalization scheme in some reports in literature[26], and it seems to get at the
material performance of the cathode despite being a difficult value to compare with
other types of battery chemistries.
Cathode Structure and Capacity
While the electrochemical surface area appears to be a clean way to anticipate
the capacity for such a battery chemistry, the system is not quite that simple. To
demonstrate this, consider the case of nitrate reduction where the cathode catalyst
is a bed of pressed nickel nanoparticles, just as in Chapter Three. By assuming a
packing density of nickel nanoparticles (here taken to be 60%, the packing factor
for randomly dispersed, pressed spheres[79]), we can calculate an upper bound on
the expected capacity for nitrate reduction by taking all of the pore space between
the nickel nanoparticles were filled with Li2 O. Equation 4.1 gives the simple series

51
of unit conversions needed to arrive at this capacity, as normalized by the mass of
nickel.
1 cm3 Ni 0.40 cm3 void 1 cm3 Li 2 O 2.01 g Li 2 O
8.91 g Ni
0.60 cm3 Ni
1 cm3 void 1 cm3 Li 2 O
(4.1)
1 mol Li 2 O
2 mol e−
96485 C 1 mAh
mAh
= 270
29.9 g Li 2 O 1 mol Li 2 O 1 mol e− 3.6 C
gNi
This simple calculation gives an upper bound on capacity significantly lower than
what is measured experimentally, even when accounting for the sparing solubility of
Li2 O in solution. A similar analysis can be performed for the lithium-oxygen system
with similar results. In order to address this apparent paradox, it is worthwhile to
consider the state of a fully discharged cathode. Figure 4.1 contains SEM images
of fully discharged cathodes from both the lithium-oxygen and nitrate reduction
systems of Chapters Two and Three.

Figure 4.1: Surface of fully discharge cathodes. (A) Li2 O2 and (B) Li2 O fully
covering the surface of the cathode after a full discharge (scale bar: 2 and 10 µm,
respectively).
Note that the surfaces of these cathodes are fully covered with the discharge product
of their respective systems. In particular, these particles which are several microns
in size are growing on top of the bed of pressed nanoparticles, not within it. Based
on a similar calculation as the one in Equation 4.1, the majority of the capacity
of these cathodes is a result of this discharge product growing out of the bed of
nanoparticles, and not the product growing within the pore space. As a result,
all the surface area of the cathode catalyst nanoparticles within the pressed bed is
essentially wasted, and the capacity of the battery cathode will not scale with its
electrochemical surface area.

52
This makes for a system vastly different than a traditional battery chemistry. Whereas
in the typical intercalation system the capacity is directly proportional to the mass
of active material, in these phase forming chemistries the capacity is determined by
the ability of the "useful" electrochemical surface area to accommodate the growth
of the discharge product. This means that not only will the cathode’s surface area
play a role in determining capacity, but its structure will too, and this is a much
harder thing to quantify.
While there is no cleanly defined material property which can be used to predict
the capacity of such a battery, the previous discussion is useful in elucidating which
attributes a cathode should have to enable a large capacity. The cathode should
have a large amount of surface that is exposed to open pore volume which can
accommodate the growth of discharge product. Figure 4.2 gives two examples of
structured electrodes with relatively large pore structures, and inverse opals have
been used in the past to study battery electrode properties[80].

Figure 4.2: Example structured nickel electrodes. (A,B) Nickel inverse opal, fabricated by electroplating nickel into the pores of settled 500 nm polymer particles,
which were then etched away (scale bar: 5 and 1 µm, respectively). (C) Nickel
nanoparticles sintered in a hydrogen environment at 500 °C (scale bar: 1 µm).
However, for the case of nitrate reduction, which we will use as a model system
in this chapter, the discharge product morphology of Li2 O is often on the order of
10 µm in size, far too large for these two structures to accommodate. The pore
size of these structures are limited by their fabrication methodology, making them
inappropriate to study this system. The following section will discuss the fabrication
of an electrode with explicit control over pore structure in the size regime we care
about.
This chapter will study the interplay between electrochemical surface area and pore
structure, and their effect on the discharge capacity of a nitrate reduction cathode.

53
In particular, we will approach this problem from two directions: first we will
design and fabricate an architected electrode with large pore volume and small
electrochemical surface area and see the effect on capacity, comparing against thin
film and nanoparticle morphologies; second we will compare two electrodes with
identical surface area but different pore volume, noting the effects on capacity and the
implications for battery energy density. Many of the conclusions drawn throughout
this chapter can be applied to other similar chemistries such as lithium-oxygen or
lithium-sulfur batteries.
4.2

Fabrication of Architected Electrodes

The previous section discussed the possible interplay of electrochemical surface area
and pore structure on the capacity of a nitrate reduction cathode (and by extension
other phase forming conversion chemistries such as lithium-oxygen and lithiumsulfur). In order to study this relation, it would be instructive to design an electrode
where all of the electrochemical surface area is "useful," in that the pore structure
cannot become clogged and render parts of the electrode inactive. The electrode
designs of Figure 4.2 point in the right direction, but better control and larger pore
sizes are desired. In particular, being able to accommodate the growth of crystals
10 µm in size would be required.
Lattice trusses are a class of structures which have been in use for centuries due
to their superior mechanical properties. Recent work in small scale fabrication has
allowed the miniaturization of these structures and led to remarkable mechanical
phenomena at the micro- and nano-scale[81–83]. This class of structures, if made
on the appropriate length scale, fulfill the requirements of an open pore cathode
described above, namely explicit control over structure and pore geometry. In
addition, lattice trusses display impressive mechanical properties which should lead
to a more mechanically resilient electrode. Due to the complexities of assembly
described later, this is beneficial for this molten salt system compared to other
potential structured electrodes.
Based on the criteria described above, a lattice with openings larger than 20 µm is
desired so that a film thickness of 10 µm would not close off the pore network. When
taking into account the beam thickness such a structure would have, but still wanting
to keep the structure as small as possible to allow a reasonable electrode thickness,
a unit cell of 30 µm is used. Figure 4.3 shows a computer rendered model of such
a lattice, with beam dimensions shown that are a result of the fabrication process

54

Figure 4.3: Model of lattice electrode unit, made using SOLIDWORKS. Unit cell:
30 µm; vertical beams: 4 µm diameter; horizontal beams: 4x10 µm ellipse.

described shortly. A simple cubic lattice is chosen for its open pore structure and
ease of fabrication. The result is a 120 µm thick electrode made of 4x4x4 unit cell
components which can be tiled in a plane to make a macroscopic electrode.
Nanoscribe Fabrication of Polymer Scaffold
In order to create such a lattice truss structure, we employ an instrument called
Nanoscribe which allows the fabrication of a polymer scaffold. This instrument has
been used to make many small scale structures for a variety of uses[83–86], and a
brief overview of its operation is provided here.
Nanoscribe employs a process called two-photon direct laser write lithography
to perform additive manufacturing of polymers on a small scale. Two-photon
lithography is a lithography process in which a photoresist is polymerized by a laser
of photons with half the energy needed for the polymerization reaction to occur (i.e.
twice the wavelength of light that would be used in a normal lithography process).
This laser is focused to a small volume, or voxel, such that the probability of two
photons interacting with the same molecule is high enough that the polymerization
process takes place. This voxel is then rastered in space, allowing the direct writing
of polymer throughout the photoresist. In the case of the Nanoscribe system, this
corresponds to a 780 nm femtosecond pulsed laser, and the resist of choice is the
proprietary IP-S due to its large interaction volume, giving a faster write time at
the sacrifice of resolution. A detailed description of the writing procedure and its
optimization for the model of Figure 4.3 is given in the appendix.
Once this structure has been written in the photoresist, it is developed in propylene

55
glycol methyl ether acetate (PGMEA) to remove the remaining photoresist, cleaned
in isopropyl alcohol (IPA), and then dried with a critical point dryer to prevent
surface tension from destroying the structure.

Figure 4.4: Fabrication of polymer lattice, as made with Nanoscribe. (A) Individual
lattice unit (scale bar: 50 µm). (B) Tiling of 880 of the lattice units from ’A’ (scale
bar: 1 mm). (C) Phone camera picture of tiled polymer lattices in ’B’ (scale: metal
disc is 1 cm in diameter).
The result of this Nanoscribe fabrication is the replication of the model of Figure
4.3 as a polymer scaffold in Figure 4.4a. This unit, a 120 µm cube, can then be tiled
across a substrate to form a macroscopic electrode 5 mm in diameter as in Figure
4.4b, a process which takes 20 hours of write time (see appendix for write time
optimization of what would have originally taken 1–2 weeks). The picture in Figure
4.4c demonstrates the macroscopic nature of the electrode, as well as showing the 1
cm polished stainless steel disc that is used as a substrate.
Sputtering a Nickel Coating
With a polymer scaffold of the desired geometry in hand, we now must coat it
with a conductive material so that it may serve as an electrode. Before coating the
electrode though, the substrate must be cleaned. Note the splotchy film surrounding
the lattice array in Figure 4.4b. This remnant photoresist will cause any deposited
film to easily peel off, as is happening to the nickel film of Figure 4.5a. This residue
seems to be a result of writing the structure on a polished metal substrate (to serve
as a current collector) as opposed to the traditional glass or silicon substrate used in
most Nanoscribe fabrications. In order to clean the stainless steel substrate, many
techniques were attempted (such as oxygen plasma cleaning among many others)
with little success. In the end, simplicity prevailed, and a sharpened wooden dowel
dipped in acetone and IPA was used to manually clean the area around the lattice
array, with the aid of an optical microscope. This method, while tedious, proved

56
surprisingly effective while maintaining the integrity of the polymer scaffold.
With the polymer scaffold made and the substrate clean enough to have a film adhere,
a sputtered film was used to coat the lattice and substrate. Sputtering has been used
in the past to coat smaller lattices[85, 86], and as long as the structure is open enough,
the film can effectively cover the polymer and function as an electrode. In the case
of nitrate reduction, this is best served by a nickel coating as detailed in Chapter
Three. Sputtering was performed with a magnetron sputter deposition system from
ATC Orion sputtering system, AJA International, Inc. with base pressure below 10-6
Torr. First, a thin (5–10 nm) chromium layer was deposited to encourage adhesion,
using a DC power supply at 100 W under 3 mTorr argon for five minutes. Following
this, a roughly 500 nm (as measured on the substrate) nickel layer was deposited,
using a DC power supply at 100 W under 5 mTorr argon for 1.5 hours.

Figure 4.5: Sputtered nickel film coating a polymer lattice. (A) Nickel film peeling
off substrate holding remains of an old lattice design iteration (scale bar: 1 mm). (B)
FIB cross-section of internal lattice beam with sputtered nickel film coating (scale
bar: 10 µm). (C) High resolution SEM micrograph showing the roughness of the
sputtered nickel surface (scale bar: 200 nm).
A FIB cross-section of an internal beam of the nickel coated lattice is shown in
Figure 4.5b. Note that while the nickel film is not as thick as it is when deposited on
the substrate, the beam is still fully covered with nickel, at an average thickness of
~250 nm. A high resolution SEM micrograph of the nickel film is shown in Figure
4.5c, demonstrating its surface roughness. This surface roughness will play a role
in determining the observed reaction kinetics, and its magnitude will be explored in
the next section.
Through this series of fabrication steps, we have made a macroscopic nickel electrode
with explicitly controlled geometry, and the final electrode is shown in Figure 4.6.
An identical stainless steel disc without a polymer scaffold was put in the sputtering

57

Figure 4.6: Nickel lattice electrode. (A,B) SEM micrographs of final iteration of
tiled nickel-coated lattices (scale bar: 200 µm and 1 mm, respectively). (C) Phone
camera picture of similar lattice electrode as those in ’A’ and ’B’ (scale: metal disc
is 1 cm in diameter).

chamber at the same time as the lattice array, providing a thin film electrode with
the same nickel surface morphology to serve as a control, such that the effect of
electrode geometry could be easily observed.
4.3

Electrochemical Surface Area of Nickel Electrodes

Before using this lattice truss array as a nitrate reduction cathode, it is important
to know its electrochemical surface area so that we can easily compare normalized
capacities across electrode geometries. There are two aspects of this cathode which
will affect the surface area: the micron-scale structural features imposed by design,
and the nano-scale features imposed by the roughness of the sputtered nickel film.
Brunauer–Emmett–Teller (BET) analysis is a common technique to measure the
surface area of nano-scale materials, but the pores in this lattice structure are too
large and the surface area too small for that method to be used here. Instead, we
turn to a recently developed electrochemical method which takes advantage of the
surface states of nickel.
Surface Chemistry of Nickel
The behavior of a polished nickel electrode in an alkaline solution over a wide
potential range is shown by CV in Figure 4.7, as performed in [87]. The electrode
was polished just before performing electrochemistry in order to remove as much
of the native oxide layer as possible. During the beginning of the scan at -1.2 V,
the hydrogen evolution reaction (HER) is taking place. As the potential increases,
the first oxidation peak around -0.7 V indicates the formation of nickel hydroxide
(Ni(OH)2 ). Moving higher in potential, a further oxidation occurs as nickel hy-

58

Figure 4.7: Surface electrochemistry of nickel, in an alkaline solution. Adapted
from [87].

droxide becomes nickel oxyhydroxide (NiOOH) around 0.5 V, and finally OER.
Reversing the scan, NiOOH reduces to Ni(OH)2 . Interestingly, this reversible redox
process drives the reaction at the positive electrode in both the nickel-cadmium and
nickel metal hydride batteries. Moving further down in potential gets back to HER,
notably not fully reversing the formation of Ni(OH)2 .
In recent years, several groups have developed techniques to measure the surface
area of nickel using electrochemical methods. In particular, there are two approaches
using the redox process described above which involve integrating the peak for either
the oxidation of nickel to Ni(OH)2 [88, 89] or the reduction of NiOOH to Ni(OH)2 [90,
91]. In this work, we will use the latter hydroxide/oxyhydroxide redox process
method, described in detail shortly. Both of these techniques require removing any
native oxide layer before doing any electrochemistry, and many different approaches
have been used. For planar electrodes, polishing is the preferred method, but
this would not work for structured electrodes for obvious reasons. For structured
electrodes, the two common approaches involve holding at the potential for HER in
hopes that the hydrogen will reduce the nickel surface or quickly etching the nickel
surface with an acid. We will use the latter method as we found that holding at the
potential for HER did not sufficiently clean the nickel surface.
In order to use the reduction of NiOOH to measure the surface area of a nickel

59
electrode, we need to know a few things. First, we must know the ratio of charge
passed to electrode surface area. This has been calculated by considering the
crystal structure of the preferred orientation of Ni(OH)2 , giving a value of 195 µC
cm-2 [90]. Second, we must be confident that the redox processes are limited to a
single monolayer of nickel. It has been shown that using an adsorbed oxalate anion
(C2 O4 2- ) will limit the Ni(OH)2 layer to a single monolayer during the first scan
when performed at a high scan rate[90, 91]. Finally, as already discussed the initial
surface must be cleaned of a native oxide layer, which can be achieved through an
acid bath.
Surface Area of Film and Lattice Electrodes
The process described in this section was first tested on commercial nickel foams,
and the values were in agreement with what was expected from literature. In order
to clean the nickel electrode before measuring its surface area, we briefly etch it
with an acid bath. The composition of the acid bath is shown in Table 4.1, and it
was found to be the best acid solution of several tested ones based on its etching rate
and selectivity for nickel[88].

Acid bath
component
Acetic Acid
(CH3 COOH)
Nitric Acid
(HNO3 )
Phosphoric Acid
(H3 PO4 )
Sulfuric Acid
(H2 SO4 )

Solution
composition
(vol %)
50
30
10
10

Table 4.1: Acid bath for nickel surface cleaning.
It is important that the acid bath be prepared before each etching session as the bath
ages fairly quickly. The nickel electrode was held in the bath for 120 seconds before
being quickly neutralized in a large bath of deionized (DI) water and finally put into
to the electrolyte for a CV similar to Figure 4.7.
All CVs were performed in a stirred aqueous electrolyte with 0.3 M potassium
hydroxide (KOH) and 0.1 M oxalic acid (H2 C2 O4 ). Due to reaction of these species,
the resulting electrolyte was 0.1 M KOH with 0.1 M potassium oxalate (K2 C2 O4 ).

60
The counter electrode was a platinum coated mesh and the reference electrode
Hg/HgO.

Figure 4.8: Electrochemical surface area of sputtered nickel, by CV in a 0.1 M KOH
and 0.1 M K2 C2 O4 electrolyte with a platinum counter electrode. Scan rate: 150
mV s-1 .
Figure 4.8 shows CVs of a sputtered nickel thin film working electrode which was
etched with the acid bath previously discussed. First, note that the peaks for both
Ni(OH)2 oxidation and NiOOH reduction grow with each cycle, suggesting that the
cycles following the first one engage more that just a monolayer. By integrating the
first scan’s NiOOH reduction peak and then calculating the electrochemical surface
area of nickel, we find a surface area of 1.6 cm2 . When integrating the peak, a CV
from a stainless steel disc was used as a background which was subtracted from the
integration, leaving just the peak area. To get a sense of the surface roughness of
the film, normalizing the electrochemical surface area value by the geometric area
of the electrode (a 1 cm disc), we find a surface area enhancement factor of 2.0.
Now that we have an understanding of the roughness of the sputtered nickel film, we
can look at the lattice electrode of Figure 4.6. In order to use this lattice electrode,
it is necessary to make sure that it can be wet. This is surprisingly non-trivial, and
a thorough discussion of lattice wetting is given in the appendix. In short, lattices
are kept in DI water overnight, and before use are checked for bubbles in an optical
microscope.
Confident that the nickel lattice can be fully wet, we put it through the same procedure
used on the nickel thin film. Unfortunately, the lattice electrodes did not survive

61
the acid bath used for cleaning the nickel surface, so a slightly different approach
was used. Instead of explicitly calculating the surface area of the electrode by first
removing the surface oxide layer, the CVs were performed on the untreated sputtered
nickel. By comparing a lattice electrode and thin film electrode, the ratio of their
surface areas can be calculated since the nickel surface states are the same.

Figure 4.9: Electrochemical surface area of lattice electrode, by CV in a 0.1 M KOH
and 0.1 M K2 C2 O4 electrolyte with a platinum counter electrode. Scan rate: 150
mV s-1 .
CVs of both the nickel thin film and lattice electrodes are shown in Figure 4.9.
Note that since the surface is untreated, the difference in peak size across each
scan is significantly smaller. While the actual value of the surface area can not be
calculated here, the ratio of NiOOH reduction peak areas gives a value of 2.1. The
calculated value from the SOLIDWORKS model of Figure 4.3 gives an expected
ratio of 2.5. When taking into account the presence of a small amount of bubbles in
the lattice electrode after performing CVs (likely due to HER and OER processes),
the discrepancy makes sense. Then we can be confident in the surface area ratio of
2.5 from the model, and using the surface area enhancement factor of the sputtered
nickel thin film, we know the electrochemical surface area of both the nickel thin film
and lattice electrodes. When attempting this process on nickel nanoparticles, the
surface area is far too large, and a meaningful value can not be calculated. Instead,
when normalizing by surface area the BET measurements mentioned in Chapter
Three are used.

62
4.4

Performance of Architected Electrodes

Now that the nickel thin film and lattice electrodes are well characterized, we can use
them as cathodes for nitrate reduction, comparing them to the nickel nanoparticle
electrode. A combination of CV and galvanostatic discharge are used to this end.
As before, these experiments were performed in the same Swagelok cells used in
Chapter Two. Nanoparticle cathodes consisted of ~6 mg cm-2 of nickel nanoparticles
pressed onto a stainless steel mesh without the use of binder. Lattice cathodes were
first wet with a 100 g L-1 LiNO3 –KNO3 in water solution by being immersed and
then briefly put in a vacuum chamber until all bubbles left the electrode. It was
then dried on a hotplate and finally overnight in a vacuum oven, impregnating the
electrode with salt so that it more easily wets when coming in contact with the
electrolyte. A detailed description of lattice wetting is given in the appendix. A
glass microfiber separator was impregnated with about 200 mg of LiNO3 –KNO3
molten salt electrolyte, and vacuum dried overnight at 200 °C. Anodes were a simple
lithium foil. All assembly took place in a glovebox, then the cells were filled with
argon after a leak test, and finally placed in an oven to be held at the 150 °C. A
detailed description of the cell design and assembly is given in the appendix
Cyclic Voltammetry
In order to compare the nickel thin film, lattice, and nanoparticle cathodes, CVs
around the nitrate reduction/nitrite oxidation redox potential are performed to understand the achievable current densities and reversibility of the redox couple.

Figure 4.10: Cyclic voltammetry of nitrate reduction on different cathodes, all of
which are 1 cm in diameter. Scan rate: 0.01 mV s-1 .

63
Figure 4.10 shows these CVs performed at a slow scan rate of 0.01 mV s-1 . The large
peak to peak separation suggests that this redox couple is not truly reversible, but
quasi-reversible in a similar vein as the CVs of Figure 2.1. As this is a fairly complex
redox process involving the growth and dissolution of solid species, one would expect
a quasi-reversible CV. For the case of nickel nanoparticles, the peak separation is
200 mV and centered around the potential observed in the GITT of Figure 3.4
for an all nitrate electrolyte. When considering the thorough characterization of
nitrate reduction with a nickel nanoparticle cathode in Chapter Three, we can be
confident that this quasi-reversible couple can be attributed to the reversible growth
and dissolution of Li2 O through nitrate reduction.
The peak to peak separation is much larger in the thin film and lattice cases, consistent
with the expectation for a much smaller surface area electrode in a quasi-reversible
system (the nanoparticle electrode has a surface area 2–3 orders of magnitude larger
than the thin film and lattice ones).In particular, the cathodic scan for nitrite oxidation
has a much larger overpotential than the anodic scan for nitrate reduction, but there
is a slowly increasing current across the cathodic scan. This wide peak separation
and small peak current suggests that the thin film and lattice electrodes will not be
able to achieve a large current on discharge. In addition, if the discharge capacity in
such a system were proportional to the cathode surface area, then we would expect
a negligible absolute capacity. This all points to these low surface area electrodes
being poor candidates for a cathode in this system.
Galvanostatic Discharge
Having an understanding of the rate capability of the low surface area electrodes,
we attempt a galvanostatic discharge. The CVs of Figure 4.10 give a sense of the
achievable currents for each cathode, and after testing a range of values, a current
of 25 µA was selected for the lattice cathode. In order to fairly compare across
electrode structures, the nanoparticle cathode was discharged at the same rate of 25
µA, and the thin film cathode was discharged at 10 µA so that the current normalized
by electrochemical surface area would be the same. These discharges are shown in
Figure 4.11.
Considering the CV behavior of the lattice and thin film electrodes, one might
have expected their discharge performance to be significantly worse than expected.
However, the discharge capacity of the lattice and nickel cathodes is quite similar
at the same current, clearly pointing to the role of structure in discharge capacity

64

Figure 4.11: First discharge of nitrate reduction on different cathodes, all of which
are 1 cm in diameter. Discharged at 25 µA for the lattice and nanoparticle cathodes,
and 10 µA for the thin film cathode.

for a chemistry such as nitrate reduction. Before more thoroughly discussing the
different normalization schemes available to compare these discharge capacities, we
first want to see if any Li2 O is actually growing within the lattice structure. As before,
the cathode was extracted from heated batteries inside an argon-filled glovebox and
rinsed with NMA. This process was extremely delicate for the lattice electrodes, and
few survived, disallowing a thorough discharge product characterization. However,
enough were salvaged to allow Figure 4.12, SEM micrographs of a discharged
lattice.

Figure 4.12: Nitrate reduction discharge product on lattice electrode, by SEM.
Discharge conditions similar to those of Figure 4.11 (scale bar: (A) 15 µm, (B) 20
µm, and (C) 10 µm).

Note two observations from the discharge product. First, it is growing within

65
the lattice structure, confirming that the pre-wetting of the electrode successfully
allowed the infiltration of electrolyte. Second, the morphology of the discharge
product is octahedral, suggesting that this is in fact Li2 O growing according to its
Wulff construction as observed for the nanoparticle cathodes in Figure 3.10. While
this is not an exhaustive confirmation of the nitrate reduction chemistry, it does seem
to confirm the growth of Li2 O.
Having established that the lattice electrode performed as desired and accommodated
the growth of Li2 O, we can turn back to the discharges of Figure 4.11. There are
several questions of interest, and the best way to address them is to normalize the
capacity in a few different ways: by projected electrode area to get a sense of cell
level capacity, by mass of nickel in a similar vein to the lithium-oxygen system, and
by the true surface area of nickel to see its interplay with electrode structure. The
projected electrode area is simply the area of the 1 cm diameter electrodes, 0.785
cm2 . The mass of nickel was simply measured for the nanoparticles, and for the
thin film and lattice it was estimated from the film thickness, and in the case of
the lattice, the calculated volumes in SOLIDWORKS. These masses were too small
(on the order of 100 µg) to measure accurately on a balance, but the substrate was
too heavy to use a micro-balance. The surface areas of nickel for the thin film and
lattice were taken from the CV measurements earlier in this chapter, and for the
nanoparticles, the BET value mentioned in Chapter Three was used.

Areal capacity
(mAh cm-2 projected )
Gravimetric capacity
(mAh g-1 nickel )
Surface capacity
(mAh m-2 nickel )

Thin film
electrode
(10 µA)
1.7

Lattice
electrode
(25 µA)
4.1

Nanoparticle
electrode
(25 µA)
5.1

3400

8200

790

8700

8300

79

Table 4.2: Normalized discharge capacity of nitrate reduction on different cathodes,
all of which are 1 cm in diameter.
These normalized capacities are displayed in Table 4.2, and they indicate a few
things. First, the lattice electrode has an areal capacity comparable to the nanoparticle one, and both of these values are similar to those found in commercial lithium-ion
batteries. This clearly demonstrates the role that "useful" surface area plays in de-

66
termining the capacity in these types of systems. Second, the thin film and lattice
electrodes have gravimetric values so large that they lose meaning. By architecting
an electrode comprised of a thin film wrapped around a polymer scaffold, one can
design an electrode with an arbitrarily small amount of nickel and achieve an inflated
capacity value, a problem in the lithium-oxygen community which has thankfully
started to be addressed in recent years. Finally, the surface capacities of the thin film
and lattice electrodes are not only much larger than that of the nickel nanoparticles
as discussed earlier, but they are also remarkably similar in value. This suggests
that if one designs a sufficiently open pore structure, the capacity will in fact scale
with surface area.
4.5

Commercial Foam Electrode Comparison

In the previous section, we addressed the role of cathode structure in a nitrate
reduction battery by designing a proof of concept electrode to demonstrate among
other things that capacity will scale with surface area given the appropriate pore
structure. Now we take the inverse approach by using two identical electrodes and
changing the pore structure of one of them, seeing the effect on performance as a
nitrate reduction cathode. To do so, we use commercially available nickel foams
(MTI Corporation), one of which is as synthesized and 1.6 mm tall, the other which
was the same but then densified by putting it through rollers to make it 80 µm tall.
These foams will be referred to as "open" and "dense" henceforth. This rolling
process allows the significant removal of pore volume while keeping the nickel mass
and surface area relatively unchanged.
The structure of these foams is shown in Figure 4.13. The open foam has pores
ranging from 200–500 µm in size, and based on the bulk foam density, a porosity
of 97.4%. Interestingly, because of the way the dense foam was densified, its pores
are no longer isotropic. The out of plane direction saw a 95% strain, and the pores
collapsed accordingly. This resulted in pores that were still a few hundred microns
for the in plane direction, but only a few microns for the out of plane direction. This
densification process resulted in a porosity of 48.2%, again as calculated from the
bulk foam density.
With a good sense of the structure of these electrodes, we can proceed with a similar
set of electrochemical measurements as the ones taken in the previous section with
thin film, lattice, and nanoparticle electrodes. Again, because the foam is not trivial
to wet (though much easier than the lattice), we pre-wet it with the same 100 g L-1

67

Figure 4.13: Foam electrode morphologies. (A,C) Open and (B,D) dense nickel
foam structures (scale bar: (A,B) 2 mm and (C,D) 200 µm).

LiNO3 –KNO3 aqueous solution. This time, due to the easy handling of the nickel
foam, we can simply sonicate it in solution, dry it on the hotplate, infiltrate with
more electrolyte, and finally dry overnight in the vacuum oven. Due to the thickness
of the open foam, a total of 450 mg of electrolyte was used instead of the typical
200 mg. Otherwise, the cell assembly procedure was the same as in the previous
section.
Cyclic Voltammetry
As in the previous section, CVs around the nitrate reduction/nitrite oxidation redox
potential are performed to understand the achievable current densities and reversibility of the redox couple. Because the surface area of a nickel foam is ~500 cm2 g-1 ,
this gives an electrode surface area about a factor of four larger than the lattice
electrode of the previous section and an order of magnitude larger than the thin film
case. As a result, we perform these CVs at a faster scan rate of 0.02 mV s-1 , twice
that of the previous section.
These Cvs are shown in Figure 4.14. Interestingly, the nitrate reduction peak current
of the open foam is more than twice that of the nanoparticle one in Figure 4.10,

68

Figure 4.14: Cyclic voltammetry of nitrate reduction on foam cathodes, both of
which are 1 cm in diameter. Scan rate: 0.02 mV s-1
but it is also shifted 200 mV lower in potential than the nanoparticle case. This
is interesting, as one would expect the lower surface area electrode to have both a
lower peak current and an increased overpotential. However, this could make sense
when considering the quasi-reversible nature of this redox couple. Due to the large
porosity of the open foam, it would not be subject to pore blocking, potentially
allowing the the electrode to reach lower potentials before it becomes passivated. In
addition, note that the peak currents of the dense foam are significantly lower than
those of the open one, further supporting the argument of pore blocking playing a
role here.
Galvanostatic Discharge
Comparing the electrochemical surface areas of the foam and lattice electrodes, in
addition to considering the CVs of Figure 4.14, gives a sense of the achievable
current for the open and dense foam electrodes. After testing a range of discharge
currents, we then attempt a deep galvanostatic discharge with a current of 50 µA,
twice that of the lattice electrode. Figure 4.15 shows this discharge profile for both
the open and dense foams.
Unlike when comparing the CV and discharge capacity for the lattice and nanoparticle cathodes, there is no apparent discrepancy in the case of the foam cathodes. The
open foam displayed larger peak currents in its CV and a larger discharge capacity
on its galvanostatic half-cycle. Interestingly, both the open and dense foams display

69

Figure 4.15: Galvanostatic discharge of nitrate reduction on foam cathodes, both of
which are 1 cm in diameter. Discharged at 50 µA

a larger discharge capacity than the nanoparticle electrode (at twice the discharge
rate too) despite having a significantly smaller electrochemical surface area.

Thickness
(mm)
Porosity
(%)
Discharge capacity
(mAh cm-2 projected )
Li2 O fraction
(%)
Filled pore space
(%)

Open foam
1.6

Dense foam
0.08

97.4

48.2

21.0

8.5

3.6

29.4

3.7

61.0

Table 4.3: Structure of discharged foam cathodes, where Li2 O fraction corresponds
to the thickness of Li2 O grown relative to the electrode thickness, and filled pore
space corresponds to the percent of porosity filled with Li2 O assuming it all grows
within the electrode.
The normalized capacities and some structural information of the two foam electrodes are shown in Table 4.3. Before beginning a detailed discussion of the role of
porosity in these foam electrodes, it is worth pointing out the absurdly large areal
capacity of the open foam cathode: 21.0 mAh cm-2 . This is about six to seven
times larger than what one can expect for a lithium-ion battery, and it corresponds

70
to stripping ~100 µm of lithium metal from the anode.
When considering the structural information of Table 4.3, there are two ways to
interpret the data. First, we can make a similar argument to the one we have made
throughout this chapter: the discharge capacity of a nitrate reduction battery, and
by extension any other phase forming conversion chemistry, is largely dependent on
the architecture of the cathode and not just its available surface area. This is clear as
we note that two cathodes with similar surface area have a significant difference in
discharge capacity which is accounted for by the larger porosity of the open foam.
This issue of pore blocking becomes clear when considering the fraction of pore
space that must be filled by Li2 O in order to accommodate the measured discharge
capacity. The open foam is only filling 3.7% of its pore space with Li2 O, while the
dense foam is filling 61%. This all suggests that the open foam is the preferable
cathode structure of the two.
However, we can look at these numbers in a different way. When considering
the utilization of pore space, the dense foam performed far better than the open
foam. It may seem strange to consider the utilization of pore space, but in a phase
forming conversion chemistry such a metric is analogous to the packing factor of
active material in a lithium-ion cathode. If the packing factor of active material
in a lithium-ion battery is low, now matter how great the loading is, it will always
have a poor cell level energy density. In the case of these foam electrodes, the same
thing is true of pore space utilization. If only a small fraction of the pore space
is used, then a significant amount of mass and volume will become attributed to
the electrolyte filling that unused pore space, resulting in a poor cell level energy
density. This concept is not often used in the lithium-oxygen field. However, it
has been explored in the lithium-sulfur one[92, 93], and simple calculations have
shown that unless a much smaller amount of electrolyte is used than what is typical
in research, lithium-sulfur batteries will never be able to compete with lithium-ion
ones[94].
When considering both of these interpretations of Table 4.3, it would appear that
the ideal foam electrode lies somewhere between the open and dense foams studied
in this section. By varying the foam thickness between these two edge cases and
performing some simple cell level energy density calculations, one could find this
ideal foam porosity for maximizing the energy density of this nitrate reduction
system.

71
4.6

Summary

We began this chapter by motivating the need to find an appropriate metric to
normalize the capacity of phase forming conversion reaction batteries, eventually
determining that one needs to understand the relationship between electrochemical
surface area and pore structure to be able to predict a cathodes discharge capacity. We
then designed and fabricated a proof of concept nickel coated lattice electrode and
determined its electrochemical surface area. Using this proof of concept electrode,
we partially decoupled the cathodes discharge capacity from its electrochemical
surface area, only seeing a direct scaling when the pore structure was explicitly designed such that the electrode behaved as a thin film. Finally, we inverted the notion
of varying an electrodes "useful" electrochemical surface area by instead varying an
electrodes porosity with a fixed surface area. This allowed us to reinforce the conclusions drawn with the lattice electrode, while also demonstrating the intricacies
of cell design. While we were able to attain very large discharge capacities by using
structured electrodes, they were still lacking in their achievable current densities,
rendering them inappropriate for most battery applications. However, these findings
provide guidelines for how one would design the ideal electrode: a hierarchical
electrode which combines large scale porosity at the micro-scale to accommodate
discharge product growth with a large surface area at the nano-scale to accommodate
a high current density.

72
Chapter 5

SUMMARY AND OUTLOOK

Throughout this thesis, we have studied battery cathodes which employ a phase
forming conversion reaction chemistry, as opposed to the more common intercalation
chemistry of lithium-ion batteries. In particular, we have stressed the difference in
how these two classes of batteries store charge and the resulting distinction in their
reaction mechanisms.
For the lithium-oxygen battery, this has meant a discussion of how an insulating
species can grow electrochemically on the cathode catalyst support. As the solution
mediated growth process for Li2 O2 was developed in literature, it became clear that
it is these intermediate species which facilitate a large capacity but are also the ones
that enable unwanted side reactions. We determined that through the use of a molten
salt electrolyte, the issue of organic solvent decomposition can be obviated. In addition, through thorough characterization we demonstrated the reversible growth and
dissolution of Li2 O2 which grew according to its Wulff construction. Furthermore,
this molten salt electrolyte resulted in a lithium-oxygen battery with enhanced intermediate solubility and reaction kinetics, leading to a large discharge capacity at
a remarkably low overpotential. However, due to degradation of the amorphous
carbon cathode catalyst support, the battery still failed due to Li2 CO3 accumulation
and passivation. While this work has served as a promising step towards realizing
a truly reversible lithium-oxygen battery, there are still many issues left unresolved,
notably the lack of a stable cathode catalyst which works in this electrolyte and the
lack of an effective protected lithium anode.
Continuing work in this molten nitrate electrolyte, we introduced a new take on an
old primary battery chemistry. Through the use of a nickel nanoparticle catalyst,
we have enabled the reversible nitrate reduction/nitrite oxidation redox couple,
carefully characterizing the growth and dissolution of Li2 O. Interestingly, this Li2 O
also grew according to its Wulff construction. In addition, we have developed a
quantitative ex situ chemical confirmation of this chemistry in a similar vein to
using mass spectrometry in the lithium-oxygen field. Confident in the reversibility
of this chemistry, we cycled a battery for over 1000 hours, and with no obvious
cathodic degradation mechanisms, it should have been able to cycle for far longer

73
given a more stable anode. However, as with the lithium-oxygen system, the lack of
a protected lithium anode stands as a crucial impediment to further development of
this chemistry.
Having examined these two phase forming conversion reaction chemistries, we
turned to the question of how to best define capacity in such a battery, using nitrate
reduction as a model system. By explicitly designing a proof of concept architected
electrode with controllable surface area and pore size, we were able to study the
interplay of these two properties and their effect on discharge capacity. In addition,
we used commercially available foam electrodes to similarly examine the role pore
structure plays in determining discharge capacity. It became clear that discharge
capacity only scales with electrochemical surface area when that surface area is able
to behave as a thin film, allowing the unobstructed growth of discharge product.
This suggests that for any real electrode, there is no simple normalization scheme
one can employ to easily define the system like there is in the lithium-ion battery.
It also lends a simple design principle if one wants to make a large capacity phase
forming chemistry: incorporate an open pore structure to accommodate discharge
product growth.
Putting all of this together, we have a good idea of what is necessary for any similar
phase forming chemistry to be realized as a practical battery. First, if a chemistry
involves the dissolution of species into the electrolyte, then it must be paired with
a protected anode to avoid crossover and the dissolved species needs to be stable
to every component in the battery. Second, if discharge product is to be grown
on the surface of some catalytic support, this structure needs to be designed to
accommodate the growth to enable a useful areal capacity. Finally, it is important
to remember that inactive components, such as a large current collector supporting
the catalyst or a large electrolyte volume enabling the product solubility, begin to
play a large role when calculating the cell level energy density. Solving all of these
problems may seem a daunting task, but it was not so long ago when the same was
said of lithium-ion batteries.

74

BIBLIOGRAPHY

1. International Energy Agency (IEA), Clean Energy Ministerial, and Electric
Vehicles Initiative (EVI). Global EV Outlook 2016: Beyond one million electric cars. IEA Publications (2016).
2. Nykvist, B. & Nilsson, M. Rapidly falling costs of battery packs for electric
vehicles. Nature Climate Change 5, 329–332. issn: 17586798 (2015).
3. Routledge, R. A popular history of science (G. Routledge and sons, 1880).
4. Dell, R. M. & Rand, D. A. J. Understanding batteries (Royal Society of
Chemistry, 2001).
5. Whittingham, M. S. Electrical Energy Storage and Intercalation Chemistry.
Science 192, 1126–1127. issn: 09476539 (1976).
6. Whittingham, M. S. Chemistry of intercalation compounds: Metal guests
in chalcogenide hosts. Progress in Solid State Chemistry 12, 41–99. issn:
00796786 (1978).
7. Mizushima, K., Jones, P. C., Wiseman, P. J. & Goodenough, J. B. LixCoO2
(0State Ionics 3-4, 171–174. issn: 01672738 (1981).
8. Yazami, R. & Touzain, P. A reversible graphite-lithium negative electrode
for electrochemical generators. Journal of Power Sources 9, 365–371. issn:
03787753 (1983).
9. Padhi, A. K., Nanjundaswamy, K. & Goodenough, J. B. Phospho-olivines as
Positive-Electrode Materials for Rechargeable Lithium Batteries. Journal of
The Electrochemical Society 144, 1188–1194. issn: 00134651 (1997).
10. Tarascon, J. M. & Armand, M. Issues and challenges facing rechargeable
lithium batteries. Nature 414, 359–367. issn: 00280836 (2001).
11. Schmuch, R., Wagner, R., Hörpel, G., Placke, T. & Winter, M. Performance and
cost of materials for lithium-based rechargeable automotive batteries. Nature
Energy 3, 267–278. issn: 2058-7546 (2018).
12. Blomgren, G. E. The Development and Future of Lithium Ion Batteries.
Journal of The Electrochemical Society 164, A5019–A5025. issn: 0013-4651
(2017).
13. Graetz, J., Ahn, C. C., Yazami, R. & Fultz, B. Highly Reversible Lithium
Storage in Nanostructured Silicon. Electrochemical and Solid-State Letters 6,
A194. issn: 10990062 (2003).
14. Kraytsberg, A., Ein-Eli, Y., Kraytsberg, A. & Ein-Eli, Y. Higher, stronger, better
... A review of 5 volt cathode materials for advanced lithium-ion batteries.
Advanced Energy Materials 2, 922–939. issn: 16146832 (2012).

75
15. Manthiram, A., Knight, J. C., Myung, S. T., Oh, S. M. & Sun, Y. K. Nickel-Rich
and Lithium-Rich Layered Oxide Cathodes: Progress and Perspectives. Advanced Energy Materials 6. issn: 16146840. doi:10.1002/aenm.201501010
(2016).
16. McCloskey, B. D., Burke, C. M., Nichols, J. E. & Renfrew, S. E. Mechanistic insights for the development of Li–O ₂ battery materials:
addressing Li ₂ O ₂ conductivity limitations and
electrolyte and cathode instabilities. Chem. Commun. 51, 12701–12715. issn:
1359-7345 (2015).
17. Abraham, K. M. & Jiang, Z. A Polymer Electrolyte - Based Rechargeable
Lithium / Oxygen Battery. Journal of the Electrochemical Society 143, 1–5.
issn: 00134651 (1996).
18. Bruce, P. G., Freunberger, S. A., Hardwick, L. J. & Tarascon, J.-M. Li–O2 and
Li–S batteries with high energy storage. Nature Materials 11, 19–29. issn:
1476-1122 (2012).
19. Luntz, A. C. & McCloskey, B. D. Nonaqueous Li-air batteries: A status report.
Chemical Reviews 114, 11721–11750. issn: 15206890 (2014).
20. Rahman, M. A., Wang, X. & Wen, C. A review of high energy density lithiumair battery technology. Journal of Applied Electrochemistry 44, 5–22. issn:
0021891X (2014).
21. Mitchell, R. R., Gallant, B. M., Shao-Horn, Y. & Thompson, C. V. Mechanisms
of Morphological Evolution of Li ₂ O ₂ Particles
during Electrochemical Growth. The Journal of Physical Chemistry Letters 4,
1060–1064. issn: 1948-7185 (2013).
22. Horstmann, B. et al. Rate-Dependent Morphology of Li 2 O 2 Growth in Li–O
2 Batteries. The Journal of Physical Chemistry Letters 4, 4217–4222. issn:
1948-7185 (2013).
23. Xu, J. J., Wang, Z. L., Xu, D., Zhang, L. L. & Zhang, X. B. Tailoring deposition
and morphology of discharge products towards high-rate and long-life lithiumoxygen batteries. Nature Communications 4, 1–10. issn: 20411723 (2013).
24. Hou, J., Yang, M., Ellis, M. W., Moore, R. B. & Yi, B. Lithium oxides
precipitation in nonaqueous Li-air batteries. Physical Chemistry Chemical
Physics 14, 13487. issn: 1463-9076 (2012).
25. Aetukuri, N. B. et al. Solvating additives drive solution-mediated electrochemistry and enhance toroid growth in non-aqueous Li-O2batteries. Nature
Chemistry 7, 50–56. issn: 17554349 (2015).
26. Xu, C., Gallant, B. M., Wunderlich, P. U., Lohmann, T. & Greer, J. R. ThreeDimensional Au Microlattices as Positive Electrodes for Li-O2
Batteries. ACS Nano 9, 5876–5883. issn: 1936086X (2015).

76
27. Johnson, L. et al. The role of LiO 2 solubility in O 2 reduction in aprotic
solvents and its consequences for Li-O 2 batteries. Nature Chemistry 6, 1091–
1099. issn: 17554349 (2014).
28. Luntz, A. C. et al. Tunneling and polaron charge transport through Li2O2 in
Li-O2 batteries. Journal of Physical Chemistry Letters 4, 3494–3499. issn:
19487185 (2013).
29. Gutmann, V. Empirical parameters for donor and acceptor properties of solvents. Electrochimica Acta 21, 661–670. issn: 00134686 (1976).
30. Cho, M. H. et al. The effects of moisture contamination in the Li-O2battery.
Journal of Power Sources 268, 565–574. issn: 03787753 (2014).
31. Meini, S., Piana, M., Tsiouvaras, N., Garsuch, A. & Gasteiger, H. A. The Effect
of Water on the Discharge Capacity of a Non-Catalyzed Carbon Cathode for LiO2 Batteries. Electrochemical and Solid-State Letters 15, A45. issn: 10990062
(2012).
32. Chen, Y., Freunberger, S. A., Peng, Z., Fontaine, O. & Bruce, P. G. Charging
a Li–O₂ battery using a redox mediator. Nature Chemistry 5,
489–494. issn: 1755-4330 (2013).
33. Chase, G. V. et al. Soluble Oxygen Evolving Catalysts for Rechargeable MetalAir Batteries. US patent 1, US 20120028137A1 (2012).
34. Fong, R., von Sacken, U. & Dahn, J. R. Studies of Lithium Intercalation into
Carbons Using Nonaqueous Electrochemical Cells. Journal of The Electrochemical Society 137, 2009. issn: 00134651 (1990).
35. Freunberger, S. A. et al. Reactions in the Rechargable Lithium-O2 Battery
with Alkyl Carbonate Electrolytes. Journal of the American Chemical Society
133, 8040–8047 (2011).
36. Kwabi, D. G. et al. Chemical instability of dimethyl sulfoxide in lithium-air
batteries. Journal of Physical Chemistry Letters 5, 2850–2856. issn: 19487185
(2014).
37. Sharon, D. et al. Reactivity of amide based solutions in lithium-oxygen cells.
Journal of Physical Chemistry C 118, 15207–15213. issn: 19327455 (2014).
38. Bryantsev, V. S. et al. Predicting solvent stability in aprotic electrolyte Liair batteries: Nucleophilic substitution by the superoxide anion radical (O2-).
Journal of Physical Chemistry A 115, 12399–12409. issn: 10895639 (2011).
39. McCloskey, B. D., Bethune, D. S., Shelby, R. M., Girishkumar, G. & Luntz,
A. C. Solvents’ Critical Role in Nonaqueous Lithium-Oxygen Battery Electrochemistry. The Journal of Physical Chemistry Letters 2, 1161–1166. issn:
1948-7185 (2011).

77
40. Black, R. et al. Screening for superoxide reactivity in Li-O₂ batteries: effect on Li₂O₂/LiOH crystallization. Journal of the American Chemical Society 134, 2902–2905. issn: 1520-5126
(2012).
41. McCloskey, B. D. et al. Combining Accurate O2 and Li2O2 Assays to Separate
Discharge and Charge Stability Limitations in Nonaqueous Li-O₂
Batteries. J. Phys. Chem. Lett. 4, 2989–2993 (2013).
42. Gallant, B. M. et al. Chemical and Morphological Changes of Li–O2 Battery
Electrodes upon Cycling. The Journal of Physical Chemistry C 116, 20800–
20805. issn: 1932-7447 (2012).
43. McCloskey, B. D. et al. Twin problems of interfacial carbonate formation
in nonaqueous Li-O₂ batteries. Journal of Physical Chemistry
Letters 3, 997–1001. issn: 19487185 (2012).
44. Peng, Z., Freunberger, S. A., Chen, Y. & Bruce, P. G. A reversible and higherrate Li-O₂ battery. Science 337, 563–566. issn: 10959203 (2012).
45. Riaz, A. et al. Carbon-free cobalt oxide cathodes with tunable nanoarchitectures for rechargeable lithium-oxygen batteries. Chemical Communications
49, 5984. issn: 1359-7345 (2013).
46. Liu, X., Wang, D. & Shi, S. Exploration on the possibility of Ni foam as
current collector in rechargeable lithium-air batteries. Electrochimica Acta 87,
865–871. issn: 00134686 (2013).
47. Aurbach, D., McCloskey, B. D., Nazar, L. F. & Bruce, P. G. Advances in
understanding mechanisms underpinning lithium-air batteries. Nature Energy
1, 1–11. issn: 20587546 (2016).
48. Lu, Y.-C. et al. Lithium-oxygen batteries: bridging mechanistic understanding
and battery performance. Energy & Environmental Science 6, 750. issn: 17545692 (2013).
49. Uddin, J., Addison, D., Giordani, V., Chase, G. V. & Walker, W. Alkali
Metal/Oxygen Batteries Employing Molten Nitrate Electrolytes 2014.
50. Raistrick, I. D., Poris, J. & Huggins, R. A. Nitrate Molten Salt Electrolytes for
Use in Intermediate Temperature Lithium Cells 1981.
51. Miles, M. H. Cation Effects on the Electrode Reduction of Molten Nitrates.
Journal of The Electrochemical Society 127, 1761. issn: 00134651 (1980).
52. Zambonin, P. G. & Jordan, J. Redox chemistry of the system O2-O2–O22–O2in fused salts. Journal of the American Chemical Society 91, 2225–2228. issn:
0002-7863 (Apr. 1969).
53. Zambonin, P. G. Reversible oxygen electrode systems in molten salts. Journal
of Electroanalytical Chemistry 33, 243–251. issn: 00220728 (1971).

78
54. Zawodzinski, T. A. & Osteryoung, R. A. Donor-Acceptor Properties of AmbientTemperature Chloroaluminate Melts. Inorganic Chemistry 28, 1710–1715.
issn: 1520510X (1989).
55. Evans, J., Vincent, C. A. & Bruce, P. G. Electrochemical measurement of
transference numbers in polymer electrolytes. Polymer 28, 2324–2328. issn:
00323861 (1987).
56. Walker, W. et al. A Rechargeable Li-O₂ Battery Using a Lithium
Nitrate/N,N-Dimethylacetamide Electrolyte. Journal of the American Chemical Society 135, 2076–2079. issn: 1520-5126 (2013).
57. Burke, C. M., Pande, V., Khetan, A., Viswanathan, V. & McCloskey, B. D.
Enhancing electrochemical intermediate solvation through electrolyte anion
selection to increase nonaqueous Li–O₂ battery capacity. Proceedings of the National Academy of Sciences 112, 9293–9298. issn: 00278424 (2015).
58. Sharon, D. et al. Oxidation of Dimethly Sulfoxide Solutions by Electrochemical Reduction of Oxygen. The Journal of Physical Chemistry Letters 4, 3115–
3119. issn: 1948-7185 (2013).
59. Strickland-Constable, R. F. Kinetics and Mechanism of Crystallization from
the Fluid Phase and of the Condensation and Evaporation of Liquids (1968).
60. Gibbs, J. W. The collected works of J. Willard Gibbs, volume I: thermodynamics
(Yale University Press, 1928).
61. Wulff, G. On the question of speed of growth and dissolution of crystal surfaces.
Z. Kristallogr 34, 449–530 (1901).
62. Herring, C. The use of classical macroscopic concepts in surface energy
problems in Structure and properties of solid surfaces (1953), 5.
63. Radin, M. D., Rodriguez, J. F., Tian, F. & Siegel, D. J. Lithium peroxide
surfaces are metallic, while lithium oxide surfaces are not. Journal of the
American Chemical Society 134, 1093–1103. issn: 00027863 (2012).
64. Mo, Y., Ong, S. P. & Ceder, G. First-principles study of the oxygen evolution
reaction of lithium peroxide in the lithium-air battery. Physical Review B Condensed Matter and Materials Physics 84, 1–9. issn: 10980121 (2011).
65. Lopez, N. et al. Reversible reduction of oxygen to peroxide facilitated by
molecular recognition. Science 335, 450–453. issn: 10959203 (2012).
66. Desimoni, E., Palmisano, F. & Zambonin, P. G. Catalytic currents in fused
salts: Discharge mechanism of nitrite in molten alkali nitrates. Journal of
Electroanalytical Chemistry 84, 315–322. issn: 00220728 (1977).
67. Huggins, R. Advanced batteries: materials science aspects (Springer Science
& Business Media, 2008).

79
68. Giordani, V. et al. A Molten Salt Lithium-Oxygen Battery. Journal of the
American Chemical Society 138, 2656–2663. issn: 15205126 (2016).
69. Bard, A. J. & Faulkner, L. R. Electrochemical methods: fundamentals and
applications (wiley New York).
70. Bansal, A., Li, X., Yi, S. I., Weinberg, W. H. & Lewis, N. S. Spectroscopic studies of the modification of crystalline si(111) surfaces with covalently-attached
alkyl chains using a chlorination/alkylation method. Journal of Physical Chemistry B 105, 10266–10277. issn: 10895647 (2001).
71. Chiang, Y.-M., Birnie, D. P., Kingery, W. D. & Newcomb, S. Physical ceramics: principles for ceramic science and engineering (Wiley New York,
1997).
72. Tietz, F., Dias, F. J., Simwonis, D. & Stöver, D. Evaluation of commercial
nickel oxide powders for components in solid oxide fuel cells. Journal of the
European Ceramic Society 20, 1023–1034. issn: 09552219 (2000).
73. Chia-Ching, W. & Cheng-Fu, Y. Investigation of the properties of nanostructured Li-doped NiO films using the modified spray pyrolysis method.
Nanoscale Research Letters 8, 1–5. issn: 19317573 (2013).
74. Nguyen, Q. M. Technological status of nickel oxide cathodes in molten carbonate fuel cells - A review. Journal of Power Sources 24, 1–19. issn: 03787753
(1988).
75. Zhong, L. et al. In Situ Transmission Electron Microscopy Observations of
Electrochemical Oxidation of Li 2 O 2. Nano Letters 13, 2209–2214. issn:
1530-6984 (May 2013).
76. Seriani, N. Ab initio thermodynamics of lithium oxides: from bulk phases to
nanoparticles. Nanotechnology 20, 445703. issn: 0957-4484 (Nov. 2009).
77. Desimoni, E., Paniccia, F. & Zambonin, P. G. Solubility and detection (down
to 30 p.p.b.) of oxygen in molten alkali nitrates. Journal of Electroanalytical
Chemistry 38, 373–379. issn: 00220728 (1972).
78. Kwak, W.-J. et al. A Mo 2 C/Carbon Nanotube Composite Cathode for
Lithium-Oxygen Batteries with High Energy Efficiency and Long Cycle Life.
ACS Nano 9, 4129–4137. issn: 1936-0851 (2015).
79. Dullien, F. A. Porous media: fluid transport and pore structure (Academic
press, 2012).
80. Pikul, J. H., Gang Zhang, H., Cho, J., Braun, P. V. & King, W. P. Highpower lithium ion microbatteries from interdigitated three-dimensional bicontinuous nanoporous electrodes. Nature Communications 4, 1732–1735. issn:
20411723 (2013).

80
81. Bauer, J., Schroer, A., Schwaiger, R. & Kraft, O. Approaching theoretical
strength in glassy carbon nanolattices. Nature Materials 15, 438–443. issn:
14764660 (2016).
82. Wendy Gu, X. & Greer, J. R. Ultra-strong architected Cu meso-lattices. Extreme Mechanics Letters 2, 7–14. issn: 23524316 (2015).
83. Meza, L. R., Das, S. & Greer, J. R. Strong, lightweight, and recoverable threedimensional ceramic nanolattices. Science (New York, N.Y.) 345, 1322–6. issn:
1095-9203 (2014).
84. Jang, D., Meza, L. R., Greer, F. & Greer, J. R. Fabrication and deformation of
three-dimensional hollow ceramic nanostructures. Nature Materials 12, 893–
898. issn: 14761122 (2013).
85. Montemayor, L. C., Meza, L. R. & Greer, J. R. Design and Fabrication of Hollow Rigid Nanolattices via Two-Photon Lithography. Advanced Engineering
Materials 16, 184–189 (2014).
86. Liontas, R. & Greer, J. R. 3D nano-architected metallic glass: Size effect
suppresses catastrophic failure. Acta Materialia 133, 393–407. issn: 13596454
(2017).
87. Seghiouer, A., Chevalet, J., Barhoun, A. & Lantelme, F. Electrochemical oxidation of nickel in alkaline solutions: a voltammetric study and modelling.
Journal of Electroanalytical Chemistry 442, 113–123. issn: 15726657 (Jan.
1998).
88. Grdeń, M., Alsabet, M. & Jerkiewicz, G. Surface Science and Electrochemical
Analysis of Nickel Foams. ACS Applied Materials & Interfaces 4, 3012–3021.
issn: 1944-8244 (June 2012).
89. Van Drunen, J. et al. Comprehensive Structural, Surface-Chemical and Electrochemical Characterization of Nickel-Based Metallic Foams. ACS Applied
Materials & Interfaces 5, 6712–6722. issn: 1944-8244 (July 2013).
90. Hall, D. S., Bock, C. & MacDougall, B. R. An Oxalate Method for Measuring
the Surface Area of Nickel Electrodes. Journal of the Electrochemical Society
161, H787–H795. issn: 0013-4651 (Sept. 2014).
91. Van Drunen, J., Barbosa, A. F. B. & Tremiliosi-Filho, G. The Formation of Surface Oxides on Nickel in Oxalate-Containing Alkaline Media. Electrocatalysis
6, 481–491. issn: 1868-2529 (Sept. 2015).
92. Chung, W. J. et al. The use of elemental sulfur as an alternative feedstock for
polymeric materials. Nature Chemistry 5, 518–524. issn: 17554330 (2013).
93. Thangavel, V. et al. A Microstructurally Resolved Model for Li-S Batteries
Assessing the Impact of the Cathode Design on the Discharge Performance.
Journal of The Electrochemical Society 163, A2817–A2829. issn: 0013-4651
(2016).

81
94. McCloskey, B. D. Attainable Gravimetric and Volumetric Energy Density of
Li-S and Li Ion Battery Cells with Solid Separator-Protected Li Metal Anodes.
Journal of Physical Chemistry Letters 6, 4581–4588. issn: 19487185 (2015).

82
Appendix A

DETAILED CELL CONSTRUCTION

All alkali metal nitrate salts were purchased from Sigma-Aldrich, vacuum dried at
120 °C for 1 week and stored inside an argon-filled glovebox. Typically, a 12 mm
diameter glass microfiber separator (Whatman) was impregnated with 200–300 mg
of eutectic mixture and then vacuum-dried at 200 °C (above the eutectic melting
point) for 1 day using an oven (LC Technology Solutions) inside the glovebox. For
trace amounts of water in the electrolyte, Karl Fischer titration was performed using
dry N-methylacetamide solvent (NMA, 5 ppm of water, Sigma-Aldrich) before and
after the addition of the dried binary or ternary mixture of nitrate/nitrite salts. Results
show that water content of 10 ppm is typically observed with these electrolytes.
The oxygen positive electrodes consisted of Super P carbon black (Timcal) mixed
with PTFE binder (Sigma-Aldrich) in a 95:5 mass ratio (no catalyst). The amorphous
carbon was used as received with no pretreatment or surface activation. Super P
carbon and PTFE were mixed together in a water–isopropanol mixture and dried in
air at room temperature. The mixture was then dry pressed on a 10 mm diameter
stainless steel mesh. Typical carbon loading in these batteries was 3–5 mg/cm2 .
The nitrate reduction positive electrodes, except for the catalyst search, consisted of
nickel nanopowder (MTI Corporation, ~30 nm particle size, ~10 m2 /g BET Surface
Area). The nanopowder was used as received with no pretreatment or surface
activation, dry pressed on a 10 mm diameter stainless steel or aluminum mesh.
Typical nickel loading in these batteries was 10–20 mg/cm2 . To improve electrolyte
infusion into the porous positive electrodes, electrolyte separators containing the
molten salts and cathode were dried together at 180 °C for 1 day under a vacuum.
Typical battery cells consisted of a hermetically sealed stainless steel fixture (Swagelok)
of known volume, which is comprised of a pressure sensor (Omega) and a valve
(Valco Instruments) that can easily be connected to a mass spectrometer for qualitative and quantitative gas analysis. See Figure A.1 for a picture of the cell assembly.
Together with pressure monitoring, precise concentrations of gas in the head-space
can be determined: consumed and evolved oxygen gas during cycling in the case
of a lithium-oxygen battery, or checking for side reaction products in the case of
a nitrate reduction battery. Batteries are routinely leak tested with helium gas at

83

Figure A.1: Picture of Swagelok cell. From top to bottom: pressure sensor (with
sticker); valve allowing access to head-space gas; hollow tube holding gas headspace (also connected to for positive electrode contact); Swagelok assembly with
spring, current collector, pressed cathode, eutectic-infused separator, lithium anode;
negative current collector rod.

150 °C prior to use. The anode consists of an 8 mm diameter lithium metal disc
(250 µm thick) used as received (MTI Corporation). Once electrolyte and cathode
were added into the battery, a stainless steel spring was used to accommodate volume changes upon electrolyte melting and to maintain good electrical contact. All
battery construction was performed in an argon-filled glovebox with oxygen and
water levels maintained below 0.1 ppm. Batteries were purged with pure oxygen or
argon (Research 5.0 grade, Airgas), for lithium-oxygen or nitrate reduction batteries
respectively, at room temperature and maintained under positive pressure, typically
around 1.1e5 –1.5e5 Pa (1.1–1.5 bar). Batteries were then transferred to an oven kept
at either 120 or 150 °C depending on the molten nitrate/nitrite salts used. Opencircuit voltage, with periods of typically 6 or 12 h, was applied to ensure both the
battery voltage and the pressure reached equilibrium.
After electrochemical testing, cells were removed from the oven, allowing the molten
salt to freeze. After purging the head-space gas with argon, they were then moved
inside an argon-filled glove box for further analysis. Typically, positive electrodes

84
were extracted from batteries after reheating to melt the electrolyte, and then allowed
to cool again before being rinsed with ultra-dry NMA solvent (3x30 minutes) in order
to remove excess nitrate salts. If extracting the electrolyte for further analysis was
desired, the entire Swagelok fixture was disassembled, allowing access to the salt
without having to remelt it.

85
Appendix B

ION EXCHANGE CHROMATOGRAPHY CALIBRATION AND
CALCULATION
Calibration Procedure
Ion exchange chromatography works by measuring the time it takes for ions to pass
through a chromatography column. This time will depend on the size and charge
of these ions. After passing through this column, the conductivity of the solution is
measured, and this conductivity is proportional to the concentration of species in the
solution. As this is not a direct measure of concentration, a calibration procedure
must be used to determine the proportionality constant.

Figure B.1: Ion exchange chromatography nitrate calibration scans.
A Dionex ICS-2000 Ion Chromatography System was used for all experiments.
Calibration curves were generated by preparing a series of solutions of know concentration in the range of 5–500 µM, and Figure B.1 contains a series of calibration
runs from prepared solutions of LiNO3 –KNO3 . Similar calibrations were performed
for nitrite salts. All alkali metal nitrate and nitrite salts were purchased from SigmaAldrich, and mQ water (ultra-pure water from a Millipore filtration machine) was
used in preparation of all solutions. Calibration curves were generated before each
set of data was collected to account for any long term drift in the instrument, and

86
any error, coming from a combination of machine error and solution measurement
error, was generally on the order of a couple percent.
Calculation Details
The concentration of ionic species was calculated based on the calibration data,
and from this a ratio of nitrate to nitrite anions could be calculated. Then, using
the initial mass of electrolyte in the cell, the number of moles of each could be
calculated. There is good reason to not just take the concentration of ionic species
and work backwards through the dilution procedure to arrive at the total amount
of a component in the electrolyte. When disassembling the battery to harvest the
electrolyte (which is done when the electrolyte is a solid), it is easy to not recover
all of the electrolyte, as some will inevitably be stuck in the pores of the cathode or
on the lithium anode. This problem is easily avoided by only taking a ratio of ions
from the chromatography data and using this with the known amount of electrolyte
originally in the battery.
Due to impurities in the electrolyte salts (i.e. nitrite salt containing nitrate impurity),
correction factors calculated from ion exchange chromatography data were applied
to arrive at the final values. When calculating the number of moles of nitrate
and nitrite to attribute to electrochemical behavior, several factors needed to be
considered.
First, when using an electrolyte that contains significant nitrate, it is important to
note that the reaction described in Equation 3.1 also occurs on the lithium anode
as the SEI forms. This process can be seen in cells that sit at OCV and is difficult
to account for, so it is hard to make an accurate correction for it when calculating
moles of nitrate consumed in an electrochemical process. However, working with
a nitrite electrolyte avoids this issue, so studying a purely nitrite electrolyte with a
cathode previously discharged is most appropriate for accurate calculations. Many
other cells were put through this ion exchange chromatography analysis, such as all
the cells of Figure 3.13, but an accurate correction factor for this extra nitrite could
not be found. The amount of time a battery sat at OCV was not a stable enough
metric to be able to generate a calibration relating the time at OCV to the amount
of nitrate generated due to SEI growth. In addition, plating and discharge processes
seemed to evolve nitrite from the lithium anode at an even faster rate, making the
use of a nitrate electrolyte untenable.
Second, nitrite is not generally pure, but it is thermodynamically favorable for ~1%

87
of it to convert to nitrate in the presence of oxygen. However, throughout all of
the nitrite calibration series performed across all of the samples run, this impurity
proved to be a constant, and it is therefore easy to account for it in calculations.

88
Appendix C

NANOSCRIBE SPEED OPTIMIZATION

Before discussing how long it takes to write a structure using the Nanoscribe, it is
necessary to give more detail on the operation of the instrument. As mentioned in
the main text, Nanoscribe employs two-photon lithography, a lithography process
in which a photoresist is polymerized by a laser of photons with half the energy
needed for the polymerization reaction to occur (i.e. twice the wavelength of light
that would be used in a normal lithography process). This laser is focused to a
small volume, or voxel, such that the probability of two photons interacting with
the same molecule is high enough that the polymerization process takes place. This
voxel is then rastered in space, allowing the direct writing of polymer throughout
the photoresist.
It is this rastering process which dictates the length of time to fabricate a sample.
The instrument is assembled on an optical table with the laser stationary, and there
are three mechanisms by which rastering can happen. For large translations, the
stage which holds the sample substrate and photoresist moves with an electric motor.
This is too rough a control to make precise features, so it is only used when necessary
to travel a significant distance. The second method also involves stage motion, but
now by use of a piezoelectric controller (piezo mode). This gives precise control
within a couple hundred micron radius but is fairly slow. Until recently, these were
the only means of effectively rastering the laser, limiting the typical sample size to
about 100 µm. However, recently (about three years ago) a new rastering method
was introduced whereby a mirror galvanometer moves the laser focal point (galvo
mode), allowing the rastering to proceed without stage motion. This is a much faster
process, but only achievable in-plane (x,y) and similarly limited to a couple hundred
micron radius.
With a good understanding of the rastering methods, it is time to describe how
all three of them are used together to make a sample. A structure no larger than
~250 µm in any dimension can be made without needing to move the stage with the
motors, so any larger sample must consist of smaller units patched together. In the
case of the lattice array of Chapter Four, this tiled unit is a 120 µm cube. This model
is broken down into a series of raster lines as defined by a MATLAB script, where

89
due to the size of the structure each beam is rastered across several times. Starting
from the substrate, all the raster lines in a plane are written in galvo mode before
moving out of plane (up in z) using piezo mode. Then the next plane of raster lines
is written, and so on until the single unit is finished. Finally, the motor takes the
stage to a new spot to begin the next unit. Because the galvo mode is so much faster
than the piezo mode, the out of plane motion plays a large role in determining the
amount of time to make a sample.
Based on the detailed description of how to make a sample, there are a few parameters
we can tune to affect the speed of writing. One is the rate of motion of the rastering,
but that only moves within a fairly small range and thus won’t play too large of a
role (this was tested, and little impact on time was seen). Another is the spacing
between the raster lines. While a close spacing of raster lines within a beam leads
to a mechanically robust structure (since the writing voxels significantly overlap),
it also takes a longer time. So in order to see what spacing was allowable, several
parameter sweeps were performed. The spacing in x,y (hatch spacing) and in z (layer
spacing) were varied from 200 nm to 1.5 µm. Because the motion in z is controlled
in piezo mode, maximizing the allowable layer spacing was crucial to minimizing
the time to fabricate a sample. In addition, the laser power was varied because it
plays a role in the extent of polymerization, and thus the mechanical properties of
the structure.

Figure C.1: Nanoscribe parameter sweep. Each group of 12 is performed at a
different laser power. Within each group, a matrix of different layer and hatch
spacings is used (scale bar: 1 mm).

After performing several parameter sweeps similar to the one shown in Figure C.1,

90
a maximum allowable raster spacing that still yields a robust structure was found.
Such parameters will be different for each photoresist used, and IP-S photoresist was
chosen because it has the largest interaction volume, allowing for the largest raster
spacing. Implementing this set of parameters for the macroscopic lattice array of
Chapter Four results in a write time of 20 hours. Trying to create such a structure
before optimization (i.e. implementing typical writing parameters used by other
group members at the time) would have taken about 2 weeks, an untenable amount
of time for making a single sample.

91
Appendix D

WETTING OF ARCHITECTED ELECTRODES
Wetting for Nitrate Reduction
When first attempting to use a lattice array as an electrode for nitrate reduction as
in Chapter Four, the electrochemical data were noisy and the achievable currents
were lower than expected, being worse than the thin film electrode. Based on
discussions with collaborators who struggled with similar problems using a molten
salt electrolyte, it was suspected that the electrolyte was not wetting the lattice.
Because the LiNO3 –KNO3 electrolyte is a solid below 125 °C, directly working with
the electrode in the electrolyte was untenable. This required a different approach
for pre-wetting electrodes before cell assembly.
At first glance, we assumed that the wetting issue was due to the viscosity of the
molten salt. To resolve this, we used a 100 g L-1 LiNO3 –KNO3 aqueous solution,
hoping the lower viscosity of water would be sufficient. Figure D.1 shows the result
of pipetting 100 µL of this solution on both the lattice and film electrodes.

Figure D.1: Wettability of lattice electrode. Aqueous solution of 100 g L-1 LiNO3 –
KNO3 dropped onto electrodes. (A) Side view of droplet on lattice and thin film
electrodes. (B) Lattice electrode viewed through droplet acting as lens (scale: metal
disc is 1 cm in diameter).
Interestingly, it was not the viscosity but that the lattice is fairly hydrophobic, confirming the assumption that wetting was likely the issue with poor nitrate reduction
performance, and the water droplet behaved like a lens, allowing observation of the
lattice array by eye. In order to encourage the solution to wet the lattice, enough
salt solution was added so that the lattice and a bubble within it were completely

92
submerged. This was then briefly put into a vacuum oven. The moment vacuum
was pulled, a bubbles emerged from the lattice, not unlike the one in Figure D.2c.
Once all the bubbles had left (after only about 30 seconds), the filled lattice was
taken out of the vacuum oven and dried on a hot plate. Once the water left, salt was
impregnated in the lattice, and when put in contact with more molten salt, it readily
filled.
Wetting for Surface Area Measurements
Having determined the viability of a lattice electrode for studying nitrate reduction,
it was necessary to know its surface area. Knowing the difficulty of wetting a lattice
from the previous section, we decided to study the wetting process, and Figure D.2
shows its evolution using a Keyence VW–9000 high speed optical microscope.

Figure D.2: Wetting of a lattice electrode, by optical microscopy. Images of lattice
electrode (A) when dry, (B) immediately after immersion in water, (C) after 15
minutes of immersion, and (D) after CVs of Figure 4.9. (scale bar: (A-C) 500 µm
and (D) 250 µm).
The entire lattice and its stainless steel substrate were submerged in a shallow bath
of DI water. Immediately upon immersion, it appeared that no bubbles were going
to leave the electrode, but suddenly a large bubble developed and began to grow over
time. Interestingly, the size of this bubble served as a good measure of the extent

93
to which air had left the electrode. By simply measuring the size of the bubble and
comparing this with porosity values calculated from the SOLIDWROKS model,
we could know what fraction of the lattice had been wet. After sitting in the bath
overnight, the lattice became fully wet.
As mentioned in Chapter Four, there were some bubbles in the lattice after running
the CVs to measure the electrode surface area, as seen in Figure D.2d. It is hard
to measure the fraction of space filled with bubble as one can not tell how many
unit cells within a column are filled with air, but this observation seems to confirm
the slightly smaller measured surface area compared to what was expected from the
SOLIDWORKS model calculation.