PHD Thesis 20150520 Rev1 Til Orbit

Downloaded from orbit.dtu.
dk on: Nov 05, 2018
Fundamental mechanisms in Li-air battery electrochemistry
Højberg, Jonathan; Vegge, Tejs; Norby, Poul; Johansen, Keld
Publication date:
2015
Document Version
Publisher's PDF, also known as Version of record
Link back to DTU Orbit
Citation (APA):
Højberg, J., Vegge, T., Norby, P., & Johansen, K. (2015). Fundamental mechanisms in Li-air battery
electrochemistry. Department of Energy Conversion and Storage, Technical University of Denmark.
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
• Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
• You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
Jonathan Højberg
Fundamental mechanisms in Li-air

battery electrochemistry
Submitted in candidacy for the degree Doctor of Philosophy
Supervisors
Prof. Tejs Vegge, DTU Energy
Senior scientist Poul Norby, DTU Energy
Senior scientist Keld Johansen, Haldor Topsøe A/S
March 24th, 2015

Title: Fundamental mechanisms in Li-air battery electrochemistry
Author
Jonathan Højberg
E-mail: jhoj@topsoe.dk
Supervisors
Prof. Tejs Vegge
Section for Atomic Scale Modeling and Materials
Department of Energy Conversion and Storage
Technical University of Denmark
E-mail: teve@dtu.dk
Senior scientist Poul Norby

E-mail: pnor@dtu.dk
Senior scientist Keld Johansen

Haldor Topsøe A/S
E-mail: kej@topsoe.dk

Frederiksborgvej 399
DK-4000 Roskilde
Denmark
www.ecs.dtu.dk
Tel: +45 46 77 58 00
E-mail: info@ecs.dtu.dk
Haldor Topsøe A/S

Nymøllevej 55
DK-2800 Kgs. Lyngby
Denmark
www.topsoe.com
Tel: +45 45 27 20 00
E-mail: topsoe@topsoe.dk
Release date: March 24th, 2015

1st revision: May 20th, 2015
Preface
This thesis is submitted in candidacy for the Ph.D. degree from the Technical University
of Denmark (DTU). The work was carried out between January 2012 and March 2015
at the section for Atomic-scale Modeling and Materials, DTU Energy and at Haldor
Topsøe A/S. A part of the work was also carried out in an external research stay at IBM
Almaden Research Center, California, USA.
The project was supervised by Prof. Tejs Vegge, Head of Section at DTU Energy, Senior
scientist Poul Norby at DTU Energy and Senior scientist Keld Johansen at Haldor
Topsøe A/S. The project was financially supported by the ReLiable project (project nr.
11-116792) funded by the Danish Council for Strategic Research Program Commission
on Sustainable Energy and Environment.
v
Abstract
The lithium-air (or Li-O2 ) batteries have received wide attention as an enabling technol-
ogy for a mass market entry of electric vehicles due to a potential capacity much higher
than current Li-ion technology. The technology is a relatively new battery concept pro-
posed in 1996, and the current research still focuses on developing an understanding of
the reactions inside the battery. This thesis is dedicated to increase this understanding
and use the knowledge to improve the performance of the battery, and the work span
from detailed investigation of the atom positions to the proposal of a system used to
manage a full size electric vehicle battery.
An automated differential electrochemical mass spectrometer (DEMS) was built to in-
vestigate the relationship between current and the consumption and release of gases,
which is important to identify and quantify degradation reactions. The setup was used
to characterize our carbon-based reference system as well as new ionic liquid-based elec-
trolytes.
Electrochemical impedance spectroscopy (EIS) has been used extensively to describe
reaction mechanisms inside the battery; the origin of the measured overpotentials; and
the onset potential for electrochemical degradation. It was confirmed that the rapid
potential loss near the end of discharge could be explained by an increase in the charge
transport resistance; that the initial Li2 O2 oxidation at 3.05 V was blocked by the for-
mation of an SEI layer; and that the voltage increase during charge was primarily due
to the formation of a mixed potential between competing oxidation reactions needed to
maintain a constant current.
The knowledge about impedance spectroscopy was used to propose and investigate a
novel battery management tool to estimate the state of charge and the state of health
of a Li-O2 battery system better than any other method available.
Finally, calculations were made to support that an open system configuration is a real-
istic option in terms of air purification, if H2 O and CO2 levels at 1 ppm are allowed.
vii
Resumé
Lithium-luft (eller Li-O2 ) batterier har fået stor opmærksomhed, fordi de med en po-
tentiel kapacitet mange gange større end nuværende Li-ion batterier, kan være med
til at sikre et kommercielt gennembrud for elbiler. Li-O2 batteriet er et relativt nyt
koncept, der første gang blev demonstreret i 1996. Denne afhandling fokuserer på at
udbygge forståelsen af hvad der sker inde i batteriet og anvende denne forståelse til at
forbedre batteriet. Arbejdet spænder fra et detaljeret studie af de atomare positioner
til udviklingen af et system, der kan bruges til at styre elbil-batterier i fuld skala.
I projektet blev der bygget et automatiseret differentielt elektrokemisk massespektrom-
eter (DEMS) for at undersøge sammenhængen mellem den strøm, der løber fra batteriet
og den mængde gas, der optages eller afgives. Dette er vigtigt for at kunne identificere
og kvantificere eventuelle degraderingsreaktionerne i batteriet. Opstillingen blev brugt
til at karakterisere vores kulstof-baserede reference system og nye elektrolytter baseret
på ioniske væsker.
Elektrokemisk impedans spektroskopi (EIS) er blevet anvendt i vid udstrækning til
at beskrive reaktionerne inde i batteriet; årsagen til det målte overpotentiale; og ved
hvilket potentiale den elektriske degradering begynder. Det blev bekræftet at det hur-
tigt faldende potentiale i slutningen af afladningen kan forklares med en stigning af
ladningsoverførselsmodstanden, at Li2 O2 oxideringen starter ved 3.05 V, men blokkeres
med det samme af et lag af kemiske degraderingsprodukter fra reaktioner mellem Li2 O2
og elektrolytten, og at stigningen i spænding gennem opladningen primært skyldes et
blandet potentiale mellem forskellige oxidative reaktioner, der aktiveres for at opretholde
en konstant strøm i takt med at overfladen dækkes af degraderingsprodukter.
Den opbyggede viden om impedansspektroskopi blev brugt til at udvikle og teste en
ny metode til at bestemme batteriets ladningstilstand og helbred. Metoden viste sig at
være markant bedre end de eksisterende løsninger.
Endelig blev der foretaget en række beregninger, der viste, at et åbent Li-O2 system
til rensning af H2 O and CO2 til et niveau under 1 ppm er et realistisk alternativ til at
benytte en oxygen tank.
ix
Acknowledgements
This thesis marks the end of three years of research. It has been a wonderful time with
many memorial experiences and a lot of lovely people that I am grateful to have met. I
would like to use this opportunity to thank all of you. My thanks go to:
My supervisor Prof. Tejs Vegge for invaluable academic discussions. For opening doors
when needed, also some of the big ones with locks and chains, and otherwise letting me
find my way.
Johan Hjelm and my co-supervisors Senior scientist Poul Norby and Senior scientist
Keld Johansen for valuable input to my scientific work throughout the project.
A special thank goes to Assistant Prof. Bryan D. McCloskey and Prof. Alan Luntz for
enabling the fantastic journey to California and the legendary IBM Almaden Reserach
Center. I am grateful for the opportunity I had to have daily discussions in the lab
and at the 4pm coffee break about Li-air battery chemistry, remodeling and the new
software update in Winfrieds Tesla.
Prof. Søren Dahl for your support when needed and for putting me in contact with
Senior scientist Poul Erik Højlund back in 2011, which led to this PhD project.
My fantastic colleagues at both Haldor Topsøe and DTU Energy. Invaluable help from
LPHA, RTIR, LAFL and the rest of the characterization crew at Haldor Topsøe. An-
dreas and Mathias for a major help with the DEMS setup. Mike for always having time,
always having a solution and never complaining when I have messed up you toolbox or
tightened the Swagelok unions too much in a 7pm quick fix.
Ane for taking so good care of me when I started working with batteries and later,
when I needed someone to share my PhD frustrations with. Kristian and Andreas
for unforgettable moments in Denmark and in Crete, for Champagne bets, and an
astonishing sprint in the end to ensure that our papers got submitted on time before my
thesis deadline. Dr. Supti Das for a great collaboration on the DEMS measurements
and a hectic effort to finish our paper. Jon for your continuous brainstorming. It has
been a pleasure and an inspiration. And great fun to start up DBS, PCLF, COMX and
all the other ones that didn’t even get a name. Nice!
I would like to thank Merete for an overwhelming support in these final months even
though you have been busy too.
And finally, I would like to express my gratitude of my beloved daughters Hedvig and
Ingeborg. Now, I can finally say to you that I have finished my book.
xi
Contents
Preface v
Abstract vii
Resumé ix
Acknowledgements xi
Contents xii
List of Figures xvii
Abbreviations xix
1 Introduction 1
1.1 Abandoning fossil fuels . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1.1 Non-fossil transportation . . . . . . . . . . . . . . . . . . . . . . 2
1.2 Batteries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.1 Gold rush of the giant leap . . . . . . . . . . . . . . . . . . . . . 3
1.3 About this work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.1 The aims of the project . . . . . . . . . . . . . . . . . . . . . . . 5
1.3.2 Outline of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . 6
2 Theory 9
2.1 The lithium-oxygen battery . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.1.1 New definitions for lithium-oxygen batteries . . . . . . . . . . . . 10
2.2 Electrochemical impedance spectroscopy . . . . . . . . . . . . . . . . . . 12
2.2.1 Correlation between overpotential and impedance . . . . . . . . . 12
2.2.2 Modeling Li-oxygen battery impedance . . . . . . . . . . . . . . 14
3 Experimental methods 17
3.1 Cell configuration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3.1.1 The Swagelok cell . . . . . . . . . . . . . . . . . . . . . . . . . . 18
3.1.2 The XC72 cathode . . . . . . . . . . . . . . . . . . . . . . . . . . 19
xiii
Contents xiv
3.1.3 Electrolyte . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.2 Differential electrochemical mass spectrometry . . . . . . . . . . . . . . 20
3.2.1 Typical operation . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
3.3 Physical characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.3.1 Quantitative optical absorption spectroscopy . . . . . . . . . . . 23
3.3.2 TEM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4 Overpotentials and degradation 27

4.1 Physical characterization . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.1.1 X-ray diffraction . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
4.1.2 Differential electrochemical mass spectrometry . . . . . . . . . . 28
4.1.3 Absorption measurements . . . . . . . . . . . . . . . . . . . . . . 29
4.1.4 Scanning electron microscopy . . . . . . . . . . . . . . . . . . . . 30
4.1.5 Transmission electron microscopy . . . . . . . . . . . . . . . . . . 31
4.2 Electrochemical characterization of the system . . . . . . . . . . . . . . 32
4.3.1 Discharge to sudden death at 0.25 mA . . . . . . . . . . . . . . . 34
4.3.2 Discharge to sudden death at 0.02 mA . . . . . . . . . . . . . . . 35
4.3.3 Supporting EIS measurements . . . . . . . . . . . . . . . . . . . 36
4.3.4 Charge at 0.25 mA . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.4 Identification of processes during discharge . . . . . . . . . . . . . . . . 39
4.5 Analysis of the overpotential during discharge . . . . . . . . . . . . . . . 41
4.6 Reaction mechanisms and SEI layer formation during charge . . . . . . 42
4.6.1 Decreasing electrical resistivity of Li2O2 in charge mode . . . . . 42
4.6.2 Identification of Li2O2 oxidation at 3.05 V . . . . . . . . . . . . 43
4.6.3 SEI layer formation . . . . . . . . . . . . . . . . . . . . . . . . . 43
4.6.4 Electrochemical degradation . . . . . . . . . . . . . . . . . . . . . 44
4.6.5 Mixed potential . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
4.7 Summary of the fundamental characterization of overpotentials and degra-
dation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5 Screening for new electrolytes 47

5.1 Investigation of ionic liquids . . . . . . . . . . . . . . . . . . . . . . . . . 48
5.1.1 Comparison between ionic liquid and organic solvent . . . . . . . 52
5.1.2 Choice of anion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.1.3 Choice of cation . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.1.4 Impurities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
5.1.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
6 Commercial applications of lithium-air batteries 55

6.1 Impedance-based battery management for Li-oxygen systems . . . . . . 55
6.1.1 The working principle of the method . . . . . . . . . . . . . . . . 57
6.1.2 Testing the method . . . . . . . . . . . . . . . . . . . . . . . . . 58
6.1.3 State of charge . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
6.1.4 Power capability . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
6.1.5 State of health . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Contents xv
6.1.6 Further development of the model . . . . . . . . . . . . . . . . . 64

6.1.7 Summary of impedance based management of Li-O2 batteries . . 64
6.2 Separation of O2 from air . . . . . . . . . . . . . . . . . . . . . . . . . . 65
6.2.1 Previous Li-O2 and Li-air strategies . . . . . . . . . . . . . . . . 66
6.2.2 A solution for an open Li-air system . . . . . . . . . . . . . . . . 66
6.2.3 Summary of separation of O2 from air . . . . . . . . . . . . . . . 67
6.3 Summary of commercial applications of lithium-air batteries . . . . . . . 68
7 Summary and outlook 69

7.1 Summary of main results . . . . . . . . . . . . . . . . . . . . . . . . . . 69
7.2 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 70
A Development of the testing equipment 73
B Differential electrochemical mass spectrometry - In situ gas analysis 77

B.1 Initial measurements and design . . . . . . . . . . . . . . . . . . . . . . 78
B.2 Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
B.3 Comparing oxygen consumption with electrochemistry . . . . . . . . . . 82
C Supporting TEM measurements 85

C.1 Beam damage and observations of lattice fringes in lithium peroxide . . 86
C.2 Selected area electron diffraction . . . . . . . . . . . . . . . . . . . . . . 86
C.3 Electron energy loss spectroscopy . . . . . . . . . . . . . . . . . . . . . . 88
C.3.1 EELS reference spectra . . . . . . . . . . . . . . . . . . . . . . . 89
D Calculations used to assess the air purification system of an open Li-O2

system 91
D.1 Design frame . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92
D.2 Pre-drying using a membrane system . . . . . . . . . . . . . . . . . . . . 93
D.3 Designing the adsorption system . . . . . . . . . . . . . . . . . . . . . . 93
D.4 Energy consumption and weight . . . . . . . . . . . . . . . . . . . . . . 95
List of publications 97
Paper I - An electrochemical impedance spectroscopy investigation of the over-
potentials in Li-O2 batteries . . . . . . . . . . . . . . . . . . . . . . . . . 98
Paper II - Reactions and SEI formation during charging of Li-O2 cells . . . . 113
Paper III - Impedance-based battery management for metal-O2 systems . . . 123
Paper IV - Rechargeability of ionic liquids in Li-O2 batteries . . . . . . . . . 135
Paper V - The influence of CO2 poisoning on overvoltages and discharge ca-
pacity in non-aqueous Li-air batteries . . . . . . . . . . . . . . . . . . . 160
Bibliography 165
List of Figures
1.1 Ragone plot of different battery technologies . . . . . . . . . . . . . . . . 3

1.2 Comparison between the specific energy density of different battery tech-
nologies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.3 Number of published papers on Li-air batteries . . . . . . . . . . . . . . 4
2.1 Schematic of a Li-battery working principle . . . . . . . . . . . . . . . . 9

2.2 Li-air battery discharge electrochemistry . . . . . . . . . . . . . . . . . . 10
2.3 Normalization of discharge capacity . . . . . . . . . . . . . . . . . . . . 12
2.4 iv -curve of Li-O2 batteries in discharge mode . . . . . . . . . . . . . . . 13
2.5 Equivalent circuit diagram of the Li-O2 battery during discharge . . . . 14
2.6 Capacitance change of Li-O2 positive electrode . . . . . . . . . . . . . . 15
3.1 Illustration of the carbon-based reference system . . . . . . . . . . . . . 17

3.2 Sketch and picture of the 10 mL Swagelok cell . . . . . . . . . . . . . . . 18
3.3 The XC72 cathode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19
3.4 Air-spraying and punched XC72 cathodes . . . . . . . . . . . . . . . . . 19
3.5 Schematic illustration of the DEMS setup . . . . . . . . . . . . . . . . . 21
3.6 Operation modes of the DEMS setup . . . . . . . . . . . . . . . . . . . . 22
3.7 SEM image of XC72 coated TEM grid . . . . . . . . . . . . . . . . . . . 24
4.1 Ex-situ and in-situ XRD measurements of Li2 O2 in the cathode . . . . . 28

4.2 DEMS measurements of the carbon XC72 based reference system . . . . 28
4.3 Comparison of O2 evolution and Li2 O2 oxidation during charge . . . . . 30
4.4 SEM images and EDS spectra of pristine and discharged cathode . . . . 30
4.5 TEM measurements performed on discharged Li-O2 cathode . . . . . . . 31
4.6 OCV map and cyclic voltammogram of a full discharge-charge cycle . . 32
4.7 Differential capacity plots of a Li-O2 charge . . . . . . . . . . . . . . . . 33
4.8 EIS measurements during discharge of a Li-O2 battery at 250 µA . . . . 34
4.9 Resistance and capacitance during discharge of a Li-O2 battery at 20 µA 35
4.10 Supporting EIS measurements . . . . . . . . . . . . . . . . . . . . . . . . 36
4.11 EIS measurements during charge of a Li-O2 battery at 250 µA . . . . . 38
4.12 Change of resistance and capacitance during charge of a Li-O2 battery . 39
4.13 Sketch of the reactions observed during charge . . . . . . . . . . . . . . 42
4.14 Overview of the primary contribution to the voltage changes during dis-
charge and charge . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
5.1 Investigated ionic liquids . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

5.2 DEMS comparison between DME and P14TFSI . . . . . . . . . . . . . . 49
5.3 DEMS analysis of IL based electrolytes with P14+ and PP13+ cations . 50
xvii
List of Figures xviii
5.4 DEMS analysis of IL based electrolytes with TFSI− and FSI− anions . . 51
5.5 XRD patterns of discharged cathodes using different electrolytes . . . . 52
5.6 Cycling with P14TFSI and P13TFSI based electrolytes. . . . . . . . . . 52
6.1 Overview of impedance-based management of Li-O2 batteries . . . . . . 56

6.2 Current loads used to test capacitance based SoC determination . . . . 58
6.3 Capacitance decrease during discharge of a Li-O2 battery . . . . . . . . 59
6.4 Correspondence between capacitance and SoC . . . . . . . . . . . . . . . 59
6.5 Capacitance changes during discharge at different current loads . . . . . 60
6.6 The effect of capacitance-based calibration of SoC . . . . . . . . . . . . 61
6.7 Effect of CO2 impurities in the Li-O2 battery . . . . . . . . . . . . . . . 65
6.8 Schematic illustration of an automotive air cleaning system . . . . . . . 67
A.1 Selected designs of the Swagelok cell . . . . . . . . . . . . . . . . . . . . 73

A.2 Sketch of the Swagelok cell . . . . . . . . . . . . . . . . . . . . . . . . . 74
A.3 Picture of the DEMS system . . . . . . . . . . . . . . . . . . . . . . . . 74
B.1 Intensity measurements with mass spectrometer . . . . . . . . . . . . . . 78

B.2 Temperature variations in DEMS setup . . . . . . . . . . . . . . . . . . 79
B.3 Mass spectrum fingerprint of DME . . . . . . . . . . . . . . . . . . . . . 80
B.4 Oxygen background in DEMS measurements . . . . . . . . . . . . . . . 80
B.5 Intensity measurements with mass spectrometer . . . . . . . . . . . . . . 82
B.6 Average oxygen per electron ratio . . . . . . . . . . . . . . . . . . . . . . 82
C.1 Test of Li2 O2 beam sensitivity . . . . . . . . . . . . . . . . . . . . . . . 86

C.2 TEM and SAED images of reference samples . . . . . . . . . . . . . . . 87
C.3 TEM-EELS reference measurements of Li2 O2 . . . . . . . . . . . . . . . 88
C.4 EELS reference spectra . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
D.1 Calculated Langmuir isotherm for H2 O and CO2 adsorption . . . . . . . 93

D.2 Schematic illustration of a plug flow reactor and concentration changes
in time and space . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94
Abbreviations
SoC State of charge

SoD State of discharge
SoH State of health
BMS Battery management system
OCV Open circuit voltage
EIS Electrochemical impedance spectroscopy
PEIS Potentiostatic electrochemical impedance spectroscopy
GEIS Galvanostatic electrochemical impedance spectroscopy
CPE Constant phase element
AC Alternating current
DC Direct current
SEI Solid-electrolyte interface
OER Oxygen evolution reaction
ORR Oxygen reduction reaction
DEMS Differential electrochemical mass spectrometry
SAED Selected area electron diffraction
EELS Electron energy loss spectroscopy
TEM Transmission electron microscopy
SEM Scanning electron microscopy
EDS Energy-dispersive X-ray spectroscopy
XRD X-ray diffraction
xix
To Hedvig and Ingeborg
xxi
Chapter 1
Introduction
1.1 Abandoning fossil fuels
Fossil fuels are still, by far, the most common source of energy in the world, constituting
79 % of the total energy production [1], but the renewable share of the energy production
is increasing. The transition is motivated by two main factors: energy security and
climate change mitigation. The dependence on oil and gas import has led to Chinese
involvement in Sudan, American involvement in the Middle East and an escalation
of the international crisis between Russia and the European countries. As the global
energy demand is expected to increase by 1.1 % each year until 2040, there is a political
incentive to decrease the amount of imported energy [2]. The latest IPCC report states
that it is extremely likely that the increase in global temperature and sea-level is caused
by the increase in CO2 and other greenhouse gasses like CH4 and N2 O, and that the
costs of a business-as-usual approach will be severe - both in terms of finance and human
well-being [3]. These combined factors result in large investments in renewable energy
sources. 144 countries have defined policy targets to increase the share of renewable
energy, and especially China has invested massively in renewable energy sources since
2011. The global amount of renewable energy sources amounted to 19 % of the total
energy production in 2012, increasing by approximately 1 % each year [1].
Most renewable energy sources produce electricity from sun or wind energy. This means
that energy is produced when the sun is shining and when the wind is blowing, and, as
that does not necessarily match the demand of energy, this creates a massive storage
need. A number of different storage methods exist with different advantages and disad-
vantages in terms of capacity, power, mobility and price. As an example, pumped hydro
is the most cost-effective way of storing large amounts of electrical energy, but it requires
1
Chapter 1. Introduction 2
huge capital costs and the presence of appropriate geography. On the other hand, bat-
teries are relatively expensive per stored kWh, but very useful for mobile applications
as the energy density is high and the unit size is very flexible.
1.1.1 Non-fossil transportation
Today, much attention is being devoted to developing green transportation, as this

accounts for 63.7 % of the global oil consumption (2012) [4]. Gasoline has a very
high energy density of approximately 12.3 kWh/kg, and the challenge is to develop
competitive energy storage methods with sufficiently high energy density. Different
technologies like batteries, fuel cells and bio-diesel are currently being investigated. All
technologies have benefits and drawbacks, and it is unlikely that any of the technologies
will be able to solve the future challenge of non-fossil transportation alone. Looking at
the use of batteries in transportation, Li-ion batteries have been implemented in different
electric cars during the last 10 years. The driving ranges of economy cars lie between 80
km and 160 km, and it is forecasted that this driving range cannot improve by more than
a factor of 2-3 using conventional battery technologies. This means that conventional
battery technologies will not be able to match the driving range of gasoline cars, which is
currently considered critical to enable a full transition to non-fossil transportation. The
need for better mobile energy storage has strongly motivated research in other battery
systems such as lithium-sulfur and lithium-oxygen batteries.
1.2 Batteries
A battery is an electrochemical cell that converts stored chemical energy into electrical
energy. It was first invented by Alessandro Volta in 1800 and has since then become a
common power source for many household and industrial applications. There are two
types of batteries: primary batteries (disposable batteries), which are designed to be
used once and discarded, and secondary batteries (rechargeable batteries), which are
designed to be recharged and used multiple times. Batteries come in many sizes, from
miniature cells used to power hearing aids and wristwatches to battery banks the size
of buildings that provide backup power for telephone exchanges and data centers or act
as frequency stabilization units in the power grid.
The wireless revolution that has dominated the western world in the late 00’s has been
powered by batteries, and especially the rechargeable lithium ion batteries. Lithium
batteries were initially proposed by M. S. Whittingham [5] in the 1970s and important
contributions have been made by John B. Goodenough among others [6]. The Li-ion
battery was commercialized by Sony in 1991 and many improvements have been made
since then. The current state Li-ion battery has the largest capacity of all commercialized
secondary batteries, and they are found everywhere. Globally, the annual production
of battery cells is astonishing 20 billion cells. Tesla alone is expected to use 0.3 billion
cells in 2015.
Specific power [W/kg at cell level]
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9
100,000 8 7 6 5 4 3 2 1
Super capacitors
18 Lead acid spirally wound
10,000 Li-ion high power
17
USCAR minimum
16 1,000 EV long term goal
15
100
14
13 Li-ion high energy
10 Ni-Cd Na/NiC12
12
Lead acid
11 1
0 50 100 150 200 250
10 Specific energy [Wh/kg at cell level]
9
Figure 1.1: Ragone plot comparing the specific energy and specific power of different
8 battery systems [7] and [8].
7
Many types 6of commercial secondary batteries exist. Figure 1.1 shows a Ragone plot of
5
the most common chemistries like Lead-acid (Pb-H), Nickel Metalhydrate (NiMH) and
4 based chemistries. Comparing with the 12.3 kWh/kg energy density of
different Li-ion
3 clear that significant improvements are needed. The energy density can
gasoline, it is
be improved2 in two ways: by increasing energy (potential) or by increasing capacity
1
(decrease weight or volume per charge). The first strategy is exemplified by
Template used lithium
to crop directly in LaTex with trim-comman
nickel manganeseoxide (LNMO), where the potential is increased from 3.6 V to 4.7 V by
substituting 25 % of the manganese with nickel, and the second strategy is exemplified
by lithium nickel manganese cobaltoxide NMC, where more lithium can be extracted per
transition metal. The Li-ion batteries are, however, limited by a heavy framework, and
disruptive technologies are needed to obtain decisive improvements in energy density.
Recently, the Li-air battery has been investigated heavily as such a disruptive technology.
1.2.1 Gold rush of the giant leap
The non-aqueous lithium-air (or Li-O2 ) battery is a relatively new battery concept
proposed by Abraham et al. in 1996 [9]. Compared with a lithium-ion battery, the
lithium-air battery is an open system that allows oxygen from the atmosphere (or from
an oxygen storage tank) to solute in the electrolyte and reduce at the positive cathode
to form lithium peroxide (Li2 O2 ) during discharge. This makes it possible to avoid
heavy transition metals in the cathode and use lighter elements like carbon instead,
which reduces the total battery weight significantly. Theoretically, and disregarding
support materials and electrolyte, the energy density of the Li-O2 battery is 3.8-11.7
kWh/kg depending on whether the oxygen is included in the calculation or not [10].
This is comparable to gasoline and, if realizable, exceeding the improvements seen in
the previous 150 years since Gaston Planté discovered the Lead-Acid battery in 1859.
Figure 1.2: Comparison between the specific energy density of different battery tech-
nologies. It is seen that the theoretical energy density of the Li-air battery is signifi-
cantly larger than all other rechargeable battery systems. The practical energy density
is however somewhat smaller [11].
Figure 1.2 shows a comparison of the Li-O2 chemistry with other battery technologies
currently investigated. The Li-air battery received only little attention until Peter Bruce
made experiments with his group demonstrating rechargeability [12]. This demonstra-
tion opened up the technology and initiated a competitive race in the scientific commu-
nity to publish results as fast as possible. IBM started up the Battery 500 project to
develop an EV Li-air battery with a driving range of 500 miles, and research groups all
over the world started up Li-O2 battery activities. Figure 1.3 shows an overview of the
number of published scientific papers within Li-air batteries.
24 23 on
Publications 22Li-air
21each20year1934318 17 16 15 14 13 12 11 10
297
18 248
157
17 99
33
1612
15
2008 2009 2010 2011 2012 2013 2014
Figure 1.3: Number of publications on lithium-air batteries. The numbers are based
14
on a Google Scholar search using allintitle: (lithium-air OR Li-air OR lithium-O2 OR
Li-O2).
13
12
11
10
In early experiments using carbonate solvents in the electrolytes, it was observed that the
cell voltage rose almost immediately to above 4 V during charge. This was incorrectly
interpreted as a high Li-O2 kinetic overpotential for charging, and has led to a strong
focus on electrocatalysis in the literature. However, as will be pointed out in Chapter 4,
this rapid rise was instead related to electrolyte decomposition that formed carbonate
and carboxylate products rather than Li2 O2 during discharge. Once formed, they do
not oxidize electrochemically until 4 V.
In general, many publications within the field of Li-O2 batteries are misinterpreted at
best, because of an insufficient characterization of the involved mechanisms. Some of
the most prominent results are the application of carbonate electrolytes, catalysts in the
positive electrode and capacity limited cycling, and today many of these measurements
and publications have been proven wrong.
A detailed review of the current state of research is found in the recent book The Lithium
Air Battery: Fundamentals by Imanishi et al. [13] and a condensed version is presented
in the excellent review by Luntz et al. [14] from 2014.
1.3 About this work
1.3.1 The aims of the project
Early in this project, it was clear that a careful study of the electrochemistry and funda-
mental reaction mechanisms was needed to ensure that we could distinguish the wanted
electrochemistry from the unwanted. This changed my work from developing new cath-
ode materials to developing a stable test system and reliable methods to characterize the
fundamental electrochemical mechanisms inside the Li-O2 battery. The focus has thus
been to understand the formation and removal of Li2 O2 , and the undesired degradation
reactions.
Today, no one knows if Li-O2 batteries will ever be commercially viable because of
the significant degradation, but as an employee at Haldor Topsøe A/S, it was also
important to assess the viability in a system perspective, which has been done under the
assumption that the challenge of degradation is solved or at least significantly reduced.
The assessment focuses on air purification in an open Li-air battery system and battery
management using an impedance based State-of-Charge (SoC) determination.
1.3.2 Outline of the thesis
The contents of this thesis are divided into seven chapters. The research results of the
project are reported in Chapters 4, 5 and 6, and the work presented in those chapters
has been carried out by me unless otherwise stated. This work resulted in two published
peer-reviewed publications, two submitted publications and one publication draft. The
five manuscripts are attached as appendices to the thesis. In addition to this, four
appendices have been attached, including work that has been of significant importance
to the project, without being a subject for publication. The thesis has been written as
an independent work and the papers have been attached to support the discussions and
conclusions presented.
2 Theory
This chapter focuses on the working principle of the Li-O2 battery and a brief introduc-
tion to the theoretical foundation of electrochemical impedance spectroscopy (EIS).
3 Experimental
This chapter focuses on the manufacturing of cathodes, the basic working principles of
the differential electrochemical mass spectrometry (DEMS) and electrochemical impedance
spectroscopy.
4 Overpotentials and degradation

This chapter presents the work on explaining the fundamental reaction mechanisms in
the Li-O2 battery. Methods like EIS, DEMS, TEM and absorption measurements give
important contributions to this work. This chapter is closely related to Paper I and
Paper II.
5 Screening for new electrolytes

This chapter presents work on finding a new electrolyte by testing promising ionic liquids
(IL) with DEMS. This chapter is closely related to Paper IV.
6 Commercializing Li-air batteries

This chapter presents a novel method to increase the accuracy of the state of charge
determination in a Li-O2 battery and work on air-purification of the inlet air in an open
Li-air battery. This chapter is closely related to Paper III.
7 Summary and outlook

The four main results of the work are summed up and six suggestions to future work
are presented.
List of included papers
Paper I
An electrochemical impedance spectroscopy investigation of the overpoten-
tials in Li-O2 batteries
Jonathan Højberg, Bryan D. McCloskey, Johan Hjelm, Tejs Vegge, Keld Johansen, Poul
Norby and Alan C. Luntz
ACS Appl. Mater. Interfaces, 7, 4039–4047 (2015)
Paper II
Reactions and SEI formation during charging of Li-O2 cells
Jonathan Højberg, Kristian B. Knudsen, Johan Hjelm and Tejs Vegge
ECS Electrochem. Lett., 4, A63–A66 (2015)
Paper III
Impedance-based battery management for metal-O2 systems
Andreas E. Christensen, Jonathan Højberg, Poul Norby and Tejs Vegge
J. Pow. Sources. Submitted (2015)
Paper IV
Rechargeability of ionic liquids in Li-O2 batteries
Supti Das, Jonathan Højberg, Kristian B. Knudsen, Poul Norby and Tejs Vegge
To be submitted
Paper V
The influence of CO2 poisoning on overvoltages and discharge capacity in
non-aqueous Li-air batteries
Yedilfana S. Mekonnen, Kristian B. Knudsen, Jon S. G. Mýrdal, Reza Younesi, Jonathan
Højberg, Johan Hjelm, Poul Norby and Tejs Vegge
J. Chem. Phys., 140 (2014)
Chapter 2
Theory
2.1 The lithium-oxygen battery
-
24 e23 22 21 20 19 18- 17 16 15 14 13 12 11 10 9 8 7 6 5
e
18
17
Negative Electrolyte Positive
16electrode electrode
15 Discharge
Li+
14 Charge
Li+
13
12
Figure 2.1: 11
Schematic of the overall Li-battery working principle.
10
Figure 2.1 shows a highly9simplified sketch of how lithium batteries work by moving Li+
ions through the electrolyte
8 from the negative electrode to the positive electrode during
discharge and vice versa 7during charge. The two electrode materials absorb, store and
release Li+ ions at two different
6 chemical potentials, and the difference between these
chemical potentials gives 5the open circuit cell voltage (OCV)
4 −

−
∆G
3 OCV = − (2.1)
zF
2
−

−
where ∆G is the change1in Gibbs free energy of the reaction, z is the charge number (1used to crop directly in LaTex
Template
in Li-batteries), and F is the Faraday constant. The battery is discharged by connecting

the two materials electrically through an external circuit. This allows electrons to move
from the negative electrode to the positive electrode with an energy corresponding to
the voltage difference multiplied by the elementary charge. The energy of the electrons
can be used in the external circuit to power a laptop or another electrical device. The
9
Chapter 2. Theory 10
battery is charged by applying a potential between the two electrodes greater than the
equilibrium potential. This forces electrons to move from the cathode to the anode,
which reverses the entire reaction. In Li-ion batteries the two electrode materials are
crystaline intercalation materials, where lithium is stored inside the crystal. In Li-O2
batteries, the electrodes are different, as both electrodes grow (shrink) during reduction
(oxidation).
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
e-
18
17
Li+ e-
16
15 Li2O2
O2 O2
14 O2
13
e-
12
11 Li+
Li
10 Li2O2 Carbon
9
8 Figure 2.2: Li-air battery discharge electrochemistry.
7
The actual mechanism of the Li-O2 battery depends on the choice of electrolyte. This
6
can be aqueous,
5 aprotic (water-free) or solid. The aprotic system has received most
attention 4in literature due to the problems with ohmic losses, rechargeability and low
energy density
3 observed in the other systems. For this reason, only the aprotic system is
considered2 in this work. Figure 2.2 shows the ideal electrochemistry of the aprotic Li-air
1 battery is open to allow oxygen inside the battery.
battery. The TemplateThe
used to oxygen
crop directly indiffuses
LaTex with trim-command
to the positive electrode, where it is reduced to form Li2 O2 . During charge, the Li2 O2
is oxidized and oxygen is released. The Li+ ions move between the two electrodes to
balance the charge and, on the negative electrode, lithium is soluted/plated to keep a
constant number of Li+ ions in the electrolyte.
2.1.1 New definitions for lithium-oxygen batteries
Rechargeability. The most important characteristic of the Li-O2 battery at the cur-
rent state is rechargeability, and because the system is fundamentally different from
the Li-ion battery, the general definitions do not apply. In Li-ion batteries, the two
electrodes are balanced such that any degradation involving either lithium or one of
the electrode materials would directly affect the capacity of the battery. All current
state Li-O2 batteries use a massive excess of both lithium and electrolyte, which means
that electrochemical degradation can occur without being detected in the electrochem-
ical measurements. This has led to the four definitions of rechargeability adapted from
Luntz et al.[14]:
1. the yield of Li2 O2 relative to that anticipated from the current and ideal cathode
reaction 2(Li+ +e− ) + O2 → Li2 O2 during discharge is YLi2 O2 = 1.00, i.e., no other
products are formed during discharge either on the cathode or in the electrolyte,
e.g., no LiOH, Li2 CO3 , LiF, carboxylates, etc
2. during discharge, the electrochemical current consumes only O2 , (e− /O2 )dis =
2.00, and during charge, all electrochemical current evolves O2 , (e− /O2 )cha = 2.00
3. no parasitic gas evolution (H2 , CO2 , etc.) occurs during the discharge-charge cycle
4. all O2 consumed during discharge (ORR) is released during charge (OER) so that
OER/ORR = 1.00
If all requirements are met, then the Li-O2 battery is perfectly rechargeable. In addition,
a long calendar life is necessary, and this requires that all thermal parasitic chemical
reactions between components of the battery are minimal or at least self-limiting, e.g.,
between Li metal or Li2 O2 and the electrolyte.
Normalization of battery capacity. Another important characteristic is the theo-

retical capacity and normalization. Li-O2 batteries do not have a fundamental limit to
the capacity in the same way as Li-ion batteries. Instead, the choice of normalization
should be chosen such that it captures the limiting factor of the battery. It is generally
accepted in the literature, that the growing, insulating Li2 O2 layer causes cell death
due to a blocking of electrons to the surface when the thickness reaches approximately
5 nm, and together with the BET area, this could give an estimate of the theoretical
capacity. As suggested by Meini et al., this number should, however, be adjusted not
to count in the micropore surface area, as they show that the capacity scales with the
non-micropore surface area rather than the full BET area [15]. Reza et al. have likewise
shown that the binder blocks the micropores [16].
Figure 2.3a shows the correlation between carbon loading and discharge capacity to 2.6
V using a current of 130 mA/gC . The correlation is clear, which suggests a normalization
to the carbon mass is appropriate for this system. Figure 2.3b shows how much elec-
trochemical oxidation has occurred during charge at a given potential. The monotonic
increase shows that a normalization to the mass is also appropriate during charge. The
red points in the two graphs are the same measurements. They have been discarded due
Capacity at 2.6V [mAh]

4.0 23 charge
24Good 22 21 20 19 18 17 163.615 24
14Good
23
13charge
22
12 21
11 20
10 19
9 18
8 17
7 16
6 15
5 14
4 13
3 12
2
Bad charge Bad charge
Potential [V]
3.5 3.5
Linear fit
183.0 183.4
172.5 173.3
2.0
16 Q(mC) = 678 mAh/gC · mC 163.2
1.5 3.1
15 2.0 2.5 3.0 3.5 4.0 4.5 5.0 5.5 6.0 15 0 100 200 300 400
Mass [mgC] Capacity [mAh/gC]
14 14
Figure 2.3: Left: Measurements of discharge capacity as a function of the carbon
13 13 below 2.85 V indicating that degra-
loading. Some cells started charging at potentials
dation products formed during discharge are oxidized at these low potentials. Such
12 12
cells are marked in red and are considered as outliers. The reason could be small leaks
11 11
or insufficient drying of the cathodes or the Swagelok cells. A linear fit through (0,0)
shows a capacity of 678 mAh/gC , independent of the carbon loading. Right: The ca-
10 10
pacity of cells charged to different voltages compared to the average capacity of 678
mAh/gC .
9 9
8
to differences in the discharge and charge curves8that indicate the presence of impurities
7 cathode.
in the 7
6 6
Paper fig. Capacity charge
2.25 Electrochemical impedance5 spectroscopy
4 4
3
Electrochemical impedance spectroscopy (EIS) 3is central in this thesis and is used ex-
2
tensively in Paper I, Paper II and Paper III. This2 section is used to highlight and explain
the 1theory relevant to the Li-O 2 system,
Paper and
fig. Capacity a 1basic understanding of EIS
normalization is fig.
Paper assumed.
Capacity normalization
Template used to crop directly in LaTex with trim-com
For additional information on the basic theory, the educational book chapter by Allen
Bard [17] (Chapter 10) and the thorough book Electrochemical Impedance Spectroscopy
by Mark Orazem and Bernard Tribollet [18] are recommended.
EIS has been used to describe the fundamental mechanisms of the Li-O2 battery by a
few other groups. The most thorough studies are presented by Adams et al. [19], Landa-
Medrano et al. [20] and Bardenhagen et al. [21]. In addition to these publications, a
number of groups have used EIS to characterize the capacity loss, hysteresis and pore
clogging in Li-O2 batteries among others [22–25] and Metha et al. have presented an
initial modeling study presenting fundamental consideration, but without good experi-
mental data to validate the model [26].
2.2.1 Correlation between overpotential and impedance
The ultimate goal of fundamental investigations such as EIS is to predict or improve

certain performance parameters. It is therefore very interesting to relate the impedance
measurement to the change in overpotential. The impedance is defined as the derivative

of the iv-curve:
∂v ∂η
Z(i) = = , (2.2)
∂i ∂i
where v is the potential, i is the current density and η is the overpotential. Therefore,
the impedance is linked closely to the Tafel plot, which has previously been used to
describe reaction mechanisms in Li-O2 batteries [27–30]. From the Tafel equation, the
overpotential is seen to be proportional to log(i) at large overpotentials (|η| RT/nF),
but as our batteries contain a porous cathode, this ideal behaviour is not valid. The
consequences of a porous electrode have been investigated by Lasia [31] and show that
the Tafel slope will increase at higher currents. This is in line with our measurements
as well as previous Li-O2 battery measurements by Viswanathan et al. [27] and Adams
et al. [19] on porous electrodes. To describe the measurements better
η = |v − OCV| = c1 · ic2 (2.3)
is applied as an empirical model, when Equation (2.2) is used to compare the measured
impedance with the overpotential. OCV is the open circuit potential, and c1 and c2 are
constants.
24 23 22 21 20 19 18 17 16 15 14 13
18
17
16
15
14
Figure 2.4: The plateau voltage dependence on current density (red dots) from 10
µA (9 µA/cm2 ) to 5 mA (4.4 mA/cm2 ). Equation (2.3) is fitted to the data (red line),
13
which is then differentiated (black line) and compared with the total resistance (black
dots) measured with impedance. Three representative discharge curves show how the
12
plateau voltage is determined. The impedance and current density have been weighed
by the carbon mass of each electrode. Reprint from Paper I.
11
Figure 2.4 from Paper I exemplifies the dependence between impedance and overpo-
10
tential. The overpotential measurements are fitted to Equation (2.3) and differentiated
9
according to Equation (2.2) to obtain the impedance. Comparing with the measured
impedances, it is seen that the correspondence is not perfect in this system.
8
7
6
2.2.2 Modeling Li-O2 battery impedance
Anode24 23 22 Cathode
21 20 19 18 17 16 15 14 13 12
resistance
CPE1 CPE2 CPE3
Serial
18
Z1 Z2 Z3
Rs
17 R1 R2 R3
16
Figure 2.5: Equivalent circuit diagrams used to model most of the impedance mea-
surements. It consists of three Voigt elements (parallel connected resistor with a con-
15
stant phase element, CPE) and a serial resistance. The contributions to the impedance
can be attributed to either the anode (Z1 ) or the cathode (Z2 and Z3 ). The impedance
is described in Equation (2.4). Reprint from Paper I.
14
13
The measured impedance response can, to a first approximation, be described using an
equivalent circuit model consisting of three Voigt elements (parallel connected resistor
12
with a constant phase element (CPE)) connected in series. The impedance of the Voigt
elements is adopted from Hirschorn et al. [32], and the total impedance, Z(ω), is thus
given by
11
10 X Ri
Z(ω) = Rs + , (2.4)
1 + (jω)ni Qi Ri
9 i=1..3
where ω is the angular frequency, and Ri , Qi and ni are parameters in Voigt element i.
8
R is the DC resistance, and Q and n are parameters of the CPE. If n = 1, the CPE is a
capacitor, and even if n is between 0.7 and 1, a pseudocapacitance, C ∗ , can be calculated.
7
As discussed in detail by Jamnik et al. [33], this capacitance is typically a double layer
6
capacitance related to the process, and by comparing with reference values, it is possible
to estimate the surface area contributing to the process. The pseudocapacitance is
5
calculated from the equivalent circuit parameters according to Hirschorn et al. [32]:
4 (1−n)/n
∗ 1/n RΩ R
C = Q , (2.5)
3 RΩ + R
RΩ is the DC resistance at the investigated frequency. As discussed by Zoltowski et al.,

2
the pseudocapacitance of a CPE element is not well defined [34], which means that the
1
surface area obtained using C ∗ might vary slightly from the actual surface area, but the
order of magnitude, and relative changes are still valid.
The surface area of the flat lithium anode and the porous cathode are in the range
of 1 cm2 and 1 m2 , respectively. The capacitance at the lithium metal surface in an
organic electrolyte is typically 10-20 µF/cm2 as reported by Aurbach et al. [35, 36] and
the capacitance of XC72 is 12.6 F/g in an organic aprotic electrolyte as reported by
1.0 24 23 22 21 20 19 18 17 16
Li2O 13 241223 11
15 14 Electrolyte 22 21
1020 919 18
8 177 16 615 145 13 412 11
3 102 9 18 7 6 5
Relative capacitance
2
Li2CO3 18 Cdl
180.8 17 C
Cdl C
dl Li2O2
CLi2O2
170.6 16 CLi2O2
15
160.4 Li2O2 Li2O2
14
150.2 13
140.0 12 Li2CO3
0 2 4 6 8 10 Cathode substrate
11
13 Layer thickness [nm]
10
12Figure 2.6: (a) The relative change of the cathode
9 capacitance as a layer of Li2 O2
or Li CO is deposited. The dotted line 8
represent a different scenario, where a 8 nm
11 2 3
7 ’8 nm’ on the x-axis correspond to a
thick layer grows along the surface. In this case
10full surface coverage. (b) Sketch of different relevant
6 scenarios for the Li-O2 battery
9 and the equivalent 5capacitance.
4
8 3 Paper fig. Capacitance change
Barbieri
7 et al. [37]. From this, it is calculated2 that the capacitances should be in the
6 of 10 µF and 25 mF for the anode and 1cathode, respectively. Furthermore, theTemplate used to crop directly in LaTe
range
5
capacitance is expected to change during discharge as the dielectric Li2 O2 is deposited.
The4 relative permittivity r of Li2 O2 has been measured to be 30-35 by Gerbig et al.
Li2O2
and3Dunst etPaper
al. fig.
[38,Capacitance
39]. Using change
a value of 30 to calculate the capacitance of the Li2 O2
2 in series with a typical electrode-electrolyte capacitance of 20 mF, a Li2 O2 layer
layer
of 51 nm will halve the cathode capacitance. This is shown in Figure
Template used to crop 2.6a. A similar
directly in LaTex with trim-command
calculation has been made for the Li2 CO3 interface layer between the cathode and the
Li2 O2 . Using the relative permittivity of Li2 CO3 of 4.9 measured by Young et al. [40],
the capacitance will be halved with a layer thickness of 0.8 nm.
The analysis of the change in capacitance will be an important part of the discussions
in Chapters 4 and 6, and, though simplified, the approach to growth of both Li2 O2
and degradation products shown in Figure 2.6 serve as a good description of the actual
change in capacitance during discharge and charge.
Chapter 3
Experimental methods
3.1 Cell configuration
Most of the work described in this work is performed using a specific combination of
components as shown in Figure 3.1. It is, intentionally, made as close to the reference
system used by McCloskey et al. described in [41] to enable a direct benchmarking and
mutual exploitation of the results obtained in the two laboratories. The measurements
Inlet 23 22 21 20 19 18 17 16 15 14 13 12 11 10
24Outlet
18 O-ring
Head space (1 mm)
17 Cathode: XC72 carbon black
Electrolyte: 1 M LiTFSI in DME
16 Polymer or glass fiber separator
Anode: Lithium foil
Quartz tubing
15
14
Figure 3.1: Schematic illustration of the reference system used in this thesis.
The battery is made 13 of a lithium metal anode, a polymer or glass fiber separa-
tor, a porous cathode made of XC72 carbon black and PTFE binder. The elec-
12
trolyte is 1 M bis(trifluoromethane)sulfonimide lithium salt (LiTFSI) dissolved in 1,2-
dimethoxyethane (DME). The Swagelok cell is described further in Section 3.1.1.
11
were made with an 11 mm
10 diameter lithium metal anode (HongKong Wisdom Tech
Company), two 12.7 mm diameter Celgard 2500 separators or one 12.7 mm Whatman
9
glass fiber separator, a 10 mm diameter porous cathode made of XC72 carbon black
Inlet Outlet
8 stainless steel 150 mesh, and 60-90 µL of 1 M LiTFSI in
and PTFE binder on a 316SS
7 on the choice of separator. Details of the cathode and the
DME electrolyte depending
electrolyte are found in Sections 3.1.2 and 3.1.3, respectively. O-ring
6 Head space (1 mm
Porous carbon cat
5 Electrolyte
Lithium foil
4 17 Quartz tubing
3
2
Chapter 3. Experimental methods 18
3.1.1 The Swagelok cell
Gas Inlet Gas Outlet

24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3
18
10 mL gas volume
17 b
16
SS current collector 15 d
PTFE ferrule 14 c
Kalrez o-ring 13
Swagelok union
12
Silica glass
11 e a
SS current collector 10
1 cm 9
8
Figure 3.2: Left: Sketch of the 10 mL
7 Swagelok cell. Right: Picture of the individual
parts of the 10 mL Swagelok cell.
6
5
The cell configuration used to test Li-O
4 2 batteries was improved throughout the project.
Pictures and descriptions of the four 3cell generations are found in Appendix A. Figure
2
3.2 shows the current generation; a Swagelok-type cell with a head-space volume of 10
1 on a maximum expected capacity of the tested
mL. The volume has been chosen based Template used to crop directly in LaTex with trim
batteries of 5 mAh, which corresponds to 2 mL O2 at 1 bar, and as the initial head-space

pressure was typically 1.8 bar, the change in oxygen partial pressure during discharge did
not exceed 10 %. The battery components are stacked in the cell between 316SS anode
and cathode tips that were sealed against a fused silica tube using FFKM Kalrez o-rings
(KZ6375, M-Seals). A 1/1600 tee with M4 thread mounted on the cathode tip enabled
connection of a gas inlet and a gas outlet to allow gases to be fed to and swept away
from the cell. Two DESO quick connects ensured an easy connection to the gas system,
while maintaining an airtight cell during the electrochemical test. The cell design has
proven very robust, but it is important to stress that the precision requirements are
significant, as the tolerance on the inner diameter of the glass tube is 50 µm. Typical
leak rates was 0.2 - 20 mbar/h, with most leak rates in the range 5 - 10 mbar/h. This
means that a 10 % change in pressure (180 mbar) from the leak alone is not reached
until after at least 9 hours in all cells, which is sufficient for most measurements.
The gas inside the Swagelok cell can be changed between argon (purity 6.0, Air Liq-
uid) and oxygen (purity 6.0, Air Liquid) using a Labview controlled automated setup
described in Appendix A.
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
a
18
17
b
18
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3
16 17
15 16
14 15
13 14
12
13
11
12
10
9 11
8 10
7 9 5 µm
6 8
5 7
4
3
Figure 3.3: (a) Picture of an XC72 cathode.6 and SEM image of an XC72 cathode.
5
(b) SEM image of the sprayed mixture of carbon and PTFE
2
1
4
Template used to crop directly in LaTex with trim-command
Carbon+PTFE
3
3.1.2 The XC72 cathode 2
1 Template used to crop directly in LaTex with trim
Figure 3.3 shows a picture of an XC72 cathode and a SEM image of the sprayed mixture
of carbon and PTFE. The cathodes were prepared by air-spraying a carbon/PTFE
dispersion onto a 316SS 150 mesh (WestCoast, Esbjerg, Denmark). The slurries were
prepared by sonicating a carbon black powder (Vulcan XC72, Cabotcorp, GA) and
PTFE (60 wt% dispersion in water, Sigma Aldrich) in a 3:1 wt/wt ratio in a 20:80
isopropanol/water mixture. Figure 3.4 shows how a Badger model 350 air-sprayer with
a heavy nozzle was used to uniformly coat the dispersion onto the SS mesh. The SS mesh
was rinsed in acetone several times prior to cathode preparation. Prior to cutting 10
mm diameter cathodes from the carbon-coated SS mesh with a Heavy Duty Disc Cutter
(MTK-T-06, MTI Corporation), the mesh was allowed to air-dry for 3-4 h. All cathodes
were dried 12 h in vacuum at 120 ◦ C, washed twice in pure 1,2-dimethoxyethane (DME)
inside a glovebox, followed by a second drying under vacuum for 10 min. Depending on
the scope of the experiment, the carbon loading of each cathode was varied between 1
mg and 5.6 mg, but typical loadings were in the range 4.7 mg - 5.6 mg.
a 24 23 22 21 20 19 18 17 16 15 14 b13 12 11 10 9 8 7 6 5 4 3 2 1
18
17
16
15
14
13
Figure 3.4: (a) Picture of air-sprayed carbon/PTFE dispersion on SS316 mesh and
12 (b) the SS316 mesh after punching out air electrodes.
11
10
9
8
7
6
3.1.3 Electrolyte
1M LiTFSI in 1,2-dimethoxyethane. DME was purchased from Novolyte (Purolyte

electrolyte grade) and LiTFSI (purity 99.9 %) was purchased at Sigma Aldrich. The
electrolytes were mixed every 1-2 months to ensure a low level of water impurities. Prior
to the mixing, the LiTFSI salt was dried under vacuum at 200 ◦ C in 12 h and DME was
dried using 4Å molecular sieves (Sigma Aldrich) for several days. The content of water
was not measured directly, but the electrolyte was used in microelectrode experiments,
were any water present would be detected immediately.
Ionic liquids. A total of six ionic liquids (IL) and the corresponding two lithium salts
were tested as described in Paper IV. The ILs were purchased from Solvonic and Sigma
Aldrich in purities between 98.5 % and 99.9 % and used as received from the suppliers.
The electrolyte salts were dried prior to preparing the electrolye. LiTFSI (purity 99.9 %,
Sigma Aldrich) was dried in vacuum for 12 h at 180 ◦ C and Lithium bis(fluorosulfonyl)
imide (LiFSI, purity 99.9 % Suzhou Fluolyte) was dried in vacuum for 12 h at 80 ◦ C.
Electrolytes with lithium salt concentrations of 0.3 M were prepared by mixing the
appropriate ratio of salt and ionic liquid and stirring at room temperature in order to
get homogenous electrolyte solution. Occasionally, the stirring was continued for several
hours to ensure solution of the salt.
3.2 Differential electrochemical mass spectrometry
Differential electrochemical mass spectrometry (DEMS) is an essential measurement

when characterizing the discharge and charge of Li-O2 batteries, and metal-O2 batteries
in general, as the method is able to determine the e- /O2 -ratio and the OER/ORR value,
which are key in defining rechargeability as discussed in Section 2.1.1. The high sensi-
tivity of the mass spectrometer enable detection of even small amounts of gas evolved
and, by using a suitable setup, it is possible to get accurate quantitative measurements
of the O2 consumption (evolution) from the reduction of O2 (oxidation of Li2 O2 ) during
discharge (charge) and CO2 and H2 from degradation reactions.
The method was originally proposed by Bruckenstein et al in 1971 [42]. They col-
lected gaseous electrochemical reaction products in a vacuum system through a teflon
membrane and detected the gases by mass spectrometry. Peter Bruce et al. reported
measurements of the evolved oxygen gas during charge of a Li-O2 battery using a similar
setup in 2006 [12]. The setup was refined significantly for analysis of the gas consumption
and release in lithium-oxygen batteries in 2011 by Bryan McCloskey and Alan Luntz
24 23 22 j21 20 19 18 17 16 15 14 13 12 11 10 9 8 7
18 i
17
MS d
l
16 m
k
15 e
c
14
13 n Pump b a
12 O2 Ar
h f
11
10 Swagelok
g
9 cell
8 Figure 3.5: Schematic illustration of the DEMS setup. The circled letters are used
to refer to specific parts of the setup in the text.
7
6 the IBM group [41], and has provided important insight to the current under-
from
5
standing of the degradation in lithium-oxygen batteries. Since then, five to ten research
groups have acknowledged the importance of this measurement and built a system in
4
their own lab.
3
I have been responsible for building such a system in our lab. Of that reason, I
2 included pictures of the experimental setups I have constructed in Appendix A
have
and
1 details of calibrations and reference measurements are included in Appendix B. Template used to crop di
The DEMS system is illustrated in Figure 3.5. Six normally closed magnetic valves
(SCG256A003NVMS 24VDC from OEM Automatic Klitsø A/S) a , b , c , i , k
and l enable automatic control of how the gases are flowing, two pressure transducers
(DMP 331i, BD Sensors) d and j monitor the pressure inside the setup, and e is an
automatic 2-position 6-way valve (EHC6WE, VICI) that enable sampling from a special
low-volume Swagelok cell g . The Swagelok cell is easily attached to the setup using
quick connects (SS-QM2-B-100 and SS-QM2-D-100, Swagelok) f and h . The mass
spectrometer (Omnistar GSD320-C, Pfeiffer Vacuum) m is used to analyze the gases
from the Swagelok cell and the vacuum pump n is used to clean the system before each
sample collection from the Swagelok cell.
3.2.1 Typical operation
As already discussed, the three important characteristics to look for in a DEMS mea-
surement are the e- /O2 -ratio, the OER/ORR value and the identification of other gases
than oxygen. To obtain these results, two modes of operation are used:
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 24
5 23
4 22
3 21
2 20
1 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
18 18 VMS
17
MS 17
MS
16 16
15 15
14 14
13 Pump 13 Pump
Vdis Vch
12 O2 Ar 12 O2 Ar
11 11
10 Swagelok 10 Swagelok
9 cell 9 cell
8 8
7 Figure 3.6: Schematic illustration of the DEMS 7 setup in open (left) and closed (right)
6
configuration. Volumes relevant for calculating6 the gas evolution and consumption are
5 5
shown in red and green color.
4 4
3 3
2 2
1 1
Template used to crop directly in LaTex with trim-command Template used to crop directly in LaTex with trim-command
Pressure mode. This method is always used in discharge, and in some cases during
charge. The change in gas pressure is measured and related to the amount of oxygen
consumed or released. This implies an assumption that oxygen is the only gas consumed
or evolved. The pressure mode is performed in the configuration shown in Figure 3.6
(left).
DEMS. This method is primarily used during charge. Evolved gases are detected by
flushing the Swagelok cell with argon, and collect the gas in a chamber VMS , that is
analyzed with the mass spectrometer. By comparing the M/Z intensity of the mass
spectrum with reference measurements, it is possible to calculate the composition (and
thereby the amount) of each gas. In the DEMS measurement, the cell is charging in
the configuration shown in Figure 3.6 (right). In this period, the magnetic valves c ,
i and k are open to remove all gases in the gas system. Just before sampling, i
is closed and a is opened. The sampling is made by turning e to the open position
and then back to the closed position. The charge continues while the sampled gas is
analyzed with the mass spectrometer by opening l . Only small amount of gas evolves
between each sampling (0.1% - 1%), so the main content of the sampled gas is argon.
The natural occurring 36 Ar isotope is used as a reference, as it constitute 0.333% of the
argon gas, which is in the same range as the oxygen that needs to be detected through
the charge.
3.3 Physical characterization
The cathode has been investigated using a number of physical characterization meth-
ods including transmission electron microscopy (TEM), scanning electron microscopy
(SEM), BET, X-ray diffraction (XRD) and quantification of Li2 O2 in the cathode using
quantitative optical absorption spectroscopy. The project has focused mostly on the ab-
sorption technique and TEM characterization and the experimental section is therefore
limited to these two measurement techniques.
3.3.1 Quantitative optical absorption spectroscopy
The amount of Li2 O2 was determined at different stages of charge, using a spectrophoto-
metric measurement. After electrochemical test of a Li-O2 battery, the cell was purged
with argon and transferred to a glovebox. The cell was carefully disassembled and the
cathode was extracted. Each cathode was washed with 1,2-dimethoxyethane (BASF)
dried using 4 Å molecular sieves (Sigma-Aldrich), and the cathodes were subsequently
dried in vacuum. The cathodes were taken from the glovebox and immediately put into
a 4 mL 0.063-0.07 % TiOSO4 aqueous solution and the colored oxidized Ti-complex was
seen immediately. The reactions occurring are
Li2 O2 + 2H2 O → 2LiOH + H2 O2 (3.1)

Ti(IV)OSO4 + H2 O2 + 2H2 O → 4H+ + H2 Ti(VI)O4 + OSO4−
4 (3.2)
H2 Ti(VI)O4 absorbs strongly at 408 nm. The solutions were left to react for 15-30 min
and to remove carbon particles, which otherwise would interfere with the spectropho-
tometric measurement, samples were centrifuged and the supernatant was extracted
yielding a clear colored liquid that was characterized using a Shimadzu UV-3600 Phar-
maSpec with 1nm resolution and medium scan in absorbance mode. Further details on
the model is found in Paper II.
3.3.2 TEM
The goal was to study the Li2 O2 deposits in the discharged battery as well as the
catalysts that most people thought would be necessary in 2011 [43–45]. As described
in Section 1.3.1, it was subsequently found that catalysts only make things worse in
Li-O2 batteries because of an increased degradation rate in the battery, and the TEM
studies shifted the focus almost entirely towards the Li2 O2 morphology and growth.
The literature contains very nice studies of this. The most prominent ones are the
identification of toroidal shaped Li2 O2 particles made up of pancake like crystals [46],
and the in situ charging of Li2 O2 particles [47]. In addition to these publications, a
few nice selected area electron diffraction (SAED) studies have been made to reveal the
structural information on the nanometer scale [48–50].
The study of Li2 O2 is challenging because it is both air- and beam-sensitive, and typ-
ical high-resolution techniques are not applicable. Furthermore, it has been shown by
Aetukuri et al. that the large toroidal shaped particles are only formed in the presence
of water impurities, and thus not interesting from a commercial point of view. This
means that the challenge is to identify and investigate thin layers of Li2 O2 with a thick-
ness of only a few nanometers located on top of a carbon substrate. The experimental
procedure is described below.
a 24 23 22 21 20 19 18
b 17 16 15 14 13 12 11 c10 9 8 7 6 5 4 3 2 1
18
17
16
15 100 µm 500 µm 100 µm
14
13Figure 3.7: SEM images of (a) plain SS316 TEM grid and (b)-(c) two magnifications
of SS316 TEM grid coated with XC72 carbon black and PTFE.
12
11
The TEM sample was made by spraying a stainless steel SS316 TEM grid in the same
10
way as the ordinary cathodes, described in Section 3.1.2. Figure 3.7 shows SEM images
9
of such a coat. The TEM grid was discharged galvanostatic together with an ordinary
8 TEM grid
cathode with the typical current density of 130 mA/gC . Comparing the discharge ca-
7
pacity with the cathode weight, approximately 20 wt% of Li2 O2 is expected compared
6
to carbon and PTFE binder.
5
The4 samples were examined in the FEI Tecnai G2 at DTU CEN in HR-TEM mode
operated
3 at 200kV. Imaging was performed at varies magnifications. The electron en-
ergy2 loss spectroscopy (EELS) acquisition was tuned by the Gatan tune and alignment
software
1 yielding an energy resolution of about 2.1eV measured
Template as
usedthe
to cropFWHM ofwith
directly in LaTex thetrim-command
zeroloss peak. The EELS was in this session only used as a qualitatively fingerprint in
order to verify the presence of lithium in the sample. The samples were briefly screened
at low magnification (for overview), and 3 areas were inspected further with EELS,
selected area electron diffraction (SAED) and higher magnification.
3.4 Electrochemical impedance spectroscopy
The electrochemical characterization of the batteries has been a central part of the work
presented in this thesis and in Paper I, Paper II and Paper III. Most measurements have
been performed using Bio-Logic MPG-2 and VMP3 potentiostats with EIS capability. In
the DEMS measurements, however, a Gamry Reference-600 was used. Several standard
techniques like OCV mapping, galvanistatic and potentiostatic discharge and charge,
and cyclic voltammetry have been used extensively and in a wide current and SoC
window. Typically, only the first cycle is investigated, because subsequent cycles will be
affected by the degradation occurring during the first cycle.
EIS measurements were conducted in galvanostatic (GEIS) and potentiostatic (PEIS)

mode, and with and without a current load depending on the purpose of the measure-
ment. More than five thousand spectra have been measured to continuously push the
boundary of what was possible to probe with the technique. In Chapter 4, Paper I and
Paper II, the focus is the discharge and charge mechanisms and thus it was beneficial
to draw a current to investigate the processes under relevant conditions as discussed
previously by Adams et al. [19]. In these measurements, the impedance was measured
at currents between 15 µA (13 µA/cm2 ) and 1 mA (0.88 mA/cm2 ) with an alternating
current (AC) amplitude of 10% of the direct current (DC) level. To investigate the
charge with a rapid increase in potential, it was however necessary to use another ap-
proach. After charging to the desired potential in galvanostatic mode, the potential was
kept an a 10 mV AC amplitude was superimposed in a PEIS measurement. This stabi-
lized the system significantly and enabled the more detailed investigation of the charge
impedance presented in Paper II. In Paper III, the focus is an accurate determination of
the double layer capacitance of the air-electrode. Since gradients inside the battery will
affect this measurement, it was desired to do these measurements at OCV. In general
EIS measurements were performed at frequencies between 10 mHz and 20 kHz, and in
some cases a wider interval. 15 points per decade is used to enable a proper fitting of
the data using equivalent circuit fitting.
Three electrode measurements were performed in the group using an EL-CELL to con-
firm the assumptions of which processes belonged to which electrode, but the majority
of measurements were performed in simple 2-electrode Swagelok cells. In the 2-electrode
cell, variations to the battery was used to identify the electrode specific reactions. Among
these tests were: (i) measuring impedance at open-circuit voltage (OCV) in argon at-
mosphere to prevent the oxygen reduction/oxidation, (ii) using a symmetrical cell of
two pre-discharged cathodes, and (iii) testing a different cathode. In the symmetrical
cell, the anode/cathode reactions are oxidation/reduction of Li2 O2 which remove any
lithium metal-related contributions from the EIS measurement. Both cathodes in the
symmetrical cell initially discharged 0.25 mAh in separate cells before they were com-
bined in a new cell. The cathodes were rinsed with DME after the individual discharge
to remove the electrolyte-salt before the cathodes were used in the symmetrical cell.
The symmetrical cell was tested in O2 gas and was made without exposing the cathodes
to air at any point.
The equivalent circuit fits are made using the scipy optimizer fmin slsqp using the soft-
ware package RAVDAV 0.9.7 [51].
Chapter 4
Overpotentials and degradation
This chapter focuses on the reference system with an XC72 cathode and is closely related
to Paper I and Paper II. A number of physical and electrochemical measurements will
be presented to qualify the discussion of the discharge and charge mechanisms and how
they affect the overpotential. All measurements were performed using a system with an
XC72 carbon black cathode, DME/LiTFSI electrolyte and lithium anode as described
in Section 3.1. The system has been chosen because it is widely studied and show low
degradation compared to other Li-O2 systems. It has been characterized extensively in
previous publications from 2011 to 2013 by McCloskey et al. [28, 41, 52–56], and these
measurements will be used to supplement the discussion. The measurements have been
conducted at DTU Energy and Haldor Topsøe A/S in Denmark and at IBM Almaden
Research Center in California.
4.1 Physical characterization
4.1.1 X-ray diffraction
X-ray diffraction provides a simple identification of Li2 O2 in the discharged cathode.

The method has mistakenly been used in literature to prove the absence of degradation
product, which is not possible, since the degradation products are not necessarily crys-
talline. In our group we have primarily used the method to study the growth of Li2 O2 .
Figure 4.1 (left) shows an XRD spectrum of a discharged XC72 cathode. The six major
Li2 O2 peaks and the peaks from the PTFE binder are clearly visible and indicate that
the primary reaction is the formation of Li2 O2 . Figure 4.1 (right) shows an in-situ study
of an XC72 cathode in a capillary cell performed by Storm et al. [57]. In this study, they
were able to observe a linear increase in the peak area of the 100 diffraction peak during
27
Chapter 4. Overpotentials and degradation 28
discharge with low uncertainty and thereby assess the amount of crystalline Li2 O2 in
the sample to supplement measurements like the absorption measurements described in
Section 3.3.1.
24 23 22 21 20 19* 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
24 23 22 21 20 19 18 17 16 15 14 13Li 12 11 10 918 8 7101 6 5 4 3 2 1
Intensity
2O2
17 100
18 Li2CO3
16
17 PTFE 102 * 110
15
Intensity
*
16 14
6 µA
13 4 µA
15
12
14 11
13 10
9
12
8 3 µA
11 30 40 50 60 70 720 25 30 35
10 2θ 6 2θ
5
9
Figure 4.1: (Left) Ex-situ XRD measurement performed 4 by sealing powder from a
8
discharged cathode in a 0.7 mm diameter capillary3 inside the glovebox without expos-
7 ing the powder to ambient air. The data is acquired 2
using a Regaku Advance X-ray
6 Diffractometer with Cu-Kα radiation (λ =0.15418 1nm). (Right) In-situ diffraction pat- Template used to crop directly in LaTex with trim-command
5 terns for the discharge of a capillary battery showing the appearance of four diffraction
4 peaks of Li2 O2 and the ones of the SS wire (*). The current density changes through
3 the measurement from 3 µA (blue) to 4 µA (red) and 6 µA (light blue). Adapted from
2 [57] with permission.
1 jonn_s025 XC7 _CV cap3 Template used to crop directly in LaTex with trim-command
4.1.2 Differential electrochemical mass spectrometry
As already discussed in Section 2.1.1, it is important to quantify the consumption and

release of oxygen and other gases during discharge and charge, and the DEMS systems
at DTU Energy and at IBM Almaden research laboratory has been essential to the
presented work. The measurements shown in this section has been conducted at DTU
Energy.
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
5.0 4.0
18 voltage
e-/O2 - ratio 3.5
17 4.5
cell voltage [V]
16 4.0 3.0
e-/O2 - ratio
15
3.5 2.5
14
2.0
13 3.0
12 2.5 1.5
11 1.0
2.0
10 0.0 0.2 0.4 0.6 0.8 1.0
9 relative capacity
8
Figure 4.2: The7 SoC dependence of the e- /O2 -ratio during discharge and charge.
The values are averages of six batteries tested at currents in the range of 100-200
6
mA/g carbon.
5
4
3
DEMS oxygen consumption and evolution average
2
1 Template used to crop directly in LaTex with trim-command
Figure 4.2 shows how the e- /O2 -ratio depends on SoC. The values are averages of six
measurements like the ones presented in Figure B.5 in Appendix B and the error-bars
indicate one standard deviation. During discharge, the error-bars are small and hidden
behind the symbols. It is seen that the process is a 2.0 e- /O2 process during the entire
discharge with very small deviations. During charge, however, the e- /O2 -ratio increases
during the entire charge ending at 3.5 e- /O2 at the end of charge. It is also noted that the
standard deviation of the e- /O2 -ratio increases significantly in charge-mode, indicating
that the exact electrochemistry may change between measurements depending on small
variations in current density or other parameters. The e- /O2 -ratio during discharge and
charge and the OER/ORR value based on the performed measurements are shown in
Table 4.1 with the corresponding values obtained by McCloskey et al. [56]. It is seen
that all key parameters are well within one standard deviation.
OER/ORR (e- /O2 )dis (e- /O2 )cha

DTU 0.78 ± 0.05 1.99 ± 0.02 2.62 ± 0.12
IBM, [56] 0.78 2.01 2.59
Table 4.1: Comparison between DEMS measurements performed at the setup built
at DTU Energy and values reported by McCloskey et al. at IBM [56].
A wider current range from 5 µA/gC to 2.5 mA/gC is explored in Paper I, and it is
found that the e- /O2 -ratio and OER/ORR-ratio are rather independent of the applied
current density at currents above 25 µA/gC .
4.1.3 Absorption measurements
The amount of Li2 O2 remaining in the cathode during discharge and charge was quan-
tified using absorption measurements of a Ti-complex oxidized by H2 O2 formed when
submerging the tested cathodes with Li2 O2 (and possibly LiO2 ) in an aqueous solution
as described in Section 3.3.1 and in Paper II.
Figure 4.3a shows the oxygen evolution (blue line) and the Li2 O2 removal (red line)
as the cathode is charged. The oxygen evolution is determined based on the DEMS
measurements presented in Figure 4.2 and the Li2 O2 removal is based on the optical
absorption measurements. The O2 evolution and, in particular, the deviation from the
theoretical value is in accordance with measurements presented by McCloskey et al.
[52], and suggests the presence of electrochemical degradation reactions, especially at
potentials above 3.5 V. The Li2 O2 is, however, disappearing more rapid than expected
from the electrochemistry, suggesting a significant chemical degradation. Figure 4.3b
shows the amount of chemical and electrochemical reactions in different potential inter-
vals, and it is clear that the chemical degradation is most significant in the potential
Relative Li2O2 oxidation 0.8

0.8
24 234.222V 21 20 19 18 Theoretical
17 16 15 14 13 0.35 24 Li23
12 11 O 22
10 921 820(no719
degradation gas) 18 17 16 15 14 13 12 11 10 9
6 5 4 3 2b 1 8 7 6 5 4 3
2 2
4.0 V Electrochemical degradation

a
Normalized molar amount

0.7
Li2O2 oxidized 18
0.300.3
18 0.6 0.6
O2 evolved Li2O2 oxidation (O2 gas)
0.25
17
0.5
17 3.5 V 2.0 V 0.200.2
0.4
0.4 16
16 3.1 V 0.15
0.3
15
3.3 V 0.100.1
0.2
0.2
15 14
0.1
0.05
14 0.0 0
13
0.000.0
0.2
0.2 0.4
0.4 0.6
0.6 0.8
0.8 1.0
1 12 2.0 2 V - 3.1 V 3.13.1 V - 3.3 V3.33.3 V - 3.5 V3.53.5 V - 4.0 V4.04.0 V - 4.2 V4.2
13 Relative capacity Voltage [V]
11
12
Figure 4.3: (a) Measurement of O2 evolution10using DEMS (blue) and Li2 O2 removal
11(either chemical or electrochemical) using absorption
9 measurements (red). The dotted
-
10line correspond to a pure 2 e /O2 oxidation of Li
8 2 O2 with no chemical degradation. (b)
The amount of Li2 O2 oxidation with and without 7 gas evolution and electrochemical
9
degradation. Values are normalized such that the sum of the electrochemical reactions
8(Blue and green) equals the relative change in 6capacity in each interval and sum up to
5
7 1 for a full charge.
4
6 3
5 2 V – 3.1 V and 3.3 V – 3.5 V. This effect
ranges 2 is somewhat more pronounced than
4
previously reported [52]. 1 Template used to crop directly in LaTex with t
3
2
4.1.4
1 Scanning electron microscopy Template used to crop directly in LaTex with trim-command
Scanning electron microscopy (SEM) was performed primarily as a preparation for the
TEM measurements, as it is a quick and easy way to get an overview of the sample. A
poor coating was used for this measurement to expose both carbon and stainless steel.
24 23 22 21 20 19 wt%
18 17 16 15 14 13 12 11 10 9 8 724 623 522 421 320 219 118 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
a
18 100%
1
O
2 3 4
18
b wt%
1 2
100%
3 4
90% F 1 90% O
17 Mn 17
80% Ni 80% Mn Wt % 1 2 3 4
16
O
16 Ni O
70% Cr F 2
Wt % 1 2 3 4 70% F F CK 48 46 61 10
Cr
15 1 60% C 15 CK 74 80 71 9 60% C
OK 16 23 18 3
S S
14 50% 14 OK 4 4 2 <1 50% FK 33 26 11 4
C Mn Mn
40% FK 5 40%
13
Fe
Ni 13 4 10 14 3 Ni SK 2 2 3 -
2 30% Cr
SK <1 <1 <1 - 30% C Fe Cr
Cr K <1 <1 2 16
12 S Fe 12 Fe
20% Cr K <1 <1 5 16 20%
11 4 10% 11 Mn K - - - 1
Mn K - - <1 2 10% S Fe K 1 1 6 58
10 0% 10 Fe K 1 1 16 61 0% Ni K - - - 8
9 100 µm 9 Ni K 3- - 2 8 100 µm
3
8 8
7
Figure 4.4: SEM images and EDS spectra 7of (a) a pristine and (b) a discharged
6
cathode. Low vacuum mode was used to6 prevent charging of the samples.
5 5
4 4
Figure
3 4.4 shows low magnification SEM images
3 and EDS analysis of a pristine and a
2 2
discharged cathode. The structure of the cathode material was very different on the two
1 1
samples, which is probably because the discharged cathode has been pressed inside the
battery. The EDS-analysis showed clear differences between the two samples. First, the
discharged cathode contained more oxygen than the pristine cathode, which is expected
in a discharged cathode containing Li2 O2 . Second, the tested cathode showed a higher
ratio of fluorine and sulfur, which is probably due to a slight amount of residual LiTFSI
salt, that was not washed out completely. The ratio between Fe, Cr, Ni and Mn matches
the content of the 316 stainless steel mesh used as support for the carbon powder.
The EDS-analysis is associated with significant uncertain since carbon and oxygen are
only emitting low energy x-rays that are easily reabsorbed before reaching the surface,
and, in addition to this, the structure is very irregular. This means that the EDS analysis
should be seen as a qualitative method rather than a quantitative one. Knowing this, it
is, however, still interesting to consider the numbers briefly and compare them with the
expected amounts of Li2 O2 . The EDS measurements detects 20 wt% of oxygen and 2
wt% of sulfur in the discharged cathode, as seen in Figure 4.4b. Oxygen is only expected
from the LiTFSI salt and Li2 O2 , and both oxygen and sulfur constitutes 22 wt% of the
LiTFSI salt. According to the EDS data, this means that the sample contain of 18 wt%
of Li2 O2 (as 2 wt% is in the LiTFSI) and 9 wt% LiTFSI (2 wt% LiTFSI divided by 22
wt% sulfur in LiTFSI). This is very close to the expected value of 20 wt% Li2 O2 in the
discharged cathode.
4.1.5 Transmission electron microscopy
24 23 22 21 20 19 18 17 16 1524
14 23
13 12
2211 21
10 20
9 8197 6 5174 16
18 3 15
2 1 14 13 12 11 10 9 8 7 6 5 4 3 2 1
18 a c d e Li2O2
Intensity
17 18200 400 Li2CO3

Li2O
16 17 300 LiOH
Carbon black
Counts [104 e-]
15
14
16150 200
13 1 µm 15 100
12 0
14100
11
10
b 13
0 2 4
1/nm
6 8 10
9 12 50 Li-K
8 f
7
11 50 100
C-K O-K
6
100 nm 10 0
5
9 0 100 200 300 400 500
4
3 8 Electron loss [eV]
2
7
1 Figure 4.5: TEM measurements performed on a discharged TEM grid prepared with
6
XC72 carbon black as described in Section 3.3.2. The measurements were performed at
5
Tecnai G2 at the Technical University of Denmark, operated in TEM mode at 200kV.
4
3
Figure 4.5a and 4.5b
2 show TEM images in two magnifications of a specimen in a dis-
charged cathode prepared
1 directly on a TEM grid as describedTemplate
in Section 3.3.2.
used to crop directly Figure
in LaTex with trim-command
4.5c shows an EELS spectrum of this area and Figure 4.5d shows the measured SAED
pattern of this area. Figure 4.5e shows the analysis of the diffractogram and reveals
three large peaks marked with black circles which correspond to scattering of carbon
black, as identified in Figure C.2 in Appendix C. In addition to this, smaller peaks are
clearly revealed, which are expected to be related to Li-structures. No bulk Li-phase was
identified, but it is important to note that the exact position of the peaks may deviate
significantly from the bulk phases when the crystals are in the nano-scale. The presence
of Li-phases is further substantiated by the EELS spectrum, where the lithium, carbon
and oxygen signals are all identified. Figure 4.5f shows the spectrum of the lithium peak
after subtraction of an exponential background estimation made just before the peak.
In conclusion, lithium-species and carbon were detected both by EELS and electron
diffraction in three investigated areas having particle morphologies resembling carbon
black particles, i.e. particles of about 20-50 nm. This means that the lithium-species
must be located within the carbon particles, and this is an indication of the expected
layer-growth of Li2 O2 . It is also noted that we did not see any big particles in the sam-
ple, and even though the absence of particles is not a proof, it is a good indication that
our system does not contain water impurities, as discussed by Aetukuri et al. [58]. We
have further measurements in preparation using an optimized EELS/EFTEM method,
described in Ref. [59], to pinpoint the location of Li-species to confirm or reject the
presence of the expected Li2 O2 layer of only a few nanometers on the carbon black
particles.
4.2 Electrochemical characterization of the system
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
400 b
a 24 23 22 21 20 19 18 1718 16 15 14 13 12 11 10 9 8 7 6 5 4
Current [mA/gC]
17 200
18 16 0
3.2 V 15
17 2.85 V 14 -200
13 -400
16 12
-600
11 2.0 2.5 3.0 3.5 4.0 4.5
15 10 Voltage [V]
9
2
Figure 4.6: (a) Measurement of OCV through a 250 µA (220 µA/cm ) discharge
14 8
and charge. The steep voltage transients occur when the battery is allowed to relax at
7
OCV. Reprint from Paper I. (b) Cyclic voltammetry
13 6 between 2.0 V and 4.6 V with a
scan rate of 0.55 mV/s.
12 4
3
Figure 4.6a shows the OCV measured as a function of the state of charge in a full
11 2
discharge-charge cycle at 250 µA (220 µA/cm21, 130 mA/gC ). The battery was allowed
jonn_s025 XC7 _CV
10 to OCV by interruption of the current a number of times during both discharge

to relax
and9charge, which is seen as steep voltage transients in Figure 4.6a. The relaxation
criteria was a change in cell voltage of less than 1 mV/h or a relaxation time of 15
8
h. The initial OCV was 3.2 V. The OCV decreased to 2.85 V after a short period of
7
discharge and stayed at this value during the entire discharge - also after reaching the
2.0 6
V cutoff at sudden death. During charge, the OCV was 2.85 V, but it increased
5
4
3
2
slightly toward the end of charge where it reached 3.2 V. Figure 4.6b shows a cyclic
voltammogram between 2.0 V and 4.6 V with a scan rate of 0.5 mV/s. The first thing
to notice is the asymmetry of the measurement, suggesting that the discharge and charge
mechanism is not a single reversible redox reaction. It seems like two processes occur
during discharge with onset potentials of 2.75 V and 2.3 V respectively. During charge, a
small peak is identified at around 3.1 V and several bumps are found in the curve as the
potential is increased further. The measurement agree with previous results published
by McCloskey et al.[41].
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 245 234 223 212 201 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5

1,200 a
2.6 V - 150 min
2.5 V - 150 min b
4.2 V
18 18 200
a
dQ/dV [mAh/gC/V]
2.2 V - 150 min

17 1,000 3.5 V
dQ/dV [mAh/gC/V]
17 2.0 V - 150 min
150
16 800 2.0 V - 0 min
3.4 V 16
15 600 3.3 V 15 100
3.85 V
14 400 3.05 V
4.3 V 14 3.05 V 3.3 V 3.4 V 3.5 V 3.85 V 4.2 V 4.3
4.5 V 50 1,200
13 13
dQ/dV [mAh/gC/V]
200 1,000
12 12 0
0 11 2.85 800
11 3.0 3.5 4.0 4.5 2.90 2.95 3.00 3.05 3.10 3.15
10 cell voltage [V] 10 cell voltage
600 [V]
9 9
Figure 4.7: (a) Differential capacity plot (dQ/dV) 8
of a typical charge.400Each peak
8 200 capacity
represents an electrochemical reaction. Reprint from 7
Paper II. (b) Differential
7
plot of the initial part of the charge following a constant
6
current of 130 mA/gC03.0discharge 3.5 4.0
6
to 2.0 V, 2.2 V, 2.5 V or 2.6 V, which was maintained 5
at 150 min before charging. The cell voltage [V]
5
measurement labeled 2.0 V – 0 min was charged4 immediately after reaching 2.0 V.
4
Reprint from Paper3 II.
3
2 2
Figure 4.7a shows a differential capacity plot (dQ/dV 1

to crop directly in)LaTex
ofwithatrim-command
typical charge curve. The Template used to crop directly in LaT
1 Template used
peaks correspond to voltage plateaus in the charge curve and thereby different processes.
This is in line with the cyclic voltammetry measurement shown in Figure 4.6b, and from
the differential capacity, eight electrochemical processes can be identified at 3.05 V, 3.3
V, 3.4 V, 3.5 V, 3.85 V, 4.2 V, 4.3 V and 4.5 V. Figure 4.7b shows how the onset potential
of the process at around 3.05 V increases with the depth of discharge and the exposure
time at low potentials. Analysis of 10 charge measurements following a discharge to
2.6 V show a capacity corresponding to 540 ± 80 µmolLi 2 O 2 /gC below 3.15 V. This
corresponds to 4.3 % of the total discharge capacity or approximately one monolayer as
calculated in Paper II.
4.3 Electrochemical impedance spectroscopy
The results presented in this section are selected among more than five thousand impedance
spectra. A wide range of current densities, potentials and techniques have been inves-
tigated and the measurements included are chosen to substantiate the discussion of the
discharge and charge mechanisms presented in Section 4.4, 4.5 and 4.6.
4.3.1 Discharge to sudden death at 250 µA
EIS measurements from the first discharge at 250 µA (220 µA/cm2 ) are shown in Fig-
ure 4.8a-d. The spectra were measured while drawing a current, which means that the
SoDs shown in Figure 4.8 and Table 4.2 are approximate values. Three arcs are dis-
tinguished in the Nyquist plot in Figure 4.8a. They were almost constant in the first
part of the discharge but changed as the potential decreased near the end of discharge.
The three identified impedance contributions are labeled Z1 , Z2 , and Z3 , and, on the
basis of a fit to the equivalent circuit given in Equation (2.4), the corresponding peak
frequencies, resistances and pseudocapacitances are given in Table 4.2 from two of these
measurements.
24 23 22
ω 21 20
Z
19 (a)18 17 16 15 14 13 ω12 11 10(b) 9 C3 = 4.0 mF 8 7 6 5 4
3 C2 = 0.45 mF
Z2
18 Z1 C3 = 10.2 mF
C1 = 2.8 µF
C1 = 2.3 µF C2 = 0.53 mF
17
(e)
16
(c) (d)
ω ω
15
14
13
Z3 Z2 Z1 Z3 Z2 Z1
12
11
Figure 4.8: Nyquist-plot (a and b) and Bode-like plot (c and d) of impedance mea-
10
surements during a 250 µA (220 µA/cm2 ) constant current discharge. The approximate
9 are shown in (e) and in the legend of (a and b). Three processes are identified and
SoDs
named Z1 , Z2 , and Z3 . The three corresponding peak frequencies are within the grey
8
intervals marked in (c and d) at all current densities and SoDs investigated. Reprint
from Paper I.
7
6 that R1 is constant through the discharge, whereas R2 and R3 increase, and
It is seen
C3∗ decreases.
5 The decrease of C3∗ and increase of R3 through the discharge, could be a
blocking of the cathode surface. The magnitudes of the pseudocapacitances indicate that
4
Z1 originates from an anode process, and Z2 and Z3 originate from cathode processes.
3
The cathode blocking and identification of reaction processes in the impedance spectra
2
are discussed further in Section 4.4.
1 frequencies changed between different current densities and close to

The peak Template
suddenused to crop directly in LaTex with
death. In all of our measurements, however, f1 was between 100 Hz and 10 kHz, f2 was
between 2 Hz and 100 Hz, and f3 was between 20 mHz and 1 Hz. These intervals are
shown in Figure 4.8b and 4.8d, and the clear separation helps in identifying the different
impedance contributions.
4.3.2 Discharge to sudden death at 20 µA
We decreased the discharge current to 20 µA (18 µA/cm2 ) to increase the stability of

the system during the impedance measurements, see Figure 4.9. When comparing this
with the previous discharge at 250 µA presented in Figure 4.8, it is important to note
that both the capacity and the polarization resistance are significantly larger in the 20
µA discharge.
24 231 22 21
2
20 319 18 17 16 15
18
17
16 700 Ω 1200 Ω 2900 Ω
15 1100 Ω 1900 Ω 4500 Ω
14
Equivalent circuit
13 Rs-(RQ)1-(RQ)2-(RQ)3
12
11
10
Figure 4.9: Resistances and normalized pseudocapacitances determined from EIS
measurements in a 20 µA (18 µA/cm2 ) constant current discharge to 2.2 V using
Equation (2.4). Nyquist plots are shown at three representative stages and the corre-
9
sponding SoDs are marked with circles on the voltage profile. R2 and C2∗ could not be
determined well at the end of discharge and are thus greyed out. Reprint from Paper
I.
8
On the basis of a fit using Equation (2.4), the resistance and pseudocapacitance param-
7
eters of Z1 , Z2 , and Z3 are presented in Figure 4.9 and summarized in Table 4.2. The
6
5
parameters of Z1 were constant through the entire discharge, and the change of param-
eters related to Z2 and Z3 are divided into three parts as indicated in Figure 4.9: (1)
At 0% - 40% SoD, only negligible change was observed, (2) at 40% - 80% SoD, R2 and
R3 increased 2-3 times, and C2∗ and C3∗ decreased by 95%, and (3) at 80% - 100% SoD,
R3 increased to 14.1 kΩ (more than 10 times the initial value), the pseudocapacitances
stayed at 5% of the initial value, and the voltage dropped. Parameters related to Z2
could not be determined in the last part of the discharge because of the overlap with
Z3 .
At 20 µA, the average relative Kramers-Kronig deviation at frequencies from 1 mHz to

10 Hz was typically 0.5% at the plateau, increasing near sudden death to 2% at 2.2 V.
4.3.3 Supporting EIS measurements
24(a)23 22 21 20 19 18 17 162415
(b)23142213211220111910189 178 1624
7 15 6 1422
(c)23 5 1321
4 1220
3 1119
2 1018
1 9 178 167 156 145 134 123 112 101
18 18 2.1 mF 18 220 Ω
18.3 mF
2.9 mF
17 280 Ω
17 17 26 Ω
1.5 mF
Li/XC72, Ar 0.2 mF 103 Ω XC72/XC72, O2 27 Ω Li/P50, O2
16 1.2 µF 16 4.3 µF 16 89 Ω
2.7 µF
Eq. circ.: R-RQ-RQ-RQ-C Eq. circ.: R-RQ-Q Eq. Circ.: R-RQ-RQ-RQ
15 15 15
14 14 14
13 13 13
Figure 4.10: Bode-plot and Nyquist-plot of supporting EIS measurements. All mea-
12 12 12and the fitted parameters
surements are fitted to the equivalent circuit listed in the plot
11
are listed in the plot (resistance
11 in blue and pseudocapacitance
11 in red). Left: Potentio-
static
10 EIS measurement at OCV
10 of a cell in Argon
R3 = 171 Ω
C*3 = 8.5 mF
atmosphere.
10 Middle: Potentiostatic
R = 27 Ω
C* = 2.1 mF C* = 4.3 µF
EIS measurement of a symmetrical cell at 0 V with an amplitude of 5 mV. The cell is
9 9 been discharged in separate9cells with a lithium anode
made of two cathodes that have
8 2
at 250 µA (220 µA/cm ). Right: 8 Galvanostatic EIS measurement8 from a 250 µA (220
2
7 µA/cm ) discharge with an7AvCarb P50 carbon cathode.7 Adapted from Paper I.
6 6 6
5
Figure 4.10 shows three supporting
5 EIS measurements; An 5EIS measurement at OCV of
4 4
a cell in argon atmosphere, an EIS 4 a different carbon cathode
measurement a cell with
3 3
and an EIS measurement of a symmetrical cathode-cathode 3cell at 0 V. All measurements
2 2 2
are fitted to the equivalent circuit listed in each plot and the fitted parameters are also
1 1 1
Temp
listed in the plot (resistance in blue and pseudocapacitance in red).
Figure 4.10 (left) shows a potentiostatic EIS measurement at OCV with 5 mV amplitude
of a fresh battery in argon atmosphere before exposure to oxygen. The purpose was to
investigate reactions not related to oxygen reduction, and it is seen that Z1 and Z2
were also present in the absence of oxygen. The low frequency tail could be modeled
with a capacitor, C3 , which means that no charge transfer reaction is present for this
process. The spectrum was modeled with the equivalent circuit R-RQ-RQ-RQ-C. The
capacitance C3 was 18.3 mF, and the pseudocapacitances C1∗ and C2∗ were 1.2 µF and
0.2 mF, respectively. The presence of Z2 suggests that this process is not related to
oxygen reduction.
Figure 4.10 (middle) shows a potentiostatic EIS measurement of a symmetrical cell

at OCV with 5 mV amplitude. The purpose was to eliminate contributions from the
lithium anode in the EIS spectrum. The cell was made of two cathodes that had been
discharged 1 hour in separate cells with a lithium anode at 250 µA (220 µA/cm2 ). The
spectrum was modeled with a R-RQ-Q circuit. The CPE element was chosen instead
of a capacitor to describe the low frequency tail, because the slope was -0.74 in the
Bode plot, rather than -1 in the case of a capacitor. The pseudo-capacitance of the low
frequency tail was 2.1 mF and the resistance and pseudo-capacitance of the RQ circuit
was 27 Ω and 4.3 µF, respectively.
Figure 4.10 (right) shows a GEIS measurement on a cell with an AvCarb P50 carbon
paper cathode prepared as described in [54]. The purpose was to investigate cathode
specific contributions to the impedance. Both Z1 and Z3 had similar parameter values
compared to XC72, whereas Z2 was very small.
4.3.4 Charge at 250 µA
In Figure 4.11, we present typical EIS measurements during a charge. To limit the
complexity of the analysis, impedance measurements are only made at voltages below
4.2 V to avoid the major decomposition reactions observed at higher potentials using
DEMS. In this measurement, the 4.2 V limit corresponded to 60% SoC.
Three impedance contributions are identified. The parameters obtained using equivalent
circuit fitting on the green (0.03 mAh) and black (0.42 mAh) spectra with Equation (2.4)
f1 [Hz] f2 [Hz] f3 [Hz] R1 [Ω] R2 [Ω] R3 [Ω] C1∗ [mF] C2∗ [mF] C3∗ [mF]
Discharge at 250 µA
0.16 mAh 733 5.4 93·10−3 96 56 145 2.3·10−3 0.53 10.2
0.51 mAh 605 3.4 184·10−3 94 92 188 2.8·10−3 0.45 4.0
Discharge at 20 µA
0.5 mAh 470 1.15 5.5·10−3 109 50 1007 3.1·10−3 2.8 19.3
1.9 mAh 464 1.12 9.9·10−3 107 158 2131 3.2·10−3 0.6 2.0
2.3 mAh 479 1.1·10−3 114 14097 2.9·10−3 0.8
Charge at 250 µA
0.03 mAh 678 9.6 267·10−3 65 166 497 3.6·10−3 76·10−3 1.0
0.42 mAh 983 14.0 19·10−3 255 105 700 0.6·10−3 99·10−3 9.0
Table 4.2: Peak frequencies, resistances and pseudo-capacitances from selected

impedance fit. The expected capacitances for the full anode and cathode are 10 µF
and 25 mF, respectively. Typical values of n are n1 = 0.77, n2 = 0.86 and n3 = 0.78.
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8
18
17
16
15
14
13
12
11
Figure 4.11: Nyquist10 (top) and Bode-like (bottom) plots of EIS measurements made
during a 250 µA (220 µA/cm2 ) constant current charge. The charge followed a discharge
9 figure 4.8e with a discharge resistance extrapolated to 3 kΩ
similar to that shown in
at 2.0V. The SoCs are shown as circles in the inset graph with the voltage profile.
The 0.03 mAh and 0.428mAh measurement were modeled using (2.4) and the obtained
parameters are found in Table 4.2. Reprint from Paper I.
7
6 three contributions are in the same frequency ranges as seen
are given in Table 4.2. The
during discharge. The polarization
5 resistance (R1 + R2 + R3 ) was almost constant
in the range 500 Ω - 1000 Ω, but the peak frequencies and the relative magnitude
4
of the different impedance-contributions changed. Looking at the pseudocapacitances,
C1∗ decreased from 3.6 µF3 to 0.6 µF, and C3∗ increased from 1.0 mF to 9.0 mF. This
2
suggests that the active cathode area is increasing and that the active area of the anode
is decreasing during charge. It is further noted that C3∗ is almost the same in the end of
1 Template used t
charge and in the beginning of the discharge (10.2 mF). Finally, it is noted that R1 was
almost constant until 3.7 V, after which it suddenly increased. This supports that the
lithium anode surface is deactivated by the formation of the solid electrolyte interface
(SEI) layer - possibly due to oxygen crossover.
To investigate the charge process further, EIS was measured at 11 different potentials
during the initial charge from 3.10 V to 3.60 V. Figure S5 in the Supporting Informa-
tion of Paper II shows a typical measurement, with the equivalent circuit fit and the
determination of the resistance, RLi 2 O 2 , and the pseudocapacitance. Both parameters
are related to the charge transfer through Li2 O2 and Li2 O2 oxidation. Figure 4.12a
shows the cathode resistance at selected voltages during charge, as determined by EIS.
It is seen that the resistance increases from 3.10 V to 3.30 V, decreases at 3.33 V, in-
creases until 3.50 V, and decreases again at 3.60 V. The resistance and corresponding
pseudocapacitances at the marked cross section are shown in Figure 4.12b for all 11
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 1.0
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
18 10 a 3.1 V
18 3
Resistance
Capacitance [mF/gC]
Li2O2 resistance [ΩgC]
3.3 V Capacitance 0.8
Resistance [ΩgC]
17 3.333 V
17
16
3.5 V
3.6 V 16 2
b 0.6
3
15 15
0.4
14 14
1 1
13 13 0.2
12 Cross section 12
11 0.3 0
11 3.0 0.0
20 40 60 80 100 3.1 3.2 3.3 3.4 3.5 3.6
10 Current [mA/gC] 10 Voltage [V]
9 9
8 Figure 4.12: (a) The resistance related to Li28O2 oxidation and charge transfer through
7 Li2 O2 , determined using EIS at different potentials.
7 The current decreases during the
6 measurement. Reprint from Paper II. (b) Resistance
6 and capacitance from charge after
5 the current has decreased to 50 % of the initial5 current. Guide lines have been inserted
4 to illustrate the stepwise increases in the4 resistance. Reprint from Paper II.
3 3
2 2
1 1 directly in LaTex with trim-command
potentials. It is seen that both the resistance and the pseudocapacitance change step-
Template used to crop Template used to crop directly in LaTex with trim-command
wise as a function of potential. The resistance increases monotonic until 2.27 V, then
it drops and continues a second monotonic increase from 3.33V to 3.50 V after which it
drops. The pseudocapacitance is high at 3.10 V, decreases at 3.20 V, increases at 3.33
V and decreases again at 3.60 V. When keeping the potential at 3.10 V, it was observed
that the capacitance decreased 60 % from 0.7 mF/gC to 0.28 mF/gC . During this EIS
measurement, the current decreased and Li2 O2 is removed, and as both changes are
expected to increase the capacitance, it clearly suggests that compounds are deposited
during this initial charge, which is in line with the absorption measurements presented
in Section 4.1.3.
4.4 Identification of processes during discharge
From Sections 4.3 and 4.2, it is seen that three impedance contributions were present
during discharge and they are referred to as Z1 , Z2 and Z3 . The five key findings
were that (i) the impedance did not change at the discharge plateau, (ii) Z2 and Z3
increased near sudden death, (iii) C3∗ decreased significantly just before sudden death,
(iv) pseudocapacitances related to Z1 , Z2 and Z3 were typically 3 µF for C1∗ , 0.1-3 mF
for C2∗ , and 1-20 mF for C3∗ , and (v) the OCV was always 2.85V during discharge.
The results support previous findings by Adams et al. and Landa-Medrano et al. [19, 20]
that Z1 originates from the anode and that Z2 and Z3 originate from the cathode. In
addition to this, our results show that Z3 is a combination of the charge transfer reac-
tion of oxygen reduction and the electronic transport through the Li2 O2 , whereas Z2
is a cathode-specific process that is not related to the oxygen reduction. The assign-
ment of anode and cathode features in the EIS is substantiated by the following three
observations.
First, the full capacitance at the lithium anode surface is expected to be in the range
of 10 µF as discussed in Section 3.4, whereas the capacitance of the XC72 electrode is
expected to be 25 mF. If only part of an electrode is active during the EIS measurement,
the capacitance will be lower. As reported, C1 was typically 3 µF, whereas C2∗ and C3∗
were in the range of 0.1 mF - 20 mF. Furthermore, the cathode capacitance per active
surface area calculated by Adams et al. was in the same range as C3∗ [19].
Second, the careful parameter study performed by Adams et al. [19] shows that relevant
cathode processes have peak frequencies below 10 Hz, which correspond with f2 and f3 in
our study, whereas the peak frequency of the anode process is 1 kHz, which corresponds
to f1 in our study.
Third, Z1 did not change during discharge, whereas Z2 and Z3 increased significantly
close to sudden death. Both electrodes change during the measurement. On the lithium
anode, Younesi et al. have previously shown that an SEI layer is forming in a combi-
nation of chemical and electrochemical reactions [60], but as shown by McCloskey et
al. [27] this is affecting neither the electrochemistry nor the measured impedance. On
the other hand, the cathode is covered with an insulating layer of mainly Li2 O2 dur-
ing discharge, and an increase in charge transfer resistance is typically captured in EIS
measurements.
Ascribing Z3 to oxygen reduction and electronic transport through Li2 O2 is based on two
observations: (i) Z3 is the only process related to oxygen reduction, as both Z1 and Z2
are present in argon, and (ii) R3 was the largest resistance during the entire discharge,
both when the cathode was limited by reaction kinetics at the discharge plateau and by
electronic conduction at sudden death.
The process related to Z2 is cathode specific and not related to oxygen reduction, as it
was present in argon, but almost absent in a measurement with P50 carbon paper. As
P50 is binder-free, this could indicate a degradation effect related to the PTFE binder.
A little surprising, Z2 was not present in the symmetrical cell measurement. This could
indicate that the parasitic reaction was chemically passivated during handling when
assembling the symmetrical cell.
The iv-curve presented in Figure 2.4 was made to ensure that all electrochemical pro-
cesses were captured in the impedance spectrum. This was indeed the case, since the
total impedance could account for the changes in the overpotential. Actually, the mea-
sured impedance seemed to overestimate the slope of the iv-curve, and the reason is
most likely a result of lower Li2 O2 formation yields, and therefore more heterogeneous
discharge electrodeposits, at lower current densities as shown in a previous publica-
tion [52]. Lastra et al. and Mekonnen et al. [61, 62] have shown that an increase of
Li2 CO3 -like inclusions in the Li2 O2 layer can change the electrical conductivity using
DFT calculations, and such changes would also change the current dependence of the
impedance and explain the deviation.
4.5 Analysis of the overpotential during discharge
The measurements show that the electrochemistry was unchanged during the entire
discharge and they support the general understanding of tunneling being the dominant
charge transport mechanism through the Li2 O2 layer at relevant current densities and
temperatures, which was initially proposed by Albertus et al. [29] and confirmed by
Luntz et al. [63] Furthermore, the discharge was initially occurring in the entire cathode,
whereas the increasing electronic transport through the growing Li2 O2 -layer passivated
large parts of the cathode during discharge.
The tunneling mechanism is supported by two observations. First, the impedance con-
tribution Z3 related to oxygen reduction and electronic conduction through Li2 O2 was
constant at the discharge plateau and increased rapidly near sudden death, which is
characteristic for the tunneling barrier that depends exponentially on the Li2 O2 layer
thickness, and second, the electrochemistry was unchanged during the discharge, as
shown by a constant 2 e- /O2 process and identification of the same three processes in
the impedance spectra at all SoDs.
The passivation of the cathode is observed in the pseudocapacitance C3∗ . At 20 µA, the
initial value is 21 mF. This is the expected value of the entire cathode, which means that
Li2 O2 deposition is occurring in the entire cathode. The decrease in stage 1, as defined
in Figure 4.9, reflects Li2 O2 formation, because the introduction of a dielectric material
in a capacitor changes the capacitance. In stage 2, the decrease is significant and cannot
be explained by the dielectric layer of Li2 O2 alone. The decrease must therefore reflect
a reduction in active surface area. The cathode passivates when the critical thickness
of Li2 O2 is reached and tunneling is no longer possible. In stage 3, the available surface
area is not sufficient to support the constant current, and the voltage drops to enable
conduction through the blocked parts of the electrode. This is seen as an increase in
cathode resistance. When fully discharged, the resistance is too large, and the current
cannot be supported within the cut-off limit of 2.2 V. This is in full agreement with
observations made by Luntz et al. using flat glassy carbon electrodes in electrolysis cells
[63].
Because of discussions in literature on the significance of oxygen diffusion in the elec-

trolyte, it is worth mentioning that the sudden death is not due to pore clogging and
increased oxygen diffusion resistance. In a typical discharge, the average thickness of

Li2 O2 is 0.5 nm - 1 nm based on the BET surface area of XC-72. This means that the
porosity and Damköhler number are almost unchanged during the entire discharge, and,
as stated by Wang et al. [64], such small changes will not give rise to the sudden death
behavior.
4.6 Reaction mechanisms and SEI layer formation during

charge
The charge of the Li-O2 battery is complicated. The DEMS and absorption measure-
ments described in Sections 4.1.2 and 4.1.3 show that both chemical and electrochemical
reactions occur during charge and the cyclic voltammogram and differential capacity plot
presented in Figures 4.6 and 4.7 both reveal a significant number of reactions. This sec-
tion will analyze the results further and propose detailed input to the charge mechanism,
degradation mechanisms and the potential increase.
4.6.1 Decreasing resistivity of Li2 O2 in charge mode
It has been reported that the onset potential of Li2 O2 oxidation on flat glassy carbon
electrodes is close to the equilibrium potential and that the charge transfer through the
Li2 O2 decreases as the voltage increases from discharge mode to charge mode [63]. The
use of impedance spectroscopy makes it possible to confirm that this is also valid in
porous electrodes.
24 23 22 21CO2 CO218 17 16 15 14 13 12 11 10 9
20 19 8 7 6 5 4 3 2 1
O2
18 Oxidation of SEI
3.3-3.5 V 4.2 V
17 3.05 V
16
SEI formation
O-rich (0001)
O2
15 3.3-3.5 V
CO2
14 3.0-3.3 V O-rich (1-100)
13 Carbonate
4.2 V
12
11
Figure 4.13: 10 Sketch of the reactions and SEI formation during charge of the Li-O2
battery. The potentials
9 in the figure are the proposed onset potentials. Li2 O2 oxidation
occurs at 3.05 V and an SEI layer is formed immediately on the freshly oxidized surface.
At 3.3 V – 3.5 V8 several reactions occur. Among these are gas evolution from the SEI
layer and
7 oxidation of other Li2 O2 surfaces. Reprint from Paper II.
6
5
4
3
The change in resistivity in charge mode is identified by comparing the impedance at the
end of discharge with the resistance in the beginning of the charge. During charge, the
polarization resistance was 500 Ω at a current of 250 µA (220 µA/cm2 ), which is much
lower than the extrapolated value of 3 kΩ at 2.0 V during discharge. Furthermore,
it is seen from Figure 4.7b and similar measurements that the charge resistance had
only little dependence on the discharge current and depth of discharge, which suggests
that the charge is not limited by the same process as the discharge. Luntz et al. have
previously explained this by a reduction of the tunneling barrier because of a change in
the Fermi energy by experiments on flat glassy carbon electrodes in an electrolysis cell
[63].
4.6.2 Identification of Li2 O2 oxidation at 3.05 V
It is argued that the process identified at 3.05 V is oxidation of Li2 O2 based on three
observations. First, Figure 4.2 and Figure S2 in the Supporting Information of Paper
II, show that the e- /O2 ratio is between 2.0 (at 2V) and 2.1 (at 3.2 V) in the beginning
of the charge, which is exactly – or at least very close to – the expected value for Li2 O2
oxidation. Second, Figure 4.7b shows how the onset potential of the process at around
3.05 V increases with the depth of discharge and the exposure time at low potentials.
To understand the shift, it is noted that DEMS measurements show that the e- /O2 -ratio
is 2.0 during the entire discharge, and McCloskey et al. show that the Li2 O2 yield is
independent of the depth of discharge [52]. This means that the thickness, and thereby
the conductivity, of the Li2 O2 layer is the only parameter expected to change between
the measurements, and as the conductivity through the Li2 O2 layer affects the onset
potential of the reaction, it suggests that the reaction is a surface reaction at the Li2 O2
surface. Third, the onset potential at the investigated current densities ( 0.1 µA/cm2
real surface area) is 2.9 V – 3.0 V which corresponds well with the onset potential of
Li2 O2 oxidation measured by Viswanathan et al. using flat glassy carbon electrodes.[27]
4.6.3 SEI layer formation
DEMS measurements show that all electrons come from the Li2 O2 oxidation at the onset
of the charge, until 3.1 V. In this interval, it was found that 4.3 % of the Li2 O2 was
oxidized electrochemically (Section 4.1.3) and in Figure 4.3b it is seen that another 4.6
% was removed without gas evolution. Since all electrons are accounted for by the gas
evolved, the reaction with no gas evolution must be chemical and it is interpreted as
the formation of an SEI layer based on three observations. First, the amount of degra-
dation is close to the amount of electrochemically oxidized Li2 O2 in the initial part of
the charge, and as this process does not continue, it suggests that an electrochemical
oxidation of Li2 O2 exposes the surface such that the oxidation is followed by a chemi-
cal degradation of Li2 O2 , forming an SEI layer. Furthermore, the amount of oxidized
and chemically degraded Li2 O2 both correspond to approximately one monolayer, which
suggest that the reaction occur on the entire surface of Li2 O2 . Second, the 60 % de-
crease in capacitance at 3.1 V suggests a significant deposition at this potential that
could be explained by the formation of an SEI layer. Third, the monotonic increase in
Li2 O2 resistance until 3.3 V suggests a decrease of available surface area or an increased
electronic transport resistance. Both could be explained by a growing SEI layer.
4.6.4 Electrochemical degradation
Identification of the lowest potential without electrochemical degradation is important

to identify a safe-voltage limit. It is proposed that at least one of the three separate
processes identified in the differential capacity plot in the voltage range from 3.3 V to 3.5
V is an electrochemical degradation reaction as the e- /O2 -ratio increases in this range.
Two observations suggest that the reaction occurs at 3.3 V, but further investigation is
needed to determine the onset potential definitively. First, EIS measurements show that
the pseudocapacitance increases and the resistance decreases at 3.3 V. A sudden change
like this suggests a new reaction pathway at this potential. Second, isotope measure-
ments presented by McCloskey et al. on an identical system show that CO2 evolution
occurs from the electrolyte-Li2 O2 interface from 3.3 V.[28] As the CO2 evolution reac-
tion depends on the potential, it is likely that this reaction is the new reaction pathway
seen in the EIS measurements.
At around 3.6 V, the resistance decreases in Figure 4.12b and the pseudocapacitance
decreases. As discussed in previous sections, the relationship between impedance and
overpotential is not straight forward, but the significant decrease in impedance as the
voltage increases, is a strong indication of a shift in equilibrium potential caused by a
mixed potential established between different oxidation reactions to maintain the con-
stant current. The theory of a mixed potential is further substantiated by measurements
at higher potentials shown in Figure 4.11. The polarization resistance is almost constant
throughout the entire charge even though the charging potential increases. As shown
by DEMS and absorption measurements, the charging reaction is not a 2 e- /O2 -process,
but rather a 2.5-3 e- /O2 -process and parasitic electrochemical reactions are thus present
during the entire charge. This is in line with previous publications by McCloskey et
al. [28, 41, 52, 56], and keeping in mind that the OCV never exceeded 3.2 V during
charge, and no significant resistance increase was seen in the impedance spectra, it
5.0
24 23 22 21 20 19 18 17 16 15 14 13
Voltage [V]
4.0
18
3.0 Resistive losses
172.0
Capacity
Figure 4.14:
16
Illustration of the primary contribution to the voltage changes during
discharge and charge based on the discussions in Sections 4.5 and 4.6. Reprint from
15 Paper I.
14
suggests that a mixed potential between these competing electrochemical reactions was
13
established during charge to support the high current.
4.6.5
12
Mixed potential
11
It is noted that these results contradicts the theory proposed by Chen et al. suggest-
10
ing that the increase in charge overpotential occurs because the Li2 O2 closest to the
electronically conducting part of the cathode oxidizes first [65]. If this was the case, an
9
increase of the charge resistance of at least an order of magnitude would be expected
8
to explain the voltage increase, but the resistance does not increase by more than a
factor of 2. Furthermore, after discharging under alternating O2 isotope atmospheres,
7
Li2 O2 oxidation was found to preferentially occur at the Li2 O2 -electrolyte interface over
the Li2 O2 -cathode interface during the initial stages of charge, as shown in a previous
publication [55]. 6
5
4.7 Summary
4 of the fundamental characterization of over-
potentials and degradation
3
2
In this chapter, the electrochemistry of the Li-O2 system with a DME-LiTFSI electrolyte
and a XC72 carbon black cathode has been studied using wide a range of physical and
1
electrochemical characterization methods, including XRD, TEM, Li2 O2 quantification
using absorption, DEMS and EIS. Based on these measurements, explanations to some
of the fundamental problems of the Li-O2 system, including the sudden death at the
end of discharge and the increase in potential during charge, has been proposed.
It was possible to assign the three identified contributions in the EIS spectra during
discharge to either the cathode or the anode. Only one of the two cathode processes
depended on the presence of oxygen. This indicates that this contribution was related
to the Li2 O2 formation. The other contribution was cathode specific and may reflect a
degradation reaction related to the PTFE binder. It was shown that the rapid potential
change near the end of discharge was due to an increase in polarization resistance,
primarily related to the charge transport through the Li2 O2 . This supports previously
published work by Luntz et al. [63], which states that the electronic transport through
Li2 O2 at relevant current densities is governed by tunneling.
In the initial part of the charge, it was shown that the impedance was low compared
to the end of discharge at sudden death, and that Li2 O2 is oxidized already at 3.05 V,
but that this facile oxidation is limited to approximately one monolayer. Analysis of
the chemical degradation and the change in double layer capacitance indicate that the
Li2 O2 surface reacts with the electrolyte to form a SEI layer as soon as the outermost
layer is oxidized. The resistance increases as the SEI layer blocks the surface and the
voltage increases to maintain the constant current.
Three reactions were identified between 3.3 V and 3.5 V. The interval is dominated
by Li2 O2 oxidation with a small amount of electrochemical degradation and significant
chemical degradation of Li2 O2 . It is expected that the reactions in this region are a gas
evolving degradation reaction in the Li2 O2 -electrolyte interface and oxidation of another
Li2 O2 crystal plane, possibly the O-rich (1-100) plane, among others.
As charging progressed, the voltage increased significantly, whereas the resistance and
OCV were almost unchanged, and DEMS measurements identified the presence of par-
asitic reactions. This suggests that the electrochemistry changed during charge and
that the voltage increase was due to a mixed potential of parasitic reactions and Li2 O2
oxidation, established to support a constant current. This was further substantiated by
a sudden decrease in resistance between 3.5 V and 3.6 V.
Knowing the exact degradation mechanisms is crucial to improve the Li-O2 system.
From the measurements presented in this chapter, the immediate formation of an SEI
layer on the oxidized surface in the initial part of the charge is highlighted as a significant
problem that needs to be resolved before a viable Li-O2 battery can be developed, and an
analysis of the very first part of the charge might serve as a suitable screening parameter
in the search for better electrolytes.
Chapter 5
Screening for new electrolytes
The development of a stable electrolyte is perhaps the greatest challenge of making a

reversible Li-O2 system. Most electrolytes investigated are actually stable until at least
4.6 V in a pristine Li-O2 battery, but as shown in Paper II and by McCloskey et al. [41],
the presence of the discharge product Li2 O2 enable alternative degradation mechanisms
with a lower onset potential. The continuous plating and stripping of lithium on the
negative electrode and Li2 O2 on the positive electrode make the system very exposed to
degradation reactions, and a stable, flexible SEI layer that remains unchanged during
thousands of cycles is difficult to imagine if the electrolyte reacts with the Li2 O2 . More-
over, a suitable electrolyte for Li-O2 cells should have low volatility to avoid solvent
evaporation, high oxygen solubility and diffusivity to enable sufficient oxygen transport
to the air electrode, low viscosity, high conductivity and a wide electrochemical window.
Although many solvents have been investigated in this regard, none of them fulfill the
above mentioned requirements.
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
Cations Anions
18
17
16
15
14
13
12
Figure 5.1: Schematic diagram of the different cations and anions used to test IL
11 based electrolytes in Li-O2 batteries. Adapted from Paper IV.
10
9
8 47
7
6
5
Chapter 5. Screening for new electrolytes 48
Ionic liquids (ILs) have been proposed by several researchers as suitable in electrolytes
for Li-O2 batteries because of their relatively high electrochemical and chemical stability
against ·O−
2 radicals [66–76]. Ionic Liquids are a class of molten organic salts which are
liquid at temperatures around 100 ◦ C or below. They are generally comprised by an
organic cation and an inorganic anion [77–81], and they have stimulated much interest
for a variety of chemical, physical and biological processes in past few decades due
to their remarkable and unique properties such as negligible volatility, thermal and
electrochemical stability, non flammability, tunable viscosity, high ionic conductivity
and superior hydrophobicity.
In 2005, Kuboki et al. presented the first results with an IL-based electrolyte in a Li-O2
system [67], and in the past few years, IL-based electrolytes have received significant
attention as safe and environmentally friendly in Li-O2 systems due to the nonvolatile
nature. [67, 69, 72, 82, 83]. Among the most promising ILs are Pyrrolidinium and
piperidinium based cations in combination with the bis(trifluoromethanesulfonyl)imide
(TFSI− ) anion as reported by Lu et al. [84], and these ionic liquids have been the focus
of our investigations.
In this study, we investigate the rechargeability in a Li-O2 cell with ionic liquid elec-
trolytes based on N-alkylmethylpyrrolidinium (P13+ and P14+ ) and N-alkylmethylpipiridinium
(PP13+ ) cations in combination with the TFSI− anion prepared as described in Section
3.1.3. An additional number of ILs are included in Paper IV for comparison.
5.1 Investigation of ionic liquids
DEMS measurements were performed as described in Section 3.2 and the gas consump-
tion and gas evolution were quantified using both pressure measurements and mass spec-
trometry. Currents of 20 µA and 50 µA were applied with lower and upper potential
Solvent Salt OER/ORR (e- /O2 )dis (e- /O2 )ch CO2 /ORR H2 /ORRb
P14TFSI LiTFSI 0.628a 2.03 2.75 0.00 0.01
P13TFSI LiTFSI 0.617a 2.00 3.09 0.00 0.01
PP13TFSI LiTFSI 0.402a 2.09 4.03 0.03 0.18
P13FSI LiFSI 0.193a 2.51 3.40 0.04 0.02
DME LiTFSI 0.796 (0.663a ) 1.99 2.62 0.08 0
Table 5.1: Differential Electrochemical Mass Spectrometry (DEMS) results of dif-

ferent ionic liquid and DME based electrolytes. a Based on charge to 4.2 V. b The H2
intensities were not calibrated and can only be used to compare H2 evolution between
the different electrolytes.
limits of 2.2 V and 4.2 V, respectively, and the cells were all tested at room tempera-
ture (21 ◦ C- 24 ◦ C). From the DEMS measurements of the IL-based electrolytes, the
following numbers were determined: (e- /O2 )dis , (e- /O2 )ch , OER/ORR, CO2 /ORR, and
H2 /ORR. The numbers are summarized Table 5.1 and the data for the DME-LiTFSI
electrolyte presented in Table 4.1 is included for comparison. As discussed in Section
2.1.1, these key characteristics are crucial to determine if a Li-O2 system is truly re-
versible [14].
4.0 24 P14TFSI
a 23 22– 5021µA 20 19 18 DME
a 17 –16 15 14 13 12 11 10 9
250 µA 8 7
ΔnO2 [µmol/h] Voltage [V]
4.0
18 Charge Charge
3.0 3.0
17 Discharge Discharge
16 2.0 2.0
0 200 400 0 200 400
15 10
14
b 10
b
13 5 5
12
0 0
11 0 200 400 0 200 400
2 e -/O 5.0
10 1.0 2 e-/O2
2
Δn [µmol/h]
c O2
c O2
9 0.5 2.5 CO2
8
CO2 H2 H2
7 0.0 0.0
0 100 200 300 0 100 200 300
6 Q [mAh/gC] Q [mAh/gC]
5
Figure 5.2: (a) Galvanostatic discharge-charge curves for cells with P14TFSI (50 µA
4
discharge-charge) and DME (250 µA discharge-charge) based electrolytes. (b) Oxygen
consumption (measured using pressure decay) during discharge and evolution (mea-
sured using3DEMS) during charge. (c) Gas evolution rates for O2 , CO2 , and H2 .
Adapted from Paper IV.
2
1 a comparison between the first galvanostatic discharge-charge cycle
Figure 5.2 shows Template used to crop di
of a DME-LiTFSI electrolyte and a P14TFSI-LiTFSI electrolyte and three important

features are noted. First, the discharge capacity of the two electrolytes are comparable
and the difference might be due to a difference in the current density. Second, the
charge voltage is not increasing as rapidly in the experiment with P14TFSI electrolyte
compared to the one with DME, and more than 15 % of the capacity is charged at a
voltage below 3.15 V with P14TFSI compared to only 4.3 % with DME as shown in
Section 4.2. Third, both electrolytes show e- /O2 ratios close to 2 during both discharge
and charge.
24 23 22 21-
20 19 18
5.05,017 16 15 14- 13 12 114,5 10 9 8 7 6
-
5 4 3 2 1
4.0 4,5
4,0
4.04,0
Voltage [V]
18 4.04,0 rge Cha

rge
3,5 Cha
ge
3,5
ar
Ch
17 Charge
E (V)
 (V)
3,5
Charge Charge
 (V)
3.0 3.0
3,0
3,0 3.03,0
16 Discharge 2,5
Discharge Discharge
2,5 2,5
Discharge Discharge Discharge
15 2.0 2,0
0 100 200 300 2.02,0
400 0 20 40 60 80 2.02,0
100
0 200
Q (mAh/g) 400 0 50
Q (mAh/g) 100 0 0 100
200
200
Q (mAhg-1)
300
400
400
14
1010
ΔnO2 [µmol/h]
2,5 12
13 8 e l 22,0 1010 e
arg µm
ol
arg mo sch .76
sch .25 µ e
arg mo
l 8 Di 1
Di sch .13 µ =1
=9
n O2 (µmol)
n O2 (µmol)
n O2 (µmol)
6 1,5 Di
12 54 nO 2
=
n O2
2
54
6
n O2
ol
11,0
11 2
n O2
e
a rg 6 µ
Ch 7.1
=
m
0,5 2
a rg
e
µm
ol
00 00,0 Cha
rge
0.68
µmo
l
0 0 Ch 4.72
=2
10 n O2=
-2
nO
0 0 100
200 200 300
400 0
400 0 20
40 40 60
80
80
00 100
100
200
200
300
300
9 Q (mAh/g)
1.01,0
Q (mAh/g)
0.4 0,4
Q (mAh/g)
1.01,0 2 e2e-/O
-
/O 2
2
22e e/O-/O2 2 e2e-/O
- -
Δn [µmol/h]
2 /O
2
8
2
0,8
0,8 O2 0,3
Charge O2 0,6 OO2 2 H2
n (µmol/h)
n (µmol/h)
7 0.50,6
n (µmol/h)
Charge
0.50,4 O2 O2 0.2
0,2
Charge
0,4
6 0,2 0,2 H2
0,1
H2 COCO2 H22
CO2
H2 H2
CO
CO2
5 0.00,0
0.00,0
CO2
0.0
0,0
2
00 50
100
100 150
200
200 250
300
300
0
0 10 20
25 30 40
50
50
00 50 100
100 150 200
200 250
4 Q (mAh/g) Q (mAh/g)
Q (mAh/g)
Q [mAh/gC] Q [mAh/g C] used to crop directly
Template Qin[mAh/g C]
LaTex with trim-command
3
2 Figure 5.3: (a) Galvanostatic discharge-charge curves for cells with P14TFSI (50 µA
discharge-charge) and PP13 (20 µA and 50 µA discharge-charge) based electrolytes. (b)
1 Oxygen consumption (measured using pressure decay) during discharge and evolution
(measured using DEMS) during charge. (c) Gas evolution rates for O2 , CO2 , and H2 .
Figure 5.3 shows a comparison between the first galvanostatic discharge-charge cycle of
electrolytes based on P14TFSI and PP13TFSI, respectively. PP13TFSI did not perform
as well as P14TFSI. At 50 µA, the battery based on PP13TFSI was not able to sustain
the current for long, showing a premature cell death, and only little oxygen was evolved
during charge. At a lower current of 20 µA, a capacity similar to P14TFSI at 50 µA was
obtained. Although the electron count per oxygen is near 2.09 during discharge, this
electrolyte showed poor reversibility with an OER/ORR ratio of 40 % and an (e- /O2 )ch
ratio of 4, which is quite similar to result obtained by McCloskey et al. [56].
Figure 5.4a shows the first galvanostatic discharge-charge cycle curve for electrolytes
based on P13TFSI and P13FSI. It is immediately seen that the FSI gives a larger
discharge capacity than all of the other electrolytes, but the rechargeable capacity is very
low and the DEMS measurements show that the extra capacity is due to a significant
electrochemical degradation during discharge. During charge, P13FSI is showing an
initial 2 e- /O2 ratio, suggesting Li2 O2 oxidation, but after charging approximately 10 %
of the full discharge capacity, the oxygen evolution decreases and the potential increases
4.0
a P13TFSI
24 50 µA20 19 18 a
23 22– 21
4.0
17 16P13FSI
15 –14
50 µA
13 12 11 10 9 8 7 6
Voltage [V]
18 Charge
17 3.0 Charge 3.0
Discharge Discharge
16
2.0 2.0
15
ΔnO2 [µmol/h] 0 200 400 0 200 400
14 15 b b
10
13 10
12 5
5
11 0 0 Charge
10 0 200 400 0 200 400
9 1.0 c 2 e-/O2 1.0 c 2 e-/O2
Δn [µmol/h]
O2
8
O2
0.5
7 0.5
6 CO2
CO2 H2 0.0
0.0 H2
5 0 100 200 300 0 50 100
4 Q [mAh/gC] Q [mAh/gC]
3 (a) Galvanostatic discharge-charge curves for cells with P13TFSI (50
Figure 5.4:
µA discharge-charge) and P13FSI (50 µA discharge-charge) based electrolytes. (b)
2
Oxygen consumption (measured using pressure decay) during discharge and evolution
1 DEMS) during charge. (c) Gas evolution rates for O2 , CO2 , and H2 .
(measured using
Template used to crop directl

rapidly. P13TFSI show a behavior very similar to P14TFSI, but with a slightly higher
(e- /O2 )ch of 3 during charge.
Figure 5.5 shows XRD measurements of discharged cathodes for pure IL and DME based
electrolytes. All cells were discharged to 2.2 V at a current of 20 µA. The six major
Li2 O2 peaks are clearly visible in all samples except in the cathode tested with an FSI−
based ionic liquid. Most of the other peaks are from the PTFE binder.
Figure 5.6 shows the e- /O2 data through six cycles at 50 µA between 2.2 V and 4.2 V for
electrolytes based on P14TFSI and P13TFSI. It is seen that the amount of electrochem-
ical degradation reactions increases during both discharge and charge as the batteries
are cycled. This shows that even though the first cycle seems promising, the system is
not stable.
24 23 22 21 20 19 18 17 16Li215
O2 14 13 12 11 10 9 8 7 6 5 4
18 Li2CO3
LiOH
17
Intensity
16 P13FSI
P13TFSI
15 P14TFSI
14 PP13TFSI
13 DME
12
11 30 40 50 60 70 80
10 2θ
9 patterns of discharged XC72 carbon cathodes with various elec-
Figure 5.5: XRD
trolytes. Adapted from Paper IV.
8
7 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
18
6
17
5
16
15
4
14
13
3
2
12
11
10
1 Template used to crop directly in LaTex with
9
8
7
6
Figure 5.6: Cycling behavior of Li-O2 batteries with P14TFSI and P13TFSI based
5
electrolytes. Adapted from Paper IV.
4
3
5.1.1 Comparison
2 between ionic liquid and organic solvent
The measurements presented in Figure 5.2 and in Table 5.1, show that the P14TFSI and
DME are relatively close to each other in terms of electrochemical stability and discharge
capacity. It is also noted that a larger part of the reaction occur at low potentials using
the P14TFSI IL compared to DME. Due to the tunability of ILs, it therefore seems
likely that some IL-based electrolytes might be able to compete with organic solvents in
terms of stability, but the ILs often have a higher viscosity and a lower oxygen solubility,
which limits the discharge current, and this should of course be taken into account as
well.
5.1.2 Choice of anion
From the measurements presented in Figure 5.4 and in Table 5.1, it is clearly seen that
the TFSI− anion is much better that the FSI− . The combination of DEMS and XRD
measurements suggests that only a small fraction of the discharge product in cells with
FSI− is Li2 O2 , whereas Li2 O2 is the primary discharge product using TFSI− . We are
currently analyzing the measurements and collecting supporting data to determine the
underlying reasons for the difference. So far we are looking primarily at the oxygen
solubility, the amount of flourine in the anion and the size of the anion.
5.1.3 Choice of cation
From the measurements presented in Table 5.1, it is seen that all of the electrolytes
having a TFSI anion showed an e- /O2 ratio of approximately 2 during discharge, but
during charge, the different cations resulted in different electrochemistry. The PP13TFSI
showed an OER/ORR of 0.40 at 4.2 V, whereas P13TFSI and P14TFSI showed a value of
0.62 - 0.63, which is close to DME at 4.2 V. The amount of electrochemical degradation
is also different, from 4.0 e- /O2 in PP13TFSI to 2.8 e- /O2 in P14TFSI. We are currently
analyzing the measurements and collecting supporting data to determine the underlying
reasons to why PP13TFSI is less stable during charging than P13TFSI and P14TFSI.
One of the major differences may be the concentration of impurities such as water
as discussed below. We are also looking into the transport limitation in PP13TFSI
indicated by the low capacity at 50 µA.
5.1.4 Impurities
One of the problems with commercially available ILs is the purity. The ILs some-
time contain a significant amount of water that is difficult to remove, and which may
have a significant impact on the electrolyte stability. In aprotic Li-ion electrolytes like
ethylenecarbonate (EC), it has been shown that a change in impurity level from 0.09 %
to 0.021 % increased the oxidation potential with 0.6 V from 4.87 V to 5.50 V [85], and
it is likely that impurities will have a similar effect on the ionic liquids. From Figure 5.3c
it is seen that PP13TFSI is developing a significant amount of H2 during charge. The
values are not calibrated, but it is a clear indication of water in the sample. This could
therefore be the reason for the difference between the cations investigated. Looking at
the cycling experiment in Figure 5.6, an increase in the e- /O2 ratio is observed. In paper
V we saw a similar situation where impurities from the first cycle affected the following
cycles. The increase in e- /O2 ratio during cycling agree well with results published by
Piana et al. [74] using P14TFSI based electrolytes.
5.1.5 Summary
A number of ILs have been tested with DEMS to investigate the electrochemical sta-
bility in the Li-O2 system. It was shown that the extent of rechargeability depends on
the choice of cation and anion. Unfortunately, none of the studied IL-based electrolytes
behaved as ideal electrolytes in a Li-O2 battery. It was found that the anion TFSI−
was much more stable in the Li-O2 system than the FSI− anion and that ILs based on
s pyrrolidium cation and the TFSI anion (P13TFSI and P14TFSI) had better recharge-
ability below 4.2 V (OER/ORR > 60%) compared to the PP13TFSI based electrolyte,
but further investigation is needed to determine whether the difference between the
cations is related to the has to do with the type of cation or the level of impurities.
The work clearly showed the benefit of the DEMS technique to qualify the discussion
of different electrolytes, as a comparison between the voltage profiles and discharge
capacities is often too simplified to distinguish the complicated reaction mechanisms
inside the Li-O2 battery during charge.
Chapter 6
Commercial applications of
lithium-air batteries
As it is clear from the results presented in Chapters 4 and 5, significant challenges need
to be overcome to enable a commercialization of the Li-O2 technology. It is, however,
relevant to consider how the technology benchmark against competing technologies, if
the degradation problem is solved, as this may decide whether it is worth the risk to
pursue this technology. A part of this is an economic assessment of the system. Recently,
two assessments of a Li-O2 system have been presented [13, 86]. Both show that Li-O2
may be significantly better than current state Li-ion batteries on a system level, but
that new types of Li-ion batteries under development have similar energy densities and
may be associated with less risks of failure in the development phase.
An important input to such calculations is a technical description of the system. This

chapter focuses on two technical issues related to a commercialization of the Li-O2
battery: 1. The air purification system to enable an open Li-air system and 2. The
determination of the State-of-Charge in a Li-O2 battery. The last part is closely related
to Paper III. The battery used for calculations in this chapter has an energy content
of 100 kWh, which is more than any EV battery on the market today. This size is in
agreement with previously published system calculations [13, 86], and further details
about the assumptions are found in the Supporting Information of Paper III.
6.1 Impedance-based management of Li-O2 batteries
A battery management system (BMS) is often needed to ensure safe and reliable perfor-
mance of any type of battery, and to predict the remaining capacity. The two important
55
Chapter 6. Commercial applications of lithium-air batteries 56
BMS functions are the calculation of the remaining capacity in the battery, the state
of charge (SoC), and the health of the battery; generally combining capacity retention
on cycling and power capability. The health parameter is referred to as the state of
health (SoH). The SoC and SoH is calculated by the BMS to predict the performance
under different scenarios, to enable optimized usage of the remaining capacity, and even
preventing dangerous situations that may occur if the battery powered device, e.g. an
electric vehicle, is suddenly without power. The SoC can be calculated in several ways,
with the most simple being a comparison of terminal voltage to previously recorded cy-
cling data of cell voltage and battery capacity. A slightly more complex approach is to
continuously monitor and integrate the current over time, also known as coulomb count-
ing. The coulomb counting method accumulates errors if calibration is not performed,
as it relies on the accuracy of the measurement and several methods for mitigating
this have been proposed for lead-acid and lithium-ion batteries [87]. In Li-O2 batteries,
main electrochemical process is unchanged during discharge and charge (assuming no
degradation). This means that the open circuit is constant and does not change with
SoC as shown in Figure 4.6a. Furthermore, constant current measurements show a flat
discharge plateau in a majority of the discharge period as shown in Figure 4.2. It has
also been observed that the current densities have a significant impact on the onset of
sudden death and thus also available capacity, due to the increase in required charge
transport through the poorly conducting Li2 O2 layer [88–90]. Since the coulomb count-
ing method relies on a known total capacity to predict sudden death, the method is not
well suited for predicting the remaining capacity in these batteries.
This section describes a method to accurately predict the SoC of Li-O2 batteries using
a single frequency EIS measurements to estimate the remaining capacity as well as the
degradation of the battery materials. EIS is used in many systems to perform in situ
Proposed Metal-O2 SOC calibration using EIS 2015-01-07
Battery capacity
Initial setup
SOC
Normal Coulomb
calculation
operation counting
routine
Resting until Perform EIS

OCV condition measurement SOC calibration
Figure 6.1: Working principle of the proposed impedance-based BMS system for
Li-O2 batteries.
determination of certain parameters like degradation of secondary Li-ion batteries [91–

93], capacity fading of Li-S batteries [94, 95], and discharge mechanisms for Si-air [96],
and the following work shows that EIS will be very suited for battery management of
Li-O2 batteries as well.
6.1.1 The working principle of the method
In Section 4.4 it was shown that the low frequency (<1 Hz) contribution of the impedance
is related to the positive electrode and an EIS measurements in this frequency regime
hold information about the double layer capacitance of the electrode. The capacitance
can be calculated under constant load, but the value depends on the current density due
to kinetic effects, and the true value is obtained at OCV. Since the measurements under
load introduces new variables, this work focus on measurements at OCV. At OCV, the
oxygen reduction reaction and Li2 O2 oxidation reactions are very slow, and if the EIS
excitation signal is sufficiently small, the impedance signal from the positive electrode
becomes capacitive at the relevant frequencies (see insert in Figure 6.3). This means
that the capacitance can be determined by the simple expression shown in Equation
(6.1) [18].
−1
C= (6.1)
2πf · Zim
where f is the AC perturbation frequency and Zim is the imaginary part of the corre-
sponding impedance. It is important that other impedance contributions from processes
with similar time constants do not overlap at the frequency used for the calculation. A
frequency of 10 mHz was chosen because it was the highest frequency (and hence shortest
measurement time) with the main contribution from the positive electrode capacitance.
During discharge, Li2 O2 is deposited on the carbon surface in the positive electrode
and as Li2 O2 is a dielectric, the capacitance of the surface will change as described in
Section 2.2.2, and the proposed SoC estimation is based on following the decrease in
capacitance as the Li2 O2 is deposited. As it will be shown later, the correlation between
the capacity and capacitance follow the same trend in all experiments. The trend is an
empirical function given by

Q
C = C0 − p2 1 − exp (6.2)
p1
where C0 is the initial capacitance, Q is the capacity, and p1 and p2 are refined param-
eters determining the exponential shape.
100 a
Current [mAgC-1]
b c d
0
-100
-200
Current
-300 EIS measurement
-400
0 2 4 6 8 10 12
Time [h]
0 e
Current [mAgC-1]
-100
-200
-300
x11
-400
0.00 0.05 0.10 0.15 0.20 0.25 0.30
Time [h]
Figure 6.2: Load profiles used in the experiments to test the applicability of SoC
determination based on the measured capacitance. (a) Discharge of 130 mA/gC to 2
V interrupted by impedance measurements every 100 mAh/gC . 13 mA/gC , 39 mA/gC
and 390 mA/gC were tested in the same way. (b) staircase discharge profile with 33
mAh/gC at 39 mA/gC , 33 mAh/gC at 390 mA/gC and 33 mAh/gC at 39 mA/gC
between each impedance measurement. (c) discharge of 130 mA/gC to 100 mAh/gC
followed by a 3.2 V voltage-limited charge of 130 mA/gC to reduce electrochemical
decomposition to a minimum. (d) Capacity limited galvanostatic cycling at 130 mA/gC
limited to 65 mAh/gC . The charge was limited to 4.5 V to avoid severe electrolyte
decomposition. (e) Drive cycle from ISEA-RWTH Aachen recorded from a Fiat eCity,
scaled to a maximum current density of 390 mA/gC . Reprint from Paper III.
6.1.2 Testing the method
Figure 6.2 shows the different current densities used to test the methods. The current
densities were selected on basis of a future scenario, where electric vehicles will be
powered by Li-O2 batteries as the only source of power. The scaling of the peak and
average current densities are based on an electric vehicle with a 100 kWh battery, a
sustained high power of 55 kW and a peak power of 105 kW, which matches most
electric vehicles today in terms of peak power and is superior in terms of capacity [97].
Details on the calculation are found in the Supporting Information of Paper III.
Figure 6.3a shows the imaginary part of the impedance spectrum at low frequencies
during discharge. It is seen that the imaginary impedance is increasing, meaning that
the capacitance decreases. The data points at 10 mHz is marked and the correspond-
ing capacitance is listed in the figure. Figure 6.3b shows the calculated capacitances
as a function of capacity, and while the potential is almost constant, the capacitance
decreases throughout the discharge. At the end of discharge, when the battery reached
2.2 V, the capacitance was 50 % of the initial value.
Capacitance 580 mAhgC-1

3 5.3 Fg-1
C 490 mAhgC-1 1.5 10
2.69 V 2.70 V
-Im(Z) [ΩgC]
390 mAhgC-1
9
-Im(Z) [ΩgC]
290 mAhgC-1 1 2.69 V
7.3 Fg-1 190 mAhgC-1
2 C
10 mHz 2.67 V
C]
1.5
8
-Im(Z) [ΩgC]
C [Fg-1
8.5 Fg-1
C 100 mAhgC-1
9.9 Fg-1
C 0 mAhg-1
C
0.5 1
7 2.63 V
1 0
0.5
0 0.5 6 0
Re(Z) [ΩgC] 0 0.5
Re(Z) [ΩgC] 2.32 V
5
Fi
1.75
0 1.50
0.01 0.03 0.1 0.3 1 0 100 200 300 400 500 600 1.25
Im(Z) [$Omega$g$_C$
Frequency [Hz] Capacity [mAhg-1
C]
1.00
0.75
0.50
0.25
Figure 6.3: Left: The imaginary part of the impedance spectrum at low frequencies 0.00
0.00
during discharge. It is seen that the imaginary impedance is increasing, meaning that
e(Z
the capacitance decreases. Right: Capacitance as a function of the capacity for a

constant current discharge at 130 mA/gC . Insert shows a Nyquist plot of the impedance
measurement. The voltage labels are measured during discharge prior to the impedance
measurements. The OCV was 2.83 V during the entire discharge. Reprint from Paper
III.
10
8
C]
C [Fg-1
Constant current
6 13 mAg-1
C
39 mAg-1
C
4 Variable current 130 mAgC-1
Drive cycle 1 Staircase 130 mAgC-1
Drive cycle 2 Charging 390 mAgC-1
2
0 400 800 1200
-1
Capacity [mAhg C]
Figure 6.4: Overview of the change in capacitance measured at OCV between different
types of discharge. Reprint from Paper III.
Figure 6.4 shows the change in capacitance as a function of capacity for all measure-
ments. It is seen that the capacitance decrease is similar in nature for all measurements,
but that the capacity of the different batteries vary significantly, up to a factor of 14.
The data has been fitted to Equation (6.2) and the fitted parameters are presented in
Table 6.1. Three observations are made based on these values: (1) all initial capacitances
fall within 10.0 Fg-1 -1
c ± 0.4 Fgc , (2) the trend of the decreasing capacity is very similar
for all measurements performed at different current densities varying a factor of 30, and
at both dynamic and constant current loads, and (3) all capacitances have decreased to
approximately 50 % in the end of discharge, except the high current measurement at
390 mA/gC and the measurement including charging. Both of these exceptions will be
discussed further below.
The purpose of a BMS is to predict the remaining capacity, and in the following we have
Normalized capacitance
1.0 a 1.0 b
0.8 0.8
13 mAgC-1 130 mAg g-1
C
39 mAgC-1 Staircase
0.6 130 mAg-1C
0.6 Charging
130 mAg-1C Drive cycle 1
390 mAgC -1
Drive cycle 2
0.4 0.4
0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0.4 0.6 0.8 1.0
Normalized capacity Normalized capacity
3.2
13 mAgC-1 130 mAg-1
C c 13 mA/g1.0
C d
3.0 39 mAgC-1 130 mAg-1
C 39 mA/gC
Voltage [V]
-1
390 mAgC 129mA/g0.9
C
2.8 129 mA/gC
387 mA/gC
2.6 0.8
Li2O2
2.4 13 mAgC-1 Li2CO3 Current collector
39 mAgC-1
2.2 129 mAg-1 0.7 Li2O2
129 mAg-1
C Cathode
Discharged
Cathode
Charged Discharged
2.0 387 mAgC
C
-1
1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.6 0 5 10 15
Normalized capacitance Cycle no.
Figure 6.5: Results from experiments on Li-O2 batteries. The data is labeled accord-
ing to current profiles in Figure 6.2. Capacitance is normalized to the initial value before
the discharge, and capacity is normalized to the value at 2.2 V. (a) Constant current
load at four current densities: 13 mA/gC , 39 mA/gC , 130 mA/gC , and 390 mA/gC , (b)
dynamic current profiles according to Figure 6.2 and data from a 130 mA/gC constant
current experiment, (c) Discharge voltage and capacitance from constant current ex-
periments, showing the relation between current density and capacitance. (d) cycling
experiment over 20 cycles showing decreasing capacitance as a function of cycle num-
ber. The insert shows the suggested reason for a decrease in capacitance, due to an
accumulation of the degradation product Li2 CO3 during cycling. Reprint from Paper
III.
therefore normalized each discharge capacity to the capacity at 2.2 V, to investigate

how well the remaining capacity is predicted by the capacitance. Even though all initial
capacitances are very similar, we have chosen to normalize these values also, as this
is what would be done in an actual BMS, where the capacitance may vary between
batteries and may change over time due to degradation.
Figure 6.5a and Figure 6.5b show the data from Figure 6.4 using this normalization to
investigate the dependence between the capacitance and the SoC at the different current
profiles seen in Figure 6.2. Figure 6.5c depicts how the voltage during discharge relates
to the capacitance for different current densities, and it is evident that the current
densities impact the discharge capacity, and thereby the capacitance at 2.2 V. Figure
6.5d shows the normalized capacitance in the fully charged state, as a function of cycle
number. Upon cycling, the capacitance decreases almost linearly to approximately 65
% after 20 cycles.
C2.2V
C0 p1 p2 Q2.2V C0
Fg-1
c mAh/gC Fg-1
c mAh/gC -
13 mA/gC 10.3 518 820 1001 53%
39 mA/gC 10.0 300 467 767 44%
130 mA/gC 9.83 169 76.9 700 52%
130 mA/gC 10.1 148 90.9 599 50%
390 mA/gC 10.2 67 465 117 78%
Staircase 10.1 241 610 519 54%
Charging 10.3 377 44.8 1614 69%
Drive cycle 1 10.0 188 54.9 812 59%
Drive cycle 2 9.64 124 21.8 652 56%
Table 6.1: Experimental values and parameters obtained by fitting experimental data
to Equation (6.2). C0 is the initial capacitance, Q2.2V is the discharge capacity at the
2.2 V cut-off, CC
2.2V
0
is the fraction of capacitance at the 2.2 V cut-off to the initial
capacitance.
6.1.3 State of charge
The constant current experiments presented in Figure 6.3 show how the capacitance
changes as a function of capacity. The correlation is decreasing and at the end of
discharge the capacitance is 50 % of the initial value. Assuming that the change is
due to a uniform layer of Li2 O2 , this corresponds to a layer thickness of ∼5 nm at
these current densities, and as discussed in Section 6.1.1, this is in line with the general
understanding of the sudden death mechanism [90]. The voltage during discharge is
shown in Figure 6.3 as labels on the capacitance data points, and remains unchanged
at ∼2.70 V between 100 % and 33 % SoC, whereas the capacitance decreases as the
battery is discharged. Seeing that all measurements behave similarly, this shows that
the capacitance holds information about the SoC, that is not possible to obtain using
With Calibration
a capacity with calibration voltage based measurements.
Without calibration
Actual SoC
100 a With calibration 500100 b BMS with calibration
Improvement [%]
Usable SoC [%]
BMS without calibration

80 400 80 Calibration
SoC [%]
60 300 60
40
40 200
20
20 100 0
0 0
0 100 200 300 400 6am 6pm 6am 6pm
Energy consumption since full charge [kWh] Time
Figure 6.6: (a) The combined accumulation of error in the estimation of SoC, in-
cluding both the coulomb counting error and the needed capacity reserve. (b) Scenario
showing the difference in estimation of useable capacity between BMS systems with
and without calibration based on capacitance.
The accuracy of the SoC estimation from capacitance is determined by evaluating the
uncertainty in the capacitance to capacity correlation from the experiments with the
130 mA/gC constant current load and the drive cycle load shown in Figure 6.5b. To
test the effect of the calibration, we have designed a BMS system based on the following
assumptions:
1. Due to high current dependency on capacity, a reserve of 10 % SoC is needed,

resulting in 90 % useable capacity.
2. Coulomb counting has a 3 % accumulation of error.
3. The SoC is calibrated at end of full charge.
4. Charging is performed at 10 kW and fast charging is not considered.
If no SoC calibration is performed, the coulomb counting error and the need for a
capacity reserve will result in a significant decrease of useable capacity as the battery is
discharge and charged. During discharge, the accumulated error corresponds to 3 % and
during charge, the error is 13 %, due to the additional need of a 10 % capacity reserve
for the following discharge. Figure 6.6a shows the effect of the combined error of 16 %
on a 100 kWh battery as a function of energy consumption of up to 400 kWh without
fully charging the battery. When the battery is fully charged, a BMS would be able to
correct the estimation of useable capacity to the nominal capacity of the battery without
the capacitance calibration, and thus reset the useable capacity to 90 %. Without the
SoC calibration, the useable SoC could decrease to 26 % over an accumulated energy
consumption of 400 kWh, whereas the SoC calibration based on capacitance would be
able to keep the SoC estimation at 90 %, thus maximizing the useable capacity.
Since the capacity of future Li-O2 batteries is expected to increase, we postulate a use-
scenario for the batteries based on ∼10 kW charging power available and no need for fast
charging with daily use. This scenario is illustrated in Figure 6.6b, where the battery is
either discharged, charged or at rest. The calibrations performed in Figure 6.6b enables
the BMS to more accurately predict the remaining capacity, whereas the estimated
capacity without calibration would become less than zero (shown on the figure as circles
when the SoC is crossing the 0 % threshold), warning the vehicle or driver to stop. The
capacitance calibration is less accurate in the less critical beginning of discharge, due to
the slope of the capacitance correlation to the capacity, and gradually more accurate as
the battery is discharged. Figure 6.6b shows that the calibrations performed below 50
% SoC have the possibility to initially correct the SoC by ∼10 %, and when performed
multiple times, able to further minimize the uncertainty of the SoC estimation, resulting
in periods with an uncertainty of less than 5 % SoC.
6.1.4 Power capability
The current density of 390 mA/gC is only able to provide 12 % of the capacity compared
to the current density of 13 mA/gC . This indicates that the available capacity depends
explicitly on the current density. For correct estimation of the available power, the
internal resistance is usually used to predict if the terminal voltage of a battery will
exceed the limit during high power demands. For Li-O2 batteries, this is not possible
since the resistance is almost unchanged during the entire discharge as shown in Paper I.
Previously, we saw that the capacitance of the 390 mA/gC experiment did not decrease
to ∼50 % of the initial capacitance. If we look at the voltage during the constant current
experiments (Figure 6.5c), we see that the high current density experiments (130 mA/gC
and 390 mA/gC ) have significantly higher overpotentials, thus the 2.2 V limit is reached
prematurely. From the measurements, it seems that current densities of 390 mA/gC are
only supported at capacitances above 78 % of the initial capacitance, while it is possible
to use 130 mA/gC until a capacitance of 50 %. This shows that the capacitance can
also be used to estimate the maximum power that the battery is capable of delivering
at the given state.
6.1.5 State of health
The cycling experiment (Figure 6.5d) shows how the capacitance of a fully charged
positive electrode decreases as a function of cycle number. From several papers [56,
62, 98, 99] it has been shown that Li2 CO3 and similar species are accumulating upon
charging and are immobilized in the positive electrode (see Figure 6.5d, insert). The
presence of Li2 CO3 in the cathode has two main effects on the battery performance.
First, the amount of Li2 CO3 will increase for each cycle, resulting in a decrease of the
available active area for Li2 O2 deposition. This results in a lower discharge capacity
for the battery. Second, the inclusion of Li2 CO3 in the deposited Li2 O2 decreases the
charge transport through the Li2 O2 , thus increasing the required overpotential [61]. It
was shown in Figure 2.6 that even a few angstrom of Li2 CO3 will decrease the capaci-
tance significantly, and using the capacitance in the fully charged state, it is possible to
track the degradation of the positive electrode, and a BMS can use this information to
determine the SoH parameter related to capacity retention, and thereby the cycle life
of the battery.
6.1.6 Further development of the model
With more knowledge about the system, it is expected that the time at OCV can be
reduced significantly and that measurements can be made even while discharging or
charging the battery. This could be done by employing impedance measurements in
the time domain to calculate the capacitance. In this case, it might even be possible
to perform SoC calibrations while driving. Another direction would be to implement
an adaptive state estimation algorithm, akin to the work done by Fleischer et al. [100],
using the capacitance measurements to improve the calibration uncertainties.
The charging experiment shown in Figure 6.5b has a much larger discharge capacity
compared to the other experiments as seen in Table 6.1 (1614 mAh/gC vs 599 mAh/gC
for the 130 mA/gC experiment), and we suspect that this is due to the accumulation
of Li2 CO3 and similar species. The reason for the capacitance not decreasing to more
than 69% is not fully understood and further studies of cycling effects on Li-O2 systems
is needed.
Many metal-O2 systems show the same type of discharge and charge curves as the Li-O2 ,
but further studies on other metal-O2 systems are needed to conclude if the proposed
method is applicable to these systems. A likely candidate for further studies is the Na-O2
battery, where the discharge product, NaO2 is considered to have surface conductivity.
The conductive nature of the NaO2 is expected to cause an increase in capacitance as
the battery is discharged, reflecting an increase in surface area as the NaO2 cubes are
grown [101].
6.1.7 Summary of impedance based management of Li-O2 batteries
A method for estimating the remaining capacity, power capability and cycle life of Li-O2
batteries has been proposed and verified through a number of tests. Experiments showed
that the capacitance of the positive electrode decreased exponentially during discharge,
and that it was possible to improve the prediction of the remaining battery capacity sig-
nificantly based on a single frequency measurement of the positive electrode capacitance.
In a typical scenario, a single SoC calibration was able to improve the available SoC by
more than 10 % of the full battery capacity, by minimizing the uncertainty of the SoC.
The capacitance was also used to estimate a degradation of the positive electrode in a
Li-O2 battery cycled 20 times. This makes the method applicable not only for electric
vehicles, but for batteries in a large range of electrical devices, as the measurements can
be performed when needed, thus maintaining a high level of accuracy for the estimation
of remaining capacity and state-of-health. The approach is furthermore expected to be
transferable to other metal-O2 systems.
6.2 Separation of O2 from air
A true Li-air system rely on the use of ambient air. Dry atmospheric air contains 78%
N2 , 20.9% O2 , 0.93% Ar and 400 ppm CO2 , and in addition to this, the air typically
contains 1% - 3% H2 O [102]. To develop a true lithium-air battery, it is therefore
important to investigate how the non-oxygen gases affect the electrochemistry. Luntz et
al. have shown that nitrogen and argon are inactive in the battery, whereas both H2 O
and CO2 have significant impact on the electrochemistry [14]. Meini et al. [103] have
shown that even a small amount of H2 O is able to cause a 10-fold increase in discharge
capacity. This effect has been investigate further by Aetukuri et al. [58], proposing that
water increase the solubility of LiO2 , to enable a solution-mediated growth of toroidal
shaped Li2 O2 particles. While improving the discharge capacity, the presence of water,
unfortunately, also speed up the degradation reactions inside the battery [58].
2.80
24 23 22 21
3.0 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
18 50 % CO2 18
4.5
17 2.75 1 % CO2 17
16 2.5 0 % CO2 16
4.0
Voltage [V]
15 2.70 15
Voltage [V]
14 14
13
2.65 2.0 13 3.5
12 0 500 1000 1500 12
11
2.60 11 3.0
10 10 50 % CO2
9 9 2.5 1 % CO2
8
2.55 8 0 % CO2
7 7
2.50 2.0
6
0 5 10 15 20 25 30 35 40 6 0 10 20 30 40 50
5
Capacity [mAh/gC] 5 Capacity [mAh/gC]
4 4
3 3
2 Figure 6.7: The effect of adding CO2 impurities
2 to the oxygen gas during discharge
1 (left) and charge (right). Adding only 1% CO2 1increases the overpotentials significantly
jonn_s025 XC7 _CV Template used to crop directly in LaTex with trim-command
jonn_s025 XC7 _CV Template used to crop directly in LaTex with trim-command
during both discharge and charge. Reprint from Paper V.
A similar effect is observed with CO2 [53, 104, 105]. In correspondence with these
observations, Figure 6.7 shows that both the overpotentials and the capacity increase
in a mixture of O2 and CO2 compared to a system with pure oxygen. If the CO2
content becomes too high, the discharge capacity decreases significantly. In Paper V
and [61], it is suggested that the increase in overpotential is caused by an increased
amount of Li2 CO3 -like species in the Li2 O2 , and the higher potentials will lead to further
degradation in the battery. The mechanism causing the increase in discharge capacity
is not fully understood, but in Paper V, Mekonnen et al. suggest that the formation
of Li2 CO3 -like species at the step valley sites of the Li2 O2 surface will alter the shape
and growth directions of the Li2 O2 in a beneficial way, but as with H2 O, the effect is
also expected to increase degradation through cycling. From a current perspective, it is
therefore crucial to remove H2 O and CO2 from the gas stream.
6.2.1 Previous Li-O2 and Li-air strategies
Several strategies have been pursued to avoid impurities in the batteries. Three of
the most investigate strategies are: 1. passive hydrophobic membranes on top of the
air electrode, 2. removal of impurities from the gas and 3. on-board oxygen storage.
Girishkumar et al. discussed that it is not possible to use hydrophobic membranes, as
the O2 permeability will always be higher than the H2 O permeability [10], and Zhang
et al. have tested such a membrane system with limited success [106]. In 2009 Excel-
latron Solid State LLC filed a patent describing the use of H2 O and CO2 scrubbers as
absorbents [107]. Since then, Bosch and others have proposed methods to reduce mois-
ture from the inlet gas either by using a regenerative dehumidifier based on a desiccant
[108–111], or by using a selectively permeable membrane [112], and a regenerative CO2
absorption system that selectively absorbs CO2 over O2 has been described in [113]. The
third strategy is having an oxygen reservoir on-board. Bosch, Ford and Tesla have filed
patents on this method in different variants. Bosch has suggested an inflatable bladder
pumped by the electrochemical reaction itself [114], whereas Ford and Tesla have looked
at more traditional systems with an oxygen tank and a compressor [115, 116].
6.2.2 A solution for an open Li-air system
The purpose of this assessment is to determine whether it is possible to make an air

purification system that is suitable for use in electric vehicles in terms of weight, volume
and energy consumption. Calculations made on a specific system, based, among others,
on the strategies mentioned above, is included in Appendix D. The main conclusions
are presented below.
The calculations are based on a 100 kWh Li-air battery with an average discharge
voltage of 2.65 V that consumes 17 m3 of oxygen through a full discharge. The scenario
is based on a calculation of 100 % humidity at 25 ◦ C and 400 ppm CO2 . In order to
dimension the system, a minimum discharge capacity time of 3 hours is assumed, which
corresponds to an average speed of 170 km/h, as a total driving range of 500 km is
expected from a 100 kWh battery.
The suggested system is cleaning the air in three steps, as sketched in Figure 6.8. First,
the air is compressed to 3 bar. This has been shown to increase the performance of
the battery [117, 118], and is thus expected to be needed anyway. As the dew-point is
almost independent of pressurization, the compression corresponds to a decrease in dew
point from 25 ◦ C to 7.9 ◦ C as described in Appendix D. Second, the gas is pre-dried in
a semipermeable membrane unit like the Permapure Drier. This decreases the amount
Filter
High pressure Low pressure
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
a b c d
18
Li+
17 e-
Li2O2
16
O2
15 e-
14
13 1.0 bar 3.0 bar 2.9 bar 2.8 bar
12 3% 1% 0.15 % <1 ppm

400 ppm 400 ppm 200 ppm <1 ppm
11
10 Figure 6.8: Schematic illustration of an air cleaning system for Li-air batteries to
9 remove H2 O and CO2 to below ppm levels. The red color represent air with impurities
and the green colors represent the allowed amounts of H2 O and CO2 .
8
7
of CO2 and H2 O significantly, and limits the size and energy consumption of the last
6
step. Third, a system of plug flow reactors with H2 O and CO2 absorbents is used to
5
reduce the impurity levels below 1 ppm. After this final step, the air is ready to use
4
in3 the battery. An important part of the cleaning design is to use the dry air from the
battery
2 in a counter flow setup in the membrane-based pre-drying step, as this will save
a1significant amount of energy. Template used to crop directly in LaTex with trim-command
Allowing 1 ppm of impurities in the gas stream, accumulates to a total amount of 40 g

H2 O and 100 g CO2 in a total lifespan of the battery corresponding to a total of 420 full
cycle equivalents similar to the calculation performed by Albertus et al. in Ref. [13].
If this is found not to be sufficient, an alkaline guard like Potassium or Sodium can be
used as a final cleaning step.
Based on the calculations performed, the system weight will be 20-30 kg in a design
based on regeneration during charging. The energy consumption is estimated to be
approximately 1 % of the battery capacity during discharge due to a pressure loss in
the cleaning system and 4% during charging to heat up the adsorbents to release the
captured H2 O and CO2 . The gas compressor itself is not considered in this calculation.
Compared to the complications and volume requirements of having an oxygen storage
tank, this looks like an attractive option. Concerning the weight, it is also noted that
the weight of the oxygen itself in the storage tank is 23 kg in a 100 kWh system.
6.2.3 Summary of separation of O2 from air
It was shown that it is possible to design a system with a relatively low weight and
energy consumption, that is able to remove H2 O and CO2 to a level below 1 ppm. A
realistic estimate of the price has been left out, as the expected economy of scale of such
a system will have a huge effect on the system price.
6.3 Summary of commercial applications of lithium-air bat-

teries
Due to significant challenges related to degradation of the electrodes and electrolyte in

Li-O2 batteries, a commercialization of the technology is not expected soon. If (or when)
the battery will be ready to enter the market, this chapter has presented solutions to
two challenges that both could have been show-stoppers. First, a method for estimating
remaining capacity, power capability and cycle life of Li-O2 batteries has been proposed
and verified through a number of tests. Second, an air purification system with accept-
able weight and energy consumption was designed to meet a demand of sub-ppm levels
of H2 O and CO2 as required in an open Li-O2 system.
Chapter 7
Summary and outlook
In this thesis I have presented the results of my research during the last three years. I
have divided the results into three chapters, and a summary of the findings has been
given at the end of each chapter. It is not my intention to fully repeat those summaries
here, but only to emphasize a few of the most important aspects of the work and briefly
discuss the outlook of this thesis.
7.1 Summary of main results
During my PhD project, I have investigated carbon based air-electrodes in Li-O2 batter-
ies to establish the necessary testing facilities and develop a theoretical framework with
DEMS and EIS as the key techniques with the ultimate goal of proposing new materials
and screening methods. The main outcome of this work can be summarized as follows:
• During the project I constructed the differential electrochemical mass spectrometer

and used it to investigate different systems, including electrolytes based on ionic
liquids. It was an important achievement, since the measurement of the oxygen
consumed and released is necessary to determine if a Li-O2 system is reversible.
• The biggest contribution of this work to the field of research is the increased
understanding of the electrochemical impedance spectroscopy (EIS) in the Li-
O2 system. Working closely with the research group at IBM Almaden Research
Center, it was possible to combine their extensive knowledge of the system with the
conducted EIS measurement. Among others, it was possible to use EIS to show
that the capacity is also limited by tunneling in porous cathodes, and that the
change in double layer capacitance at sudden death corresponds to a Li2 O2 layer
69
Chapter 7. Summary and outlook 70
of approximately 5 nm in porous electrodes, which is in line with flat electrode

measurements and DFT calculations.
• Further development of the EIS technique enabled a detailed probing of the impedance
during charge and combined with DEMS and other supplementary techniques, it
was possible to show that the Li2 O2 oxidizes at very low potentials (3.05 V) during
charge, but that the reaction is blocked by the formation of an SEI layer. The
EIS measurements were important to identify the existence of a mixed potential
during charge.
• It was possible to use the knowledge obtained working with EIS to develop a new
and simple tool for battery management systems to estimate the state of charge
and state of health of Li-O2 batteries.
7.2 Outlook
Having worked within the same field of research in three years naturally brings up many
ideas for further research. I have here chosen the six suggestions that I consider most
interesting.
1. The ultimate goal of investigating the fundamental electrochemistry with DEMS,
EIS and supplementary methods has been to find new materials. I suggest to use the
established competences to look for new materials. The search for new electrolytes using
DEMS is well documented in Chapter 5, and an additional screening parameter could
be to use the capacity in charge below 3.15 V to measure the formation rate of the SEI
layer. Cathodes are more complicated as the electrolyte degradation is often dominat-
ing, but the use of EIS and differential capacity measurements might reveal reactions at
certain potentials, that is only related to electrode degradation, which could then serve
as a screening test.
2. Disregarding the stability issues, carbon is indisputably the best cathode material
because it is light, conducting and can be made with a very high surface area. Investiga-
tions of ALD coating of carbon cathodes with a protective layer might therefore provide
the optimal solution.
3. Further development of the EELS/EFTEM method to probe lithium and oxygen on
the sub-nanometer scale. TEM studies often presents nice pictures of toroidal shaped
particles formed in the presence of water impurities, but no one has probed the thin lay-
ers on top of the carbon and this might very well give important inside into the reaction
mechanisms and the SEI layer formation among others.
4. The use of the capacitance to determine the state of charge should be investigated
further and, if possible, expanded to other metal-air systems. Techniques to reliably
Chapter 7. Summary and outlook 71
determine the capacitance in the time domain and away from OCV, will strengthen the
method significantly.
5. Having the entire framework of DEMS and EIS ready, it would be interesting to look
at Na-O2 and other similar systems to qualify and quantify the discussion. Especially
EIS could be useful to describe the differences between Li-O2 and Na-O2
6. Finally, the DEMS setup can be used to study many different processes. One option
is to investigate high voltage behavior of electrolytes and measure the gas evolution as
a function of potential.
Appendix A
Development of the testing

equipment
In the project I have been main responsible for building up the basic testing equipment
for Li-O2 batteries. Among others, I have designed and developed
• the differential electrochemical mass spectrometer (DEMS)

• the 2-electrode Swagelok cells
• the equipment used to automatically change gas in the cells from oxygen to argon
and vice verse
• the cell holders to mount the Swagelok cells inside the furnace
• the cathode fabrication using an airbrush
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
a
18
b c
17
16
15
14
13
d
12
11
10
9
8
Figure A.1: (a)-(b) The initial cell design with manual valves. (c) First Li-O2 battery
7 pack in the ReLiable project. Tested in 2012 using the second cell design with quick-
6 connect valves. (d) Final cell design with 10 mL volume.
5
4
73
3
2
Appendix A. Development of the testing equipment 74
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
a
18
b
17
16
15
14
13
12
11
10
9
8
7 Figure A.2: (a) Gas filling and changing system. (b) Five Swagelok cells in the
temperature controlled furnace.
6
5
Figure A.1 shows three of the four design versions of the Swagelok cell. As it may be
4
seen, we increased the gas volume inside the cells in each generation. As we improved
3
our competences in the preparation of the active materials, our capacity in the cathode
2
increased and we needed a larger gas volume in the cells. In our initial design we had a
gas volume of 0.5 mL, which was insufficient to support a full discharge. The volumen
was increased to 2 mL, which was still a little too low to be able to neglect the pressure
decrease during discharge. The choice of welding a larger volume onto the Swagelok
cells was made to ensure that the pressure did not decrease by more than 15 % during
discharge, while maintaining the rather compact design.
Figure A.2a shows the Labview controlled gas-change station used to automatically
change gas in the cells from oxygen to argon and vice verse. Based on a wide number
of experiments, we found this necessary to ensure a consistent gas-change routine in
terms of flow time and gas pressure inside the cell after gas-change. In addition to this,
the program was designed to prevent a back-flow of ambient air, which ensured that we
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
a
18
b c
17
16
15
14
Figure A.3: The DEMS setup was constructed in two phases. To gain experience, we
13 constructed a manual DEMS setup to determine appropriate volumes, flushing times
12 etc. After this, we automated the setup in a Labview interface, basically making DEMS
measurements a one-click experiment.
11
10
9
8
7
Appendix A. Development of the testing equipment 75
always had the pure 6.0 gas available. Figure A.2b shows five Swagelok cell being tested
inside the temperature controlled furnace.
Figure A.3a shows the first version of the DEMS setup. It was a manual system designed
to gain experience with the method before the more complicated and expensive system
was built. I co-supervised Andreas Hansen Poulsen, who did most of the work in a
close collaboration with me. Figures A.3b and A.3c show the automatic DEMS system
under and after construction. I designed and constructed the setup with invaluable help
from research technician Mike Wichmann and student assistant Mathias Kjærgaard
Christensen. It was automated by using magnetic valves, an automatic 2-position 6-way
valve and a much more advanced Labview interface, controlling the mass spectrometer,
potentiostat, magnetic valves and pressure measurements.
Appendix B
Differential electrochemical mass

spectrometry - In situ gas
analysis
77
Appendix B. Differential electrochemical mass spectrometry 78
B.1 Initial measurements and design
The first setup in our lab was built in the spring 2012 with manual valves and without
automation to enable fast changes as we learned more about the method and the lithium-
air system. Andreas Hansen Poulsen was responsible for this development under my
supervision. From September 2012 to March 2013, I built a fully automated DEMS
setup based on the experiences we learned from the manual setup. This would not have
been possible without the invaluable help from Mathias Kjærgaard Christensen who
helped me setup the Labview framework and Mike Wichmann, who assembled the pipes
and valves.
To ensure a reliable measurement, a wide number of parameters need to be measured

and characterized. Some of the most important steps are included below.
24 23 22 21 20 19 18 17 16 15 14 13 12
32 11
m/z 10 9 8 7 0.5
6 245 234 223 212 201 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
18 12 36 m/z 18
ΔnO2 / [µmol/sample]
10
17 17
Ion current
16 16
15 15
14 14 0.1
13 13
12 120.05
13
1110 0.0 0.2 0.4 0.6 0.8 1.0 11 0.0 0.2 0.4 0.6 0.8 1.0
10 Time [h] 10 Time [h]
9 9
8 Figure B.1: Left: Measurement of ion currents 8 on m/z values of 32 and 36 corre-
36
7 sponding to O2 and Ar. Right: The ion currents
7 have been used to calculate the
6 molar amount of O2 in the sampled volume. 6 The time dependence reflect that the
5 gases are not equally distributed in the capillary
5 inlet to the mass spectrometer.
4 4
MS Experiment no 167 Intensity
3 3 MS Experiment no 167 Gas evolution
2 2
The
1 timing 1
of the measurement of ion currents turned out to be very important.
Figure B.1 (left) shows a 1 h measurement with the mass spectrometer of a gas sample
in the beginning of the first charge of a Li-O2 battery. The time axis start just as
magnetic valve l in Figure 3.5 was opened. It is seen that the ion currents of m/z 32
and 36 behave differently, and as the relative current between these channels are used
to calculate the molar amount of gas evolved, it is clear that timing is important. In
Figure B.1 (right) the molar amount of O2 calculated based on the ion currents is shown
without the proper calibration. This emphasizes that it is important to fix the time from
the magnetic valve l opens, to the measurement of ion currents is performed. From
these measurements, it was decided to use the data after 10 minutes. Since then, the
pumping procedure between each sampling has improved and the measurement is now
used after only 2 minutes. Every time this changes, a new calibration of the ion currents
is needed.
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
27.0
12 pm
18
Temperature [°C]
17 26.5
16 26.0
15
14 25.5
13 25.0
12 6 am
11 24.5
0.0 0.5 1.0 1.5 2.0 2.5 3.0
10 Time [day]
9
Figure B.2: Temperature
8 variations in the room of the DEMS setup.
7
6
Midnight
Temperature Figure B.2
5 shows the temperature variations in the room in a 3-days pe-
Midday
4
riod. The temperature varies
3 periodically with a minimum temperature in the morning
Temperature. Graphite. Experiment no 309.
2
and a maximum temperature at noon. As the temperature changes affect the pressure
inside the DEMS setup, it is important measure the temperature throughout the entire
measurement to correct the pressure measurements. Using the ideal gas law, a change
in temperature of 2◦ C from 24.7◦ C to 26.7◦ C corresponds to a change in pressure of
((273.15+26.7)/(273.15+24.7)-1)*1.8 bar = 12 mbar. In the DEMS setup this pres-
sure change is equivalent to 0.14 mAh of the 2e- /O2 discharge or charge reaction. For
cathodes with a small carbon loading this corresponds to one third of the total capacity.
Calibration of ion current A number of calibration experiments are needed to en-

able a quantitative translation of M/Z intensities and pressure changes to moles of gas
consumed or evolved. We used a 5 % O2 , 5 % CO2 and 90 % Ar to calibrate the corre-
spondence between the measured intensities of O2 , CO2 and Ar and the corresponding
partial pressures. Using measurements of the intensities after 2 minutes, the following
correction factors were obtained:
pO2 IO2
= 1.273 (B.1)
p36 Ar I36 Ar
pCO2 ICO2
= 1.152 , (B.2)
p36 Ar I36 Ar
By diluting the gas with pure argon, it was shown that this ratio was unchanged in the
range 0.1 % to 5 % of O2 and CO2 .
Volume calibration The internal volumes are important to calculate the molar amount
of gas consumed or evolved. Relative volumes in the setup was determined by moni-
toring the pressure changes of the two pressure transducers as different combinations of
valves were opened or closed. The actual volumes were determined by replacing a part
of the pipe with two different calibration volumes (500 µL and 1.00 mL) and measuring
the effect on the pressure changes as valves were opened or closed.
The two volumes, essential to calculate the molar amounts of gas, were determined to
be
Vdis = 4.98 mL (B.3)

VMS = 1.03 mL (B.4)
where Vdis and VMS are specified in Figure 3.6.
10024 23Reference
22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
Measurement
18
Relative intensity
80
17
16 60
15
40
14
13 20
12
11 0 16 20 24 28 32 36 40 44 48 52 56 60
10 Mass [amu]
9
Figure B.3: Mass spectrum
8 of 1,2-dimethoxyethane as measured by the mass spec-
trometer. The measurement
7 is compared with a reference spectrum from NIST Chem-
6 istry Webbook [119]
5
4
DME fingerprint
3
Overlapping peaks in 2the mass spectrum The electrolyte of our reference system
in the ReLiable project is1based on DME, which is very volatile. Template
It isusedtherefore important
to crop directly in LaTex with trim-command
to ensure that the m/z peaks of DME in the mass spectrum does not overlap with the
peaks of O2 and CO2 . Figure B.3 shows a measured and a reference spectrum of DME
and it is seen that the contributions at m/z values of 32 (O2 ) and 44 (CO2 ) are very
small.
Oxygen background [µmol/h]
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9
0.25 8 7 6 5 4 3 2 1
18 0.20 Charge start
17
0.15
16
15 0.10
14
0.05
13
12 0.0 0.5 1.0 1.5 2.0
11 Time [h]
10
Figure B.4: The oxygen
9
background in DEMS measurements as a function of time.
8
7
6
Flushing O2 out of the
5
system When the DEMS setup is measuring the evolved
gas inside the cell, it flushes
4 the cell with argon. Most of the gas in the cell is changed,
3
but not everything. In the beginning of the charge, the cell is initially filled with pure
2 DEMS Experiment no 406 - Oxygen flush
O2 and even small amounts
1 of O2 left in the cell will be significant
Template used compared totrim-command
to crop directly in LaTex with the
evolved amount of O2 during charge. Therefore, the cell needs to be flushed sufficiently.
Figure B.4 shows four measurements of how the amount of oxygen decreases between
each measurement after changing the gas to argon. These measurements were made in a
dummy cell without O2 evolution, but the onset time of the charge in the real batteries
is shown in the figure. A typical discharge current in the DME-XC72 system is 0.5 mA,
which correspond to 9.3 µmole/h. In this case the oxygen left in the system is insignif-
icant, but in measurements with a lower current density, as in the measurements with
ionic liquids, described in Chapter 5, the O2 background must be taken into account.
B.2 Calculations
The key calculations are:
Q
∆ne- = (B.5)
F
∆p · V
∆nO2 = , (B.6)
R·T
where ∆ne- and ∆nO2 are the moles of electrons and oxygen molecules consumed in a
given period of time, Q is the integrated charge, F = 9.65 · 104 C/mole is the Faradays
constant, ∆p is the gas pressure inside the Swagelok cell, V is the volume shown in red in
Figure 3.6 left, R is the gas constant and T is the temperature. As described in Section
B.1, the volume has been calibrated to be 4.98 mL, and as most of the volume is in the
gas system and not in the Swagelok cell, small differences between the different cells
does not affect the measurements. The temperature is measured continuously during
the measurement.
pO2 · V
nO 2 = (B.7)
R·T
As shown in Section B.1, measurements with a calibration gas with O2 and CO2 partial
pressures in the range of 0.1% to 5% showed that
pO2 IO2
= 1.273 (B.8)
p36 Ar I36 Ar
pCO2 ICO2
= 1.152 , (B.9)
p36 Ar I36 Ar
and using that p36 Ar = 0.00333 · pAr = pAr /300.3, Equation B.7 can be rewritten as:
IO2 p36 Ar · V 1.273 IO2 pAr · V

nO 2 = 1.273 · = · (B.10)
I36 Ar R·T 300.3 I36 Ar R·T
ICO2 p36 Ar · V 1.152 ICO2 pAr · V
nCO2 = 1.152 · = · (B.11)
I36 Ar R·T 300.3 I36 Ar R·T
As argon constitutes 99-99.9% of the gas, the pAr is approximated with the total pressure
of the analyzed gas.
B.3 Comparing oxygen consumption with electrochemistry
The DEMS measurement can be used in two modes: Pressure mode and DEMS. Figure
B.5 shows measurements using the XC72 reference system using both of these modes
for comparison. First of all it is noted that they agree well. The pressure measurements
does not have information about which gases are evolved, and is thus only applicable
when the main gas evolved is oxygen.
4.524 23 22 21 20 19 18 17 16 15 14 13 12 11 10 2.02
9 8 7 6 5 4 324 223 Voltage
122 21 20 19
O2 18 17
CO16
2
15 14 13 12 11- 10 149
2.0 e /O2
8 7 6 5 4 3 2 1
Voltage
18 Pressure 18 4.5 2.3 e-/O2 12
4.0 2.01
cell voltage [V]

Fit
ΔnO2 / [µmol/h]
17 17 2.6 e-/O2 10
cell voltage [V]
Pressure [bar]
2.00 16 4.0
16 3.5 8
15 2.0 e-/O2 1.99 15
3.0 2.4 e-/O2 6
14 1.98 14
3.5 4
13 2.5 13
1.97 2
12 12
11 2.0 1.96 3.0 0
0.0 0.1 0.2 0.3 0.4 0.5 11 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5
10 Capacity [mAh] 10 capacity [mAh]
9 9
8 Figure B.5: Left: Pressure change in a Li-O82 battery during discharge and charge.
7 Right: A quantitative analysis of the evolved 7gases during charge measured by mass
6 6
5
spectrometry.
5
4
4
3
DEMS pressure Experiment no 232 3
2
To 1validate the DEMS setup, a number of measurements
2 have been performed
DEMS Experiment no 359 on the Template used to crop directly in LaTex with trim-command
reference system for comparison with similar measurements performed by Bryan D. DEMS pressure Experiment no 359
4.8
DEMS Experiment no 359 Oxygen evolution
Voltage 1.75
Pressure 4.6
McCloskey et al. at IBM. The measurements are simple charges and discharges like the 3.0
Fit 1.955
1.70 4.4
2.8
xygen evolution [e$^-$/O$_2$

ones shown in Figure B.5 with current densities of 100-200 mA/g carbon. 1.65
4.2
cell voltage [V]
cell voltage [V]

pressure [bar]
2.6 4.0
1.60
3.8
2.4
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1 3.6
5.0 4.0 1.55
18 2.2 voltage 3.4

e-/O2 - ratio 3.5 1.50
17 4.5 2.0
3.2
cell voltage [V]
1.45
16 4.0 3.0
e-/O2 - ratio
3.0
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 0.0 0.5 1.0 1.5 2.0
Capacity [mAh] Capacity [mAh]
15
3.5 2.5
14
2.0
13 3.0
12 2.5 1.5
11 1.0
2.0
10 0.0 0.2 0.4 0.6 0.8 1.0
9 relative capacity
8
-
Figure B.6: The SOC 7 dependence of the e /O2 -ratio. The values are averages of six
batteries tested
6 at currents in the range of 100-200 mA/g carbon.
5
4
3
- DEMS oxygen consumption and evolution average
Figure B.6 shows how the
2 e /O2 -ratio depends on SoC. The values are averages of six
measurements and the error-bars indicate one standard deviation. It is seen that the
process is a 2.0 e- /O2 process during the entire discharge with very small deviations.
During charge, however, the e- /O2 -ratio increases during the entire charge ending at 3.5
e- /O2 at the end of charge. It is also noted that the standard deviation of the e- /O2 -
ratio increases significantly in charge-mode, indicating that the exact electrochemistry
may change between measurements depending on small variations in current density
or other parameters. Based on the performed measurements the discharge process is a

1.99 e- /O2 with a standard deviation of 0.02 e- /O2 . The average value during charge is
2.62 e- /O2 with a standard deviation of 0.12 e- /O2 . The OER/ORR ratio is 0.78 with
a standard deviation of 0.05. These values are compared with the values reported by
McCloskey et al. in [56] in Table B.1, and it is noted that all values are well within one
standard deviation.
OER/ORR (e- /O2 )dis (e- /O2 )cha

DTU 0.78 ± 0.05 1.99 ± 0.02 2.62 ± 0.12
IBM, [56] 0.78 2.01 2.59
Table B.1: Comparison between DEMS measurements performed at the setup built
at DTU and values reported by McCloskey et al. at IBM [56].
Appendix C
Supporting TEM measurements
85
Appendix C. Supporting TEM measurements 86
C.1 Beam damage and observations of lattice fringes in

Li2 O2
a 24 e23
30 -/nm22
2s-121 20 19
b 30 e18-/nm172s-116 15 14 13 12 11 10 9
2 min 10 min
18
17
16
100 nm 100 nm
15
14
c 300 e-/nm2s-1 d 300 e-/nm2s-1
13 1 min 10 min
12
11
100 nm 100 nm
10
9
Figure C.1: Test of Li2 O2 beam sensitivity by monitoring visual changes to the
particles. The same particle is subjected to (a)-(b) 30 e− /nm2 s−1 in 10 min and
8
then (c)-(d) 300 e− /nm2 s−1 in 10 min. The measurements were performed at Titan
microscope operated in TEM-mode at 300kV.
7
A Li2 O2 particle subjected to increasing exposure time and dose rate is shown in Figure
6
C.1. It is seen that the particle can be imaged and spectroscopically inspected by TEM-
EELS for at least510 minutes without any visible damage at 300keV using a low electron
dose rate of 30 e- /nm2 s-1 . A ten times higher electron dose rate of 300 e- /nm2 s-1 , results
4
in morphological changes. This is on the same order of magnitude as recently reported
3 (20 e-/nm2s-1 at 200 keV) to study Li2O2 precipitates.
by Zhong et al. [47]
2
C.2 Selected
1 area electron diffraction Templat
The cathode used the most in the experiments described in this thesis contains XC72
carbon black and PTFE binder, and, if discharged, it also contains Li2 O2 . As lithium
is a very light element, it is important to find a suited strategy to distinguish this from
the rest of the cathode. Selected area electron diffraction (SAED) and electron energy
loss spectroscopy (EELS) are measured on reference systems to identify the best suited
method. The morphology is also useful for quick identification, but it cannot stand
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5
18
17
Figure 1. T
16
at two ma
15 and select
14 diffraction
(a,b,c) Li2O
13
Carbon bl
12 (g,h,i) PTF
11
10
9
8
7
6
5
4
3
2
Figure C.2: TEM images at two magnifications and selected area diffraction patterns
1of (a,b,c) Li2 O2 , (d,e,f) XC72 carbon black, and (g,h,i) PTFE binder. Template used to crop directly in L
The measure-
ment time on Li2 O2 was kept below 10 min with a doserate of 140 e- /nm2 s-1 to avoid
visual structural and crystalline changes. The measurements were performed at CM200
(Haldor Topsøe A/S) in the HR-TEM mode. The beam damage is investigated further
in Figure C.1.
alone. The aim of the reference measurements is to test the TEM-EELS technique to
study Li2 O2 precipitates on the nm-scale, i.e. species that may be formed during the
discharging of Li-air batteries. In particular, the beam stability of Li2 O2 material is
addressed, and low-loss EELS is used to spectroscopically identify lithium species. The
possibilities of imaging crystalline Li2 O2 at high-resolution, i.e. imaging of lattice fringes
and eventually sub-nm surface layers are estimated from a theoretical/practical point of
view.
Figure C.2 shows TEM images in two magnifications and SAED patterns for Li2 O2 ,
carbon XC72 and PTFE binder. The compounds were dispersed on a standard Cu
TEM-grid with lacey C-film. The XC72 was ultrasonicated in ethanol for 1 min and
dispersed in ambient conditions, the PTFE was dispersed using an air-brush using the
same procedure as cathode fabrication, but without adding carbon to the dispersion,
and the Li2 O2 was crushed and dispersed dry and transferred to the CM200 microscope.
The Li2 O2 was exposed to ambient conditions for less than 2 min.
The samples were examined in CM200 in the HR-TEM mode at different magnification.
Approximately 20 agglomerates from each sample were viewed and 5 of the agglom-
erates were recorded at low magnification (for overview) and two agglomerates were
further inspected at higher magnification and by SAED. Electron diffraction patterns
were recorded with a fixed camera length of 330 mm, i.e. the same magnification. Figure
C.2shows TEM and SAED images of the three samples. The SAED patterns show that
PTFE and XC72 are amorphous, whereas the Li2 O2 particle is polycrystalline.
C.3 Electron energy loss spectroscopy
EELS measurements were performed on commercial Li2 O2 in the Titan microscope

operated in TEM-mode at 300kV at a nominal TEM magnification of SA 6,300x. In
addition, a post-magnification was introduced by the GIF (about x15). The length
scale of the images is not calibrated, but the image size is estimated. The zero loss
peak for the EELS was continuously adjusted to allow an energy resolution of 1.2-1.4
eV measured as the FWHM of the zero loss peak. All EELS data were recorded with a
dispersion of 0.2eV/channel and a selected entrance aperture (SEA) of 2 mm inserted.
The incident beam dose rate on the sample was measured on the retractable current-
readout-stick (viewing stick) inserted in the electron beam. Electron beam dose rates of
30-300 e- /nm2 s-1 were used. Imaging was done prior to and after each EELS acquisition
to record eventual visual changes of the sample during EELS acquisition. Also images
with and without the selected entrance aperture inserted were recorded to identify the
localized sample area giving rise to the EELS signals.
40 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4 3 2 1
24a23 22 21 20 19 18 17 16 b15 14 13 12 11 10 9 c 8 7 6 5 4 3 2 1
40
18
Counts [103 e-]
400 18
17
30 17
FWHM = 1.4 eV
16
200 1630
Counts [103 e-]
Counts [103 e-]
15 Li-K
15
14 20 0
0 100 200 300 1420
13
Electron loss [eV] 13 Li-K
12
12
11 10
1110
10
10
9
0 9 0
8 -50 0 50 100 150 200 250 300 350
8 -50 0 50 100 150 200 250 300 350
7 Electron loss [eV] Electron loss [eV]
7
6
Figure 4. TEM images of the same area of Li2O2 (Figure 3) recorded at an increased 6
5 Figure 5. TEM images of the an area containing both amorphous carbon film and Li2O2 recorded at the higher doserate of 300 e-
4
Figure C.3: Low-loss TEM-EELS spectrum of commercial Li2 O2 in an area (a) with-
- 2 -1
doserate of 300 e /nm s after additional (a) 1 min of exposure and (b) 10 min. (c) Low-loss 5
EELS spectrum recorded between the acquisitions (a) and (b), revealing clear signature of
/nm2 s-1 after (a) 2 min of exposure and (b) 7 min. (c) Low-loss EELS spectrum recorded between the acquisitions (a) and (b),
revealing clear signature of lithium K-edge (Li-K, near 60eV), whereas characteristics features of Li in the 20-60 eV range is
4
3 out lithium
lacey-C film and (b) with lacey-C film recorded at a dose rate of 300 e- /nm2 s-1 .
K-edge (Li-K, near 60eV). A visual damage is observed on the particle. Image size is
not calibrated, but is estimated to about 420 nm x 420 nm based on the thickness of the rim 3
smeared out by the predominant carbon signal. A visual damage is observed on the particle. Image size is not calibrated, but is
estimated to about 420 nm x 420 nm based on the thickness of the rim on the carbon film of ~15 nm (Figure 2).
2
1 The spectra show clear signature of the lithium
on the carbon film of ~15 nm (Figure 2). 2 K-edge (Li-K, near 55eV). A visual
damage was observed on the particle at this
Figure 3. TEM images of Li O recorded at a low electron doserate of 30 e /nm s after (a) 2
2 2
min of total exposure time and (b) 10 min. (c) Low-loss EELS spectrum recorded between the
dose rate. Image size is about 420 nm x
- 2 -1
420 nm.
acquisitions (a) and (b), revealing clear signature of lithium K-edge (Li-K, near 60eV). No
visual damage is observed on the particle. Image size is not calibrated, but is estimated to
about 420 nm x 420 nm based on the thickness of the rim on the carbon film of ~15 nm
(Figure 2).
Figure C.3 shows the EELS signal from as-prepared Li2 O2 sample. The spectrum demon-
strates distinctive EELS signatures of lithium such as Li K edge near 55eV in the low
loss region, which are not present in EELS from the carbon and/or oxygen, shown in
Figure C.4. This verifies the presence of Li. Figure C.3b shows the EELS signal from
an area of both carbon-film and lithium-sample. The determination of the crystal phase
(e.g. Li2 O2 , Li2 O, or LiO2 ) cannot be readily deduced from the present EELS data and
complementary techniques such as SAED should be used.
In conclusion, Li2 O2 precipitates can be imaged and spectroscopically inspected by

TEM-EELS for at least 10 minutes without any visible damage at 300keV using a
low electron dose rate of 30 e- /nm2 s-1 . A ten times higher electron dose (rate) of 300
e− /nm2 s−1 , results in clearly visually damaged/changed particles during acquisitions. Li
and C can readily be identified by EELS on the nm-scale, although a drift stable EELS
setup is required due to the weak signals. The oxygen EELS signal was not addressed
in this study. Atomic resolution imaging of Li2 O2 without damaging the structure by
the electron beam is lofty as extremely long acquisition times of about 18 minutes are
needed (as compared to 1s for typical high resolution imaging).
C.3.1 EELS reference spectra
Figure C.4a shows the electron energy loss spectra (EELS) of the low loss and core loss
energy windows for lithium, carbon, and oxygen. For the present study, only the energy
losses of 0-300eV were recorded to reveal the zero-loss peak (to determine the energy
resolution), the low loss features (plasmonic region, ca. 10-50eV) and core loss of Li-K
( 55 eV) and C-K (onset at 284eV).
a24 23 22 21 LiF
20 19 c18 17 16 15 14 C13 12e 11 10 9 8 O72 6 5 4 3 2
18
Counts
Counts
Counts
17
Li-K
16
15 0 20 40 60 80 0 25 50 75 100 -10 0 10 20
14
b LiF d C-K
C f Al2O3
13
Counts
O-K
Counts
Counts
12
Li-K
11
10 0 200 400 600 800 0 100 200 300 400 520 600 680
9 Figure C.4: References of electron energy loss spectra. From the EELS atlas.
8
7
6
5
Appendix D
Calculations used to assess the air

purification system of an open
Li-O2 system
91
Appendix D. Calculations of an air purification system 92
D.1 Design frame
To assess the viability of an open system based on known technology, a specific calcu-
lation on dimensions and weight of such a system is performed in this section. The key
numbers are presented in this section. The design criteria of the filter are
• Energy: 100 kWh

• Discharge potential: 2.65 V
• Minimum discharge time: 3h
• Required purity: <1 ppm H2 O and <1 ppm CO2
Using these numbers, the total consumption of oxygen in a full discharge is calculated
ne - Q E 100 kWh
nO 2 = = = = = 704 mole (D.1)
2 2F 2U F 2 · 2.65 V · 96.5 kC/mole
At standard conditions, this correspond to a volume of
nO2 RT 704 mole · 8.314 J/mole/K · 298 K

VO2 = = = 17 m3 (D.2)
pO2 1 bar
With 20.9 % oxygen in the air, this corresponds to 82 m3 air with up to 1.82 kg H2 O
(3 %) and 59 g CO2 (400 ppm).
As discussed in Section 6.2.2, it is reasonable to assume that compression of the gas is

needed. If the air is compressed to 3 bar, this corresponds to a decrease in dew point
from 25 ◦ C to 7.9 ◦ C or a change in maximum water content from 1.82 kg to 650 g,
using the equation of the dew point defined by Buck [120]

−6 17.502 · T
pH2 O = (1.0007 + 3.46 · 10 · ptotal ) · 6.1121 · exp (D.3)
240.97 + T
where pH2 O and ptotal is the water and total pressure in millibar, and T is the temper-
ature in degrees Celsius.
The best zeolite absorbents reach an absorption capacity of 30% [121], which means
that 6.1 kg zeolite could absorb the water from a full discharge, but as it will cost a
lot of energy to release the water again, a pre-drying step is added. By using a water
permeable Nafion membrane like the Permapure Drier in a counter flow setup with
the dry gas from the battery, the energy consumption of the system can be reduced
significantly.
D.2 Pre-drying using a membrane system
The relative dimensioning between zeolites and the passive pre-drying unit depends
on the need for energy savings. If energy is not a concern, the amount of adsorbents
should be increased, whereas the pre-drying unit is better, if energy savings on the
regeneration is important. This calculation is based on a Permapure 24” PD-200T,
Ø1”. Using a maximum discharge time of 3 h, 82 m3 air corresponds to 460 l/min, and
with a maximum flow rate in the Permapure tubes of 40 l/min, 12 pipes are needed.
Using the dimensions of the pipes, they can be fiited into a box of 66 cm x 12 cm x 8
cm. The pressure drop can be calculated to be 0.1 bar and the output concentration od
H2 O is calculated to be 1550 ppm. As the Permapure dryer is also permeable to CO2 ,
this concentration is assumed to be halved to 200 ppm.
D.3 Designing the adsorption system
The typical design of an absorption system is a plug-flow reactor. Two considerations

are important when designing such a system: capacity and kinetics. The capacity should
be sufficient to store all impurities in the inlet gas until regeneration and the kinetics
should be fast enough such that impurities are not able to slip through the reactor
without being absorbed.
0.25
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 9 80,02024
7 236 22 521 420 19
3 182 17 116 15 14 13 12 11 10 9 8 7 6 5
Water adsorbed (g/g)
18 0.20 18
Adsorption capacity (g/g)
25°C
17 0,015
17 0.15 0 oC 25 oC 50 oC
16
16 15 0,010
0.10
15 14
13 0,005
14 0.05 125 oC 12
13 0.00 90°C
11 0,000
12 1 10 100 1000 10000 10 0 100 200 300 400 500
H2O concentration (ppm) 9
11 Concentration CO2 (ppm)
8
10
Figure D.1: Langmuir isotherm for H2 O adsorption7 in zeolites and CO2 adsorption
9 in a solid amine. 6
8 5
7 4
3
6
The capacity is typically calculated using the Langmuir
2 isotherm, that describes
5 1 Template used to crop directly in LaT
the
4 ability of the material to adsorb a given molecule as a function of pressure and
temperature.
3 It is defined as
2
Kpa
θa = (D.4)
1 Kpa + 1
where θa is the fractional occupancy of the adsorption sites, pa is the partial pressure of
−

−
the impurity ’a’ and K is the equilibrium constant given by K = k exp(−∆G /(RT )) =
k exp(−(∆H − T ∆S)/(RT )). In addition to describing the capacity, the Langmuir

isotherm is important to define how easy it is to regenerate the adsorbent by heating.
Figure D.1 shows the Langmuir isotherm for a proposed adsorbent for H2 O and one for
CO2 . The important thing to note is that the storage capacity of both materials are
high at room temperature and decreases significant at 125 ◦ C for the H2 O adsorbent
and at 90 ◦ C for the CO2 adsorbent. This means that heating the adsorbent 100 ◦ C,
will release almost all of the adsorbed H2 O and CO2 molecules.
1.0 9
24 23 22 21 20 19 18 17 16 15 14 13 12 11 10 8 7 6 5 4 3 2 1
Adsorbent
Impurities in outlet
18
Saturation level
17
16 time
15 h
15 1.0 ppm H2O 16.8 h
14 0.0 30
Position in reactor Time [h]
13
12 Figure D.2: Schematic illustration of a plug flow reactor with an adsorbent. The
graph below show the change in impurity concentrations in the adsorbents during oper-
11 ation and the graph to the right is adapted from [122] and show the H2 O concentration
10 in the output stream of a H2 O adsorption reactor as a function of time. The steep
transition of ¡1 ppm level to no adsorption, show that the kinetics are very fast.
9
8 kinetics
The Figure D.2 shows an actual measurement of the water adsorption. The
7 flow reactor and saturation profile is sketched in the left part of the figure. As an
plug
6
increasing amount of impurities are adsorbed, the saturation profile moves to the right.
5
When the profile approaches the outlet of the reactor, it is no longer possible to adsorb
4 molecules. Figure D.2 (right) shows the impurities in the outlet during operation
all
3 the zeolite considered as H2 O adsorbent. It is seen that the concentration is below
using
1 2ppm until a total test time of 15 h, and as the full capacity corresponds to 16.8 h, an
1
excess capacity of 12 % is needed to ensure sufficient adsorption.
Calculating the system size Based on the fast kinetics and a capacity of 0.1-0.2
g/g and fast kinetics, an excess capacity of 30 % is chosen with an average capacity
utilization of 0.15 g/g. To remove 1550 ppm H2 O and 200 ppm CO2 from 82 m3 of air,
corresponding to 94 g H2 O and 30 g CO2 , a total of 815 g H2 O adsorbent and 260 g
CO2 adsorbent are needed.
D.4 Energy consumption and weight
Pressure drop The pressure drop in the in the Permapure dryer is 0.2 bar (0.1 bar
entering the battery and 0.1 bar exiting the battery. The pressure drop in the plug flow
reactors are assumed to be 0.1 bar. A total pressure drop of 0.3 bar in a 82 m3 of gas
corresponds to 2.5 MJ or 0.7 kWh, which is 0.7 % of the total battery capacity.
Heating during regeneration A reasonable measure of the energy required to heat

the adsorbents 100 ◦ C is to compare the system with sand. If the system weight is
20 kg, the energy required is approximately 2 MJ. The desorption of H2 O and CO2
further needs 2.5 MJ and including heating of the gasses and other losses, the energy
requirements may be somewhat bigger than during discharge. 4 kWh (or 4 % of the
battery capacity) is used to be on the safe side, and it is noted that the exact value is
not as important during charge, because the car is connected to the grid.
Weight of the system The permapure dryers can fit into a 6.6 L box which means
that the weight is probably 10-15 kg. The active materials in the adsorption system
weigh a little more than 1 kg and it is safe to assume that the entire system with
heaters do not exceed a weight of 10 kg. The total weight of the system will therefore
be 20-30 kg.
List of publications
Paper I
An electrochemical impedance spectroscopy investigation of the overpoten-
tials in Li-O2 batteries
Jonathan Højberg, Bryan D. McCloskey, Johan Hjelm, Tejs Vegge, Keld Johansen, Poul
Norby and Alan C. Luntz
ACS Appl. Mater. Interfaces, 7, 4039–4047 (2015)
Paper II
Reactions and SEI formation during charging of Li-O2 cells
Jonathan Højberg, Kristian B. Knudsen, Johan Hjelm and Tejs Vegge
ECS Electrochem. Lett., 4, A63–A66 (2015)
Paper III
Impedance-based battery management for metal-O2 systems
Andreas E. Christensen, Jonathan Højberg, Poul Norby and Tejs Vegge
J. Pow. Sources. Submitted (2015)
Paper IV
Rechargeability of ionic liquids in Li-O2 batteries
Supti Das, Jonathan Højberg, Kristian B. Knudsen, Poul Norby and Tejs Vegge
To be submitted
Paper V
The influence of CO2 poisoning on overvoltages and discharge capacity in
non-aqueous Li-air batteries
Yedilfana S. Mekonnen, Kristian B. Knudsen, Jon S. G. Mýrdal, Reza Younesi, Jonathan Højberg,
Johan Hjelm, Poul Norby and Tejs Vegge
J. Chem. Phys., 140 (2014)
97
Paper I - EIS investigation of overpotentials in Li-O2 batteries 98
Research Article
www.acsami.org
An Electrochemical Impedance Spectroscopy Investigation of the

Overpotentials in Li−O2 Batteries
Jonathan Højberg,*,†,‡,§ Bryan D. McCloskey,†,⊥ Johan Hjelm,‡ Tejs Vegge,‡ Keld Johansen,§
Poul Norby,‡ and Alan C. Luntz†,∥
†
IBM Almaden Research Center, San Jose, California 95120, United States
‡
DTU Energy, Frederiksborgvej 399, DK-4000 Roskilde, Denmark
§
Haldor Topsøe A/S, Nymøllevej 55, DK-2800 Kgs. Lyngby, Denmark
∥
SUNCAT, SLAC National Accelerator Laboratory, Menlo Park, California 94025, United States
*
S Supporting Information
ABSTRACT: Lithium−O2 (Li−O2) batteries are currently

limited by a large charge overpotential at practically relevant
current densities, and the origin of this overpotential has been
heavily debated in the literature. This paper presents a series of
electrochemical impedance measurements suggesting that the
increase in charge potential is not caused by an increase in the
internal resistance. It is proposed that the potential shift is
instead dictated by a mixed potential of parasitic reactions and
Li2O2 oxidation. The measurements also confirm that the rapid
potential loss near the end of discharge (“sudden death”) is
explained by an increase in the charge transport resistance. The findings confirm that our theory and conclusions in ref 1, based
on experiments on smooth small-area glassy carbon cathodes, are equally valid in real Li−O2 batteries with porous cathodes. The
parameter variations performed in this paper are used to develop the understanding of the electrochemical impedance, which will
be important for further improvement of the Li−air battery.
KEYWORDS: Li−O2 batteries, electrochemical impedance spectroscopy, overpotential, mixed potential, battery performance
1. INTRODUCTION it is a powerful tool to obtain noninvasive in situ information

Lithium−air and Li−O2 batteries have attracted much attention on degradation mechanisms and possible bottlenecks in the
in recent years because of a potentially high specific energy electrochemical reactions. Very recently, Adams et al. and
density and low cost. Furthermore, the fundamental electro- Landa-Medrano et al. have also used EIS to measure the
chemistry has very low reaction barriers, which, in principle, internal resistances of a Li−O2 battery using a two-electrode
enables a high cycle efficiency. The low reaction barriers have configuration.8,9 They varied parameters like cathode morphol-
been predicted by Hummelshøj et al.2,3 using density functional ogy, oxygen partial pressure, salt concentration, and state-of-
theory (DFT) and proven recently by Viswanathan et al.4 using charge (SOC), and they succeed in assigning the different
experiments on flat glassy carbon cathodes in an electrolysis impedance contributions to the processes of either the anode or
cell. the cathode. The batteries investigated were, however, not
However, more realistic batteries with porous electrodes characterized by DEMS and Li2O2 titration, which are
show large overpotentials,5 which significantly reduces the cycle important complementary methods necessary to link EIS
efficiency. Understanding this is crucial to develop a results to the fundamental electrochemistry.
commercially viable Li−O2 technology. The origin of the In this work, we used an intensely studied and well-
overpotential has been investigated intensively. We have characterized in-house reference system used in a number of
previously used DFT modeling,2,4 differential electrochemical previous publications.5,10−15 It consists of an XC-72 carbon
mass spectrometry (DEMS), and Li2O2 titration.5 In addition black and poly(tetrafluoroethylene) (PTFE) binder cathode,
to this, and among others, Zhong et al. used in situ transmission lithium metal anode, and 1 M LiTFSI/1,2-dimethoxyethane
electron microscopy (TEM) to study growth mechanisms,6 and electrolyte. A series of electrochemical impedance spectra was
Chen et al. used a redox-mediator to investigate limitations in measured at different state-of-discharge (SOD), SOC, and at
the electronic conductivity.7 different current densities with a focus on three states of the
In this study, we use electrochemical impedance spectrosco-
py (EIS) that is often used as a diagnostic tool to identify the Received: November 6, 2014
underlying mechanisms of the polarization curves in electro- Accepted: January 27, 2015
chemical systems such as lithium-ion batteries and fuel cells, as Published: January 27, 2015
© 2015 American Chemical Society 4039 DOI: 10.1021/am5083254

ACS Appl. Mater. Interfaces 2015, 7, 4039−4047
ACS Applied Materials & Interfaces Research Article
Li−O2 battery electrochemistry: the discharge plateau, sudden To distinguish impedance contributions from the anode and
death at the end of discharge, and the initial stage of the the cathode in a two-electrode cell, it is often necessary to vary
charging process. physical parameters that will affect the two electrodes
By combining the measurements with previous results differently. We used three methods: (i) measuring impedance
presented by McCloskey and Luntz et al.,1,4,5 the EIS at open-circuit voltage (OCV) in argon atmosphere to prevent
measurements are related to the measured potential. the oxygen reduction/oxidation, (ii) using a symmetrical cell of
We show that the overpotential during discharge is caused by two predischarged cathodes, and (iii) testing a different cathode
internal resistance and is dominated by the charge transport (see Figure 4 below and Figures S2 and S3 in the Supporting
through the deposited Li2O2 at the end of discharge. During Information). In the symmetrical cell, the anode/cathode
charge, however, the potential increase reflects a mixed reactions are oxidation/reduction of Li2O2, which remove any
potential of parasitic reactions and Li2O2 oxidation. lithium metal-related contributions from the EIS measurement.
Both cathodes in the symmetrical cell initially discharged 0.25
2. EXPERIMENTAL SECTION mAh in separate cells before they were combined in a new cell.
Most of the experimental setup and procedures have been The cathodes were rinsed with DME after the individual
described in detail previously in ref 11 and are only briefly discharge to remove the electrolyte salt before the cathodes
described here. The cathodes were prepared by air-spraying a were used in the symmetrical cell. The symmetrical cell was
carbon/PTFE dispersion onto a 316SS 100 mesh (TWP, Inc., tested in O2 gas and was made without exposing the cathodes
Berkeley, CA). The slurries were prepared by sonicating a to air at any point.
carbon black powder (Vulcan XC72, www.fuelcellstore.com) 2.2. Modeling Li−O2 Impedance. The impedance is
and PTFE (60 wt % dispersion in water, Sigma-Aldrich) in a defined as the derivative of the iv curve:
3:1 wt/wt ratio in a 20:80 isopropanol/water mixture. A Badger ∂v ∂η
model 350 air sprayer was used to uniformly coat the SS mesh Z (i ) = =
∂i ∂i (1)
(the SS mesh was rinsed in acetone several times prior to
cathode preparation). Prior to cutting 12 mm diameter where v is the potential, i is the current density, and η is the
cathodes from the carbon-coated SS mesh, the mesh was overpotential. Therefore, the impedance is linked closely to the
allowed to air-dry for 1 h. All cathodes were dried in vacuum at Tafel plot, which has previously been used to describe reaction
120 °C for at least 12 h, washed in pure 1,2-dimethoxyethane mechanisms in Li−O2 batteries.4,10,16,17 From the Tafel
(DME) in a glovebox, followed by a second drying under equation, the overpotential is seen to be proportional to log(i)
vacuum for 10 min, and then at 200 °C in the glovebox for at at large overpotentials (|η| ≫ RT/nF), but as our batteries
least 1 h. A typical carbon loading was two milligrams per contain a porous cathode, this ideal behavior is not applicable.
cathode. The consequences of a porous electrode have been investigated
All solvents and salts in this study were purchased from by Lasia and show that the Tafel slope will increase at higher
Novolyte (Purolyte electrolyte grade), stored in an argon currents.18 This is in line with our measurements as well as
glovebox (0.1 ppm of O2 and H2O) and used without further previous Li−O2 battery measurements by Viswanathan et al.
purification. The H2O content of solvents was periodically and Adams et al. on porous electrodes.4,8 To describe the
checked with a Karl Fischer titrator (Metrohm Inc.) and found measurements better
to be no more than 20 ppm. The measurements were made
with an 11 mm diameter lithium metal anode, a 12.5 mm η = |v − OCV| = c1·i c2 (2)
diameter Celgard 2500 separator, a 12 mm diameter cathode, is applied as an empirical model, when eq 1 is used to compare
and 60 μL of 1 M bis(trifluoromethane)sulfonimide lithium salt the measured impedance with the overpotential. OCV is the
(LiTFSI) dissolved in DME. open circuit potential, and c1 and c2 are constants. As c2 is found
The test cell used in this study is shown in Figure S1 in the to be less than 1 in Section 3.4, Z(i) is expected to be larger at
Supporting Information. The cell components were stacked small currents according to eq 1.
between SS alloy 20 anode and cathode tips that were The measured impedance response can, to a first
hermetically sealed against a quartz tube using compressed approximation, be described using an equivalent circuit model
Markez O-rings (Marco Rubber). Capillaries were silver consisting of three Voigt elements (parallel connected resistor
soldered into the cathode tip to allow gases to be fed to and with a constant phase element (CPE)) connected in series, see
swept away from the cell. Gases swept away from the cell could Figure S4a in the Supporting Information. The impedance of
be quantitatively identified using the DEMS setup described in the Voigt elements is adopted from Hirschorn et al.,19 and the
detail in ref 10. total impedance, Z(ω), is thus given by
2.1. Electrochemical Impedance Spectroscopy. All
electrochemical measurements were made with a BioLogic Ri
Z(ω) = R s + ∑
VMP3 potentiostat. Electrochemical impedance spectra were i = 1,2,3
1 + (jω)ni Q iR i (3)
measured while a current was drawn (GEIS) to investigate the
processes under relevant conditions as discussed previously by where ω is the angular frequency, and Ri, Qi, and ni are
Adams et al.8 Impedance was measured at currents between 15 parameters in Voigt element i. Even though the model is not
μA (13 μA/cm2) and 1 mA (0.88 mA/cm2). Frequencies anchored in an electrochemical model, it is likely that key
between 4 mHz and 100 kHz were investigated with 15 points physical processes like charge transfer reactions, diffusion, and
per decade and an alternating current (AC) amplitude of 10% electronic transport through the Li2O2 layer will dominate one
of the direct current (DC) level. This typically gave an AC or more of the observed features in the spectra. Therefore,
potential response amplitude of 2−5 mV, which was found to parametrization using the simplified model makes it possible to
be within the linear regime, while still ensuring a sufficient determine the magnitude of these processes, although each
signal-to-noise ratio. feature may contain contributions from multiple physical
4040 DOI: 10.1021/am5083254
processes, and anode and cathode processes may overlap to with hindered water removal during operation.32 This
some extent. The sum of the frequency-dependent resistances approach, however, requires interpolation, which is difficult to
R1, R2, and R3 is the polarization resistance, Rp. The equivalent apply in this case with a dramatically decreasing cell voltage
circuit fits are made using the scipy optimizer fmin_slsqp using toward the end of discharge of the battery. Without use of such
the software package RAVDAV 0.9.7.20 methods, one can reduce drift problems by decreasing the
The Voigt elements have three parameters: R, Q, and n. R is measurement time or decreasing the change of the system. To
the DC resistance, and Q and n are parameters of the CPE. If n do this, we optimized the frequency range and compared
= 1, the CPE is a capacitor, and even if n is between 0.7 and 1, a impedance spectra from stable low-current measurements (18
pseudocapacitance, C*, can be calculated. This criterion is met μA/cm2 cathode) with impedance spectra from measurements
in all measurements presented in this work, except for the very at less stable but more realistic current densities (>0.2 mA/cm2
end of the 20 μA discharge presented in Figure 3. As discussed cathode).
in detail by Jamnik et al.,21 this capacitance is typically a double- The Kramers−Kronig relation was used to evaluate the
layer capacitance related to the process, and by comparing with causality of all measurements. The largest deviations occur at
reference values, it is possible to estimate the surface area low frequencies as the electrochemistry changes during the
contributing to the process. This can be used to distinguish measurement. To minimize this effect, a frequency cutoff level
reactions at the flat lithium anode from reactions in the porous of 5% deviation of the Kramer−Kronig transform was used.
cathode, since the surface areas of these are ∼1 cm2 and 1 m2,
respectively. The capacitance at the lithium metal surface in an 3. RESULTS
organic electrolyte is typically 10−20 μF/cm2 as reported by All measurements were performed using a system with an
Aurbach et al.,22,23 and the capacitance of XC72 is 12.6 F/g in XC72 carbon black cathode, DME/LiTFSI electrolyte, and
an organic aprotic electrolyte as reported by Barbieri et al.24 lithium anode. This system has been characterized extensively
From this, it is calculated that the capacitances should be in the in previous publications from 2011 to 2013 by McCloskey et
range of 10 μF and 25 mF for the anode and cathode, al.5,10−15 The most important methods used in these studies are
respectively. These values should then be compared with the differential electrochemical mass spectrometry (DEMS),
pseudocapacitance, calculated from the equivalent circuit peroxide titration, and X-ray diffraction.
parameters according to Hirschorn et al.19 In this work, we used DEMS to quantify gas consumption
and release during discharge and charge at all investigated
⎛ R R ⎞(1 − n)/ n current densities to verify that the measured impedance can be
C* = Q1/ n⎜ Ω ⎟ related to previous work. A ratio of 2.0 e−/O2 was observed
⎝ RΩ + R ⎠ (4)
during discharge at all investigated currents between 10 μA (8.8
where R, Q, and n are fitting parameters from the Voigt μA/cm2) and 5 mA (4.4 mA/cm2). During charge, the amount
elements, and RΩ is the DC resistance at the investigated of oxygen released corresponded to 2.5 e−/O2 until a potential
frequency. As discussed by Zoltowski et al., the pseudocapa- of 3.7 V. Above this potential, the ratio changed to 3 e−/O2.
citance of a CPE element is not well-defined,25 which means CO2 was evolved at potentials above 4.2 V.
that the surface area obtained using C* might vary slightly from The OCV was measured as a function of discharge and
the actual surface area, but the order of magnitude and relative charge to ensure an accurate determination of the over-
changes are still valid. potential. A full discharge−charge cycle at 250 μA (220 μA/
The capacitance is expected to change during discharge as cm2) is seen in Figure 1. The battery was allowed to relax to
the dielectric Li2O2 is deposited. The relative permittivity εr of
Li2O2 has been measured to be 30−35 by Gerbig et al. and
Dunst et al.26,27 Using a value of 30 to calculate the capacitance
of the Li2O2 layer in series with a typical electrode−electrolyte
capacitance of 20 mF, a Li2O2 layer of 8 nm will halve the
cathode capacitance. A similar calculation can be made for the
Li2CO3 interface layer between the cathode and the Li2O2.
Using the relative permittivity of Li2CO3 of 4.9 measured by
Young et al.,28 the capacitance will be halved with a layer
thickness of 1 nm.
The role of oxygen diffusion in the electrolyte has been
discussed in several papers.29−31 To evaluate the significance of
diffusion, the Damköhler number, Da, can be used as a quick
comparison between the oxygen consumption/evolution with
the diffusion rate.29 Using typical values for our system, Da is Figure 1. Measurement of OCV through a 250 μA (220 μA/cm2)
0.3 at a current of 250 μA, which means that the diffusion rate discharge and charge. The steep voltage transients occur when the
is ∼3 times higher than the consumption rate. Oxygen diffusion battery is allowed to relax at OCV.
is therefore not expected to be dominating, but it will have
some significance. OCV by interruption of the current a number of times during
Basic requirements for carrying out EIS measurements are both discharge and charge, which is seen as steep voltage
that the system is stable, causal, and linear. Among other things, transients in Figure 1. The relaxation criteria were a change in
this implies that no (or only a negligible) change in voltage and cell voltage of less than 1 mV/h or a relaxation time of 15 h.
impedance characteristics is allowed during the measurement The initial OCV was 3.2 V. The OCV decreased to 2.85 V after
period. Methods have been proposed to deal with impedance a short period of discharge and stayed at this value during the
measurements in nonstationary systems such as a PEM fuel cell entire dischargealso after reaching the 2.0 V cutoff at sudden
4041 DOI: 10.1021/am5083254
Figure 2. Nyquist plot (a) and (b) and Bode-like plot (c) and (d) of impedance measurements during a 250 μA (220 μA/cm2) constant current
discharge. The approximate SODs are shown in (e) and in the legends of (a) and (b). Three processes are identified and named Z1, Z2, and Z3, and
the corresponding peak frequencies are within the gray intervals marked in (c) and (d) at all current densities and SODs investigated.
Table 1. Peak Frequencies, Resistances, and Pseudocapacitances from Selected Impedance Fita
f1 [Hz] f 2 [Hz] f 3 [Hz] R1 [Ω] R2 [Ω] R3 [Ω] C1* [mF] C2* [mF] C3* [mF]
discharge at 250 μA
0.16 mAh 733 5.4 93 × 10−3 96 56 145 2.3 × 10−3 0.53 10.2
0.51 mAh 605 3.4 184 × 10−3 94 92 188 2.8 × 10−3 0.45 4.0
discharge at 20 μA
0.5 mAh 470 1.15 5.5 × 10−3 109 50 1007 3.1 × 10−3 2.8 19.3
1.9 mAh 464 1.12 9.9 × 10−3 107 158 2131 3.2 × 10−3 0.6 2.0
2.3 mAh 479 1.1 × 10−3 114 14 097 2.9 × 10−3 0.8
charge at 250 μA
0.03 mAh 678 9.6 267 × 10−3 65 166 497 3.6 × 10−3 76 × 10−3 1.0
0.42 mAh 983 14.0 19 × 10−3 255 105 700 0.6 × 10−3 99 × 10−3 9.0
a
The expected capacitances for the full anode and cathode are 10 μF and 25 mF, respectively. Typical values of n are n1 = 0.77, n2 = 0.86, and n3 =
0.78.
death. During charge, the OCV was 2.85 V, but it increased indicate that Z1 originates from an anode process, and Z2 and
slightly toward the end of charge where it reached 3.2 V. Z3 originate from cathode processes. The cathode blocking and
3.1. Discharge to Sudden Death at 250 μA. EIS identification of reaction processes in the impedance spectra are
measurements from the first discharge at 250 μA (220 μ/cm2) discussed further in Section 4.1.
are shown in Figure 2a−d. The spectra were measured while The peak frequencies changed between different current
drawing a current, which means that the SODs shown in Figure densities and close to sudden death. In all of our measurements,
2 and Table 1 are approximate values. Three arcs are however, f1 was between 100 Hz and 10 kHz, f 2 was between 2
distinguished in the Nyquist plot in Figure 2a. They were and 100 Hz, and f 3 was between 20 mHz and 1 Hz. These
almost constant in the first part of the discharge but changed as intervals are shown in Figure 2c,d, and the clear separation
the potential decreased near the end of discharge. The three helps in identifying the different impedance contributions.
identified impedance contributions are labeled Z1, Z2, and Z3, 3.2. Discharge to Sudden Death at 20 μA. We decreased
and, on the basis of a fit to the equivalent circuit given in eq 3 the discharge current to 20 μA (18 μA/cm2) to increase the
and shown in Figure S4a in the Supporting Information, the stability of the system during the impedance measurements, see
corresponding peak frequencies, resistances, and pseudocapa- Figure 3. When comparing this with the previous discharge at
citances are calculated. The values from two of these 250 μA presented in Figure 2, it is important to note that both
calculations are given in Table 1. the capacity and the polarization resistance, Rp, are significantly
It is seen that R1 is constant through the discharge, whereas larger in the 20 μA discharge.
R2 and R3 increase and C3* decreases. The decrease of C3* and On the basis of a fit using eq 3, representing the equivalent
increase of R3 through the discharge could be a blocking of the circuit presented in Figure S4a in the Supporting Information,
cathode surface. The magnitudes of the pseudocapacitances the resistance and pseudocapacitance parameters of Z1, Z2, and
4042 DOI: 10.1021/am5083254
Figure 4. Bode plot and Nyquist plot (inset) of a potentiostatic EIS

measurement at OCV in argon atmosphere. The spectrum is modeled
with an R-RQ-RQ-RQ-C circuit, shown in Figure S4c in the Supporting
Information.
3.4. iv Curve at the Discharge Plateau. As discussed in

Section 2.2, the impedance is the slope of the iv-curve at a given
current. To obtain a full understanding of the relationship
between the impedance and the overpotential, it is necessary to
investigate the current dependence of the impedance. We did
this by measuring the impedance at the plateau at ∼40% of the
total capacity at different current densities and compared this
Figure 3. Resistances and normalized pseudocapacitances determined with the corresponding iv curve. As the impedance is almost
from EIS measurements in a 20 μA (18 μA/cm2) constant current constant in the first part of a discharge, the exact time of
discharge to 2.2 V using eq 3. Nyquist plots are shown at three measurement was of less importance. To avoid effects of
representative stages, and the corresponding SODs are marked with degradation in the battery, each point in the plot was made with
circles on the voltage profile. R2 and C2* could not be determined well a fresh battery. To eliminate variations due to different masses
at the end of discharge and are thus greyed out. of the cathodes, both currents and impedances were weighed
with the mass of carbon.
Z3 are presented in Figure 3 and summarized in Table 1. The Figure 5 shows the iv curve of the plateau potential as a
parameters of Z1 were constant through the entire discharge, function of current density (red dots). The values are fitted
and the change of parameters related to Z2 and Z3 are divided
into three parts as indicated in Figure 3: (1) At 0−40% SOD,
only negligible change was observed, (2) at 40−80% SOD, R2
and R3 increased 2−3 times, and C2* and C3* decreased by 95%,
and (3) at 80−100% SOD, R3 increased exponentially to 14.1
kΩ at 90% SOD (more than 10 times the initial value), the
pseudocapacitances stayed at ∼5% of the initial value, and the
voltage dropped. Parameters related to Z2 could not be
determined in the last part of the discharge because of an
overlap with Z3.
At 20 μA, the average relative Kramers−Kronig deviation at
frequencies from 1 mHz to 10 Hz was typically 0.5% at the
plateau, increasing near sudden death to 2% at 2.2 V. The n3
value got below 0.7 in the end of the discharge to typical values
of 0.62. This means that the pseudocapacitance C*3 is less
meaningful to calculate.
3.3. EIS Measurement in Argon. A potentiostatic EIS Figure 5. Plateau voltage dependence on current density (red dots)
measurement at OCV with 5 mV amplitude was made on a from 10 μA (9 μA/cm2) to 5 mA (4.4 mA/cm2). Equation 2 is fitted
fresh battery in argon atmosphere before exposure to oxygen to to the data (red line), which is then differentiated (black line) and
investigate reactions not related to oxygen reduction. The compared with the total resistance (black dots) measured with
impedance. Three representative discharge curves show how the
measurement is shown in Figure 4, and clearly, reactions were
plateau voltage is determined. The impedance and current density
still taking place in the absence of oxygen, as Z1 and Z2 were were weighed by the carbon mass of each electrode.
still present. The low frequency tail could be modeled with a
capacitor, C3, which means that no charge transfer reaction is
present for this process. The spectrum was modeled with the with eq 2 (red line) with OCV, c1, and c2 as fitting parameters.
equivalent circuit R-RQ-RQ-RQ-C, shown in Figure S4c in the The result is OCV = 2.78 V, c1 = 12 mV·(gC/mA)0.44 and c2 =
Supporting Information. The capacitance C3 was 18.3 mF, and 0.44, which correspond to a Tafel slope of 120 millivolts per
the pseudocapacitances C1* and C2* were 1.2 μF and 0.2 mF, decade at 180 mA/gC. This is in line with previous publications
respectively. The presence of Z2 suggests that this process is not by Viswanathan et al. and Lu et al.4,17 The iv curve fit is
related to oxygen reduction. differentiated (black line) and compared with the measured
4043 DOI: 10.1021/am5083254
impedance (black dots) in accordance with eq 1. It is seen that is further noted that C3* is almost the same in the end of charge
the measured impedance followed the same trend, but that it and in the beginning of the discharge (10.2 mF). Finally, it is
was higher than expected based on the iv curve. The reason for noted that R1 was almost constant until 3.7 V, after which it
this discrepancy is likely related to a chemically induced suddenly increased. This supports that the lithium anode
parasitic side reaction that becomes more pronounced at lower surface is deactivated by an accelerated formation of the solid
discharge rates. This will be discussed further in Section 4.1. electrolyte interface (SEI) layer − possibly due to oxygen
3.5. Charge at 250 μA. In Figure 6, we present typical EIS crossover.
measurements during a charge. To limit the complexity of the
4. DISCUSSION
To gain electrochemical knowledge of the fundamental
reactions and bottlenecks during discharge and charge of the
Li−O2 battery, a series of electrochemical impedance spectra at
different current densities and SOCs has been measured. It was
seen that three impedance contributions were present during
both discharge and charge, and they are referred to as Z1, Z2,
and Z3. The five key findings were that
(i) the impedance did not change at the discharge plateau,
(ii) Z2 and Z3 increased near sudden death,
(iii) C3* decreased significantly just before sudden death,
(iv) pseudocapacitances related to Z1, Z2, and Z3 were
typically 3 μF for C1*, 0.1−3 mF for C2*, and 1−20 mF for
C3*, and
(v) the OCV was always 2.85 V during discharge and the
initial stages of charge, then slowly increased to 3.2 V at
the end of charge.
4.1. Identification of Processes during Discharge. Our
results support previous findings by Adams et al. and Landa-
Medrano et al. in refs 8 and 9 that Z1 originates from the anode
and that Z2 and Z3 originate from the cathode. In addition to
this, our results show that Z3 is a combination of the charge
transfer reaction of oxygen reduction and the electronic
transport through the Li2O2, whereas Z2 is a cathode-specific
Figure 6. Nyquist (a) and Bode-like (b) plots of EIS measurements process that is not related to the oxygen reduction. The
made during a 250 μA (220 μA/cm2) constant current charge. The
assignment of anode and cathode features in the EIS is
charge followed a discharge similar to that shown in Figure 2 with a
discharge resistance extrapolated to 3 kΩ at 2.0 V. The SOCs are substantiated by the following three observations.
shown as circles in the inset graph of the voltage profile. The 0.03 mAh First, the full capacitance at the lithium anode surface is
and 0.42 mAh measurement was modeled using eq 3, and the obtained expected to be in the range of 10 μF as discussed in Section 2.2,
parameters are presented in Table 1 whereas the capacitance of the XC72 electrode is expected to
be 25 mF. If only part of an electrode is active during the EIS
measurement, the capacitance will be lower. As reported, C*1
analysis, impedance measurements are only made at voltages was typically 3 μF, whereas C*2 and C*3 were in the range of
below 4.2 V to avoid the major decomposition reactions 0.1−20 mF. Furthermore, the cathode capacitance per active
observed at higher potentials using DEMS. In this measure- surface area calculated by Adams et al. was in the same range as
ment, the 4.2 V limit corresponded to 60% SOC. C3*.8
Three impedance contributions are identified. The correla- Second, the careful parameter study presented by Adams et
tion between the impedance and SOC is more complex than al. in ref 8 shows that relevant cathode processes have peak
during discharge. The spectrum is dominated by a high- and a frequencies below 10 Hz, which correspond with f 2 and f 3 in
low-frequency response similar to Z1 and Z3 during discharge. It our study, whereas the peak frequency of the anode process is
seems like the frequencies between 1 and 100 Hz are ∼1 kHz, which corresponds to f1 in our study.
dominated by a mix of different processes appearing at certain Third, Z1 did not change during discharge, whereas Z2 and Z3
SOCs and then disappearing at higher SOC, but further studies increased significantly close to sudden death. Both electrodes
are needed to qualify this. The parameters obtained using changed during the measurement. On the lithium anode,
equivalent circuit fitting on the green (0.03 mAh) and black Younesi et al. have previously shown that an SEI layer is
(0.42 mAh) spectra with eq 3 are given in Table 1. The three forming in a combination of chemical and electrochemical
contributions are in the same frequency ranges as seen during reactions,33 but as shown by McCloskey et al. this is affecting
discharge. The polarization resistance, Rp, was almost constant neither the electrochemistry nor the measured impedance.4 On
in the range of 500−1000 Ω, but the peak frequencies and the the other hand, the cathode is covered with an insulating layer
relative magnitude of the different impedance contributions of mainly Li2O2 during discharge, and an increase in charge
changed. Looking at the pseudocapacitances, C*1 decreased transfer resistance is typically captured in EIS measurements.
from 3.6 μF to 0.6 μF, and C3* increased from 1.0 mF to 9.0 Ascribing Z3 to oxygen reduction and electronic transport
mF. This suggests that the active cathode area is increasing and through Li2O2 is based on two observations: (i) Z3 is the only
that the active area of the anode is decreasing during charge. It process related to oxygen reduction, as both Z1 and Z2 are
4044 DOI: 10.1021/am5083254
present in argon, and (ii) R3 was the largest resistance during and the current cannot be supported within the cutoff limit of
the entire discharge, both when the cathode was limited by 2.2 V. This is in full agreement with observations made by
reaction kinetics at the discharge plateau and by electronic Luntz et al. using flat glassy carbon electrodes in electrolysis
conduction at sudden death. cells.1
The process related to Z2 is cathode specific and not related Because of discussions in literature on the significance of
to oxygen reduction, as it was present in argon but almost oxygen diffusion in the electrolyte, it is worth mentioning that
absent in a measurement with P50 carbon paper, shown in the sudden death is not due to pore clogging and increased
Figure S3 in the Supporting Information. As P50 is binder-free, oxygen diffusion resistance. In a typical discharge, the average
this could indicate a degradation effect related to the PTFE thickness of Li2O2 is 0.5−1 nm based on the BET surface area
binder. A little surprising, Z2 was not present in the symmetrical of XC-72. This means that the porosity and Damköhler number
cell measurement. This could indicate that the parasitic reaction are almost unchanged during the entire discharge, and as stated
was chemically passivated during handling when assembling the by Wang et al.,29 such small changes will not give rise to the
symmetrical cell. sudden death behavior.
The iv curve in Figure 5 was made to ensure that all 4.3. Analysis of the Overpotential during Charge. The
electrochemical processes were captured in the impedance EIS measurements from the charge confirm that the electrical
spectrum. This was indeed the case, since the total impedance resistivity through Li2O2 decreases in charge mode as proposed
could account for the changes in the overpotential. Actually, the by Luntz et al. using flat glassy carbon electrodes1 and show
measured impedance seemed to overestimate the slope of the iv that the voltage increase during charge is a mixed potential
curve, and the reason is most likely a result of lower Li2O2 rather than an increase in internal resistance, as McCloskey et
formation yields, and therefore more heterogeneous discharge al. have also suggested based on modeling.10
electrodeposits, at lower current densities as shown in a The change in resistivity in charge mode is identified by
previous publication.5 Garcia-Lastra et al. and Mekonnen et al. comparing the impedance at the end of discharge with the
have shown that an increase of Li2CO3-like inclusions in the resistance in the beginning of the charge. During charge, the
Li2O2 layer can change the electrical conductivity using DFT polarization resistance was ∼500 Ω at a current of 250 μA (220
calculations,34,35 and such changes would also change the μA/cm2), which is much lower than the extrapolated value of 3
current dependence of the impedance and explain the kΩ at 2.0 V during discharge. Furthermore, the charge
deviation. resistance had only little dependence on the discharge current
4.2. Analysis of the Overpotential during Discharge. and depth of discharge, which suggests that the charge is not
The measurements show that the electrochemistry was limited by the same process as the discharge. Luntz et al. have
unchanged during the entire discharge, and they support the previously explained this by a reduction of the tunneling barrier
general understanding of tunneling being the dominant charge because of a change in the Fermi energy by experiments on flat
transport mechanism through the Li2O2 layer at relevant glassy carbon electrodes in an electrolysis cell.1
current densities and temperatures, which was initially The mixed potential is identified because the impedance was
proposed by Albertus et al.16 and confirmed by Luntz et al.1 not increasing as the charging potential increased, which
Furthermore, the discharge was initially occurring in the entire indicates a change of reaction mechanisms. As shown by DEMS
cathode, whereas the increasing electronic transport through and Li2O2 titration, the charging reaction is not a 2 e−/O2
the growing Li2O2 layer passivated large parts of the cathode process but rather a 2.5−3 e−/O2 process, and parasitic
during discharge. electrochemical reactions are thus present during the entire
The tunneling mechanism is supported by two observations. charge.5,10,11,15 Keeping in mind that the OCV never exceeded
First, the impedance contribution Z3 related to oxygen 3.2 V during charge, and no significant resistance increase was
reduction and electronic conduction through Li2O2 was seen in the impedance spectra, it suggests that a mixed potential
constant at the discharge plateau and increased rapidly near between these competing electrochemical reactions was
sudden death, which is characteristic for the tunneling barrier established during charge to support the high current.
that depends exponentially on the Li2O2 layer thickness, and These results contradict the theory proposed by Chen et al.
second, the electrochemistry was unchanged during the suggesting that the increase in charge overpotential occurs
discharge, as shown by a constant 2 e−/O2 process and because the Li2O2 closest to the electronically conducting part
identification of the same three processes in the impedance of the cathode oxidizes first.7 If this was the case, an increase of
spectra at all SODs. the charge resistance of at least an order of magnitude would be
Passivation of the cathode is observed in the pseudocapa- expected to explain the voltage increase, but the resistance does
citance C*3 . At 20 μA, the initial value is 21 mF. This is the not increase by more than a factor of 2. Furthermore, after
expected value of the entire cathode, which means that Li2O2 discharging under alternating O2 isotope atmospheres, Li2O2
deposition is occurring in the entire cathode. The decrease in oxidation was found to preferentially occur at the Li2O2/
stage 1, as defined in Figure 3, reflects Li2O2 formation, because electrolyte interface over the Li2O2/cathode interface during
the introduction of a dielectric material in a capacitor changes the initial stages of charge, as shown in a previous publication.14
the capacitance. In stage 2, the decrease is significant, cannot be Interestingly, R1 increased four times when the battery
explained by the dielectric layer of Li2O2 alone, and must reached 4 V, and C*1 decreased to 20%. This suggests significant
therefore reflect a reduction in active surface area. The cathode anode degradation and is in line with previous work by Younesi
passivates when the critical thickness of Li2O2 is reached and et al. showing how the SEI layer changes on the lithium metal
tunneling is no longer possible. In stage 3, the available surface during charge of a Li−O2 battery.33 At this point it is not
area is not sufficient to support the constant current, and the possible to determine whether this change is caused by
voltage drops to enable conduction through the blocked parts degradation of the anode or an overlapping cathode process,
of the electrode. This is seen as an increase in cathode but if the increase is because of anode degradation, this will be
resistance. When fully discharged, the resistance is too large, important to prevent in a commercial Li−O2 battery.
4045 DOI: 10.1021/am5083254
5. CONCLUSION (2) Hummelshøj, J. S.; Blomqvist, J.; Datta, S.; Vegge, T.; Rossmeisl,
J.; Thygesen, K. S.; Luntz, A. C.; Jacobsen, K. W.; Nørskov, J. K.
By measuring EIS spectra at different current densities and Communications: Elementary Oxygen Electrode Reactions in the
different SOCs it was possible to assign the three identified Aprotic Li-Air Battery. J. Chem. Phys. 2010, 132, 071101−1−071101−
contributions to either the cathode or the anode. Only one of 4.
the two cathode processes depended on the presence of (3) Hummelshøj, J. S.; Luntz, A. C.; Nørskov, J. K. Theoretical
oxygen. This indicates that this contribution was related to the Evidence for Low Kinetic Overpotentials in Li-O2 Electrochemistry. J.
Li2O2 formation. The other contribution was cathode specific Chem. Phys. 2013, 138, 034703−1−034703−12.
and may reflect a degradation reaction related to the PTFE (4) Viswanathan, V.; Nørskov, J. K.; Speidel, A.; Scheffler, R.; Gowda,
binder. S. R.; Luntz, A. C. Li−O2 Kinetic Overpotentials: Tafel Plots from
During discharge, the rapid potential change near the end of Experiment and First-Principles Theory. J. Phys. Chem. Lett. 2013, 4,
discharge was due to an increase in polarization resistance, 556−560.
(5) McCloskey, B. D.; Valery, A.; Luntz, A. C.; Gowda, S. R.; Wallra,
primarily related to the charge transport through the Li2O2.
G. M.; Garcia, J. M.; Mori, T.; Krupp, L. E. Combining Accurate O2
This supports previously published work by Luntz et al. in ref 1, and Li2O2 Assays to Separate Discharge and Charge Stability
which states that the electronic transport through Li2O2 at Limitations in Nonaqueous Li-O2 Batteries. J. Phys. Chem. Lett.
relevant current densities is governed by tunneling. 2013, 4, 2989−2993.
In the initial part of the charge, the impedance was low (6) Zhong, L.; Mitchell, R. R.; Liu, Y.; Gallant, B. M.; Thompson, C.
compared to the end of discharge at sudden death. This V.; Huang, J. Y.; Mao, S. X.; Shao-Horn, Y. In Situ Transmission
supports that the electronic conductivity is improved when Electron Microscopy Observations of Electrochemical Oxidation of
changing to charge mode, which has also been shown in a Li2O2. Nano Lett. 2013, 13, 2209−2214.
previous work on smooth glassy carbon cathodes in an (7) Chen, Y.; Freunberger, S. A.; Peng, Z.; Fontaine, O.; Bruce, P. G.
electrolysis cell.1 During charge, the voltage increased Charging a Li-O2 Battery Using a Redox Mediator. Nat. Chem. 2013, 5,
significantly, whereas the resistance and OCV were almost 489−494.
(8) Adams, J.; Karulkar, M.; Anandan, V. J. Evaluation and
unchanged, and DEMS measurements identified the presence
Electrochemical Analyses of Cathodes for Lithium-Air Batteries. J.
of parasitic reactions. This suggests that the electrochemistry Power Sources 2013, 239, 132−143.
changed during charge and that the voltage increase was due to (9) Landa-Medrano, I.; Ruiz de Larramendi, I.; Ortiz-Vitoriano, N.;
a mixed potential of parasitic reactions and Li2O2 oxidation, Pinedo, R.; Ignacio Ruiz de Larramendi, J.; Rojo, T. In Situ Monitoring
established to support a constant current.
■
of Discharge/Charge Processes in Li-O2 Batteries by Electrochemical
Impedance Spectroscopy. J. Power Sources 2014, 249, 110−117.
ASSOCIATED CONTENT (10) McCloskey, B. D.; Speidel, A.; Scheffler, R.; Miller, D.;
*
S Supporting Information Viswanathan, V.; Hummelshøj, J. S.; Nørskov, J. K.; Luntz, A. C. Twin
Schematic illustration of the test cell, equivalent circuit Problems of Interfacial Carbonate Formation in Nonaqueous Li-O2
Batteries. J. Phys. Chem. Lett. 2012, 3, 997−1001.
diagrams, and EIS measurements on modified systems that
(11) McCloskey, B. D.; Bethune, D. S.; Shelby, R. M.; Girishkumar,
suppress different features in the EIS spectrum made to support G.; Luntz, A. C. Solvents’ Critical Role in Nonaqueous Lithium-
the conclusions made in the article. These modifications are (1) Oxygen Battery Electrochemistry. J. Phys. Chem. Lett. 2011, 2, 1161−
using a symmetrical carbon−carbon cell and (2) using a 1166.
cathode with a different type of carbon. This material is (12) Gowda, S. R.; Brunet, A.; Wallraff, G. M.; McCloskey, B. D.
available free of charge via the Internet at http://pubs.acs.org. Implications of CO2 Contamination in Rechargeable Nonaqueous Li-
■ AUTHOR INFORMATION
Corresponding Author
O2 Batteries. J. Phys. Chem. Lett. 2013, 4, 276−279.
(13) McCloskey, B. D.; Scheffler, R.; Speidel, A.; Bethune, D. S.;
Shelby, R. M.; Luntz, A. C. On the Efficacy of Electrocatalysis in
Nonaqueous Li-O2 Batteries. J. Am. Chem. Soc. 2011, 133, 18038−
*E-mail: jhoj@topsoe.dk. Phone: (+45) 27292175. Fax: (+45)
18041.
45272999. (14) McCloskey, B. D.; Scheffler, R.; Speidel, A.; Girishkumar, G.;
Present Address Luntz, A. C. On the Mechanism of Nonaqueous Li-O2 Electro-
⊥
Bryan D. McCloskey: Department of Chemical and chemistry on C and Its Kinetic Overpotentials: Some Implications for
Biomolecular Engineering, University of California and Li-Air Batteries. J. Phys. Chem. C 2012, 116, 23897−23905.
Environmental Energy Technologies Division, Lawrence (15) McCloskey, B. D.; Luntz, A. C.; Bethune, D. S.; Shelby, R. M.;
Berkeley National Laboratory, Berkeley, CA 94720, United Mori, T.; Scheffler, R.; Speidel, A.; Sherwood, M. Limitations in
States. Rechargeability of Li-O2 Batteries and Possible Origins. J. Phys. Chem.
Lett. 2012, 3, 3043−3047.
Notes (16) Albertus, P.; Girishkumar, G.; McCloskey, B. D.; Sánchez-
The authors declare no competing financial interest.
■
Carrera, R. S.; Kozinsky, B.; Christensen, J.; Luntz, A. C. Identifying
Capacity Limitations in the Li/Oxygen Battery Using Experiments and
ACKNOWLEDGMENTS Modeling. J. Electrochem. Soc. 2011, 158, A343.
The authors acknowledge support of this work from the (17) Lu, Y.-C.; Gasteiger, H. A.; Shao-Horn, Y. Method Develop-
ReLiable project (Project No. 11-116792) funded by the ment to Evaluate the Oxygen Reduction Activity of High-Surface-Area
Catalysts for Li-Air Batteries. Electrochem. Solid-State Lett. 2011, 14,
Danish Council for Strategic Research Program Commission A70−A74.
■
(18) Lasia, A. Impedance of Porous Electrodes. J. Electroanal. Chem.
1995, 397, 27−33.
REFERENCES (19) Hirschorn, B.; Orazem, M. E.; Tribollet, B.; Vivier, V.; Frateur,
(1) Luntz, A. C.; Viswanathan, V.; Voss, J.; Varley, J. B.; Speidel, A.; I.; Musiani, M. Determination of Effective Capacitance and Film
Nørskov, J. K.; Scheffler, R. Tunneling and Polaron Charge Transport Thickness from Constant-Phase-Element Parameters. Electrochim. Acta
Through Li2O2 in Li-O2 Batteries. J. Phys. Chem. Lett. 2013, 4, 3494− 2010, 55, 6218−6227.
3499. (20) Graves, C. RAVDAV Data Analysis Software, Version 0.9.7.
4046 DOI: 10.1021/am5083254

(21) Jamnik, J.; Maier, J. Generalised Equivalent Circuits for Mass

and Charge Transport: Chemical Capacitance and Its Implications.
Phys. Chem. Chem. Phys. 2001, 3, 1668−1678.
(22) Aurbach, D.; Weissman, I.; Zaban, A.; Chusid, O. Correlation
Between Surface Chemistry, Morphology, Cycling Efficiency and
Interfacial Properties of Li Electrodes in Solutions Containing
Different Li Salts. Electrochim. Acta 1994, 39, 51−71.
(23) Aurbach, D.; Zaban, A.; Schechter, A.; Ein-eli, Y.; Zinigrad, E.;
Markovsky, B. The Study of Electrolyte Solutions Based on Ethylene
and Diethyl Carbonates for Rechargeable Li Batteries. I. Li Metal
Anodes. J. Electrochem. Soc. 1995, 142, 2873−2882.
(24) Barbieri, O.; Hahn, M.; Herzog, a.; Kötz, R. Capacitance Limits
of High Surface Area Activated Carbons for Double Layer Capacitors.
Carbon 2005, 43, 1303−1310.
(25) Zoltowski, P. J. On the Electrical Capacitance of Interfaces
Exhibiting Constant Phase Element Behaviour. J. Electroanal. Chem.
1998, 443, 149−154.
(26) Gerbig, O.; Merkle, R.; Maier, J. Electron and Ion Transport in
Li2O2. Adv. Mater. 2013, 25, 3129−3133.
(27) Dunst, A.; Epp, V.; Hanzu, I.; Freunberger, S. A.; Wilkening, M.
Short-Range Li Diffusion vs. Long-Range Ionic Conduction in
Nanocrystalline Lithium Peroxide Li2O2 - the Discharge Product in
Lithium-Air Batteries. Energy Environ. Sci. 2014, 7, 2739−2752.
(28) Young, K. F.; Frederikse, H. P. R. Compilation of the Static
Dielectric Constant of Inorganic Solids. J. Phys. Chem. Ref. Data 1973,
2, 313.
(29) Wang, Y.; Cho, S. C. Analysis of Air Cathode Perfomance for
Lithium-Air Batteries. J. Electrochem. Soc. 2013, 160, A1847−A1855.
(30) Mehta, M.; Mixon, G.; Zheng, J. . P.; Andrei, P. Analytical
Electrochemical Impedance Modeling of Li-Air Batteries under D.C.
Discharge. J. Electrochem. Soc. 2013, 160, A2033−A2045.
(31) Sandhu, S. S.; Fellner, J. P.; Brutchen, G. W. Diffusion-Limited
Model for a Lithium/Air Battery with an Organic Electrolyte. J. Power
Sources 2007, 164, 365−371.
(32) Schiller, C. A.; Richter, F.; Gülzow, E.; Wagner, N. Validation
and Evaluation of Electrochemical Impedance Spectra of Systems with
States that Change with Time. Phys. Chem. Chem. Phys. 2001, 3, 374−
378.
(33) Younesi, R.; Hahlin, M.; Roberts, M.; Edström, K. The SEI
Layer Formed on Lithium Metal in the Presence of Oxygen: A Seldom
Considered Component in the Development of the Li-O2 Battery. J.
Power Sources 2013, 225, 40−45.
(34) Garcia-Lastra, J. M.; Myrdal, J. S. G.; Christensen, R.; Thygesen,
K. S.; Vegge, T. DFT+U Study of Polaronic Conduction in Li2O2 and
Li2CO3: Implications for Li-Air Batteries. J. Phys. Chem. C 2013, 117,
5568−5577.
(35) Mekonnen, Y. S.; Knudsen, K. B.; Mýrdal, J. S. G.; Younesi, R.;
Højberg, J.; Hjelm, J.; Norby, P.; Vegge, T. Communication: The
Influence of CO2 Poisoning on Overvoltages and Discharge Capacity
in Non-aqueous Li-Air Batteries. J. Chem. Phys. 2014, 140, 121101.
4047 DOI: 10.1021/am5083254

Paper I - SI - EIS investigation of overpotentials in Li-O2 batteries 107
Supporting Information: An Electrochemical
Impedance Spectroscopy Investigation of the
Overpotentials in Li-O2 Batteries
Jonathan Højberg†,‡,§,*, Bryan D. McCloskey†, Johan Hjelm‡, Tejs Vegge‡, Keld Johansen§, Poul
Norby‡, Alan C. Luntz†,ǁ
†
IBM Almaden Research Center, San Jose, California 95120, United States
‡
DTU Energy Conversion, Frederiksborgvej 399, DK-4000 Roskilde, Denmark
§
ǁ
SUNCAT, SLAC National Accelerator Laboratory, Menlo Park, California 94025
*Corresponding author
Tel.: (+45) 27292175
Fax: (+45) 45272999
1
Figure S1. Schematic figure of the test cell used in this study. The inlet and outlet allow gasses
to be fed to and swept away from the cell and analyzed in a mass spectrometer.
2
Figure S2. Bode-plot and Nyquist-plot (inset) of a potentiostatic EIS measurement of a
symmetrical XC72 cell at 0 V with an ac amplitude of 5 mV. The cell was made of two cathodes
that had been discharged in separate cells with a lithium anode at 250 μA (220 μA/cm2). The
spectrum is modeled with a 𝑅-𝑅𝑄-𝑄 circuit, shown in Figure S4b.
A potentiostatic EIS measurement of a symmetrical cell at OCV with 5 mV amplitude was made
to eliminate contributions from the lithium anode in the EIS spectrum. The cell was made of two
cathodes that had been discharged 1 hour in separate cells with a lithium anode at 250 µA (220
µA/cm2). The measurement is shown in Figure S2. The spectrum has been modeled with a 𝑅-
𝑅𝑄-𝑄 circuit. The CPE element was chosen instead of a capacitor to describe the low frequency
tail, because the slope was -0.74 in the Bode plot, rather than -1 in the case of a capacitor. The
pseudo-capacitance of the low frequency tail was 2.1 mF and the resistance and pseudo-
capacitance of the 𝑅𝑄 circuit was 27 Ω and 4.3 µF, respectively.
3
Figure S3. Bode-like plot and Nyquist-plot (inset) of a 250 μA (220 μA/cm2) discharge GEIS of
an AvCarb P50 cathode carbon paper. The spectrum is modeled with a 𝑅-𝑅𝑄-𝑅𝑄-𝑅𝑄 circuit,
shown in Figure S4a.
A GEIS measurement on AvCarb P50 carbon paper prepared as described in Ref. 1 is shown in
Figure S3. Both 𝑍1 and 𝑍3 have similar parameter values compared to XC72 and 𝑍2 is very
small.
4
Figure S4. Equivalent circuit diagrams used to model the impedance measurements presented in
the paper. The most common element is a Voigt element (parallel connected resistor with a
constant phase element, CPE). (a) The equivalent circuit used the most. As discussed in the
paper, the contributions to the impedance can be attributed to either the anode (𝑍1) or the cathode
(𝑍2 and 𝑍3). (b) is used to model the data in Figure S2 and (c) is used to model data in Figure 4.
5
References
( 1) McCloskey, B. D.; Scheffler, R.; Speidel, A.; Bethune, D. S.; Shelby, R. M.; Luntz, A. C.
On the Efficacy of Electrocatalysis in Nonaqueous Li-O2 Batteries. J. Am. Chem. Soc.
2011, 133, 18038–18041.
6
Paper II - Reactions and SEI formation during charging of Li-O2 cells 113
ECS Electrochemistry Letters, 4 (7) A63-A66 (2015) A63

2162-8726/2015/4(7)/A63/4/$33.00 © The Electrochemical Society
Reactions and SEI Formation during Charging of Li-O2 Cells

Jonathan Højberg,a,b,z Kristian Bastholm Knudsen,b Johan Hjelm,b,∗ and Tejs Veggeb
a Haldor Topsøe A/S, DK-2800 Kgs Lyngby, Denmark
b Department of Energy Conversion and Storage, Technical University of Denmark, DK 4000 Roskilde, Denmark
In this letter we combine detailed electrochemical impedance measurements with quantitative measurements of O2 evolution and
Li2 O2 oxidation to describe the charge mechanisms during charge of Li-O2 batteries with porous carbon electrodes. We identify
Li2 O2 oxidation at 3.05 V and an apparent chemical formation of a solid electrolyte interface (SEI) layer as the first monolayer
of Li2 O2 is oxidized, leading to a voltage increase. The first electrochemical degradation reaction is identified between 3.3 V and
3.5 V, and the chemical degradation is limited above 3.5 V, suggesting that a chemically stable SEI layer has been formed.
© 2015 The Electrochemical Society. [DOI: 10.1149/2.0051507eel] All rights reserved.
Manuscript submitted March 31, 2015; revised manuscript received May 4, 2015. Published May 13, 2015.
The non-aqueous Li-O2 battery has received significant attention in the current decreased to 13 mA/gcarbon . 130 mA/gcarbon was chosen
the past years due to its potentially high specific energy and low cost, because it is within a commercially interesting range and comparable
which makes it ideal for future electric vehicles. The combination of to previous studies.7–9 The voltage profile is presented in Figure S3 in
metallic lithium as the negative electrode and reduction of molecular the Supporting Information.
oxygen at the positive electrode enable a theoretical energy density of DEMS measurements were performed at 130–260 mA/gcarbon and
3.8 kWh/kg including the weight of lithium and oxygen. the gas consumption and gas evolution were quantified using both
During discharge, oxygen is consumed to form the insoluble pressure measurements and mass spectrometry. The applied in-house
Li2 O2 ; a high bandgap product that will limit conduction of elec- DEMS setup is similar in design to the setup used by McCloskey
trons and holes to the surface when the growing layer reaches a critical et al.9
thickness.1,2 Low overpotentials (<0.2 V) have been predicted for dis- To assess the amount of Li2 O2 in the air electrode at different stages
charge and charge using density functional theory,3–6 and supported of charge, we used a spectrophotometric measurement to determine
experimentally by Luntz et al. using flat glassy carbon electrodes.2 the concentration of a Ti-complex. Li2 O2 was allowed to react with
However, practical batteries with large surface area cathodes display water to form H2 O2 that oxidizes TiOSO4 in the solution to form the
large overpotentials.7 This decreases the cycle efficiency significantly Ti-complex. The concentration of the Ti-complex was determined by
and open up potential dependent parasitic reactions during charge.8 measuring the absorbance at around 408 nm. A detailed description
Here, we focus on the initial part of the charge until a potential of the method is included in the Supporting Information.
of 3.6 V to understand why the potential increases as the battery is
charged. Understanding and ultimately solving this problem is an im- Results and Discussion
portant step toward commercialization of the Li-O2 technology. We
have analyzed Li-O2 batteries identical to a carbon based reference Figure 1b shows a typical charge curve, and Figure 2 shows differ-
system used in a number of previous publications.8–12 A differential ential capacity plots (dQ/dV) of such curves. The peaks correspond to
capacity plot of a galvanostatic charge is used to identify the onset voltage plateaus in the charge curve and thereby different processes,
of at least eight electrochemical reactions during a full charge and and from this, eight electrochemical processes can be identified at
using differential electrochemical mass spectrometry (DEMS), elec- 3.05 V, 3.3 V, 3.4 V, 3.5 V, 3.85 V, 4.2 V, 4.3 V and 4.5 V. These
trochemical impedance spectroscopy (EIS), and quantitative optical potentials form the basis of the following discussion. Analyses of 10
absorption spectroscopy; it was possible to explain why the initial charge measurements following a discharge to 2.6 V show a charge
low-overpotential oxidation of Li2 O2 does not continue and why the capacity below 3.15 V corresponding to 540 ± 80 μmolLi2 O2 /gcarbon .
voltage increases. The findings are illustrated in Figure 1a. This corresponds to 4.3% of the total discharge capacity or approxi-
mately one monolayer as calculated in the Supporting Information.
Experimental Quantification of Li2 O2 and O2 evolution.— Figure 3a shows the

All electrochemical measurements were performed using a oxygen evolution (blue line) and the Li2 O2 removal (red line) as the
2-electrode Swagelok cell with XC72 carbon black cathodes (Vul- cathode is charged. The oxygen evolution is determined based on the
can XC72, Cabotcorp, GA), 1 M LiTFSI (Sigma-Aldrich) in 1,2- DEMS measurements presented in Figure S1 in the Supporting In-
dimethoxyethane (BASF) electrolyte, a Whatman glass fiber sepa- formation and the Li2 O2 removal is based on the optical absorption
rator and lithium anode. LiTFSI was dried at 180◦ C for 12 h and measurements presented in Figure S2 in the Supporting Information.
1,2-dimethoxyethane was dried using 4 Å molecular sieves (Sigma- The O2 evolution and, in particular, the deviation from the theoretical
Aldrich). The carbon cathodes were manufactured by air-spraying a value is in accordance with measurements presented by McCloskey
slurry of XC72 Carbon Black and PTFE (60 wt% dispersion in water) et al.,7 and suggests the presence of electrochemical degradation re-
in a wt/wt ratio of 3:1 as described in Ref. 8. actions, especially at potentials above 3.5 V. The Li2 O2 is, however,
Electrochemical impedance spectroscopy (EIS) and galvanostatic disappearing more rapidly than expected from the electrochemistry,
discharge-charge curves were measured using Bio-Logic VMP3 and suggesting a significant chemical degradation. Figure 3b shows the
MPG-2 potentiostats. Potentiostatic EIS measurements were per- amount of chemical and electrochemical reactions in different poten-
formed at different charge potentials, Uch , from 3.1 V to 3.6 V. Fre- tial intervals, and it is clear that the chemical degradation is most
quencies between 20 kHz and 10 mHz were investigated with 15 significant in the potential ranges 2 V–3.1 V and 3.3 V–3.5 V. This
points per decade and an alternating current (AC) amplitude of 5 mV. effect is somewhat more pronounced than previously reported.7
All impedance measurements followed the procedure: a) discharge to
2.6 V at 130 mA/gcarbon followed by 150 min at 2.6 V, b) charge to Uch Electrochemical impedance spectroscopy (EIS).— EIS was mea-
at 130 mA/gcarbon , and c) continuous EIS measurements at Uch until sured at 11 different potentials during the initial charge from 3.10 V to
3.60 V. Figure S4 in the Supporting Information shows a typical mea-
surement, with the equivalent circuit fit and the determination of the
∗
Electrochemical Society Active Member. resistance, RLi2 O2 , and the pseudocapacitance. Both parameters are re-
z
E-mail: jhoj@topsoe.dk lated to a combination of the charge transfer through Li2 O2 and Li2 O2
A64 ECS Electrochemistry Letters, 4 (7) A63-A66 (2015)
Figure 1. (a) Sketch of the reactions and SEI formation during charge of the Li-O2 battery as discussed in this letter. The potentials in the figure are the proposed
onset potentials. Li2 O2 oxidation occurs at 3.05 V and an SEI layer is formed immediately on the freshly oxidized surface. At 3.3 V–3.5 V several reactions occur.
Among these are gas evolution from the SEI layer and oxidation of other Li2 O2 surfaces. (b) Charge of a Li-O2 battery after a discharge to 2.6 V.
Figure 2. (a) Differential capacity plot (dQ/dV) of a representative battery charge. Each peak represents the onset of an electrochemical reaction. (b) Differential
capacity plot of the initial part of the charge following a constant current 130 mA/gcarbon discharge to 2.0 V, 2.2 V, 2.5 V or 2.6 V, which was maintained for 150
min before charging. The measurement labeled 2.0 V–0 min corresponds to charging immediately after discharge to 2.0 V.
oxidation.8 Figure 4a shows the cathode resistance at selected volt- pounds are deposited during this initial charge, which is in line with
ages during charge, as determined by EIS. It is seen that the resistance the absorption measurements.
increases from 3.10 V to 3.30 V, decreases at 3.33 V, increases until
3.50 V, and decreases again at 3.60 V. The resistance and correspond- Identification of Li2 O2 oxidation at 3.05 V.— We argue that the
ing pseudocapacitances at the marked cross section at 65 mA/gcarbon process identified at 3.05 V is oxidation of Li2 O2 based on three
are shown in Figure 4b for all 11 potentials. Both the resistance and observations. First, Figure S1 in the Supporting Information, shows
the pseudocapacitance change stepwise as a function of potential. The that the e− /O2 ratio is between 2.0 (at 2 V) and 2.1 (at 3.2 V) in the
resistance increases monotonic until 2.27 V, then it drops and con- beginning of the charge, which is exactly – or at least very close to
tinue a second monotonic increase from 3.33 V to 3.50 V after which – the expected value for Li2 O2 oxidation. Second, Figure 2b shows
it drops. The pseudocapacitance is high at 3.10 V, decreases at 3.20 how the onset potential of the process at around 3.05 V increases with
V, increases at 3.33 V and decreases again at 3.60 V. When keeping the depth of discharge and the exposure time at low potentials. To
the potential at 3.10 V, it was observed that the capacitance decreased understand this shift, it is noted that DEMS measurements show that
60% from 0.7 mF/gcarbon at 130 mA/gcarbon to 0.28 mF/gcarbon at 13 the e− /O2 ratio is 2.0 during the entire discharge, and McCloskey et al.
mA/gcarbon . During the measurement, the current decreases and Li2 O2 show that the Li2 O2 yield is independent of the depth of discharge.7
is removed. Both changes are expected to increase the capacitance, This means that the thickness, and thereby the conductivity, of the
and the decreasing capacitance therefore clearly suggests that com- Li2 O2 layer is the only parameter expected to change between the
Figure 3. (a) Measurement of O2 evolution using DEMS (blue) and Li2 O2 removal (either chemical or electrochemical) determined spectrophotometrically (red).
The dotted line corresponds to a pure electrochemical 2 e− /O2 oxidation of Li2 O2 without any chemical degradation. (b) The amount of Li2 O2 oxidation with and
without gas evolution and electrochemical degradation in different potential intervals during charge. Values are normalized such that the sum of the electrochemical
reactions (blue and green) equals the relative change in capacity in each interval and sum up to 1 for a full charge.
ECS Electrochemistry Letters, 4 (7) A63-A66 (2015) A65
Figure 4. (a) The resistance related to the charge transfer through Li2 O2 and Li2 O2 oxidation measured at different potentials determined using EIS. The current
decreases during the measurement. (b) Resistance and capacitance values at different potentials during charge. Guide lines have been inserted to illustrate the
monotonic increases in the resistance.
measurements, and as the conductivity through the Li2 O2 layer affects decrease in impedance as the voltage increases, is a strong indication
the onset potential of the reaction, it suggests that the reaction occurs of a shift in equilibrium potential caused by a mixed potential estab-
at the Li2 O2 surface. Third, the onset potential at the investigated lished between different oxidation reactions to maintain the constant
current densities (∼0.1 μA/cm2 real surface area) is 2.9 V–3.0 V current. The theory of a mixed a potential is further substantiated by
which corresponds well with the onset potential of Li2 O2 oxidation measurements at higher potentials shown in a previous publication.8
measured by Viswanathan et al. using flat glassy carbon electrodes.13
Summary
SEI layer formation.— DEMS measurements show that all elec-
trons come from the Li2 O2 oxidation at the onset of the charge, until The main results of this work are shown in Figure 1a. We have
3.1 V. In this interval, it was found from Figure 2b that 4.3% of the showed that Li2 O2 is oxidized already at 3.05 V in porous carbon
Li2 O2 was oxidized electrochemically and in Figure 3b it is seen that cathodes, but that this facile oxidation is limited to approximately one
another 4.6% was removed without gas evolution. Since all electrons monolayer. Analysis of the chemical degradation and the change in
are accounted for by the gas evolved, the reaction with no gas evolu- double layer capacitance indicate that the Li2 O2 surface reacts with
tion must be chemical and it is interpreted as the formation of an SEI the electrolyte to form a SEI layer as soon as the outermost layer is
layer based on three observations. First, the amount of Li2 O2 degra- oxidized. The resistance increases as the SEI layer blocks the surface
dation is close to the amount of electrochemically oxidized Li2 O2 in and the voltage increases to maintain the constant current.
the initial part of the charge, and as the oxidation does not continue, Three reactions were identified between 3.3 V and 3.5 V. The
it suggests that the electrochemical oxidation of Li2 O2 exposes the interval is dominated by Li2 O2 oxidation with a small amount of
surface such that the oxidation is followed by a chemical degradation electrochemical degradation and significant chemical degradation of
of Li2 O2 , forming an SEI layer. Furthermore, the amount of oxidized Li2 O2 . It is expected that the reactions in this region are a gas evolving
and chemically degraded Li2 O2 both correspond to approximately degradation reaction in the Li2 O2 -electrolyte interface and oxidation
one monolayer, which suggest that the reaction occur on the entire of another Li2 O2 crystal plane, possibly the O-rich (1–100) plane,
surface of Li2 O2 . Second, the 60% decrease in capacitance at 3.1 V among others. Above 3.5 V the chemical and electrochemical reac-
suggests a significant deposition of a dielectric compound that could tions become more complicated and a shift in equilibrium potential
be explained by the formation of an SEI layer. Third, the monotonic due to the establishment of a mixed potential is indicated as previ-
increase in Li2 O2 resistance until 3.3 V suggests a decrease of avail- ously reported,8 but further work would be needed to understand and
able surface area or an increased electronic transport resistance. Both distinguish these reactions fully.
options could be explained by a growing SEI layer. In conclusion, the immediate formation of an SEI layer on the
oxidized Li2 O2 surface in the initial part of the charge is a significant
Electrochemical degradation.— Identification of the lowest po- problem that needs to be resolved before a viable Li-O2 battery can
tential without electrochemical degradation is important to identify be developed and an analysis of the very first part of the charge
a safe-voltage limit. We propose that at least one of the three sepa- might serve as a suitable screening parameter in the search for better
rate processes identified in the differential capacity plot in the voltage electrolytes.
range from 3.3 V to 3.5 V is an electrochemical degradation reaction as
the e− /O2 ratio increases in this range. Two observations suggest that
Acknowledgments
the reaction occurs at 3.3 V, but further investigation is needed to de-
termine the onset potential definitively. First, EIS measurements show The authors acknowledge support from the ReLiable project
that the pseudocapacitance increases and the resistance decreases at funded by the Danish Council for Strategic Research – Programme
3.3 V. A sudden change like this suggests a new reaction pathway at Commission on Sustainable Energy and Environment (project #11-
this potential. Second, isotope measurements presented by McCloskey 116792).
et al. using an identical system show that CO2 evolution occurs from
the electrolyte-Li2 O2 interface from 3.3 V.9 As the CO2 evolution
reaction depends on the potential, it is likely that this reaction is the References
new reaction pathway seen in the EIS measurements. To explain that 1. V. Viswanathan, K. S. Thygesen, J. S. Hummelshøj, J. K. Nørskov, G. Girishkumar,
three processes are identified in this voltage range, it is noted that DFT B. D. McCloskey, and A. C. Luntz, J. Chem. Phys., 135, 214704 (2011).
calculations from different groups show that onset potentials in this 2. A. C. Luntz, V. Viswanathan, J. Voss, J. B. Varley, A. Speidel, J. K. Nørskov, and
R. Scheffler, J. Phys. Chem. Lett., 4, 3494 (2013).
range could also be oxidation of another Li2 O2 crystal plane like the 3. J. S. Hummelshøj, J. Blomqvist, S. Datta, T. Vegge, J. Rossmeisl, K. S. Thygesen,
oxygen rich (1–100) surface.14,15 A. C. Luntz, K. W. Jacobsen, and J. K. Nørskov, J. Chem. Phys., 132, 071101 (2010).
4. J. S. Hummelshøj, A. C. Luntz, and J. K. Nørskov, J. Chem. Phys., 138, 034703
(2013).
Charge above 3.5 V.— At around 3.6 V, the resistance and 5. J. Chen, J. S. Hummelshøj, K. S. Thygesen, J. S. G. Myrdal, J. K. Nørskov, and
the pseudocapacitance decrease again. The correspondence between T. Vegge, Catal. Today, 165, 2 (2011).
impedance and overpotential is not straight forward, but the significant 6. D. J. Siegel and M. D. Radin, Energy Environ. Sci., 6, 2370 (2013).
A66 ECS Electrochemistry Letters, 4 (7) A63-A66 (2015)
7. B. D. McCloskey, A. Valery, A. C. Luntz, S. R. Gowda, G. M. Wallra, J. M. Garcia, 11. S. R. Gowda, A. Brunet, G. M. Wallraff, and B. D. McCloskey, J. Phys. Chem. Lett.,
T. Mori, and L. E. Krupp, J. Phys. Chem. Lett., 4, 2989 (2013). 4, 276 (2013).
8. J. Højberg, B. D. McCloskey, J. Hjelm, T. Vegge, K. Johansen, P. Norby, and 12. B. D. McCloskey, R. Scheffler, A. Speidel, D. S. Bethune, R. M. Shelby, and
A. C. Luntz, ACS Appl. Mater. Interfaces, 7, 4039 (2015). A. C. Luntz, J. Am. Chem. Soc., 133, 18038 (2011).
9. B. D. McCloskey, A. Speidel, R. Scheffler, D. Miller, V. Viswanathan, 13. V. Viswanathan, J. K. Nørskov, A. Speidel, R. Scheffler, S. R. Gowda, and
J. S. Hummelshøj, J. K. Nørskov, and A. C. Luntz, J. Phys. Chem. Lett., 3, 997 A. C. Luntz, J. Phys. Chem. Lett., 4, 556 (2013).
(2012). 14. M. D. Radin, F. Tian, and D. J. Siegel, J. Mater. Sci., 47, 17
10. B. D. McCloskey, D. S. Bethune, R. M. Shelby, G. Girishkumar, and A. C. Luntz, J. (2012).
Phys. Chem. Lett., 2, 1161 (2011). 15. J. S. G. Myrdal and T. Vegge, Rsc Adv., 4, 15671 (2014).
Paper II - SI - Reactions and SEI formation during charging of Li-O2 cells 117
Supporting Information: Reactions and SEI formation during
charging of Li-O2 cells
Jonathan Højberg1,2,*, Kristian Bastholm Knudsen2, Johan Hjelm2, Tejs Vegge2
1
Haldor Topsøe A/S, Nymøllevej 55, DK-2800 Kgs Lyngby, Denmark
2
Department of Energy Conversion and Storage, Technical University of Denmark, Frederiksborgvej 399, DK-
4000 Roskilde, Denmark
*Corresponding author
Tel.: (+45) 27292175
Fax: (+45) 45272999

DEMS measurements
Figure S1. Left: Charge of a Li-O2 battery after a discharge to 2.6 V. The e-/O2 ratio was calculated using an
average of six DEMS measurements. The errorbars indicate one standard deviation. Right: Headspace pressure
increase during charge at 130 mA/gcarbon constant current charge until a potential Uch (noted in the figure) is
reached. The charging is continued potentiostatic at Uch, resulting in a decreasing current with time.
Calculation of monolayers
It is not possible to estimate the thickness of the Li2O2 layer, because the full BET area may
not reflect the accessible surface area, as the use of binder has been shown to block the
micropores of the carbon electrode.1,2 Another approach is to use that it is generally accepted
that the sudden death occur as the insulating Li2O2 layer reaches a thickness of ~5 nm at
relevant current densities, and, using this thickness, the initial oxidation of 4.3 % correspond
to a removal of 2.1 Å Li2O2. This is approximately one monolayer of Li2O2.
2
Chemical quantification of Li2O2
After electrochemical test of a Li-O2 battery, the cell was purged with argon and transferred to
a glovebox. The cell was carefully disassembled and the cathode was extracted. Each cathode
was washed with 1,2-dimethoxyethane (BASF) dried using 4 Å molecular sieves (Sigma-
Aldrich), and the cathodes were subsequently dried in vacuum. The cathodes were taken from
the glovebox and immediately put into a 4 mL 0.063-0.07 % TiOSO4 aqueous solution and
the colored oxidized Ti-complex was seen immediately. The reactions occurring are listed in
(S1) and (S2).3
Li2O2 + 2H2O  2LiOH + H2O2 (S1)
Ti(IV)OSO4 + H2O2 + 2H2O  4H+ + H2Ti(VI)O4 + OSO44- (S2)
H2Ti(VI)O4 absorbs strongly at 408 nm. The solutions were left to react for 15-30 min and to
remove carbon particles, which otherwise would interfere with the spectrophotometric
measurement, samples were centrifuged and the supernatant was extracted yielding a clear
colored liquid that was characterized using a Shimadzu UV-3600 PharmaSpec with 1 nm
resolution and medium scan in absorbance mode.
The results from the absorption measurements are illustrated in Figure S2.
Figure S2. Optical absorption spectroscopy for the washed Li2O2 coated electrodes. Left: The extinction of the
H2Ti(VI)O4 complex in aqueous solutions illustrating the amount of detected Li2O2. Right: Lambert-Beer type
3
calibration curve used to determine the amount of Li2O2 in the cathodes. The curve is made by measuring the
absorbance of solutions with a known amount of peroxide. In the graph, the amount of Li2O2 is represented by a
capacity equivalent.
The chemical quantification of Li2O2 using the Ti-complex presumes that Li2O2 reacts with
water forming LiOH and H2O2. Another reaction is, however, possible
Li2O2 + H2O  ½O2 + 2LiOH (S3)
If this reaction takes place, the amount of Li2O2 would be underestimated, since the Ti()
complex is only oxidized by H2O2 that is not formed during this reaction.
Experimentally we did not observe any O2 evolution from cathodes submerged in H2O and
previous McCloskey et al. used a similar method to convert Li2O2 and LiO2 from identical
cathodes into H2O2, and detailed tests of the method showed that all Li2O2 was converted to
H2O2.4
4
Electrochemical impedance spectroscopy
Figure S3. Typical voltage profile used to measure the impedance at specific potentials during charge. All
impedance measurements follow the procedure: a) discharge to 2.6 V at 130 mA/gcarbon followed by 150 min at
2.6 V, b) charge to Uch at 130 mA/gcarbon, and c) continuous EIS measurements at Uch until the current decreased
to 13 mA/gcarbon. The voltage profile show a measurement with Uch = 3.30 V.
Figure S4. Left: Typical EIS measurement performed at a constant potential Uch. The spectrum is dominated by
the low frequency arc that has been shown to relate to the Li 2O2 reduction during discharge and Li2O2 oxidation
during charge.5 Right: The equivalent circuit used to describe the impedance during charge. The allowed peak-
frequency intervals of all processes are listed in the figure. As previously shown, Z4 is related to the oxygen
oxidation and double layer capacitance of the cathode.5
As discussed in a previous publication, the pseudocapacitance C* is calculated using5
∗ 𝟏/𝒏
𝑹Ω 𝑹 (𝟏−𝒏)/𝒏
𝑪 =𝑸 ( ) ,
𝑹Ω + 𝑹
where 𝑅, 𝑄 and 𝑛 are fitting parameters from the Voigt elements, and 𝑅Ω is the DC resistance
at the investigated frequency.
5
References
1. S. Meini, M. Piana, H. Beyer, J. Schwammlein, and H. A. Gasteiger, J. Electrochem. Soc.,

159, A2135–A2142 (2012).
2. R. Younesi, N. Ingh, S. Urbonaite, and K. Edström, in ECS transaction,, p. 121–127

(2010).
3. G. Eisenberg, Ind. Eng. Chem. Anal. Ed., 15, 327–328 (1943).
4. B. D. McCloskey, A. Valery, A. C. Luntz, S. R. Gowda, G. M. Wallra, J. M. Garcia, T.

Mori, and L. E. Krupp, J. Phys. Chem. Lett., 4, 2989–2993 (2013).
5. J. Højberg, B. D. McCloskey, J. Hjelm, T. Vegge, K. Johansen, P. Norby, and A. C. Luntz,

ACS Appl. Mater. Interfaces, 7, 4039–4047 (2015).
6
Paper III - Impedance-based battery management for metal-O2 systems 123
Impedance-based Battery Management for Metal-O2 Systems
Andreas E. Christensena,c,∗, Jonathan Højbergb,c , Poul Norbyc , Tejs Veggec

a Lithium
Balance A/S, Baldershøj 26C, 2635 Ishøj, Denmark
b HaldorTopsøe A/S, Nymøllevej 55, 2800 Kgs. Lyngby, Denmark
c DTU Energy, Frederiksborgvej 399, 4000 Roskilde, Denmark
Abstract
Li-O2 batteries have received wide attention as an
enabling technology for a mass-market entry of - - - - - -
electric vehicles due to a potential capacity much - - - - - - Li2O2

- - -
- - -
Li2O2
higher than current Li-ion technology. In electric +
Cathode
+ + + + +
Cathode
+ + + + + +
Cathode
+ + + + + +
vehicles, reliable estimation of the state-of-charge

(SoC) is crucial to determine the remaining capacity,
Capacitance OCV
but Li-O2 batteries are very different to Li-ion
batteries, and current SoC-estimation methods Udischarge
prove insufficient. In Li-O2 batteries, the capacity
is highly dependent on the discharge rate, since
different current densities enable different growth
SoC
mechanisms of Li2 O2 , and an on-board calibration
of the SoC is therefore needed. Such a calibration
is typically performed by measuring the open-circuit voltage (OCV), but as the OCV of the Li-O2 battery does
not change as a function of capacity, this method cannot be used. In this manuscript, we propose a method, based
on a single-frequency electrochemical impedance measurement, to estimate the remaining capacity and assess the
state-of-health by calculating the capacitance of the positive electrode where the discharge products are formed. The
results show that the capacitance is a good measure of the remaining capacity and that the SoC estimation can be
improved significantly by the calibration.
Highlights
− A new method to estimate capacity and degradation in metal-O2 batteries is proposed
− The method uses impedance measurement to determine the positive electrode capacitance
− Extensive testing on Li-O2 batteries show a decreasing capacitance during discharge
− The proposed method is able to reduce the SoC estimation uncertainty to less than 10 %
Keywords: Battery management, Metal-O2 , Li-O2 , Impedance spectroscopy, SoC, State-of-health
1. Introduction that are much higher than the battery technology pow-
ering today’s devices. For electric vehicles, the Li-O2
The continued research in improving the performance battery technology has the potential of increasing the
of batteries for portable devices and electric vehicles driving range by up to 4 times, compared to the elec-
has led to an increasing interest in metal-O2 battery tric vehicles on the market in 2014 [3]. To ensure safe
technologies such as Li-O2 , Al-O2 , Mg-O2 , and Na-O2 and reliable performance from any battery, and to pre-
[1, 2], because these technologies offer specific energies dict remaining capacity, a battery management system
(BMS) is needed. The two important BMS function-
alities are the calculation of the remaining capacity in
Tel.: +45 5851 5104 Fax.: +45 5851 5098
∗ Corresponding author.
Email address: andreas@lithiumbalance.com (Andreas E.

the batteries, the state-of-charge (SoC), and the health
Christensen) of the battery; generally combining capacity retention
Preprint submitted to Journal of Power Sources March 23, 2015
on cycling and power capability. The health parame- is not limited to Li-O2 , we will also discuss the applica-
ter is referred to as the state-of-health (SoH). The SoC bility in other metal-O2 systems.
and SoH is calculated by the BMS to predict the per-
formance under different scenarios, to enable optimised
usage of the remaining capacity, and even preventing 2. Experimental
dangerous situations that may occur if the battery pow-
Electrochemical measurements were performed on
ered device, e.g. an electric vehicle, is suddenly with-
Li-O2 batteries using 2-electrode Swagelok cells that
out power. The SoC can be calculated in numerous
were assembled inside an Ar-filled glovebox (<3 ppm
ways, with the most simple being a comparison of ter-
O2 and H2 O). The batteries used an XC72 carbon
minal voltage to previously recorded cycling data of
black positive electrode (Vulcan XC72, Cabotcorp,
cell voltage and battery capacity. A slightly more com-
GA), and 1 M LiTFSI (99.95 % Sigma-Aldrich) in 1,2-
plex approach is to continuously monitor and integrate
dimethoxyethane (BASF) electrolyte was used with a
the current over time, also known as coulomb count-
Whatman GF/A glass fiber separator (Whatman) and
ing [4]. The coulomb counting method accumulates er-
lithium anode (HongKong Wisdom Tech Company).
rors if calibration is not performed, as it relies on the
The carbon electrodes were made by air-spraying a
accuracy of the measurement and several methods for
slurry of XC72 Carbon Black and PTFE (60 wt%
mitigating this have been proposed for lead-acid and
dispersion in water) in a wt/wt ratio of 3:1 as de-
lithium-ion batteries [5]. Common for the secondary
scribed in Mekonnen et al. [14]. All experiments were
metal-O2 technologies is that the main electrochemi-
performed in a temperature controlled environment at
cal process is unchanged during discharge and charge
25 ◦ C. Electrochemical impedance spectroscopy (EIS)
(assuming no degradation). This means that the open
and discharge-charge curves were measured using Bio-
circuit voltage (OCV) does not change as a function of
Logic VMP3 and MPG2 potentiostats. Impedance mea-
SoC. Furthermore, constant current measurements show
surements were performed throughout a number of dif-
a flat discharge plateau in a majority of the discharge pe-
ferent current profiles. All reported impedance mea-
riod. This has been shown for Li-O2 and Na-O2 batter-
surements were performed at OCV after at least 3 h
ies [6, 7], as well as for Mg-O2 and Al-O2 [8, 9]. Taking
rest. The full impedance spectra included frequencies
the well studied Li-O2 battery as an example, the dom-
between 20 kHz and 10 mHz with 15 points per decade
inating process during discharge is reduction of oxygen
and an alternating current (AC) amplitude of 10 mV
to deposit Li2 O2 on top of an existing Li2 O2 layer [10].
(potentiostatic mode) or a current density of 13 mA g-1
c
Initially Li2 O2 is deposited on the pristine positive elec-
(galvanostatic mode). In measurements replicating an
trode and later the process changes to Li2 O2 depositing
actual BMS system, only one frequency (10 mHz) was
on Li2 O2 . As this process continues during the entire
investigated in order to establish a fast and simple as-
discharge, both OCV and discharge potential is constant
sessment. The validity of using only this single fre-
until the end of discharge, where other processes be-
quency was confirmed by comparison with a full spec-
come limiting, resulting in a rapid voltage drop referred
trum. For all experiments on Li-O2 batteries, the dis-
to as “sudden death”. It has also been observed that the
charge capacity was calculated as the capacity between
current densities have a significant impact on the on-
fully charged and the the onset of sudden death; indi-
set of sudden death and thus available capacity, due to
cated by the voltage decreasing to 2.2 V. The current
the increase in required electron transport through the
densities were calculated based on the carbon-loading
poorly conducting Li2 O2 layer [11, 12, 13]. Since the
of the electrodes, which typically was ∼6.5 mg cm-2 for
coulomb counting method relies on a known total ca-
an electrode with a diameter of 10 mm.
pacity to predict sudden death, the method is not well
suited for predicting the remaining capacity in these bat-
teries. 2.1. Impedance analysis
New methods have to be developed to overcome the EIS is used in many systems to perform in situ de-
constant OCV and flat discharge plateau that otherwise termination of certain parameters like degradation of
would complicate accurate online prediction of remain- secondary Li-ion batteries [15, 16, 17], capacity fad-
ing capacity in Li-O2 batteries. In the following, we ing of Li-S batteries [18], and discharge mechanisms
propose a method to accurately predict the SoC of Li-O2 for Si-air [19]. In Li-O2 batteries, it has previously
batteries using a single frequency impedance measure- been shown that the low frequency (<1 Hz) contribu-
ments to estimate the remaining capacity as well as the tion of the impedance is related to the positive elec-
degradation of the battery materials. Since the method trode [20, 21, 22]. EIS measurements in this frequency
2
regime gives a capacitance value that relates to the dou- has been determined to be 35 by Gerbig et al. [25],
ble layer capacitance of the electrode [20]. The best and using these values it can be calculated that going
measurements of this value is performed at OCV, as from a pristine carbon surface without Li2 O2 to ∼6 nm
this gives a uniform distribution of charged species in of Li2 O2 will halve the capacitance (a detailed descrip-
the entire electrode, but our initial experiments showed tion of this is found in Supporting Information S1). The
that it is also possible to measure the capacitance under proposed SoC estimation is based on analysis of the de-
constant load. Because the measurements under load crease in capacitance as the Li2 O2 is deposited.
introduces new variables, we are focusing on measure- As it will be shown later, the correlation between
ments at OCV in this manuscript. At OCV, the oxy- the capacity and capacitance is following the same
gen reduction reaction and Li2 O2 oxidation reactions trend for all the performed Li-O2 experiments and that
are very slow, and if the EIS excitation signal is suf- trend can be described by the simple functional form in
ficiently small, the impedance signal from the positive Equation 3.
electrode becomes capacitive at the relevant frequencies
Q
!!
(see insert in Figure 2). This means that the capacitance C = C0 − p2 1 − exp (3)
can be determined by the simple expression shown in p1
Equation 1 [23]. where C0 is the initial capacitance, Q is the capacity,
−1 and p1 and p2 are refined parameters determining the
C= (1) exponential shape.
2π f · Zim
where f is the AC perturbation frequency and Zim is 2.3. Scaling the current density
the imaginary part of the corresponding impedance. It
The experiments are divided into two categories, (i)
is important that other impedance contributions from
constant current discharges at different current densities,
processes with similar time constants do not overlap
and (ii) dynamic current densities with different profiles.
at the frequency used for the calculation. A frequency
Figure 1 shows the applied techniques. Current den-
of 10 mHz was chosen because it was the highest fre-
sities are selected on basis of a future scenario, where
quency (and hence shortest measurement time) with the
electric vehicles will be powered by metal-O2 batteries
main contribution from the positive electrode capaci-
as the sole source of power. The scaling of the peak
tance.
and average current densities are based on an electric
vehicle with a 100 kWh battery, a sustained high power
2.2. Relating capacitance to the discharge processes
of 55 kW and a peak power of 105 kW, which matches
The capacitance of the positive electrode is deter- most electric vehicles today in terms of peak power and
mined by the electrochemical double layer capacitance is superior in terms of capacity [27]. See Supporting
that reflects the ability of the system to store electrical Information S2 for details.
charge by rearranging the charges in the carbon (elec-
trons) and in the electrolyte (ions). The value is pro-
portional to the surface area and depends on the choice
of electrolyte, the concentration of charged species in
the electrolyte (salt concentration) and deposits on the
carbon surface. For a Li-O2 battery, Li2 O2 is deposited
on the surface during discharge and as Li2 O2 is a di-
electric, the capacitance of the surface will change. The
capacitance of XC72 is 12.6 F g-1 c in an organic aprotic
electrolyte as reported by Barbieri et al., which corre-
sponds to 5.4 µF cm-2 , using a measured BET area of
235 m2 g-1 [24]. The capacitance of the Li2 O2 layer is
calculated using Equation 2.
A
C = r 0 (2)
d
where r is the relative permittivity, 0 is the vacuum
permittivity, A is the area of the layer and d is the thick-
ness of the layer. The relative permittivity of Li2 O2
3
100 a
Current [mAgC-1]
b c d
0
-100
-200
Current
-300 EIS measurement
-400
0 2 4 6 8 10 12
Time [h]
0 e
Current [mAgC-1]
-100
-200
-300
x11
-400
0.00 0.05 0.10 0.15 0.20 0.25 0.30
Time [h]
Figure 1: Current profiles used in the experiments. (a) discharge at 130 mA g-1 c to 2.0 V interrupted by impedance measurements every
100 mAh g-1 -1 -1 -1 -1 -1
c . 13 mA gc , 39 mA gc , and 390 mA gc were tested in similar manner. (b) staircase discharge profile with 33 mAh gc at 39 mA gc ,
33 mAh g-1c at 390 mA g -1 and 33 mAh g-1 at 39 mA g-1 between each impedance measurement. (c) discharge of 130 mA g-1 for 100 mAh g-1
c c c c c
with 3.20 V voltage-limited charge at 130 mA g-1 c to reduce electrochemical decomposition to a minimum. (d) capacity limited galvanostatic
-1 -1
cycling at 130 mA gc limited to 65 mAh gc . The charge was limited to 4.5 V to avoid severe electrolyte decomposition. (e) drive cycle from
ISEA-RWTH Aachen recorded from a Fiat 500 eCity [26], scaled to a maximum of 390 mA g-1 c (55 kW).
3. Results of capacity and it is clear how the capacitance changes

even though the voltage is unchanged. The capacitance
To investigate the correlation of the capacitance dur- is calculated according to Equation 1 and the insert in
ing discharge and charge to the SoC, SoH and power ca- Figure 2 shows how the impedance measurement is re-
pability of the Li-O2 battery, we have performed a num- lated to the capacitance. For every data point, a unique
ber of experiments based on the current profiles listed in impedance measurement was made and the capacitance
Figure 1. All experiments with Li-O2 batteries show a was determined. At the end of discharge, when the bat-
decrease in the capacitance as the battery is discharged. 1.5
tery reached sudden death, the capacitance was 50 % of
-Im(Z) [ΩgC]
Figure 2 shows the capacitance decrease as a function the initial value.1

Figure 3a and Figure 3b show additional measure-
0.5
10 ments of the dependence between the capacitance and
2.69 V 2.70 V the SoC at the different
0 current profiles seen in Figure
9 2.69 V 0 0.5
1. When testing Li-O
Re(Z) [Ωg2 Cbatteries
] using different current
10 mHz 2.67 V
C]
1.5
8 profiles, the capacity varies significantly. From Table 1
-Im(Z) [ΩgC]
C [Fg-1
7
1
2.63 V it is seen that the discharge capacity can vary up to a fac-
0.5 tor of ∼14 between the conducted measurements. The
6 0 purpose of a BMS is to predict the remaining capacity,
0 0.5
Re(Z) [ΩgC] 2.32 V and in the following we have therefore normalised each
5
Fig4 inset
1.75
1.50
0 100 200 300 400 500 600 discharge capacity to the capacity at 2.2 V, to investigate
1.25
Im(Z) [$Omega$g$_C$
Capacity [mAhg-1
C] how well the remaining capacity is predicted by the ca-
1.00
0.75
pacitance. Even though all initial capacitances are very

0.50
0.25
Figure 2: Capacitance as a function of the capacity for a constant similar, we have chosen to normalise this value also, as
0.00
current discharge at 130 mA g-1

0.00 0.25 0.50 0.75
c . Insert shows a Nyquist plot of the

e(Z) [$Omega$g$_C$
impedance measurement, where the capacity data is highlighted. The this is what would be done in an actual BMS, where the
voltage labels are measured during discharge prior to the impedance capacitance may vary between batteries and may change
measurements. The OCV was 2.83 V during the entire discharge. over time due to degradation. The original data is pre-
4
sented in Supporting Information S3. All measurements

have been fitted to Equation 3 and it is seen that they
all follow the same trend. Table 1 shows the experi-
1.0 a mental values and derived parameters, and three obser-

vations are made: (i) all initial capacitances fall within
0.8 10.0 F g-1 -1
c ± 0.4 F gc , (ii) the trend of the decreasing ca-
13 mAgC-1 pacity is very similar for all measurements performed at
0.6
39 mAgC-1 current densities varying a factor of 30, and at both dy-
130 mAg-1
130 mAg-1
C
C
namic and constant current experiments, and (iii) all ca-
390 mAg-1C pacitances have decreased to approximately 50 % in the
0.4 end of discharge, except the high current measurement
0.0 0.2 0.4 0.6 0.8 1.0
Normalized capacity at 390 mA g-1 c and the measurement including charging;
Both these exceptions will be discussed further below.
1.0 b
Figure 3c depicts how the voltage during discharge
relates to the capacitance for different current densities,
0.8 and it is evident that the current densities impact the dis-
130 mAg g-1
C charge capacity, and thereby the capacitance at 2.2 V.
Staircase
0.6 Charging Figure 3d shows the normalised capacitance in the fully
Drive cycle 1 charged state, as a function of cycle number. Upon cy-
Drive cycle 2
cling, the capacitance decreases almost linearly to ap-
0.4
0.0 0.2 0.4 0.6 0.8 1.0 proximately 65 % after 20 cycles.
Normalized capacity
C2.2V
3.2 C0 p1 p2 Q2.2V C0
13 mAgC-1 130 mAg-1 c 13 mA/gC
F g-1 mAh g-1 F g-1 mAh g-1
C
3.0 39 mAgC-1 130 mAg-1
C
39 mA/gC c c c c -
Voltage [V]
2.8
390 mAg-1
C
129mA/gC 13 mA g-1
c 10.3 518 820 1001 53%
129 mA/gC
387 mA/gC 39 mA g-1
c 10.0 300 467 767 44%
2.6 130 mA g-1 9.83 169 76.9 700 52%
c
2.4 13 mAgC-1130 mA g-1c 10.1 148 90.9 599 50%
39 mAgC-1390 mA g-1 10.2 67 465 117 78%
2.2 129 mAg-1 c
C
129 mAg-1 Staircase 10.1 241 610 519 54%
2.0 C
387 mAg-1C Charging 10.3 377 44.8 1614 69%
1.0 0.9 0.8 0.7 0.6 0.5 0.4 Drive cycle 1 10.0 188 54.9 812 59%
Normalized capacitance Drive cycle 2 9.64 124 21.8 652 56%
1.0 d Li CO
Table 1: Experimental values and 2
parameters
3
obtained by fitting ex-
0.9 perimental data to Equation 3. C0 is the initial capacitance, Q2.2V
C
is the discharge capacity at the 2.2 V cut-off, C2.2V0
is the fraction of
Li2O2
capacitance at the 2.2 V cut-off to the initial capacitance.
0.8
Li2CO3 Current collector Current collector
0.7 Li2O2
Cathode
Discharged
Cathode
Charged Discharged
4. Discussion
Charged
0.6 0 5 10 15
Cycle no. In the following, we show the correlation between
the capacitance and the capacity, which qualifies the
Figure 3: Results from experiments on Li-O2 batteries, data labeled method for further studies targeting commercial use in
according to current profiles in Figure 1. Capacitance is normalised a Li-O2 BMS.
to the initial value before discharge, and capacity is normalised to
the value at 2.2 V. (a) constant current at four current densities:
13 mA g-1 -1 -1 -1 4.1. State-of-charge
c , 39 mA gc , 130 mA gc , and 390 mA gc , (b) dynamic
current profiles according to Figure 1 and data from a 130 mA g-1 c
The constant current experiments presented in Fig-
constant current experiment to make comparison easier, (c) discharge ure 2 show how the capacitance changes as a function
voltage and capacitance from constant current experiments, showing of capacity. The correlation is decreasing and at the
the relation between current density and capacitance. (d) cycling ex-
periment over 20 cycles showing decreasing capacitance as a function end of discharge the capacitance is 50 % of the initial
of cycle number, insert showing possible degradation mechanism by value, which is in very good agreement with the esti-
Li2 CO3 accumulation. mates based on Equation 2. Assuming that the change
5
is due to a uniform layer of Li2 O2 , this corresponds to 3. The SoC is calibrated at end of full charge.
a layer thickness of ∼6 nm at these current densities, 4. Fast charging is not considered, charging performed at ∼10 kW.
and as discussed in Section 2.2, this is in line with the
If no SoC calibration is performed, the coulomb count-
general understanding of the sudden death mechanism
ing error and the need for a capacity reserve will result
[13]. The voltage during discharge is shown in Fig-
in a signicant decrease of useable capacity as the bat-
ure 2 as labels on the capacitance data points, and re-
tery is discharge and charged. During discharge, the ac-
mains unchanged at ∼2.70 V between 100 % and 33 %
cumulated error corresponds to 3 % and during charge,
SoC, whereas the capacitance decreases as the battery is
the error is 13 %, due to the additional need of a 10 %
discharged. Seeing that all measurements behave simi-
capacity reserve for the following discharge. Figure 4a
larly, this shows that the capacitance holds information
shows the effect of the combined error of 16 % on a
about the SoC, that is not possible to obtain using volt-
With Calibration 100 kWh battery as a function of energy consumption
age based measurements. Without calibration
Extra capacity with calibration of up to 400 kWh without fully charging the battery.
When the battery is fully charged, a BMS would be able
100 a With calibration 500 Improvement [%] to correct the estimation of useable capacity to the nom-
Usable SoC [%]
80 400 inal capacity of the battery without the capacitance cal-

ibration, and thus reset the useable capacity to 90 %.
60 300
Without the SoC calibration, the useable SoC could de-
40 200 crease to 26 % over an accumulated energy consump-
20 100 tion of 400 kWh, whereas the SoC calibration based on
capacitance would be able to keep the SoC estimation
0 0 at 90 %, thus maximising the useable capacity.
0 100 200 300 400
Energy consumption since full charge [kWh] Since the capacity of future metal-O2 batteries is ex-
b Actual SoC pected to increase, we postulate a use-scenario for the
100 BMS with calibration
BMS without calibration
batteries based on ∼10 kW charging power available
80 Calibration and no need for fast charging with daily use. This sce-
SoC [%]
60 nario is illustrated in Figure 4b, where the battery is re-

40 spectively discharged, charged or at rest. The calibra-
20 tions performed in Figure 4b enables the BMS to more
0 accurately predict the remaining capacity, whereas the
estimated capacity without calibration would become
6am 6pm 6am 6pm less than zero (shown on the figure as circles when the
Time SoC is crossing the 0 % threshold), warning the vehicle
or driver to stop. The capacitance calibration is less ac-
Figure 4: (a) Effect of error accumulation in coulomb counting and curate in the less critical beginning of discharge, due to
capacity reserve, without reaching a full charge. (b) Scenario showing
difference in estimation of useable capacity between a BMS using the slope of the capacitance correlation to the capacity,
calibration based on capacitance, and no calibration other than full- and gradually more accurate as the battery is discharged.
charge correction, shown as a vertical line at the end. The circles Figure 4b shows that the calibrations performed below
at 0 % SoC highlights places were the capacitance calibration would
50 % SoC have the possibility to initially correct the
enable further discharging of the battery.
SoC to an uncertainty of less than ∼10 %, and when
performed multiple times, able to further minimise the
To estimate the error of the SoC estimation based
uncertainty of the SoC estimation, resulting in periods
on capacitance measurements, the two drive cycle mea-
with an uncertainty of less than 5 % SoC.
surements and the measurement at 130 mA g-1 c in Figure
3b are used to give a minimum relative capacity based
on a given capacitance. The points of calibration used 4.2. Power capability
in the drive scenario are presented in Supporting Infor- The current density of 390 mA g-1 c is only able to
mation S4, Table S2. provide 12 % of the capacity compared to the current
To test the effect of the calibration, we have designed density of 13 mA g-1 c . This indicates that the avail-
a BMS system based on the following assumptions: able capacity depends explicitly on the current density.
1. Due to high current dependency on capacity, a reserve of ∼10 % For correct estimation of the available power, the inter-
SoC is needed, resulting in 90 % useable capacity. nal resistance is usually used to predict if the terminal
2. Coulomb counting has a ∼3 % accumulation of error. voltage of a battery will exceed the limit during high
6
power demands. For Li-O2 batteries, this is not possible The charging experiment shown in Figure 3b has a
since the resistance is almost unchanged during the en- much larger discharge capacity compared to the other
tire discharge [20]. Previously, we saw that the capac- experiments (Table 1: 1614 mAh g-1 -1
c vs 599 mAh gc for
itance of the 390 mA g-1 c experiment did not decrease the 130 mA g-1 c experiment), and we suspect that this is
to ∼50 % of the initial capacitance. If we look at the due to the accumulation of Li2 CO3 and similar species.
voltage during the constant current experiments (Figure The reason for the capacitance not decreasing to more
3c), we see that the high current density experiments than 69% is not fully understood and further studies of
(130 mA g-1 -1
c and 390 mA gc ) have significantly higher cycling effects on Li-O2 systems is needed.
overpotentials, thus the 2.2 V limit is reached prema- Since the capacitance is expectedly one of the only
turely. From the measurements, it seems that current parameters changing during large parts of the discharge,
densities of 390 mA g-1 c are only supported at capaci- it is likely that the proposed method will be an impor-
tances above 78 % of the initial capacitance, while it is tant part of any metal-O2 BMS. Many metal-O2 systems
possible to use 130 mA g-1 c until a capacitance of 50 %. show the same type of discharge and charge curves as
This shows that the capacitance can also be used to esti- the Li-O2 , but further studies on other metal-O2 systems
mate the maximum power that the battery is capable of are needed to conclude if the proposed method is appli-
delivering at the given state. cable to these systems. A likely candidate for further
studies is the Na-O2 battery, where the discharge prod-
4.3. State-of-health uct, NaO2 is considered to have surface conductivity.
The cycling experiment (Figure 3d) shows how the The conductive nature of the NaO2 is expected to cause
capacitance of a fully charged positive electrode de- an increase in capacitance as the battery is discharged,
creases as a function of cycle number. From several reflecting an increase in surface area as the NaO2 cubes
papers [14, 28, 29, 30] it has been shown that Li2 CO3 are grown [32].
and similar species are accumulating upon charging and
are immobilised in the positive electrode (see Figure
3d, insert). The presence of Li2 CO3 in the cathode has 5. Conclusion
two main effects on the battery performance. First, the
A method for estimating remaining capacity, power
amount of Li2 CO3 will increase for each cycle, result-
capability and cycle life of metal-O2 batteries has been
ing in a decrease of the available active area for Li2 O2
proposed. The method was verified using Li-O2 bat-
deposition. This results in a lower discharge capac-
teries, and the approach is expected to be transferrable
ity for the battery. Second, the inclusion of Li2 CO3
to other metal-O2 systems. Experiments showed that
in the deposited Li2 O2 decreases the charge transport
the capacitance of the positive electrode decreased ex-
through the Li2 O2 , thus increasing the required overpo-
ponentially during discharge, and that it was possible to
tential [31]. It has previously been shown that even a
improve the prediction the remaining battery capacity
few angstrom of Li2 CO3 will decrease the capacitance
significantly based on a single frequency measurement
significantly [20], and using the capacitance in the fully
of the positive electrode capacitance. In a typical sce-
charged state, it is possible to track the degradation of
nario, a single SoC calibration is able to improve the
the positive electrode, and a BMS can use the informa-
available SoC by more than 10 % of the full battery ca-
tion to determine the SoH parameter related to capacity
pacity, by minimising the uncertainty of the SoC. The
retention, and thereby the cycle life of the battery.
capacitance was also used to estimate a degradation of
4.4. Further development of the model the positive electrode in a Li-O2 batteries cycled over
20 times. This makes the method applicable not only
With more knowledge about the system, it is expected
for electric vehicles, but for batteries in a large range
that the time at OCV can be reduced significantly and
of electrical devices, as the measurements can be per-
that measurements can be made even while discharging
formed when needed, thus maintaining a high level of
or charging the battery. This could be done by employ-
accuracy for the estimation of remaining capacity and
ing impedance measurements in the time domain to cal-
state-of-health.
culate the capacitance. In this case, it might even be
possible to perform SoC calibrations while driving. An-
other direction would be to implement an adaptive state 6. Acknowledgements
estimation algorithm, akin to the work done by Fleis-
cher et al. [26], using the capacitance measurements to The authors acknowledge support of this work from
improve the calibration uncertainties. the ReLiable project (Project No. 11-116792) funded
7
by the Danish Council for Strategic Research Program [15] F. Huet, A review of impedance measurements for determination
Commission on Sustainable Energy and Environment. of the state-of-charge or state-of-health of secondary batteries,
Journal of Power Sources 70 (1998) 59–69. doi:10.1016/
S0378-7753(97)02665-7.
[16] M. H. Hung, C. H. Lin, L. C. Lee, C. M. Wang, State-of-charge
References and state-of-health estimation for lithium-ion batteries based on
dynamic impedance technique, Journal of Power Sources 268
[1] M. M. Thackeray, C. Wolverton, E. D. Isaacs, Electrical energy (2014) 861–873. doi:10.1016/j.jpowsour.2014.06.083.
storage for transportationapproaching the limits of, and going [17] S. Rodrigues, N. Munichandraiah, a. K. Shukla, Review of state-
beyond, lithium-ion batteries, Energy & Environmental Science of-charge indication of batteries by means of a.c. impedance
5 (7) (2012) 7854. doi:10.1039/c2ee21892e. measurements, Journal of Power Sources 87 (1) (2000) 12–20.
[2] F. Cheng, J. Chen, Metalair batteries: from oxygen reduction doi:10.1016/S0378-7753(99)00351-1.
electrochemistry to cathode catalysts, Chemical Society Re- [18] Z. Deng, Z. Zhang, Y. Lai, J. Liu, J. Li, Y. Liu, Electrochem-
views 41 (6) (2012) 2172. doi:10.1039/c1cs15228a. ical Impedance Spectroscopy Study of a Lithium/Sulfur Bat-
[3] K. G. Gallagher, S. Goebel, T. Greszler, M. Mathias, tery: Modeling and Analysis of Capacity Fading, Journal of
W. Oelerich, D. Eroglu, V. Srinivasan, Quantifying the promise the Electrochemical Society 160 (4) (2013) A553–A558. doi:
of lithiumair batteries for electric vehicles, Energy & Environ- 10.1149/2.026304jes.
mental Science 7 (2014) 1555. doi:10.1039/c3ee43870h. [19] G. Cohn, R. a. Eichel, Y. Ein-Eli, Supplement: New insight into
[4] S. Piller, M. Perrin, A. Jossen, Methods for state-of-charge de- the discharge mechanism of silicon-air batteries using electro-
termination and their applications, Journal of Power Sources chemical impedance spectroscopy., Physical chemistry chemi-
96 (1) (2001) 113–120. doi:10.1016/S0378-7753(01) cal physics : PCCP 15 (9) (2013) 3256–63. doi:10.1039/
00560-2. c2cp43870d.
[5] K. S. Ng, C. S. Moo, Y. P. Chen, Y. C. Hsieh, Enhanced coulomb [20] J. Hø jberg, B. D. McCloskey, J. Hjelm, T. Vegge, K. Jo-
counting method for estimating state-of-charge and state-of- hansen, P. Norby, A. C. Luntz, An Electrochemical Impedance
health of lithium-ion batteries, Applied Energy 86 (9) (2009) Spectroscopy Investigation of the Overpotentials in LiO 2
1506–1511. doi:10.1016/j.apenergy.2008.11.021. Batteries, ACS Applied Materials & Interfaces 7 (7) (2015)
[6] P. Hartmann, C. L. Bender, M. Vračar, A. K. Dürr, A. Garsuch, 150211151515002. doi:10.1021/am5083254.
J. Janek, P. Adelhelm, A rechargeable room-temperature sodium [21] I. Landa-Medrano, I. Ruiz De Larramendi, N. Ortiz-Vitoriano,
superoxide (NaO2) battery., Nature materials 12 (3) (2013) 228– R. Pinedo, J. Ignacio Ruiz De Larramendi, T. Rojo, In situ mon-
32. doi:10.1038/nmat3486. itoring of discharge/charge processes in Li-O2 batteries by elec-
[7] B. D. McCloskey, J. M. Garcia, A. C. Luntz, Chemical and Elec- trochemical impedance spectroscopy, Journal of Power Sources
trochemical Differences in Nonaqueous LiO 2 and NaO 2 Bat- 249 (c) (2014) 110–117. doi:10.1016/j.jpowsour.2013.
teries, The Journal of Physical Chemistry Letters 5 (7) (2014) 10.077.
1230–1235. doi:10.1021/jz500494s. [22] J. Adams, M. Karulkar, V. Anandan, Evaluation and electro-
[8] T. Shiga, Y. Hase, Y. Kato, M. Inoue, K. Takechi, A rechargeable chemical analyses of cathodes for lithium-air batteries, Journal
non-aqueous MgO2 battery, Chemical Communications 49 (80) of Power Sources 239 (0) (2013) 132–143. doi:10.1016/j.
(2013) 9152–9154. jpowsour.2013.03.140.
[23] M. Orazem, B. Tribollet, Electrochemical impedance spec-
[9] R. Revel, T. Audichon, S. Gonzalez, Non-aqueous aluminium-
air battery based on ionic liquid electrolyte, Journal of Power troscopy, John Wiley & Sons, 2008.
[24] Barbieri, Hahn, Herzog, Kotz, Capacitance limits of high sur-
Sources 272 (c) (2014) 415–421.
[10] J. S. Hummelshø j, J. Blomqvist, S. Datta, T. Vegge, J. Ross- face area activated carbons for double layer capacitors, Carbon
43 (6) (2005) 1303–1310. doi:10.1016/j.carbon.2005.
meisl, K. S. Thygesen, Communications : Elementary oxygen
electrode reactions in the aprotic Li- air battery Communica- 01.001.
[25] O. Gerbig, R. Merkle, J. Maier, Electron and ion transport in
tions : Elementary oxygen electrode reactions in the aprotic Li-
air battery, The Journal of Chemical Physics 071101 (7) (2010) Li2O2, Advanced Materials 25 (2013) 3129–3133. doi:10.
1002/adma.201300264.
–. doi:10.1063/1.3298994.
[11] J. Chen, J. S. Hummelshø j, K. S. Thygesen, J. S. G. Myrdal, [26] C. Fleischer, W. Waag, H. M. Heyn, D. U. Sauer, On-line
adaptive battery impedance parameter and state estimation con-
J. K. Nø rskov, T. Vegge, The role of transition metal interfaces
on the electronic transport in lithium-air batteries, Catalysis To- sidering physical principles in reduced order equivalent circuit
day 165 (1) (2011) 2–9. doi:10.1016/j.cattod.2010.12. battery models: Part 2. Parameter and state estimation, Jour-
022. nal of Power Sources 262 (2014) 457–482. doi:10.1016/j.
[12] M. D. Radin, D. J. Siegel, Charge transport in lithium perox- jpowsour.2014.03.046.
ide: relevance for rechargeable metalair batteries, Energy & En- [27] Compare Electric Vehicles Side-by-Side, (accessed 2015-03-
vironmental Science 6 (8) (2013) 2370. arXiv:1305.2904, 19).
doi:10.1039/c3ee41632a. URL http://www.fueleconomy.gov/feg/evsbs.shtml
[13] V. Viswanathan, K. S. Thygesen, J. S. Hummelshj, J. K. Nrskov, [28] B. D. McCloskey, D. S. Bethune, R. M. Shelby, T. Mori,
G. Girishkumar, B. D. McCloskey, a. C. Luntz, Electrical con- R. Scheffler, A. Speidel, M. Sherwood, a. C. Luntz, Limita-
ductivity in Li 2O 2 and its role in determining capacity lim- tions in Rechargeability of Li-O2Batteries and Possible Origins,
itations in non-aqueous Li-O 2 batteries, Journal of Chemical The Journal of Physical Chemistry Letters 3 (2012) 3043–3047.
Physics 135 (21) (2011) 214704. doi:10.1063/1.3663385. doi:10.1021/jz301359t.
[14] Y. S. Mekonnen, K. B. Knudsen, J. S. G. Mýrdal, R. Younesi, [29] M. M. Ottakam Thotiyl, S. a. Freunberger, Z. Peng, P. G. Bruce,
J. Hø jberg, J. Hjelm, P. Norby, T. Vegge, Communication: The carbon electrode in nonaqueous Li-O2 cells, Journal of the
The influence of CO2 poisoning on overvoltages and discharge American Chemical Society 135 (1) (2013) 494–500. doi:10.
capacity in non-aqueous Li-Air batteries, Journal of Chemical 1021/ja310258x.
Physics 140 (12) (2014) 121101. doi:10.1063/1.4869212. [30] B. M. Gallant, R. R. Mitchell, D. G. Kwabi, J. Zhou, L. Zuin,
8
C. V. Thompson, Y. Shao-Horn, Chemical and Morphological

Changes of LiO2 Battery Electrodes upon Cycling, The Journal
of Physical Chemistry C 116 (2012) 20800–20805. arXiv:10.
1021/jp308093b, doi:10.1021/jp308093b.
[31] J. M. Garcia-Lastra, J. S. G. Myrdal, R. Christensen, K. S.
Thygesen, T. Vegge, DFT+U study of polaronic conduction in
Li2O2 and Li2CO3: Implications for Li-air batteries, Journal
of Physical Chemistry C 117 (11) (2013) 5568–5577. doi:
10.1021/jp3107809.
[32] P. Hartmann, C. L. Bender, J. Sann, A. K. Dürr, M. Jansen,
J. Janek, P. Adelhelm, A comprehensive study on the cell chem-
istry of the sodium superoxide (NaO2) battery., Physical chem-
istry chemical physics : PCCP 15 (28) (2013) 11661–72. doi:
10.1039/c3cp50930c.
9
Paper III - SI - Impedance-based battery management for metal-O2 systems 132
Supporting Information: Impedance-based Battery Management for Metal-O2

Systems
Andreas E. Christensena,c,∗, Jonathan Højbergb,c , Poul Norbyc , Tejs Veggec

a LithiumBalance A/S, Baldershøj 26C, 2635 Ishøj, Denmark
b Haldor Topsøe A/S, Nymøllevej 55, 2800 Kgs. Lyngby, Denmark
c DTU Energy, Frederiksborgvej 399, 4000 Roskilde, Denmark
S1. Li2 O2 capacitance
14
Capacitance [F/g]
12 100 %
10
8
50 %
6
4
2
0
0 2 4 6 8 10 12
Li2O2 layer [nm]
Figure S1: Calculation of capacitance depending on Li2 O2 layer thickness, showing where capacitance is 50 % of the initial 12.6 F g-1
c
The calculation of capacitance is based on a simple model with two capacitances: the capacitance of the pris-
tine carbon electrode, and the capacitance of the Li2 O2 layer deposited on the carbon electrode. Equation 2 in the
manuscript and a calculation of two capacitances in series is used to calculate the total capacitance depending on the
Li2 O2 layer thickness. The final formula is:
!−1
1 1
Ctotal (d) = + (1)
r 0 A/d CXC72
Where r is the relative permittivity, 0 is the vacuum permittivity, A is the area of the layer, d is the thickness of
the Li2 O2 layer, and CXC72 is the capacitance of XC72 in an aprotic electrolyte. For the case where d = 0 only
the capacitance of XC72 is used, all other calculations use: r = 35 [1], 0 = 8.9 · 10−12 , A = 235 m2 g-1 , and
CXC72 = 12.6 F g-1c [2].
∗ Correspondingauthor.
Email address: andreas@lithiumbalance.com (Andreas E. Christensen)
Preprint submitted to Journal of Power Sources March 23, 2015

S2. Scaling of current densities
Target energy Vdischarge Q2.2V mc Ihigh Phigh Ipeak Ppeak

kWh V mAh g-1c kg mA g-1 c kW mA g-1 c kW
100 2.65 700 53.9 390 55.7 736 105.2
Table S1: Calculation based on target energy of 100 kWh.
The current densities are calculated based on a target energy of 100 kWh, a discharge voltage plateau of 2.65 V,
and a capacity of 700 mAh g-1
c . Peak power of 105 kW corresponds to expected peak motor power for electric vehicles,
based on comparison of 8 vehicles, model year 2013 [3].
14
250
Peak power [kW]
12 100 %
10 200
8 150
50 %
6
100
4
2 50
0 0
Smart EV
Scion iQ EV
Ford Focus EV
Mitsubishi i-MiEV
Nissan Leaf
Tesla S 85 kWh
Honda Fit EV
Fiat 500e
0 2 4 6 8 10 12
Li2O2 layer [nm]
Figure S2: Comparison of peak motor power for selected electric vehicles, model year 2013.
2
S3. Original capacitance data

Capacitance data before normalisation, from all experiments.
10
8
C]
C [Fg-1
Constant current
6 13 mAgC-1
39 mAgC-1
Variable current 130 mAg-1
4 Drive cycle 1 Staircase
C
130 mAg-1C
Drive cycle 2 Charging 390 mAg-1C
2
0 400 800 1200
-1
Capacity [mAhgC ]
Figure S3: Capacitance data from all Li2 O2 experiments.
S4. SoC estimation error

To estimate the error of the SoC estimation based on capacitance measurements, the two drive cycle measurements
and the measurement at 130 mA g-1 c in Figure 3b in the manuscript are used to give a minimum relative capacity based
on a given capacitance. The points of calibration used in the drive scenario are presented in Table S2.
Actual capacity Capacitance Minimum capacity

100% 100% 79%
85% 99% 69%
68% 98% 62%
51% 94% 39%
35% 90% 29%
18% 83% 18%
1% 61% 1%
Table S2: Calibration data.
It is clearly seen that the capacity is not estimated well in the beginning of the discharge, but in this regime
coulomb counting will be sufficient. In the end of the discharge, where the exact knowledge of the remaining capacity
is much more important, the estimation is, however, much more accurate. This enables a precise determination of the
SoC and thereby a better utilisation of the battery capacity as shown in Figure 4 in the manuscript.
References
[1] O. Gerbig, R. Merkle, J. Maier, Electron and ion transport in Li2O2, Advanced Materials 25 (2013) 3129–3133. doi:10.1002/adma.
201300264.
[2] Barbieri, Hahn, Herzog, Kotz, Capacitance limits of high surface area activated carbons for double layer capacitors, Carbon 43 (6) (2005)
1303–1310. doi:10.1016/j.carbon.2005.01.001.
[3] Compare Electric Vehicles Side-by-Side, (accessed 2015-03-19).
URL http://www.fueleconomy.gov/feg/evsbs.shtml
3
Paper IV - Rechargeability of ionic liquids in Li-O2 batteries 135
Rechargeability/Reversibility of Ionic Liquids in Li-O2

Battery/Cells
Supti Das1, Jonathan Højberg1,2, ………
1
Department of Energy Conversion and Storage, Technical University of Denmark, Frederiksborgvej
399, DK-4000 Roskilde, Denmark

2
Abstract
Keywords:
1. Introduction
A quest of alternative energy storage systems has led to significant attention towards Li-air (or Li-O2)
battery during the past decade.1-6 Exceptionally high theoretical energy density comparable to gasoline
makes Li-O2 battery more appealing than other types of metal-ion or metal-air battery systems. But there
are many challenges that are essential to overcome in order to bring this technology into practical and
commercially viable applications. In the Li-O2 battery, oxygen is reduced during discharge to form
Li2O2 at the air electrode while the lithium electrode is oxidized. During charge, Li2O2 is oxidized at the
air electrode and lithium is plated on the lithium electrode. To ensure full rechargeability, it is important
that the reaction is completely reversible with an insignificant loss in degradation reactions. The
development of an electrolyte with sufficient stability towards Li2O2 and intermediate reaction products
like the superoxide radical has been described as the biggest challenge in the development of the Li-O2
system,6 and, so far, no stable electrolytes have been identified.1, 7-9 Moreover a suitable electrolyte for
Li−O2 cells should also have superior anodic stability; low volatility to avoid solvent evaporation in
open cell system; high oxygen solubility and diffusivity to enable sufficient oxygen transport to the air
electrode to support the required currents; low viscosity; high conductivity and a wide electrochemical
window. Although many solvents have been investigated in this regard, none of them fulfill the above
mentioned requirements. Ionic liquids (ILs) have been proposed by several researchers as suitable
1
electrolyte because of their relatively high electrochemical and chemical stability against O2 radicals.3-4,
10-18
Ionic Liquids (ILs), generally comprised by an organic cations and inorganic anions, are a class
of molten organic salts which are fluids at temperatures around 100 °C or below.19-23 They have
stimulated much interest for a variety of chemical, physical and biological processes in past few decades
due to their remarkable and unique properties such as negligible volatility, thermal & electrochemical
stability, non flammability, tunable viscosity, high ionic conductivity and superior hydrophobicity.
Recently ILs have received significant attention as safe and environmentally friendly electrolytes in
electrochemical devices such as Li-O2 battery due to the nonvolatile nature.4, 11, 18, 24-25
Another
interesting fact about these ILs is that the physicochemical properties can be tuned just by varying the
combinations of cations and anions. However, one should take into account that not all ILs are suitable
as electrolyte. As an example, ILs based on imidazolium families is unstable against peroxide radical
attack.26 On the other hand pyrrolidinium and piperidinium based cations with combination of
bis(trifluoromethanesulfonyl)imide (TFSI-) anion has been reported to meet the criteria of good
electrolytes.2 There are many discussions in literature regarding the use of ionic liquid-based electrolytes
in Li-oxygen cells. Bresser et al.27 have recently reviewed the utilization of ionic liquids in Li-O2
battery. The first report on ionic liquid as electrolye in Li-air battery was published by Kuboki et al. in
2005.4 In their study hydrophobic Imidazolium based ionic liquids were investigated as electrolyte
solvent in primary lithium-O2 and lithium-air cells. Mizuno et al.28 studied the applicability of N-
methyl-N-propyl piperidinium bis(trifluoromethanesulfonyl)imide (PP13TFSI) as electrolyte solvent for
lithium-O2 batteries, where they found retention of reversible capacity about 60% of the initial capacity
after 30 cycles. Recently Monaco et al.14 reported the use of N-butyl-N-methylpyrrolidinium bis-(tri
fluoromethane sulfonyl) imidelithium-bis-(trifluoromethane-sulfonyl) imide (P14TFSI-LiTFSI, 9:1) as
electrolyte in a novel Li-O2 flow cell configuration by circulating the oxygen-saturated electrolyte
through the cell. In a more recent study, Elia et al.11 have demonstrated the reversibility of,
P14TFSI−LiTFSI) electrolyte with energy efficiency in the order of 82%, by using capacity-limited
2
galvanostatic cycling. In another report the stability of P14TFSI−LiTFSI electrolytes was investigated by
Piana et al.15 in a different Li-O2 cell configurations. They showed that P14TFSI gets reduced on
metallic lithium but it works well in specially designed configuration. They also mentioned that poor
cyclability in the particular cell design might be due to the insufficient long-term stability against the
attack of O2. Based on these new promising results, the research in developing stable IL based
electrolytes increases in order to develop a reversible Li-O2 battery for energy storage devices.
In our study, we investigate the rechargeability in a Li-O2 cell with ionic liquid electrolytes based
on five different cations and two different anions as shown in Figure 1. The main focus is N-
alkylmethylpyrrolidinium (P13+ and P14+) and N-alkylmethylpipiridinium (PP13+) based cations in
combination with the TFSI- anion, and other IL based electrolytes will be used primarily for comparison.
As mentioned before, these ILs have shown promising results as stable electrolytes in Li-O2 battery, and
we analyse the stability further using Differential Electrochemical Mass Spectrometry (DEMS) to
quantitatively assess the amount of oxygen consumed during discharge (ORR) and the types and amount
of gas evolved during charge (OER). Complementary measurements with the
bis(trifluoromethanesulfonyl)imide (FSI-) anion, or a cation based on either imidazolium or quaternary
ammonium are used to describe the desired features of an IL-based electrolyte for Li-O2 batteries.
2. Experimental materials and methods
The ILs used in this study are: N-methyl-N-propylpyrrolidinium bis(trifluoromethanesulfonyl) imide
(P13TFSI, purity 99.9 %, Solvionic), N-methyl-N-propylpyrrolidinium bis(fluorosulfonyl) imide (P13FSI,
purity 99.9 %, Solvionic), N-methyl-N-propylpiperidinium bis(trifluoromethanesulfonyl) imide
(PP13TFSI, purity 99.9 %, Solvionic), N,N-Diethyl-N-methyl-N-propylammonium
bis(fluorosulfonyl)imide(N1223FSI, purity 99.9 %, Solvionic), 1-butyl-2,3-dimethyllimidazolium
bis(trifluoromethanesulfonyl)imide (BdImTFSI, purity 99.9 %, Solvionic), N-methyl-N-
butylpyrrolidinium bis(trifluoromethanesulfonyl) imide (P14TFSI, purity 98.5 %, Sigma-Aldrich). All
ILs were used as received from suppliers. Lithium bis(trifluoromethanesulfonyl) imide (LiTFSI, purity
3
99.9 %, Sigma-Aldrich), was preheated at 180 C and Lithium bis(fluorosulfonyl) imide (LiFSI, purity
99.9 %, Suzhou Fluolyte) was preheated at 80 C prior to preparing the electrolye. The 0.3 M LiX-IL
(X=TFSI, FSI) electrolytes were prepared by adding the appropriate amount of salt and ionic liquid and
stirring over night at room temperature in order to get homogenous electrolyte solution.
Homemade carbon cathodes were manufactured by air-spraying a slurry of XC72 carbon black
(Vulcan XC72, Cabotcorp, GA) and PTFE (60 wt% dispersion in water, Sigma-Aldrich) in a wt/wt ratio
of 3:1 onto a 316SS stainless steel 150 mesh (Westcoast, Denmark) as described in.8 After air-drying for
at least one hour, the coated SS mesh was cut in 10 mm diameter cathodes. The cathodes were rinsed
carefully. First, an acetone and isopropanol wash removed any loose particles and secondly, the cathodes
were in vacuum for at least 30 minutes at room temperature followed by at least 12 h at 200 °C inside a
glove box. A homebuild airtight swagelok Li-O2 cell was used for all studies. The cell was assembled
with lithium as anode (HongKong Wisdom Tech Company), Whatmann glass fibre separators and XC72
cathode. ~60 μl of electrolyte was used for each experiment.
The presence of crystalline products on electrodes after first cycle (for discharge and charge,
separately) was analyzed using a Regaku Advance X-ray Diffractometer (2θ = 20-80°) working with Cu-
Kα radiation (λ =0.15418 nm). For this purpose, cathode materials were scratched from the electrode
after disassembling inside the glove box and inserted into a 0.7mm diameter capillary. The capillary was
sealed with glue and measured in the diffractometer.
DEMS measurements were performed at currents of 20 µA and 50 µA and the gas consumption and
gas evolution were quantified using both pressure measurements and mass spectrometry. The applied in-
house DEMS setup is similar in design to the setup used by McCloskey et al..29 From the DEMS
measurements it is possible to determine the e/O2 ratio during both discharge and charge, the amount of
oxygen consumed (ORR) compared to the oxygen evolved (OER) and the amount of CO2 and H2
30
evolved. These key characteristics are crucial to determine if a Li-O2 system is truly reversible and
Table 1 summarizes this information from all investigated electrolytes.
4
The concentration and diffusion of oxygen, denoted as [O2] and DO2, respectively, in the RTILs
were determined by a chronoamperometric technique adapted from Shoup and Szabo.31-32 Prior to this,
the ionic liquid was saturated with O2 for 12 h. A 33 µm glassy-carbon (GC) microdisc electrode (Bio-
Logic) was polished using 0.05 µm Al2O3 (Buehler). The electrochemical determinations were
conducted using a minimal amount of the IL (a few drops) in a borosilicate glass vial that prior to each
experiment had been cleaned in boiling HNO3 (Sigma-Aldrich) and then heated to a 120 oC and brought
into a glovebox where the IL was added. The glass vial was taken out of the glovebox and rapidly put
under an O2 with a Pt wire acting as both pseudo-reference- and counter electrode. The chronoampero
metric experiments were performed using a Bio-Logic VMP3 potentiostat, with a sample time of 0.01 s,
by stepping the potential from a voltage from zero current, held for 30 s, to a chosen potential after the
reduction of oxygen, measured for 5 s, that was identified by previous cyclic voltammograms.
3. Results and Discussions:
Molecular formulas of the ionic liquid cations and anions investigated in this study are shown in Figure
1. DEMS results for (e/O2)dis, (e/O2)ch, OER/ORR, CO2/ORR, and H2/ORR for all studied IL based
electrolytes and the DME based electrolyte are summarized Table 1.
Figure 2(A) shows the first galvanostatic discharge-charge cycle curve for electrolytes based on
P13TFSI and P13FSI in Li-oxygen cells at room temperature. The cells were discharged at 50µA to a
voltage cutoff of 2.2 V and charged to 4.2V at the same current. Figure 2(B) presents the total gas (O2)
consumed during discharge (ORR) and gas evolved during charge (OER). The value of the ratio
(OER/ORR) should be exactly 1 if the cell is fully rechargeable. The measurement shows that P13TFSI
and P13FSI are only 60 % and 20 % rechargeable at potentials below 4.2 V. Figure 2(C) shows
DEMS measurements during charge. Oxygen (m/z = 32) is identified as the main gaseous charging
product in both the electrolytes. For P13TFSI, the initial O2 evolution rate is close to 2 e/O2, which is
consistent with Li2O2 oxidation. But as charging continues, the oxygen evolution rate gradually
decreases to 3 e/O2. P13FSI is also showing an initial 2 e/O2 ratio, suggesting Li2O2 oxidation, but
5
after charging approximately 10 % of the full discharge capacity, the oxygen evolution decreases and the
potential increases rapidly.
Results for cells with P14TFSI and PP13TFSI based electrolytes are displayed in Figure 3. The
oxygen gas evolution rate of P14TFSI was close to 2e/O2 throughout most of the charging process and
decreased slightly at the end of charge. PP13TFSI, on the other hand, did not perform well. At 50 µA,
the battery was not able to sustain the current for long and only little oxygen was evolved. At a lower
current of 20 µA, a capacity similar to P14TFSI (measured at 50 µA) was obtained. This measurement
showed poor reversibility with an OER/ORR ratio of 40%, which is quite similar with the result by
McCloskey et al.1 Although the electron count per oxygen is 2.09 during discharge, the charge reveal
significant electrochemical degradation corresponding to more than half of measured the current.
XRD results of discharged cathodes for pure ionic liquids and DME based electrolyte was
depicted in Figure 4. All cells were discharged to 2.2 V at a current of 20 µA. Li2O2 crystalline phases
were observed in the all discharged cathodes except the electrolytes with FSI based ionic liquids (e.g.
P13FSI, N1223FSI; Figure 4, Figure S3). This indicates that no detectable Li2O2 or other crystalline
products are formed during discharge. Although the presence of Li2CO3 phases was observed from
XRD but we could not see any CO2 gas evolved till 4.2V in DEMS.
Figure 5 shows the e/O2 data through six cycles for electrolytes based on P14TFSI and
P13TFSI. It is seen that the amount of electrochemical degradation reactions increases during both
discharge and charge as the batteries are cycled which shows that even though the first cycle seems
promising, the system is not stable.
Discussion of P13FSI and P13TFSI
DEMS result in Figure 2 clearly indicates some parasitical electrochemistry attributed to
oxidation/degradation of the electrolyte or to the oxidation electrolyte degradation products.1 Here we
should mention that the error bar in O2 evolution rate in case of P13FSI-0.3MLiFSI is related to the
background correction. DEMS result clearly shows the difference in performance in Li-O2 cell just by
6
changing the anion. Hence the stability could be because of different structure of those two anions.
33
Literature shows that O2-solubility of IL might be enhanced by choosing a suitable anion although
oxygen mobility may decrease. Monaco et al.33 demonstrated in this study that the reason for the
increase in O2-solubility is the increase in fluorine content in the anion part. Other reason of the
improvement in O2-solubility could be attributed to the increase in the size of anions.34 Our solubility
measurement using the method described above also show higher O2 solubility in TFSI- based IL than
FSI- one (Table 2) which suggest better reversibility in P13TFSI based electrolytes due to better O2
stability of TFSI- anion.
Discussion of P14TFSI and PP13TFSI
It might be interesting to look into oxygen evolution rate behavior during charge of piperidinium
(PP13TFSI) based electrolytes at different current density. Here oxygen evolved at a rate higher than 2
(e/O2) (3) at the beginning but it decreases gradually to 4 electron processes at low current (20µA)
while at relatively higher current (50µA) it is showing value 4 throughout the charging process. This
clearly indicates other chemical/electrochemical reactions are taking part extensively and this electrolyte
is not stable in Li-O2 cell.
The stability of P14TFSI based electrolytes are extensively studied recently by Piana et al. in different
Li-O2 cell configurations.15 They observed decomposition of P14TFSI via reduction on metallic lithium
when they assemble cell with pure lithium as anode and vulcan XC72 carbon as cathode. They found
alkene and amines as the degradation products, detected by OEMS and 1H-NMR experiments. We too
used the similar type cell configuration and the DEMS result show only  62% reversibility in the first
cycle. That indicates some other electrochemistry are taking part during discharge and charge but we
could not detect any alkene or amine in DEMS.
General discussion
7
The same result for DME and other ionic liquid based electrolyte compositions are presented in the
Supporting Information (Figures S1, S2). Significantly small parasitical reactions are observed in case of
P13TFSI, P14TFSI and PP13TFSI where the (e/O2)dis are fairly near 2 while the value is really high in
case of BdImTFSI, P13FSI and N1223FSI based electrolytes (Table1). For comparison, 1MLiTFSI -
DME based electrolytes also display quite high degree of reversibility with the (e/O2)dis value very near
to 2 (table1 and SI). In general all electrolytes show much higher (e/O2)ch (≫ 2.0) values which implies
that some parasitical electrochemistry always occurs during the charging process. The other gas evolved
(e.g., CO2, H2) during the cell operation for all electrolytes was very low (below detection level) except
for the electrolyte based on PP13TFSI ionic liquid. We have observed substantial amounts of H2 (∼27
%, although non calibrated*) in pipiredinium based electrolytes from DEMS study. Also we have
detected trace amount of water in PP13TFSI and BdImTFSI ionic liquids while conducting the solubility
measurement.
Although P13TFSI and P14TFSI have been observed to be the best interms of stability in Li-O2
cells but we could see that these are not sufficient for long term stability purpose (Figure 5). An increase
in the e/O2 ratio during cycling agrees well with results published by McCloskey et al. in DME based
electrolytes and Piana et al. in P14TFSI based electrolytes.
4. Conclusions
We have demonstrated using DEMS that the extent of rechargeability is much dependent on the choice
of cation and anion. Unfortunately none of the studied ionic liquid based electrolytes could behave as
true ideal electrolytes in Li-O2 battery. Although the pyrrolidium-cation and TFSI-anion (P13TFSI,
P14TFSI) based ionic liquids have better rechargrability (OER/ORR>60%) than pyrrolidium-cation and
FSI-anion based ionic liquids; still these are quite far for fulfilling the requirement of perfect
electrolytes. Other electrolytes based on piperidinium, imidazoliun and quaternary ammonium are
unstable in Li-O2 battery.
8
Acknowledgements
The authors acknowledge support the ReLiable project funded by the Danish Council for
Strategic Research – Programme Commission on Sustainable Energy and Environment (project
#11-116792).
9
References
1. McCloskey, B.; Bethune, D.; Shelby, R.; Mori, T.; Scheffler, R.; Speidel, A.; Sherwood, M.;
Luntz, A., Limitations in Rechargeability of Li-O2 Batteries and Possible Origins. J. Phys. Chem. Lett.
2012, 3, 3043-3047.
2. Lu, J.; Li, L.; Park, J.-B.; Sun, Y.-K.; Wu, F.; Amine, K., Aprotic and Aqueous Li–O2 Batteries.
Chem. Rev. 2014, 114, 5611-5640.
3. Katayama, Y.; Sekiguchi, K.; Yamagata, M.; Miura, T., Electrochemical Behavior of
Oxygen/Superoxide Ion Couple in 1-Butyl-1-Methylpyrrolidinium Bis (Trifluoromethylsulfonyl) Imide
Room-Temperature Molten Salt. J. Electrochem. Soc. 2005, 152, E247-E250.
4. Kuboki, T.; Okuyama, T.; Ohsaki, T.; Takami, N., Lithium-Air Batteries Using Hydrophobic
Room Temperature Ionic Liquid Electrolyte. J. Power Sources 2005, 146, 766-769.
5. Ogasawara, T.; Débart, A.; Holzapfel, M.; Novák, P.; Bruce, P. G., Rechargeable Li2o2
Electrode for Lithium Batteries. J. Am. Chem. Soc. 2006, 128, 1390-1393.
6. McCloskey, B.; Bethune, D.; Shelby, R.; Girishkumar, G.; Luntz, A., Solvents’ Critical Role in
Nonaqueous Lithium–Oxygen Battery Electrochemistry. J. Phys. Chem. Lett. 2011, 2, 1161-1166.
7. Mekonnen, Y. S.; Knudsen, K. B.; Mýrdal, J. S. G.; Younesi, R.; Højberg, J.; Hjelm, J.; Norby,
P.; Vegge, T., Communication: The Influence of Co2 Poisoning on Overvoltages and Discharge
Capacity in Non-Aqueous Li-Air Batteries. J. Chem. Phys. 2014, 140, 121101.
8. Højberg, J.; McCloskey, B. D.; Hjelm, J.; Vegge, T.; Johansen, K.; Norby, P.; Luntz, A. C., An
Electrochemical Impedance Spectroscopy Investigation of the Overpotentials in Li–O2 Batteries. ACS
Appl. mater. Inter. 2015, 7, 4039-4047.
9. Younesi, R.; Urbonaite, S.; Edström, K.; Hahlin, M., The Cathode Surface Composition of a
Cycled Li–O2 Battery: A Photoelectron Spectroscopy Study. J. Phys. Chem. C 2012, 116, 20673-20680.
10. Balaish, M.; Kraytsberg, A.; Ein-Eli, Y., A Critical Review on Lithium–Air Battery Electrolytes.
Phys. Chem. Chem. Phys. 2014, 16, 2801-2822.
10
11. Elia, G. A.; Hassoun, J.; Kwak, W.-J.; Sun, Y.-K.; Scrosati, B.; Mueller, F.; Bresser, D.;
Passerini, S.; Oberhumer, P.; Tsiouvaras, N., An Advanced Lithium–Air Battery Exploiting an Ionic
Liquid-Based Electrolyte. Nano Lett. 2014, 14, 6572-6577.
12. Ara, M.; Meng, T.; Nazri, G.-A.; Salley, S. O.; Ng, K. S., Ternary Imidazolium-Pyrrolidinium-
Based Ionic Liquid Electrolytes for Rechargeable Li-O2 Batteries. J. Electrochem. Soc. 2014, 161,
A1969-A1975.
13. Allen, C. J.; Mukerjee, S.; Plichta, E. J.; Hendrickson, M. A.; Abraham, K., Oxygen Electrode
Rechargeability in an Ionic Liquid for the Li–Air Battery. J. Phys. Chem. Lett. 2011, 2, 2420-2424.
14. Monaco, S.; Soavi, F.; Mastragostino, M., Role of Oxygen Mass Transport in Rechargeable
Li/O2 Batteries Operating with Ionic Liquids. J. Phys. Chem. Lett. 2013, 4, 1379-1382.
15. Piana, M.; Wandt, J.; Meini, S.; Buchberger, I.; Tsiouvaras, N.; Gasteiger, H. A., Stability of a
Pyrrolidinium-Based Ionic Liquid in Li-O2 Cells. J. Electrochem. Soc. 2014, 161, A1992-A2001.
16. Soavi, F.; Monaco, S.; Mastragostino, M., Catalyst-Free Porous Carbon Cathode and Ionic
Liquid for High Efficiency, Rechargeable Li/O 2 Battery. J. Power Sources 2013, 224, 115-119.
17. Allen, C. J.; Hwang, J.; Kautz, R.; Mukerjee, S.; Plichta, E. J.; Hendrickson, M. A.; Abraham,
K., Oxygen Reduction Reactions in Ionic Liquids and the Formulation of a General Orr Mechanism for
Li–Air Batteries. J. Phys. Chem. C 2012, 116, 20755-20764.
18. Lewandowski, A.; Świderska-Mocek, A., Ionic Liquids as Electrolytes for Li-Ion Batteries—an
Overview of Electrochemical Studies. J. Power Sources 2009, 194, 601-609.
19. Galiński, M.; Lewandowski, A.; Stępniak, I., Ionic Liquids as Electrolytes. Electrochim. Acta
2006, 51, 5567-5580.
20. Fukaya, Y.; Ohno, H., Hydrophobic and Polar Ionic Liquids. Phys. Chem. Chem. Phys. 2013, 15,
4066-4072.
21. MacFarlane, D.; Meakin, P.; Sun, J.; Amini, N.; Forsyth, M., Pyrrolidinium Imides: A New
Family of Molten Salts and Conductive Plastic Crystal Phases. J. Phys. Chem. B 1999, 103, 4164-4170.
11
22. Forsyth, C.; MacFarlane, D.; Golding, J.; Huang, J.; Sun, J.; Forsyth, M., Structural
Characterization of Novel Ionic Materials Incorporating the Bis (Trifluoromethanesulfonyl) Amide
Anion. Chem. Mater. 2002, 14, 2103-2108.
23. Golding, J.; Hamid, N.; MacFarlane, D.; Forsyth, M.; Forsyth, C.; Collins, C.; Huang, J., N-
Methyl-N-Alkylpyrrolidinium Hexafluorophosphate Salts: Novel Molten Salts and Plastic Crystal
Phases. Chem. Mater. 2001, 13, 558-564.
24. Grande, L.; Paillard, E.; Kim, G.-T.; Monaco, S.; Passerini, S., Ionic Liquid Electrolytes for Li–
Air Batteries: Lithium Metal Cycling. Int. J. Mol. Sci. 2014, 15, 8122-8137.
25. Kar, M.; Simons, T. J.; Forsyth, M.; MacFarlane, D. R., Ionic Liquid Electrolytes as a Platform
for Rechargeable Metal–Air Batteries: A Perspective. Phys. Chem. Chem. Phys. 2014, 16, 18658-18674.
26. Hayyan, M.; Mjalli, F. S.; Hashim, M. A.; AlNashef, I. M., An Investigation of the Reaction
between 1-Butyl-3-Methylimidazolium Trifluoromethanesulfonate and Superoxide Ion. J. Mol. Liq.
2013, 181, 44-50.
27. Bresser, D.; Paillard, E.; Passerini, S., Ionic Liquid-Based Electrolytes for Li Metal/Air Batteries:
A Review of Materials and the New 'Labohr' Flow Cell Concept. J. Electrochem. Sci. Tech. 2014, 5, 37-
44.
28. Mizuno, F.; Takechi, K.; Higashi, S.; Shiga, T.; Shiotsuki, T.; Takazawa, N.; Sakurabayashi, Y.;
Okazaki, S.; Nitta, I.; Kodama, T., Cathode Reaction Mechanism of Non-Aqueous Li–O 2 Batteries with
Highly Oxygen Radical Stable Electrolyte Solvent. J. Power Sources 2013, 228, 47-56.
29. McCloskey, B. D.; Speidel, A.; Scheffler, R.; Miller, D. C.; Viswanathan, V.; Hummelshøj, J. S.;
Nørskov, J. K.; Luntz, A. C., Twin Problems of Interfacial Carbonate Formation in Nonaqueous Li–O2
Batteries. J. Phys. Chem. Lett. 2012, 3, 997-1001.
30. Luntz, A. C.; McCloskey, B. D., Nonaqueous Li–Air Batteries: A Status Report. Chem. Rev.
2014, 114, 11721-11750.
31. Shoup, D.; Szabo, A., Chronoamperometric Current at Finite Disk Electrodes. J. Electroanal.
Chem. Inter. Electrochem. 1982, 140, 237-245.

12
32. Xiong, L.; Aldous, L.; Henstridge, M. C.; Compton, R. G., Investigation of the Optimal
Transient Times for Chronoamperometric Analysis of Diffusion Coefficients and Concentrations in
Non-Aqueous Solvents and Ionic Liquids. Anal. Method 2012, 4, 371-376.
33. Monaco, S.; Arangio, A. M.; Soavi, F.; Mastragostino, M.; Paillard, E.; Passerini, S., An
Electrochemical Study of Oxygen Reduction in Pyrrolidinium-Based Ionic Liquids for Lithium/Oxygen
Batteries. Electrochim. Acta 2012, 83, 94-104.
34. Hu, Y.-F.; Liu, Z.-C.; Xu, C.-M.; Zhang, X.-M., The Molecular Characteristics Dominating the
Solubility of Gases in Ionic Liquids. Chem. Soc. Rev. 2011, 40, 3802-3823.
Figure Captions
Table 1: Differential Electrochemical Mass Spectrometry (DEMS) result of different liquid and DME
based electrolytes.
Figure 1: Schematic diagram of different cations and anions of ionic liquid used in this study.
Figure 2: (A) Galvanostatic discharge−charge curves for cells utilizing P13TFSI-0.3MLiTFSI and
P13FSI-0.3MLiFSI. (B) Oxygen consumption (measured using pressure decay) during discharge and
evolution (measured using DEMS) during charge. (c) Gas evolution rates for O2, CO2, and H2.
Figure 3: (A) Galvanostatic discharge−charge curves for cells employing P14TFSI-0.3MLiTFSI (50µA
discharge-charge) and PP13TFSI-0.3MLiTFSI (50µA and 20µA discharge-charge). (B) Oxygen
consumption (measured using pressure decay) during discharge and evolution (measured using DEMS)
during charge and (c) Gas evolution rates for O2, CO2, and H2.
Figure 4: XRD patterns of discharged XC72 carbon cathodes of various electrolytes.
Figure 5: Cycling behavior of P14TFSI and P13TFSI based electrolytes.
Table 2: The concentration [O2] and diffusion of oxygen, DO2, in the RTILs.
13
Manuscript Figures
Table 1: Differential Electrochemical Mass Spectrometry (DEMS) result of different ionic liquid and
DME based electrolytes (* indicates non calibrated data).
Cathode Solvent Salt OER/ORR (e-/O2)dis (e-/O2)ch CO2/ORR H2/ORR*
XC72 P13TFSI LiTFSI 0.6173 2.0048 3.0914 0.00493 0.0103

(±0.0018)
(±0.081) (±0.1413) (±0.2717) (±0.0011)
P14TFSI LiTFSI 0.6284 2.0309 2.7495 0.00415 0.01438
(±0.1368) (±0.1271) (±0.4071)
PP13TFSI LiTFSI 0.40214 2.093 4.025 0.0316 0.18

(±0.086)
(±0.1236) (±0.049) (±0.078) (±0.0011)
BdImTFSI LiTFSI 0.45 2.144 8.264 0.0155 0.008

(±0.1077) (±0.0098)
(±0.15) (±0.0068)
P13FSI LiFSI 0.1929 2.51 3.404 0.044 0.015

(±0.098)
(±0.0042) (±0.032) (±0.011)
N1223FSI LiFSI 0.1888 2.45 4.059 0.0046 0.0026

(±0.0569) (±0.117) (±0.0018) (±0.0008)
DME LiTFSI 0.796 1.897158 0.079475 0.027453

(±0.01) (±0.0468)
14
Figure 1: Schematic diagram of different cations and anions of ionic liquid used in this study.
15
Figure 2: (A) Galvanostatic discharge−charge curves for cells utilizing P13TFSI-0.3MLiTFSI
LiTFSI and P13FSI-0.3MLiFSI (both at 50µA discharge-charge). (B) Oxygen consumption
(measured using pressure decay) during discharge and evolution (measured using DEMS) during
charge. (c) Gas evolution rates for O2, CO2, and H2.
16
Figure 3: (A) Galvanostatic discharge−charge curves for cells employing P14TFSI-0.3MLiTFSI
(50µA discharge-charge) and PP13TFSI-0.3MLiTFSI (50µA and 20µA discharge-charge). (B)
Oxygen consumption (measured using pressure decay) during discharge and evolution (measured
using DEMS) during charge and (c) Gas evolution rates for O2, CO2, and H2.
17
Figure 4: XRD patterns of discharged XC72 carbon cathodes of various electrolytes.
18
Figure 5: Cycling behavior of P14TFSI and P13TFSI based electrolytes.
19
IL DO2 [cm2/s] [O2] [mM] η [mPa s]
P13FSI 2.57e-6 8.17 52.70
P13TFSI 9.17e-7 11.71 71.23
PP13TFSI 1.78e-6 10.58 151
BdIm-TFSI 1.22e-6 18.81 115.22
P14TFSI (1.8 ±0.2) e-6 14 89*
N1223FSI 1.22e-6 7.71
Table 2: The concentration [O2] and diffusion of oxygen, DO2, in the RTILs.
20
Paper IV - SI - Rechargeability of ionic liquids in Li-O2 batteries 155
Supporting information
Figure S1:
Figure S2:
Figure S3:
FigureS4: solubility data

Paper V - Influence of CO2 on overvoltages and capacity in Li-air batteries 160
THE JOURNAL OF CHEMICAL PHYSICS 140, 121101 (2014)
Communication: The influence of CO2 poisoning on overvoltages

and discharge capacity in non-aqueous Li-Air batteries
Yedilfana S. Mekonnen,1,2 Kristian B. Knudsen,1 Jon S. G. Mýrdal,1 Reza Younesi,1
Jonathan Højberg,1 Johan Hjelm,1 Poul Norby,1 and Tejs Vegge1,a)
1
Department of Energy Conversion and Storage, Technical University of Denmark, Frederiksborgvej 399,
DK-4000 Roskilde, Denmark
2
Center for Atomic-scale Materials Design, Technical University of Denmark, DK-2800 Lyngby, Denmark
(Received 31 January 2014; accepted 11 March 2014; published online 24 March 2014)
The effects of Li2 CO3 like species originating from reactions between CO2 and Li2 O2 at the cathode
of non-aqueous Li-air batteries were studied by density functional theory (DFT) and galvanostatic
charge-discharge measurements. Adsorption energies of CO2 at various nucleation sites on a stepped
(11̄00) Li2 O2 surface were determined and even a low concentration of CO2 effectively blocks the
step nucleation site and alters the Li2 O2 shape due to Li2 CO3 formation. Nudged elastic band calcu-
lations show that once CO2 is adsorbed on a step valley site, it is effectively unable to diffuse and im-
pacts the Li2 O2 growth mechanism, capacity, and overvoltages. The charging processes are strongly
influenced by CO2 contamination, and exhibit increased overvoltages and increased capacity, as a
result of poisoning of nucleation sites: this effect is predicted from DFT calculations and observed
experimentally already at 1% CO2 . Large capacity losses and overvoltages are seen at higher CO2
concentrations. © 2014 AIP Publishing LLC. [http://dx.doi.org/10.1063/1.4869212]
I. INTRODUCTION during discharge they will limit the electronic conduction

and lead to sudden death during discharge within 5–10 nm
As energy storage needs are growing rapidly, there is also
thick Li2 O2 deposits.13, 14 However, recent DFT calculations
an increase in research into high energy density materials for
found that hole and electron polaronic transports at the sur-
energy storage. Significant attention has been given to metal-
face and in bulk Li2 O2 and Li2 CO3 can take place. Using
air batteries, particularly Li-air batteries, as future environ-
a PBE+U (Hubbard-corrected Perdew–Burke–Ernzerhof) ex-
mentally friendly high energy density storage for vehicles,
change correlation functional, Garcia-Lastra et al.11 revealed
where the capacity offered by existing Li-ion technology is
that the hole polarons have higher mobility than electron po-
too low to solve the increasing demands on batteries.1 The
larons and Li2 CO3 exhibits lower conduction than Li2 O2 . Re-
Li-O2 couple is particularly attractive and could have ∼5–
cent works by Luntz et al. have shown that hole tunneling
10 times greater specific energies than currently available Li-
should dominate and polaronic transport is only expected to
ion batteries, though there are severe scientific and technical
be significant in Li2 O2 at elevated temperatures and low cur-
challenges that need to be addressed.2, 3 Such as a clear un-
rent densities.15, 16
derstanding of the Li2 O2 growth mechanisms, transport pro-
Li2 CO3 like crystalline species are formed by parasitic
cesses, interfacial phenomena, air impurities, and stability of
side reactions between the Li2 O2 or LiO2 and carbon sources
the key components are vital parts of non-aqueous recharge-
from air impurities such as CO and CO2 gases,17 the graphite
able Li-air cell research.4
itself, or the decomposition of aprotic electrolytes. Younesi
As first reported by Abraham and Jiang in 1996, the
et al.18, 34 reported the degradation of various electrolytes by
Li-O2 battery with aprotic solvent is shown to be recharge-
Li2 O2 and documented Li2 CO3 as a decomposition product
able, when Li2 O2 is formed during discharge at the cathode.5
from aprotic electrolytes. Likewise, McCloskey et al.3 have
Detailed understanding of the Li2 O2 growth mechanism is im-
shown that carbonates accumulate at the C-Li2 O2 and Li2 O2 -
portant to solve the problem associated with the practical lim-
electrolyte interfaces and are responsible for a large poten-
itations of the battery. Previous theoretical works by Hum-
tial increase during recharge and a huge decrease in exchange
melshøj et al.6 and Radin et al.7, 8 showed that steps on a
current density. This makes growth of Li2 O2 on Li2 CO3 an
reconstructed (11̄00)surface could act as nucleation sites for
equally important process to investigate, but this is beyond
low discharge overvoltage and facets such as (0001), (11̄00),
the scope of this communication. As reported by Siegfried
and (112̄0) have similar surface energies. Hummelshøj et al.9
et al.19 and Myrdal and Vegge20 adsorption of sulfur contain-
have also shown that surfaces are potential dependent and
ing compounds on oxide surfaces could also control the elec-
vary during discharge and charge. According to G0 W0 cal-
trochemical growth mechanism. Adsorbed species at surfaces
culations, both Li2 O2 and Li2 CO3 are insulating materials
can potentially block the nucleation sites, and therefore, alter
with wide band gap of 4.9 and 8.8 eV, respectively.10–12
the growth directions, overvoltages, and capacities.
Therefore, as these materials deposit at the cathode surface
In this communication, we address the influence of
CO2 contamination on the Li2 O2 growth mechanism, dis-
a) E-mail: teve@dtu.dk charge/charge overvoltages, and capacity in non-aqueous
0021-9606/2014/140(12)/121101/5/$30.00 140, 121101-1 © 2014 AIP Publishing LLC
This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to IP:
192.38.67.112 On: Thu, 15 May 2014 10:34:44
121101-2 Mekonnen et al. J. Chem. Phys. 140, 121101 (2014)
TABLE I. Adsorption energies of CO2 in the gas phase at (11̄00) Li2 O2 sites of the stepped (11̄00) Li2 O2 surface are favorable nucle-
surface. ation sites for a low overvoltage Li2 O2 growth mechanism.
The influence of CO2 poisoning on the Li2 O2 growth mecha-
Species Sites Adsorption energy (eV)
nism is studied while CO2 is already adsorbed at step valley
CO2 Step valley −0.73 site (Fig. 1(b)).
Terrace valley −0.21 The free energy diagram in Fig. 2 shows a four steps,
Step ridge −0.02 two formula units Li2 O2 growth mechanism on the stepped
(11̄00) Li2 O2 surface with and without CO2. The first step in
the presence of CO2 is adsorption of LiO2 species (Fig. 1(c)),
Li-air batteries using density functional theory (DFT) and and which reduces the binding energy by 0.44 V compared
galvanostatic measurements. Among other air contaminants, to the pure discharge. The next step is the addition of a sec-
CO2 is the most critical subject due to its high solubility in ond LiO2 species (Fig. 1(d)), which is the potential limiting
aprotic electrolytes and high reactivity with Li2 O2 to form an charge step that raises the binding energy by 0.20 V com-
insulating material Li2 CO3 . pared to pure Li2 O2 . This is followed by subsequent additions
of two Li (Figs. 1(e) and 1(f)) with relatively small binding
energies with respect to a pure discharge. In the pure O2 dis-
II. COMPUTATIONAL RESULTS AND ANALYSIS
charge mechanism, unlike in the presence of CO2 , addition of
DFT21–23 as implemented in the GPAW (grid-based the first Li is the limiting charge potential step. The 2Li2 O2
projector-augmented wave method) code24 is used to per- growth at the step surface effectively displaces CO2 from the
form the presented calculations through the atomic simu- step to the less stable terrace site.
lation environment (ASE).25 GPAW is built on real space Hummelshøj et al. have reported that the pure Li2 O2
grids and non-valence electrons are described by PAW (pro- growth mechanism follows a 4 steps reaction mechanism,
jector augmented-wave method).26, 27 Electron exchange and where all reaction steps are electrochemical, similar to what
correlation is approximated by the revised Perdew–Burke– is seen in the presence of CO2 . The equilibrium potential can
Ernzerhof (RPBE) functional.28 The stepped (11̄00) Li2 O2 be obtained as U0 = −G/2e. The effective equilibrium po-
surface with a super cell consisting of a 56–64 atoms slab tential on a pure surface becomes 2.73 V (experimental value,
with a 18 Å vacuum layer between periodic images along U0,Exp = 2.85 V), while in the presence of CO2 , this is effec-
the z-axis, see Fig. S1 in the supplementary material.35 Since tively reduced to 2.53 V for the first cycle due to the shift in
the oxygen rich (0001) facet will also be exposed, in particu- binding energy of CO2 from a step valley to terrace site. As
lar under charging conditions,9 and subsequent investigations a result, discharge at other facets may become activate.9 At
should be performed to analyze the detailed mechanisms of neutral bias all reaction steps are downhill, but at an applied
CO2 bonding to this facet. Recent computational DFT results potential, the free energy difference changes for each step cal-
for SO2 adsorption on stepped (0001) and (11̄00) surfaces do, culated as
however, show preferential bonding to the (11̄00) facets,20
which is investigated here. The k-points are sampled with a Gi,U = Gi − eU. (1)
(4,4,1) Monkhorst-Pack mesh and 0.15 grid points is used. The lowest free energy step, Gi,min , along the reaction path
Atomic energy optimization calculations are performed until becomes uphill first at an applied potential called limited dis-
all forces are less than 0.01 eV/Å. Energy barriers are cal- charge potential, Udischarge , while the largest free energy step,
culated by the climbing image nudged elastic band (CINEB) Gi,max , that is last to become downhill for the reversed re-
method.29–31 action at an applied potential called limited charge potential,
Adsorption energies of CO2 at various nucleation sites on Ucharge , obtained as
a stepped (11̄00) Li2 O2 surface were determined, see Table I.
CO2 binds preferentially at the step valley site and weakly Udischarge = min [−Gi /e] and Ucharge = max [−Gi /e].
binds at the step ridge site. NEB calculations show that once (2)
CO2 is adsorbed at step valley site, it is bound by barriers up-
wards of 3 eV, see Fig. S2 in the supplementary material,35 In the presence (absence) of a single CO2 molecule, this dis-
since the CO2 molecule is required to desorb from the surface charge occurs as described in Fig. 1, resulting in Udischarge
prior to re-adsorbing at the step site. The detailed nature of a = 2.21 V (2.66 V), and Ucharge = 2.97 V (2.81 V) and the
conversion of adsorbed CO2 to Li2 CO3 warrants further inves- discharge and charge overvoltages in the presence (absence)
tigations, but we find the adsorption of a single CO2 molecule of CO2 are ηdischarge = 0.31 V (0.07 V), and ηcharge = 0.44 V
forms a Li∼3 CO3 -type complex (Fig. 1(b)), which could act (0.08 V). The calculated 0.44 V overvoltage for charge corre-
as a nucleation site for further growth of Li2 CO3 . sponds to low CO2 concentrations, where only a single CO2
The computational lithium electrode approach is used in molecule is adsorbed on the Li2 O2 step forming a Li∼3 CO3
the free energy calculations.6, 32 Defined as, U = 0, when bulk type complex (see Fig. 1). Here, the charging process follows
Li anode and Li ions in solution (Li+ + e− ) are at equilib- the same reaction steps as the discharge, but in reverse (from
rium. The free energy change of the reaction is shifted by right to left in Fig. 2), i.e., the first two steps are desorption
−neU at an applied bias, where n is the number of transferred of two Li and followed by desorption of 2 LiO2 species: in
electrons; other assumptions are listed in the supplementary total desorbing 2 Li2 O2 units from the surface and returning
material.35 As reported by Hummelshøj et al., kinks and steps to the configuration in Fig. 1(b). Quantitative agreement with
192.38.67.112 On: Thu, 15 May 2014 10:34:44
FIG. 1. Stepped Li2 O2 (11̄00) surface before and after adsorption of CO2 and 4 steps Li2 O2 growth pathways during discharge. (a) Pure stepped Li2 O2 surface.
(b) CO2 adsorbs to step valley site forming a Li∼3 CO3 type complex. (c) 1st LiO2 adsorbs. (d) 2nd LiO2 adsorbs. (e) 1st Li. (f) 2nd Li adsorbs to the surface
completing growth of 2 Li2 O2 formula units. Atoms labeled as: C (gray), Li (purple), and O (red). Deposited atoms shown as: Li (yellow) and O (green).
experimental overvoltages can therefore only be expected for 10 mm diameter lithium foil (99.9%, Sigma-Aldrich) was
low concentrations of CO2 (e.g., 1%). For higher CO2 con- used as anode. Two Celgard separators 2500 (Celgard) were
centrations, the formation of crystalline Li2 CO3 would be ex- placed in between the two electrodes. The separators were
pected, resulting in significantly larger overvoltages.3 sonicated in EtOH (99.9%, Sigma-Aldrich), transferred to a
glovebox, and rinsed with DME before drying in vacuum at
80 ◦ C for 12 h. Experiments were performed using a Bio-
III. EXPERIMENTAL RESULTS AND ANALYSIS Logic VMP3 Multichannel galvanostat (Bio-Logic, Claix,
Li-air batteries were constructed using a Swagelok de- France). Batteries were operated in two galvanostatic modes:
sign and assembled inside an Ar-filled glovebox (≤3 ppm First, at 100 μA (127.3 μA/cm2 ) where cells were discharged
O2 and H2 O). Each battery contained a 200 μl 1 M LiTFSI to 2 V and charged to 4.6 V vs. Li+ /Li. Second, at 50 μA
(99.95%, Sigma-Aldrich) and 1,2-dimethoxymethane, DME, (63.6 μA/cm2 ) using the same potential limits.
(H2 O < 20 ppm, BASF) electrolyte. Cathodes consisted of To investigate the effect of gaseous CO2 , the assembled
P50 AvCarb carbon paper (Fuel cell store), which were son- cells were purged with three different atmospheres: 0/100
icated using 2-propanol (99.5%, Sigma-Aldrich) and acetone CO2 /O2 , 1/99 CO2 /O2 , and 50/50 CO2 /O2 . Three individ-
(≥99.8%, Sigma-Aldrich), introduced into a glovebox where ual batteries were assembled and investigated for each atmo-
they were rinsed with DME before drying in vacuum at 80 ◦ C sphere and each curve presented in Figs. 3 and 4 is there-
for 12 h. Cathodes were supported by a 316 steel mesh. A fore an average of three cells with the equal atmosphere as
shown in Fig. S3 in the supplementary material.35 The lowest
discharge capacity was observed for the 50% CO2 cells and
is likely caused by the high concentration of electrochemi-
cally inactive CO2 . A similar effect was observed, by Gowda
et al.17 for a pure CO2 cell, where the cell potential immedi-
ately dropped. It should however be noted that Takechi et al.33
observed, quite to the contrary of our observations, higher
discharge capacities up to 70% CO2 with respect to pure O2
cells. Interestingly, a higher discharge capacity was observed
for the 1% CO2 cells in respect to the pure O2 cells as shown
in Fig. 3 (inset). A possible explanation is the dissolution of
Li2 CO3 species in DME and/or, as also suggested by Gowda
et al., or a change in deposition morphology compared to that
deposited in the pure O2 cells as suggested by Myrdal and
Vegge.20 Such morphological changes could increase the to-
tal electrodeposited layer and lead to higher capacities.
All CO2 cells have higher discharge overvoltages com-
FIG. 2. Calculated free energy diagrams for a four steps discharge mecha-
pared to cells with pure O2 at a discharge rate of 127.3
nism on a stepped (11̄00)Li2 O2 surface with and without adsorbed CO2 . μA/cm2 , which may be caused by the blocking of the
192.38.67.112 On: Thu, 15 May 2014 10:34:44
by the Butler-Volmer model, while the larger overvoltage for

the 1% CO2 cells than 0% CO2 cells is expectedly caused
by the formation and oxidation of the carbonate like species
(Fig. 1(b)). A second charge at 63.6 μA/cm2 shows identical
results for 1% and 0% CO2 . This can be ascribed to the evo-
lution of CO2 observed during the initial charge cycle, where
CO2 is released at 4.5 V, as shown in Fig. S5 in the sup-
plementary material,35 resulting in residual CO2 in the elec-
trolyte causing blocking of the step sites in subsequent charg-
ing experiments.
IV. CONCLUSIONS
Influences of CO2 poisoning at a stepped (11̄00) Li2 O2
surface in non-aqueous Li-air battery were studied using DFT
FIG. 3. Galvanostatic discharge profiles at 127.3 μA/cm2 discharge at three calculations and cells were characterized by electrochemical
different atmospheres: 50% CO2 , 1% CO2 , and 0% CO2 . Inset shows the charge-discharge measurements. CO2 preferentially binds at
increase in discharge capacity in 1% CO2 . step valley site at the Li2 O2 surface and the Li2 O2 growth
mechanism consists of four electrochemical steps, following
the same sequence for both pure and contaminated systems.
active nucleation sites by solubilized CO2 , forcing the
Accordingly, the first step of the growth mechanism is the ad-
reactions to follow pathways with higher overvoltages. This
sorption of two LiO2 species and followed by addition of two
effect can even be seen at 1% CO2 , as illustrated in Fig. 3
Li to form 2 Li2 O2 at the cathode surface. For charge in the
above. The charge capacity, as seen in Fig. 4 and Fig. S4 in
low CO2 limit, a similar reaction will occur, but in reverse
the supplementary material,35 is very dependent on the CO2
order.
concentration, with high concentrations limiting charge ca-
Low concentrations of CO2 (1%) effectively block the
pacity and thereby the cell reversibly. The 50% CO2 cells
surface-active nucleation sites and alter the shape and growth
reach the lower potential limit (2.0 V) early, at approximately
directions of Li2 O2 on the surface; resulting in an increased
35 mAh/g, while 1% CO2 cells and pure O2 cells continued
capacity of the battery at the expense of an increase in the
until capacities in the range 1150–1600 mAh/g were reached
overvoltage in the presence of CO2 . A similar behavior is seen
depending on current density. The low charge capacity at high
in pure oxygen following charging to 4.5 V, resulting from
CO2 contaminations should be attributed to the poor Li-CO2
decomposition reactions. The effective discharge potential is
electrochemistry, also reported by Gowda et al. The charging
reduced by 0.20 V on a stepped (11̄00) Li2 O2 surface, shifting
overvoltages are a function of both current density and the
the reaction to alternate nucleation sites. In general, the DFT
level of CO2 contamination. While there is no significant dif-
calculations and experimental results show that the recharging
ference in overvoltages between cells charge at 127.3 and 63.6
process is strongly influenced by CO2 contamination, and ex-
μA/cm2 for 50% CO2 cells, which again can be attributed to
hibits significantly increased charging overvoltage, which is
the poor Li-CO2 electrochemistry. At 127.3 μA/cm2 , there is
observed already with 1% CO2 contamination, while at 50%
an increase in overvoltage of about 0.4 and 0.3 V for 1% CO2
CO2 a large capacity loss is also seen.
cells and 0% CO2 cells, respectively. The general increase in
overvoltages with increasing current density can be explained
ACKNOWLEDGMENTS
The authors acknowledge support of this work from the
ReLiable project (Project No. 11-116792) funded by the Dan-
ish Council for Strategic Research Programme Commission
1 D. Linden and T. Reddy, Hand Book of Batteries, 3rd ed. (McGraw Hill,
New York, 2001).
2 T. Ogasawara, A. Débart, M. Holzapfel, P. Novák, and P. G. Bruce, J. Am.
Chem. Soc. 128, 1390 (2006).

3 B. D. McCloskey, A. Speidel, R. Scheffler, D. C. Miller, V. Viswanathan,
J. S. Hummelshøj, J. K. Nørskov, and A. C. Luntz, J. Phys. Chem. Lett. 3,

997 (2012).
4 G. Girishkumar, B. D. McCloskey, A. C. Luntz, S. Swanson, and W.
Wilcke, J. Phys. Chem. Lett. 1, 2193 (2010).

5 K. M. Abraham and Z. Jiang, J. Electrochem. Soc. 143, 1 (1996).
6 J. S. Hummelshøj, J. Blomqvist, S. Datta, T. Vegge, J. Rossmeisl, K. S.
Thygesen, A. C. Luntz, K. W. Jacobsen, and J. K. Nørskov, J. Chem. Phys.

132, 071101 (2010).
FIG. 4. Galvanostatic charge profiles at 127.3 (solid) and 63.6 (dotted) 7 M. D. Radin, J. F. Rodriguez, F. Tian, and D. J. Siegel, J. Am. Chem. Soc.
μA/cm2 at three different atmospheres: 50% CO2 , 1% CO2 , and 0% CO2 . 134, 1093 (2011).
192.38.67.112 On: Thu, 15 May 2014 10:34:44
8 M. D. Radin, F. Tian, and D. J. Siegel, J. Mat. Sci. 47, 7564 (2012). 23 J. J. Mortensen, L. B. Hansen, and K. W. Jacobsen, Phys. Rev. B 71, 035109
9 J. S. Hummelshøj, A. C. Luntz, and J. K. Nørskov, J. Chem. Phys. 138, (2005).
034703 (2013). 24 J. Enkovaara, C. Rostgaard, J. J. Mortensen, J. Chen, M. Dulak, L. Ferrighi,
10 P. Albertus, G. Girishkumar, B. D. McCloskey, R. S. Sanchez-Carrera, B. J. Gavnholt, C. Glinsvad, V. Haikola, H. A. Hansen, H. H. Kristoffersen, M.
Kozinsky, J. Christensen, and A. C. Luntz, J. Electrochem. Soc. 158(3), Kuisma, A. H. Larsen, L. Lehtovaara, M. Ljungberg, O. Lopez-Acevedo,
A343 (2011). P. G. Moses, J. Ojanen, T. Olsen, V. Petzold, N. A. Romero, J. Stausholm-
11 J. M. Garcia-Lastra, J. S. G. Myrdal, K. S. Thygesen, and T. Vegge, J. Phys. Moller, M. Strange, G. A. Tritsaris, M. Vanin, M. Walter, B. Hammer, H.
Chem. C 117, 5568 (2013). Hakkinen, G. K. H. Madsen, R. M. Nieminen, J. K. Nørskov, M. Puska,
12 J. M. Garcia-Lastra, J. D. Bass, and K. S. Thygesen, J. Chem. Phys. 135, T. T. Rantala, J. Schiøtz, K. S. Thygesen, and K. W. Jacobsen, J. Phys.
121101 (2011). Condens. Matter 22, 253202 (2010).
13 V. Viswanathan, K. S. Thygesen, J. S. Hummelshøj, J. K. Nørskov, G. 25 S. R. Bahn and K. W. Jacobsen, Comput. Sci. Eng. 4, 56 (2002).
Girishkumar, B. D. McCloskey, and A. C. Luntz, J. Chem. Phys. 135, 26 P. E. Blöchl, Phys. Rev. 50, 17953 (1994).
214704 (2011). 27 P. E. Blöchl, C. J. Först, and J. Schimpl, Bull. Mater. Sci. 26, 33 (2003).
14 J. Chen, J. S. Hummelshøj, K. S. Thygesen, J. S. G. Myrdal, J. K. Nørskov, 28 B. Hammer, L. B. Hansen, and J. K. Nørskov, Phys. Rev. B 59, 7413
and T. Vegge, Catal. Today 165, 2 (2011). (1999).

15 J. B. Varley, V. Viswanathan, J. K. Nørskov, and A. C. Luntz, Energy Env- 29 H. Jonsson, G. Mills, and K. W. Jacobsen, Classical and Quantum Dynam-
iron. Sci. 7, 720 (2014). ics in Condensed Phase Systems, edited by B. J. Berne, G. Cicotti, and D.
16 A. C. Luntz, V. Viswanathan, J. Voss, J. B. Varley, J. K. Nørskov, R. Schef- F. Coker (World Scientific, 1998).
fler, and A. Speidel, J. Phys. Chem. Lett. 4, 3494 (2013). 30 G. Henkelman and H. Jónsson, J. Chem. Phys. 113, 9978 (2000).
17 S. R. Gowda, A. Brunet, G. M. Wallraff, and B. D. McCloskey, J. Phys. 31 G. Henkelman, B. Uberuaga, and H. A. Jónsson, J. Chem. Phys. 113, 9901
Chem. Lett. 4, 276 (2013). (2000).

18 R. Younesi, M. Hahlin, F. Björefors, P. Johansson, and K. Edström, Chem. 32 J. K. Nørskov, J. Rossmeisl, A. Logadottir, L. Lindqvist, J. R. Kitchin, T.
Mater. 25, 77 (2013). Bligaard, and H. Jonsson, J. Phys. Chem. B 108, 17886 (2004).
19 M. J. Siegfried and K. S. Choi, Adv. Mat. 16, 1743 (2004). 33 K. Takechi, T. Shiga, and T. Asaoka, Chem. Commun. 47, 3463 (2011).
20 J. S. G. Myrdal and T. Vegge, “DFT study of selective poisoning of Li-Air 34 R. Younesi, P. Norby, and T. Vegge, ECS Electrochem. Lett. 3, A15
batteries for increased discharge capacity,” RSC Adv. (to be published). (2014).
21 P. Hohenberg and W. Kohn, Phys. Rev. 136, B864 (1964). 35 See supplementary material at http://dx.doi.org/10.1063/1.4869212 for
22 W. Kohn and L. Sham, Phys. Rev. 140, A1133 (1965). Figs. S1–S5.
192.38.67.112 On: Thu, 15 May 2014 10:34:44
Bibliography
[1] REN21. Renewables 2014: Global Status Report - Key Findings. Production,
pages 1–13, 2014. doi: ISBN9783981593426.
[2] IEA. World Energy Outlook 2014, 2014. doi: 10.1787/weo-2014-en.
[3] IPCC. Summary for Policymakers. In Climate Change 2013: The Physical Science
Basis. Contribution of Working Group I to the Fifth Assessment Report of the
Intergovernmental Panel on Climate Change, page 33. 2013. ISBN 9781107661820.
doi: 10.1017/CBO9781107415324.
[4] International Energy Agengy. 2014 Key World Energy Statistics, 2014.
URL http://www.iea.org/publications/freepublications/publication/
KeyWorld2014.pdf.
[5] M.Stanley Whittingham. Chemistry of intercalation compounds: Metal guests in

chalcogenide hosts, 1978. ISSN 00796786. doi: 10.1016/0079-6786(78)90003-1.
[6] K. Mizushima, P. C. Jones, P. Wiseman, and John B. Goodenough. LixCoO2

(0¡x¡1): A new cathode material for batteries of high energy density, 1981. ISSN
01672738. doi: 10.1016/0167-2738(81)90077-1.
[7] International Energy Agency (IEA). Technology roadmap: Electric and plug-in
hybrid electric vehicles. International Energy Agency, Tech. Rep, (June):52, 2011.
doi: 10.1109/IEMBS.2004.1403974.
[8] USCAR. Goals for Advanced Batteries for EVs (2013) United States Council for
Automotive Research document, 2013. URL www.uscar.org/commands/files_
download.php?files_id=27.
[9] K. M. Abraham and Z. Jiang. A Polymer Electrolyte-Based Rechargeable Lithi-

um/Oxygen Battery. Journal of The Electrochemical Society, 143(1):1–5, 1996.
doi: 10.1149/1.1836378.
165
Bibliography 166
[10] G. Girishkumar, Bryan D. McCloskey, Alan C. Luntz, S. Swanson, and W. Wilcke.

Lithium-Air Battery: Promise and Challenges. Journal of Physical Chemistry
Letters, 1(14):2193–2203, 2010. doi: 10.1021/jz1005384.
[11] Jang-Soo Lee, Sun Tai Kim, Ruiguo Cao, Nam-Soon Choi, Meilin Liu, Kyu Tae
Lee, and Jaephil Cho. Metal-Air Batteries: Metal-Air Batteries with High Energy
Density: Li-Air versus Zn-Air. Advanced Energy Materials, 1(1):2–2, 2011. doi:
10.1002/aenm.201190001.
[12] Takeshi Ogasawara, Aurélie Débart, Michael Holzapfel, Petr Novák, and Peter G.
Bruce. Rechargeable Li2O2 Electrode for Lithium Batteries. Journal of the Amer-
ican Chemical Society, 128(4):1390–1393, 2006. doi: 10.1021/ja056811q.
[13] Nobuyuki Imanishi, Alan C. Luntz, and Peter G. Bruce. The Lithium Air Battery:
Fundamentals. 2014. ISBN 9781489980618. doi: 10.1007/978-1-4899-8062-5.
[14] Alan C. Luntz and Bryan D. McCloskey. Nonaqueous Li-Air Batteries : A Status
Report. Chemical Reviews, 114:11721–11750, 2014.
[15] Stefano Meini, M. Piana, H. Beyer, J. Schwammlein, and Hubert A. Gasteiger.

Effect of Carbon Surface Area on First Discharge Capacity of Li-O2 Cathodes and
Cycle-Life Behavior in Ether-Based Electrolytes. Journal of the Electrochemical
Society, 159(12):A2135–A2142, October 2012. doi: 10.1149/2.011301jes.
[16] Reza Younesi, Neelam Ingh, Sigita Urbonaite, and Kristina Edström. The Effect
of Pore Size on the Performance of the Li-O. In ECS transaction, pages 121–127,
2010. doi: 10.1149/1.3414010.
[17] Allen J. Bard and Larry R. Faulkner. Electrochemical Methods - Fundamentals

and Applications. 2001. ISBN 0471043729.
[18] Mark E. Orazem and Bernard Tribollet. Electrochemical Impedance Spectroscopy.

The Electrochemical Society, 2008. ISBN 9780470041406.
[19] James Adams, Mohan Karulkar, and Venkataramani Anandan. Evaluation and
electrochemical analyses of cathodes for lithium-air batteries. Journal of Power
Sources, 239:132–143, 2013. doi: 10.1016/j.jpowsour.2013.03.140.
[20] Imanol Landa-Medrano, Idoia Ruiz de Larramendi, Nagore Ortiz-Vitoriano, Ri-

cardo Pinedo, José Ignacio Ruiz de Larramendi, and Teófilo Rojo. In situ
monitoring of discharge/charge processes in Li-O2 batteries by electrochemical
impedance spectroscopy. Journal of Power Sources, 249:110–117, March 2014.
doi: 10.1016/j.jpowsour.2013.10.077.
Bibliography 167
[21] Ingo Bardenhagen, Olga Yezerska, Matthias Augustin, Daniela Fenske, Arne Witt-
stock, and Marcus Bäumer. In situ Investigation of Pore Clogging During Dis-
charge of a Li/O2 Battery by Electrochemical Impedance Spectroscopy. Journal
of Power Sources, 278:255–264, December 2014. doi: 10.1016/j.jpowsour.2014.12.
076.
[22] M. Eswaran, N. Munichandraiah, and L. G. Scanlon. High Capacity Li-O2 Cell

and Electrochemical Impedance Spectroscopy Study. Electrochemical and Solid-
State Letters, 13(9):A121–A124, 2010. doi: 10.1149/1.3447867.
[23] Mojtaba Mirzaeian and Peter J. Hall. Characterizing capacity loss of lithium
oxygen batteries by impedance spectroscopy. Journal of Power Sources, 195(19):
6817 – 6824, 2010. doi: 10.1016/j.jpowsour.2010.04.064.
[24] Ruben Nelson, Mark H. Weatherspoon, Jamie Gomez, Egwu E. Kalu, and Jim P.
Zheng. Investigation of a Li-O2 cell featuring a binder-free cathode via impedance
spectroscopy and equivalent circuit model analysis. Electrochemistry Communi-
cations, 34:77–80, September 2013. doi: 10.1016/j.elecom.2013.04.021.
[25] Ali Rinaldi, Olivia Wijaya, Harry E. Hoster, and Denis Y W Yu. History effects
in lithium-oxygen batteries: How initial seeding influences the discharge capacity.
ChemSusChem, 7:1283–1288, 2014. doi: 10.1002/cssc.201300986.
[26] M. Mehta, G. Mixon, J. .P. Zheng, and P. Andrei. Analytical Electrochemical

Impedance Modeling of Li-Air Batteries under D.C. Discharge. Journal of the
Electrochemical Society, 160(11):A2033–A2045, September 2013. doi: 10.1149/2.
046311jes.
[27] Venkat Viswanathan, Jens K. Nø rskov, Angela Speidel, Rouven Scheffler, San-
keth R. Gowda, and Alan C. Luntz. Li-O2 Kinetic Overpotentials Tafel Plots from
Experiment and First-Principles Theory. Journal of Physical Chemistry Letters,
4:556–560, 2013.
[28] Bryan D. McCloskey, Angela Speidel, Rouven Scheffler, D. Miller, Venkat

Viswanathan, Jens S. Hummelshø j, Jens K. Nø rskov, and Alan C. Luntz. Twin
Problems of Interfacial Carbonate Formation in Nonaqueous Li-O2 Batteries. The
Journal of Physical Chemistry Letters, 3:997–1001, 2012.
[29] Paul Albertus, G. Girishkumar, Bryan D. McCloskey, Roel S. Sánchez-Carrera,

Boris Kozinsky, Jake Christensen, and Alan C. Luntz. Identifying Capacity Lim-
itations in the Li/Oxygen Battery Using Experiments and Modeling. Journal of
The Electrochemical Society, 158(3):A343, 2011. doi: 10.1149/1.3527055.
Bibliography 168
[30] Yi-Chun Lu, Hubert A. Gasteiger, and Yang Shao-Horn. Method Development
to Evaluate the Oxygen Reduction Activity of High-Surface-Area Catalysts for
Li-Air Batteries. Electrochemical and Solid-State Letters, 14(5):A70–A74, 2011.
doi: 10.1149/1.3555071.
[31] Andrzej Lasia. Impedance of porous electrodes. Journal of Electroanalytical Chem-

istry, 397(1-2):27–33, November 1995. doi: 10.1016/0022-0728(95)04177-5.
[32] Bryan Hirschorn, Mark E. Orazem, Bernard Tribollet, Vincent Vivier, Isabelle
Frateur, and Marco Musiani. Determination of effective capacitance and film
thickness from constant-phase-element parameters. Electrochimica Acta, 55(21):
6218–6227, August 2010. doi: 10.1016/j.electacta.2009.10.065.
[33] J. Jamnik and J. Maier. Generalised equivalent circuits for mass and charge
transport: chemical capacitance and its implications. Physical Chemistry Chemical
Physics, 3(9):1668–1678, 2001. doi: 10.1039/b100180i.
[34] Piotr Zoltowski. On the electrical capacitance of interfaces exhibiting constant

phase element behaviour. Journal of Electroanalytical Chemistry, 443(1):149–154,
February 1998. doi: 10.1016/S0022-0728(97)00490-7.
[35] Doron Aurbach, I. Weissman, A. Zaban, and O. Chusid. Correlation between

surface chemistry, morphology, cycling efficiency and interfacial properties of Li
electrodes in solutions containing different Li salts. Electrochimica Acta, 39(1):
51–71, 1994.
[36] Doron Aurbach, Arie Zaban, Alexander Schechter, Yair Ein-eli, Ella Zinigrad,
and Boris Markovsky. The Study of Electrolyte Solutions Based on Ethylene and
Diethyl Carbonates for Rechargeable Li Batteries. I. Li metal Anodes. Journal of
The Electrochemical Society, 142(9):2873–2882, 1995.
[37] O. Barbieri, M. Hahn, a. Herzog, and R. Kötz. Capacitance limits of high surface
area activated carbons for double layer capacitors. Carbon, 43(6):1303–1310, May
2005. doi: 10.1016/j.carbon.2005.01.001.
[38] Oliver Gerbig, Rotraut Merkle, and Joachim Maier. Electron and Ion Transport
In Li2O2. Advanced materials, 25(22):3129–3133, May 2013. doi: 10.1002/adma.
201300264.
[39] A. Dunst, V. Epp, I. Hanzu, S. A. Freunberger, and M. Wilkening. Short-range

Li diffusion vs. long-range ionic conduction in nanocrystalline lithium peroxide
Li2O2 - the discharge product in lithium-air batteries. Energy & Environmental
Science, 7:2739–2752, 2014. doi: 10.1039/c4ee00496e.
Bibliography 169
[40] K. F. Young and H. P. R. Frederikse. Compilation of the Static Dielectric Constant

of Inorganic Solids. Journal of Physical and Chemical Reference Data, 2(2):313,
1973. doi: 10.1063/1.3253121.
[41] Bryan D. McCloskey, Donald S. Bethune, Robert M. Shelby, G. Girishkumar, and

Alan C. Luntz. Solvents’ Critical Role in Nonaqueous Lithium-Oxygen Battery
Electrochemistry. Journal of Physical Chemistry Letters, 2(10):1161–1166, 2011.
doi: 10.1021/jz200352v.
[42] Stanley Bruckenstein and R. Rao Gadde. Use of a Porous Electrode for in Situ
Mass Spectrometric Determination of Volatile Electrode Reaction Products. Jour-
nal of the American Chemical Society, 93(3):793–794, 1971.
[43] Yi-Chun Lu, David G. Kwabi, Koffi P. C. Yao, Jonathon R. Harding, Jigang Zhou,
Lucia Zuin, and Yang Shao-Horn. The discharge rate capability of rechargeable
Li-O2 batteries. Energy and Environmental Science, 4(8):2999–3007, 2011. doi:
10.1039/C1EE01500A.
[44] Yi-Chun Lu, Zhichuan Xu, Hubert A. Gasteiger, Shuo Chen, Kimberly Hamad-
Schifferli, and Yang Shao-Horn. Platinum-Gold Nanoparticles: A Highly Active
Bifunctional Electrocatalyst for Rechargeable Lithium-Air Batteries. Journal of
the American Chemical Society, 132:12170–12171, 2010. doi: 10.1021/ja1036572.
[45] Robert R. Mitchell, Betar M. Gallant, Carl V. Thompson, and Yang Shao-Horn.
All-carbon-nanofiber electrodes for high-energy rechargeable Li-O2 batteries. En-
ergy and Environmental Science, 4(8):2952–2958, 2011.
[46] Robert R. Mitchell, Betar M. Gallant, Yang Shao-horn, and Carl V. Thompson.
Mechanisms of Morphological Evolution of Li2O2 Particles During Electrochemi-
cal Growth. The Journal of Physical Chemistry Letters, 4:1060–1064, 2013.
[47] Li Zhong, Robert R. Mitchell, Yang Liu, Betar M Gallant, Carl V Thompson,
Jian Yu Huang, Scott X Mao, and Yang Shao-Horn. In situ transmission electron
microscopy observations of electrochemical oxidation of Li2O2. Nano letters, 13
(5):2209–14, May 2013. doi: 10.1021/nl400731w.
[48] Yin Yang, Qian Sun, Yue-Sheng Li, Hong Li, and Zheng-Wen Fu. Nanostructured
Diamond Like Carbon Thin Film Electrodes for Lithium Air Batteries. Journal
of The Electrochemical Society, 158(10):B1211, 2011. doi: 10.1149/1.3623431.
[49] Yin Yang, Qian Sun, Yue-Sheng Li, Hong Li, and Zheng-Wen Fu. A CoOx/carbon
double-layer thin film air electrode for nonaqueous Li-air batteries. Journal of
Power Sources, 223:312–318, February 2013. doi: 10.1016/j.jpowsour.2012.09.052.
Bibliography 170
[50] Hun-Gi Jung, Hee-Soo Kim, Jin-Bum Park, In-Hwan Oh, Jusef Hassoun,
Chong Seung Yoon, Bruno Scrosati, and Yang-Kook Sun. A transmission electron
microscopy study of the electrochemical process of lithium-oxygen cells. Nano
letters, 12(8):4333–5, August 2012. doi: 10.1021/nl302066d.
[51] Christopher Graves. RAVDAV data analysis software, version 0.9.7, 2012.
[52] Bryan D. McCloskey, Alexia Valery, Alan C. Luntz, Sanketh R. Gowda, Gre-
gory M. Wallra, Jeannette M. Garcia, Takashi Mori, and Leslie E. Krupp. Com-
bining Accurate O2 and Li2O2 Assays to Separate Discharge and Charge Stability
Limitations in Nonaqueous Li-O2 Batteries. The Journal of Physical Chemistry
Letters, 4:2989–2993, 2013.
[53] Sanketh R. Gowda, A. Brunet, G. M. Wallraff, and Bryan D. McCloskey. Impli-

cations of CO2 Contamination in Rechargeable Nonaqueous Li-O2 Batteries. The
Journal of Physical Chemistry Letters, 4:276–279, 2013.
[54] Bryan D. McCloskey, Rouven Scheffler, Angela Speidel, Donald S. Bethune,

Robert M. Shelby, and Alan C. Luntz. On the Efficacy of Electrocatalysis in
Nonaqueous Li-O2 Batteries. Journal of American Chemical Society, 133(45):
18038–18041, 2011. doi: 10.1021/ja207229n.
[55] Bryan D. McCloskey, Rouven Scheffler, Angela Speidel, G. Girishkumar, and

Alan C. Luntz. On the Mechanism of Nonaqueous Li-O 2 Electrochemistry on
C and Its Kinetic Overpotentials : Some Implications for Li-Air Batteries. The
Journal of Physical Chemistry C, 116:23897–23905, 2012.
[56] Bryan D. McCloskey, Alan C. Luntz, Donald S. Bethune, Robert M. Shelby,

T. Mori, Rouven Scheffler, Angela Speidel, and M. Sherwood. Limitations in
Rechargeability of Li-O2 Batteries and Possible Origins. The Journal of Physical
Chemistry Letters, 3:3043–3047, 2012.
[57] Mie Mø ller Storm, Rune E. Johnsen, Reza Younesi, and Poul Norby. Capillary
based Li–air batteries for in situ synchrotron X-ray powder diffraction studies. J.
Mater. Chem. A, 3:3113–3119, 2015. doi: 10.1039/C4TA04291C.
[58] Nagaphani B. Aetukuri, Bryan D. McCloskey, Jeannette M. Garcı́a, Leslie E.

Krupp, Venkatasubramanian Viswanathan, and Alan C. Luntz. Solvating addi-
tives drive solution-mediated electrochemistry and enhance toroid growth in non-
aqueous Li–O2 batteries. Nature Chemistry, (January), 2014. doi: 10.1038/nchem.
2132.
Bibliography 171
[59] Feng Wang, Jason Graetz, M. Sergio Moreno, Chao Ma, Lijun Wu, Vyacheslav
Volkov, and Yimei Zhu. Chemical distribution and bonding of lithium in inter-
calated graphite: Identification with optimized electron energy loss spectroscopy.
ACS Nano, 5:1190–1197, 2011. doi: 10.1021/nn1028168.
[60] Reza Younesi, Maria Hahlin, Matthew Roberts, and Kristina Edström. The SEI
Layer Formed on Lithium Metal in the Presence of Oxygen: A Seldom Considered
Component in the Development of the Li-O2 battery. Journal of Power Sources,
225:40–45, October 2013. doi: 10.1016/j.jpowsour.2012.10.011.
[61] Juan Maria Garcia-Lastra, Jon S.G. Myrdal, Rune Christensen, Kristian S. Thyge-
sen, and Tejs Vegge. DFT+U Study of Polaronic Conduction in Li2O2 and
Li2CO3: Implications for Li-Air Batteries. The Journal of Physical Chemistry
C, 117(11):5568–5577, March 2013. doi: 10.1021/jp3107809.
[62] Yedilfana S. Mekonnen, Kristian B. Knudsen, Jon S G Mýrdal, Reza Younesi,

Jonathan Hø jberg, Johan Hjelm, Poul Norby, and Tejs Vegge. Communication:
The influence of CO2 poisoning on overvoltages and discharge capacity in non-
aqueous Li-Air batteries. Journal of Chemical Physics, 2014. doi: 10.1063/1.
4869212.
[63] Alan C. Luntz, Venkat Viswanathan, J. Voss, J. B. Varley, Angela Speidel, Jens K.
Nø rskov, and Rouven Scheffler. Tunneling and Polaron Charge Transport through
Li2O2 in Li-O2 Batteries. The Journal of Physical Chemistry Letters, 4:3494–3499,
2013.
[64] Y. Wang and S. C. Cho. Analysis of Air Cathode Perfomance for Lithium-Air
Batteries. Journal of the Electrochemical Society, 160(10):A1847–A1855, August
2013. doi: 10.1149/2.092310jes.
[65] Yuhui Chen, Stefan A. Freunberger, Zhangquan Peng, Olivier Fontaine, and Pe-
ter G. Bruce. Charging a Li-O2 battery using a redox mediator. Nature chemistry,
5(June):489–494, 2013. doi: 10.1038/NCHEM.1646.
[66] Yasushi Katayama, Kaori Sekiguchi, Masaki Yamagata, and Takashi Miura.
Electrochemical Behavior of Oxygen/Superoxide Ion Couple in 1-Butyl-1-
methylpyrrolidinium Bis(trifluoromethylsulfonyl)imide Room-Temperature
Molten Salt. Journal of The Electrochemical Society, 152:E247, 2005. doi:
10.1149/1.1946530.
[67] Takashi Kuboki, Tetsuo Okuyama, Takahisa Ohsaki, and Norio Takami. Lithium-
air batteries using hydrophobic room temperature ionic liquid electrolyte. Journal
of Power Sources, 146(1-2):766–769, 2005. doi: 10.1016/j.jpowsour.2005.03.082.
Bibliography 172
[68] Moran Balaish, Alexander Kraytsberg, and Yair Ein-Eli. A critical review on
lithium-air battery electrolytes. Physical chemistry chemical physics : PCCP, 16:
2801–22, 2014. doi: 10.1039/c3cp54165g.
[69] G A Elia, J Hassoun, W.-J. Kwak, Y.-K. Sun, B Scrosati, F Mueller, D Bresser,
S Passerini, P Oberhumer, N Tsiouvaras, and J Reiter. An Advanced Lithium–Air
Battery Exploiting an Ionic Liquid-Based Electrolyte. Nano Letters, 14:6572–6577,
2014. doi: 10.1021/nl5031985.
[70] M. Ara, T. Meng, G.-a. Nazri, S. O. Salley, and K. Y. Simon Ng. Ternary
Imidazolium-Pyrrolidinium-Based Ionic Liquid Electrolytes for Rechargeable Li-
O2 Batteries. Journal of the Electrochemical Society, 161(14):A1969–A1975, 2014.
doi: 10.1149/2.0031414jes.
[71] Chris J. Allen, Sanjeev Mukerjee, Edward J. Plichta, Mary A. Hendrickson, and
K. M. Abraham. Oxygen Electrode Rechargeability in an Ionic Liquid for the Li-
Air Battery. Journal of Physical Chemistry Letters, 2(19):2420–2424, 2011. doi:
10.1021/jz201070t.
[72] Andrzej Lewandowski and Agnieszka Świderska Mocek. Ionic liquids as electrolytes
for Li-ion batteries-An overview of electrochemical studies. Journal of Power
Sources, 194:601–609, 2009. doi: 10.1016/j.jpowsour.2009.06.089.
[73] Simone Monaco, Francesca Soavi, and Marina Mastragostino. Role of oxygen mass
transport in rechargeable Li/O2 batteries operating with ionic liquids. Journal of
Physical Chemistry Letters, 4:1379–1382, 2013. doi: 10.1021/jz4006256.
[74] Michele Piana, Johannes Wandt, Stefano Meini, Irmgard Buchberger, Nikolaos
Tsiouvaras, and Hubert A Gasteiger. Stability of a Pyrrolidinium-Based Ionic
Liquid in Li-O2 Cells. Journal of the Electrochemical Society, 161(14):1992–2001,
2014. doi: 10.1149/2.1131412jes.
[75] Francesca Soavi, Simone Monaco, and Marina Mastragostino. Catalyst-free porous
carbon cathode and ionic liquid for high efficiency, rechargeable Li/O2 battery.
Journal of Power Sources, 224(2013):115–119, 2013. doi: 10.1016/j.jpowsour.2012.
09.095.
[76] Chris J. Allen, Jaehee Hwang, Roger Kautz, Sanjeev Mukerjee, Edward J. Plichta,
Mary a. Hendrickson, and K. M. Abraham. Oxygen Reduction Reactions in Ionic
Liquids and the Formulation of a General ORR Mechanism for Li–Air Batteries.
The Journal of Physical Chemistry C, 116(39):20755–20764, October 2012. doi:
10.1021/jp306718v.
Bibliography 173
[77] Maciej Galiński, Andrzej Lewandowski, and Izabela Stepniak. Ionic liquids as
electrolytes. Electrochimica Acta, 51:5567–5580, 2006. doi: 10.1016/j.electacta.
2006.03.016.
[78] Y Fukaya and H Ohno. Hydrophobic and polar ionic liquids. Phys Chem Chem
Phys, 15:4066–4072, 2013. doi: 10.1039/c3cp44214d.
[79] Dr MacFarlane and P Meakin. Pyrrolidinium imides: a new family of molten salts
and conductive plastic crystal phases. The Journal of . . . , 103:4164–4170, 1999.
doi: 10.1021/jp984145s.
[80] C. M. Forsyth, E. D. MacFarlane, J. J. Golding, J Huang, J Sun, and

M Forsyth. Structural characterization of novel ionic materials incorporating the
bis(trifluoromethanesulfonyl)amide anion. Chemistry of Materials, 14(7):2103–
2108, 2002. doi: 10.1021/cm0107777.
[81] J. Golding, N. Hamid, D. R. MacFarlane, M. Forsyth, C. Forsyth, C. Collins,

and J. Huang. N-methyl-N-alkylpyrrolidinium hexafluorophosphate salts: Novel
molten salts and plastic crystal phases. Chemistry of Materials, 13(9):558–564,
2001. doi: 10.1021/cm000625w.
[82] Lorenzo Grande, Elie Paillard, Guk Tae Kim, Simone Monaco, and Stefano
Passerini. Ionic liquid electrolytes for Li-air batteries: Lithium metal cy-
cling. International Journal of Molecular Sciences, 15:8122–8137, 2014. doi:
10.3390/ijms15058122.
[83] Mega Kar, Tristan J Simons, Maria Forsyth, and Douglas R MacFarlane. Ionic
liquid electrolytes as a platform for rechargeable metal-air batteries: a perspective.
Phys. Chem. Chem. Phys., 16:–, 2014. doi: 10.1039/C4CP02533D.
[84] Jun Lu, Li Li, Jin-Bum Park, Yang-Kook Sun, Feng Wu, and Khalil Amine.
Aprotic and aqueous Li-O2 batteries. Chemical reviews, 114:5611–40, 2014. doi:
10.1021/cr400573b.
[85] Masaki Yoshio, Ralph J. Brodd, and Akiya Kozawa. Lithium-Ion Batteries.
Springer New York, New York, NY, 2009. ISBN 978-0-387-34444-7. doi:
10.1007/978-0-387-34445-4.
[86] Kevin G. Gallagher, Steven Goebel, Thomas Greszler, Mark Mathias, Wolfgang
Oelerich, Damla Eroglu, and Venkat Srinivasan. Quantifying the promise of
lithium–air batteries for electric vehicles. Energy & Environmental Science, 7:
1555, 2014. doi: 10.1039/c3ee43870h.
Bibliography 174
[87] Kong Soon Ng, Chin Sien Moo, Yi Ping Chen, and Yao Ching Hsieh. Enhanced
coulomb counting method for estimating state-of-charge and state-of-health of
lithium-ion batteries. Applied Energy, 86(9):1506–1511, 2009. doi: 10.1016/j.
apenergy.2008.11.021.
[88] Jingzhe Chen, Jens S. Hummelshø j, Kristian S. Thygesen, Jon S.G. Myrdal,
Jens K. Nø rskov, and Tejs Vegge. The role of transition metal interfaces on the
electronic transport in lithium-air batteries. Catalysis Today, 165(1):2–9, 2011.
doi: 10.1016/j.cattod.2010.12.022.
[89] Maxwell D Radin and Donald J Siegel. Charge transport in lithium peroxide:
relevance for rechargeable metal–air batteries. Energy & Environmental Science,
6(8):2370, 2013. doi: 10.1039/c3ee41632a.
[90] Venkat Viswanathan, K. S. Thygesen, Jens S. Hummelshø j, Jens K. Nø rskov,

G. Girishkumar, Bryan D. McCloskey, and Alan C. Luntz. Electrical conductivity
in Li2O2 and its role in determining capacity limitations in non-aqueous Li-O2
batteries. The Journal of chemical physics, 135(21):214704, December 2011. doi:
10.1063/1.3663385.
[91] F Huet. A review of impedance measurements for determination of the state-of-

charge or state-of-health of secondary batteries. Journal of Power Sources, 70:
59–69, 1998.
[92] Min-hsuan Hung, Chang-hua Lin, Liang-cheng Lee, and Chien-ming Wang. State-
of-charge and state-of-health estimation for lithium-ion batteries based on dynamic
impedance technique. Journal of Power Sources, 268:861–873, 2014. doi: 10.1016/
j.jpowsour.2014.06.083.
[93] Shalini Rodrigues, N Munichandraiah, and A K Shukla. Review of state-of-charge

indication of batteries by means of a.c. impedance measurements. Journal of
Power Sources, 87(1):12–20, 2000. doi: 10.1016/S0378-7753(99)00351-1.
[94] Zhaofeng Deng, Zhian Zhang, Yanqing Lai, Jin Liu, Jie Li, and Yexiang Liu. Elec-
trochemical Impedance Spectroscopy Study of a Lithium/Sulfur Battery: Model-
ing and Analysis of Capacity Fading. Journal of the Electrochemical Society, 160
(4):A553–A558, January 2013. doi: 10.1149/2.026304jes.
[95] Holger Blanke, Oliver Bohlen, Stephan Buller, Rik W. De Doncker, Birger
Fricke, Abderrezak Hammouche, Dirk Linzen, Marc Thele, and Dirk Uwe Sauer.
Impedance measurements on lead–acid batteries for state-of-charge, state-of-
health and cranking capability prognosis in electric and hybrid electric vehicles.
Bibliography 175
Journal of Power Sources, 144(2):418–425, June 2005. doi: 10.1016/j.jpowsour.

2004.10.028.
[96] Gil Cohn, Rüdiger A Eichel, and Yair Ein-Eli. New insight into the discharge
mechanism of silicon-air batteries using electrochemical impedance spectroscopy.
Physical chemistry chemical physics, 15(9):3256–63, March 2013. doi: 10.1039/
c2cp43870d.
[97] Compare Electric Vehicles Side-by-Side. URL http://www.fueleconomy.gov/

feg/evsbs.shtml.
[98] Muhammed M Ottakam Thotiyl, Stefan A. Freunberger, Zhangquan Peng, and

Peter G. Bruce. The carbon electrode in nonaqueous Li-O2 cells. Journal of
the American Chemical Society, 135(1):494–500, January 2013. doi: 10.1021/
ja310258x.
[99] Betar M. Gallant, Robert R. Mitchell, David G. Kwabi, Jigang Zhou, Lucia Zuin,
Carl V. Thompson, and Yang Shao-Horn. Chemical and Morphological Changes
of Li-O2 Battery Electrodes upon Cycling. The Journal of Physical Chemistry C,
116:20800–20805, 2012.
[100] Christian Fleischer, Wladislaw Waag, Hans Martin Heyn, and Dirk Uwe Sauer.
On-line adaptive battery impedance parameter and state estimation considering
physical principles in reduced order equivalent circuit battery models: Part 1.
Requirements, critical review of methods and modeling. Journal of Power Sources,
260:276–291, 2014. doi: 10.1016/j.jpowsour.2014.01.129.
[101] Pascal Hartmann, Conrad L Bender, Joachim Sann, Anna Katharina Dürr, Martin
Jansen, Jürgen Janek, and Philipp Adelhelm. A comprehensive study on the cell
chemistry of the sodium superoxide (NaO2) battery. Physical chemistry chemical
physics : PCCP, 15(28):11661–72, June 2013. doi: 10.1039/c3cp50930c.
[102] World Meteorological Organization. Technical Regulations. Volume I - General

Meteorological Standards and Recommended Practices. World Meteorological Or-
ganization, Geneva, 2011 edition, 2012. ISBN 9789263100498.
[103] Stefano Meini, Michele Piana, Nikolaos Tsiouvaras, Arnd Garsuch, and Hubert a.
Gasteiger. The Effect of Water on the Discharge Capacity of a Non-Catalyzed
Carbon Cathode for Li-O2 Batteries. Electrochemical and Solid-State Letters, 15
(4):A45, 2012. doi: 10.1149/2.005204esl.
[104] Kensuke Takechi, Tohru Shiga, and Takahiko Asaoka. A Li–O2/CO2 battery.
Chemical Communications, 47(12):3463–3465, February 2011. doi: 10.1039/
C0CC05176D.
Bibliography 176
[105] Hyung-Kyu Lim, Hee-Dae Lim, Kyu-Young Park, Dong-Hwa Seo, Hyeokjo Gwon,
Jihyun Hong, William a Goddard, Hyungjun Kim, and Kisuk Kang. Toward a
Lithium-”Air” Battery: The Effect of CO2 on the Chemistry of a Lithium-Oxygen
Cell. Journal of the American Chemical Society, 135(26):9733–42, July 2013. doi:
10.1021/ja4016765.
[106] Jian Zhang, Wu Xu, Xiaohong Li, and Wei Liu. Air Dehydration Membranes for
Nonaqueous Lithium–Air Batteries. Journal of The Electrochemical Society, 157
(8):A940–A946, 2010. doi: 10.1149/1.3430093.
[107] L G Johnson. Oxygen battery system, September 2009. URL https://www.

google.dk/patents/US20090239132.
[108] R Hilse, M Schiemann, H G Schweiger, and S D Tillmann. Battery system for use
in electrochemical energy generating, storing or consuming apparatus, has drying
unit and housing with opening for exchange of gas between housing inner space,
housing outer space and drying unit, January 2011. URL https://www.google.
dk/patents/DE102009032463A1?cl=en.
[109] M Heger, H Thomas, D J Morton, and I Rottler. Battery system for e.g. hybrid car,
has air interchange device that allows exchange of air between interior of battery
housing and exterior environment to prevent or minimize ingress of water from
exterior environment into interior, October 2012. URL https://www.google.dk/
patents/DE102011015925A1?cl=en.
[110] A Gleiter. Vorrichtung zur Reduzierung von Feuchtigkeit in einem Innenraum

eines Gehäuses einer Energieeinheit A device for reducing moisture in an internal
space of a housing of a power unit, July 2014. URL https://www.google.dk/
patents/DE102013200796A1?cl=en.
[111] D G Duff, J C Weintraub, and G Arcenio. Battery pack dehumidifier with active
reactivation system, December 2012. URL https://www.google.dk/patents/
EP2533325A1?cl=en.
[112] Paul Albertus, John F. Christensen, and Boris Kozinsky. Metal/oxygen battery
with an oxygen supply system, January 2014. URL https://www.google.dk/
patents/WO2014018842A1?cl=en.
[113] K Noda and T Tomita. Metal-air battery system including co2 selective ab-
sorber and operating method therefor, May 2013. URL https://www.google.
dk/patents/US20130106359.
Bibliography 177
[114] Paul Albertus, John F. Christensen, Timm Lohmann, Roel S. Sanchez-Carrera,

and Boris Kozinsky. Metal/Air Battery with Electrochemical Oxygen Compres-
sion, December 2013. URL https://www.google.dk/patents/US20130344401.
[115] A Pulskamp, A R Drews, J Yang, S Hirano, and M A Tamor. Metal Oxygen

Battery Containing Oxygen Storage Materials, June 2011. URL https://www.
google.dk/patents/US20110143228.
[116] W A Hermann and J B Straubel. Control, Collection and Use of Metal-Air

Battery Pack Effluent, February 2012. URL https://www.google.dk/patents/
US20120041628.
[117] G HÜBNER and A K Speidel. Lithium-air battery with high oxygen satura-
tion, February 2012. URL https://www.google.dk/patents/WO2012016606A1?
cl=en.
[118] Paul Albertus, John Christensen, Timm Lohmann, and Boris Kozinsky. Met-
al/oxygen battery with oxygen pressure management, August 2014. URL https:
//www.google.dk/patents/US20140234730.
[119] P J Linstrom and W G Mallard. NIST Chemistry WebBook, 2011. ISSN

0009-4978. doi: 10.5860/CHOICE.35-2709. URL http://webbook.nist.gov/
chemistry/.
[120] Arden L. Buck. New Equations for Computing Vapor Pressure and Enhancement
Factor, 1981. ISSN 0021-8952. doi: 10.1175/1520-0450(1981)020h1527:NEFCVPi
2.0.CO;2.
[121] Zeolite Store, 2015. URL http://www.zeolitestore.com/en/zeolites.
[122] Alan L Myers and Sir. Gas Separation by Zeolites. Handbook of Zeolite Science
and Technology, 2003. doi: 10.1201/9780203911167.ch22.

PHD Thesis 20150520 Rev1 Til Orbit

Uploaded by

Copyright:

Available Formats

PHD Thesis 20150520 Rev1 Til Orbit

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

PHD Thesis 20150520 Rev1 Til Orbit

Uploaded by

Copyright:

Available Formats

Downloaded from orbit.dtu.

dk on: Nov 05, 2018

Fundamental mechanisms in Li-air battery electrochemistry

Højberg, Jonathan; Vegge, Tejs; Norby, Poul; Johansen, Keld

Link back to DTU Orbit

Fundamental mechanisms in Li-air

Submitted in candidacy for the degree Doctor of Philosophy

March 24th, 2015

Senior scientist Poul Norby

Senior scientist Keld Johansen

Technical University of Denmark

Haldor Topsøe A/S

Release date: March 24th, 2015

List of Figures xvii

4 Overpotentials and degradation 27

5 Screening for new electrolytes 47

6 Commercial applications of lithium-air batteries 55

6.1.6 Further development of the model . . . . . . . . . . . . . . . . . 64

7 Summary and outlook 69

A Development of the testing equipment 73

B Differential electrochemical mass spectrometry - In situ gas analysis 77

C Supporting TEM measurements 85

D Calculations used to assess the air purification system of an open Li-O2

1.1 Ragone plot of different battery technologies . . . . . . . . . . . . . . . . 3

2.1 Schematic of a Li-battery working principle . . . . . . . . . . . . . . . . 9

3.1 Illustration of the carbon-based reference system . . . . . . . . . . . . . 17

4.1 Ex-situ and in-situ XRD measurements of Li2 O2 in the cathode . . . . . 28

5.1 Investigated ionic liquids . . . . . . . . . . . . . . . . . . . . . . . . . . . 47

6.1 Overview of impedance-based management of Li-O2 batteries . . . . . . 56

A.1 Selected designs of the Swagelok cell . . . . . . . . . . . . . . . . . . . . 73

B.1 Intensity measurements with mass spectrometer . . . . . . . . . . . . . . 78

C.1 Test of Li2 O2 beam sensitivity . . . . . . . . . . . . . . . . . . . . . . . 86

D.1 Calculated Langmuir isotherm for H2 O and CO2 adsorption . . . . . . . 93

SoC State of charge

1.1 Abandoning fossil fuels

1.1.1 Non-fossil transportation

Today, much attention is being devoted to developing green transportation, as this

1.2.1 Gold rush of the giant leap

1.3 About this work

1.3.1 The aims of the project

1.3.2 Outline of the thesis

4 Overpotentials and degradation

5 Screening for new electrolytes

6 Commercializing Li-air batteries

7 Summary and outlook

List of included papers

2.1 The lithium-oxygen battery

in Li-batteries), and F is the Faraday constant. The battery is discharged by connecting

2.1.1 New definitions for lithium-oxygen batteries

Normalization of battery capacity. Another important characteristic is the theo-

Capacity at 2.6V [mAh]

2.2.1 Correlation between overpotential and impedance

The ultimate goal of fundamental investigations such as EIS is to predict or improve

measurement to the change in overpotential. The impedance is defined as the derivative

η = |v − OCV| = c1 · ic2 (2.3)

2.2.2 Modeling Li-O2 battery impedance

RΩ is the DC resistance at the investigated frequency. As discussed by Zoltowski et al.,

3.1 Cell configuration

3.1.1 The Swagelok cell