Aging Mechanisms of Li-Ion Batteries: Seen From An Experimental and Simulation Point of View

Aging mechanisms of Li-ion batteries : seen from an
experimental and simulation point of view

Citation for published version (APA):
Li, D. (2017). Aging mechanisms of Li-ion batteries : seen from an experimental and simulation point of view
Eindhoven: Technische Universiteit Eindhoven
Document status and date:

Published: 16/03/2017
Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)
Please check the document version of this publication:
• A submitted manuscript is the version of the article upon submission and before peer-review. There can be
important differences between the submitted version and the official published version of record. People
interested in the research are advised to contact the author for the final version of the publication, or visit the
DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page
numbers.
Link to publication
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 the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please
follow below link for the End User Agreement:
www.tue.nl/taverne
Take down policy

If you believe that this document breaches copyright please contact us at:
openaccess@tue.nl
providing details and we will investigate your claim.
Download date: 01. Jun. 2019

Aging Mechanisms of Li-ion Batteries

Seen From an Experimental and Simulation Point of View
PROEFSCHRIFT
ter verkrijging van de graad van doctor aan de Technische Universiteit

Eindhoven, op gezag van de rector magnificus prof.dr.ir. F.P.T. Baaijens,
voor een commissie aangewezen door het College voor Promoties, in het
openbaar te verdedigen op donderdag 16 maart 2017 om 16:00 uur
door
Dongjiang Li
geboren te Shandong, China

Dit proefschrift is goedgekeurd door de promotoren en de samenstelling van de
promotiecommissie is als volgt:
voorzitter: prof.dr.ir. E.J.M. Hensen

1e promotor: prof.dr. P.H.L. Notten
2e promotor: prof.dr. Y. Yang (Xiamen University)
copromotor(en): dr. D.L. Danilov
leden: prof.dr. R.-A Eichel (RWHT Aachen University)
prof.dr. D.U. Sauer (RWHT Aachen University)
prof.dr.ir. H.J. Bergveld
prof.dr. F. Roozeboom
Het onderzoek of ontwerp dat in dit proefschrift wordt beschreven is

uitgevoerd in overeenstemming met de TU/e Gedragscode
Wetenschapsbeoefening.

Aging Mechanisms of Li-ion Batteries

Seen from an Experimental and Simulation Point of View

Dongjiang Li

gÉ yâàâÜx ‹

《愚公移山》
河曲智叟笑而止之曰：“甚矣，汝之不
惠。以残年余力，曾不能毁山之一毛，
其如土石何？”北山愚公长息曰：“汝
心之固，固不可彻，曾不若孀妻弱子。
虽我之死，有子存焉；子又生孙，孙又
生子；子又有子，子又有孙；子子孙孙
无穷匮也，而山不加增，何苦而不平？”
是故
愚公移山，
不惟志坚。
前仆后继，
薪火相传。
---寄语新能源之研究、开发与推广

The work represented in this thesis has been carried out in the group of Energy Material and
Device (EMD), Eindhoven University of Technology, the Netherlands. Finance support was
from China Scholarship Council (CSC).
Cover design: Dongjiang Li

The front cover illustrates aging models of Li-ion batteries under cycling condition and
corresponding simulations of the battery capacity losses. The back cover shows an example of
the renewable energy source, windmill.
Copy right © 2017 by Dongjiang Li

All rights are reserved. No part of this thesis is allowed to be reproduced, stored in a retrieval
system or transmitted in any form or by any means, electronic, mechanical, photocopying,
recording, scanning or otherwise without the prior written permission of the Author.
A catalogue record is available from the Eindhoven University of Technology Library

ISBN: 978-90-386-4168-3
Printed by Gildeprint − the Netherlands

Table of contents

Table of contents
List of Abbreviations ............................................................................................I
List of Symbols .................................................................................................. III
Chapter 1 Introduction ....................................................................................... 1
1.1 Energy storage ....................................................................................................... 1
1.1.1 Primary batteries ........................................................................................................ 2
1.1.2 Secondary batteries .................................................................................................... 3
1.2 Li-ion batteries ....................................................................................................... 6
1.2.1 Historical development .............................................................................................. 6
1.2.2 Structure and working principles ............................................................................... 7
1.3 Challenges of Li-ion batteries.............................................................................. 13
1.3.1 Energy and power densities ..................................................................................... 13
1.3.2 Battery performance ................................................................................................. 14
1.3.3 Battery safety ........................................................................................................... 16
1.3.4 Cost ......................................................................................................................... 16
1.4 Scope of this thesis .............................................................................................. 17
1.5 References ........................................................................................................... 19
Chapter 2 Overview of Degradation Mechanisms of Li-ion Batteries ......... 23
2.1 Capacity degradation ........................................................................................... 24
2.2 SEI formation ...................................................................................................... 27
2.2.1 What is SEI?............................................................................................................. 27
2.2.2 SEI formation mechanisms ...................................................................................... 29
2.2.3 Experimental characterization of SEI ...................................................................... 38
2.2.4 SEI formation models .............................................................................................. 41
2.3 Cathode electrode decay ...................................................................................... 44
2.3.1 LiFePO4 electrode .................................................................................................... 44
2.3.2 LiCo1/3Ni1/3Mn1/3O2 (NMC) ..................................................................................... 49
2.4 Graphite (C6) electrode decay.............................................................................. 52
2.4.1 Physical properties of graphite ................................................................................. 52
2.4.2 Li intercalation into graphite electrode .................................................................... 53
2.4.3 Graphite electrode degradation ................................................................................ 55
2.5 References ........................................................................................................... 56
I

Table of contents

Chapter 3 Experimental ................................................................................... 69
3.1 Batteries selected ................................................................................................. 70
3.2 Storage experiments ............................................................................................ 70
3.2.1 Storage experiments of prismatic 50 Ah batteries ................................................... 71
3.2.2 Storage experiments of A123 batteries .................................................................... 71
3.3 Cycling experiments ............................................................................................ 72
3.3.1 Full cycling measurements....................................................................................... 72
3.3.2 Partial cycling measurements................................................................................... 74
3.4 X-ray photoelectron spectroscopy (XPS) ............................................................ 75
3.5 Raman spectroscopy ............................................................................................ 75
3.6 Scanning electron microscopy (SEM) ................................................................. 77
3.7 Inductively Coupled Plasma-Optical Emission Spectrometry (ICP-OES) ......... 77
3.8 References ........................................................................................................... 78
Chapter 4 Methodologies and Terminologies ................................................. 81
4.1 EMF determination .............................................................................................. 82
4.2 Parameters and definitions................................................................................... 83
4.3 Overpotential and resistance determinations ....................................................... 86
4.4 Non-destructive quantification of QC6 ................................................................. 88
4.5 Analyses of dV/dQEMF curves .............................................................................. 89
4.6 References ........................................................................................................... 93
Chapter 5 Degradation Mechanisms of LFP Batteries: Experimental
Analyses of Calendar Aging ............................................................................. 95
5.1 Introduction ......................................................................................................... 96
5.2 Results.................................................................................................................. 97
5.2.1 Aging of prismatic batteries ..................................................................................... 97
5.2.2 Aging of A123 Batteries .......................................................................................... 98
5.2.3 Graphite electrode characterization ....................................................................... 107
5.2.4 Quantification of the inaccessibility of the graphite electrode .............................. 111
5.3 Calendar ageing model ...................................................................................... 113
5.4 Conclusions ....................................................................................................... 116
5.5 References ......................................................................................................... 116
Chapter 6 Degradation Mechanisms of LFP Batteries: Experimental
Analyses of Cycling-induced Aging ............................................................... 121
II

Table of contents

6.1 Introduction ....................................................................................................... 122

6.2 Results................................................................................................................ 123
6.2.1 Cycling prismatic batteries (50Ah) ........................................................................ 123
6.2.2 Cycling cylindrical batteries (2.3Ah) in the full SoC range .................................. 124
6.2.3 Cycling cylindrical batteries in various SoC ranges .............................................. 138
6.3 Discussion .......................................................................................................... 146
6.3.1 SEI formation model .............................................................................................. 146
6.3.2 Evolution of the individual graphite electrode plateaus......................................... 149
6.3.3 Influence of graphite degradation on battery capacity loss.................................... 150
6.4 Conclusions ....................................................................................................... 152
6.5 References ......................................................................................................... 153
Chapter 7 Degradation Mechanisms of LFP Batteries: Modeling Calendar
and Cycling-induced Aging ............................................................................ 157
7.1 Introduction ....................................................................................................... 158
7.2 Model development ........................................................................................... 159
7.2.1 Aging mechanisms of LFP batteries ...................................................................... 159
7.2.2 SEI formation model .............................................................................................. 163
7.2.3 Cathode dissolution model ..................................................................................... 170
7.2.4 Summary of the aging model ................................................................................. 174
7.3 Results and discussion ....................................................................................... 175
7.3.1 Irreversible capacity losses under storage .............................................................. 179
7.3.2 Irreversible capacity losses during cycling ............................................................ 185
7.3.3 Influence of graphite parameters on the irreversible capacity losses..................... 196
7.4 Conclusions ....................................................................................................... 197
7.5 References ......................................................................................................... 198
Chapter 8 Degradation Mechanisms of NMC Batteries: Experimental
Analyses of Cycling-induced Aging ............................................................... 203
8.1 Introduction ....................................................................................................... 204
8.2 Results and discussion ....................................................................................... 205
8.3 Conclusions ....................................................................................................... 214
8.4 References ......................................................................................................... 215
Summary .......................................................................................................... 219
Appendix I ........................................................................................................ 227
Appendix II ...................................................................................................... 229
III

Table of contents

Appendix III ..................................................................................................... 235
List of publications .......................................................................................... 237
Acknowledgement............................................................................................ 239
Curriculum Vitae............................................................................................. 243

IV

List of Abbreviations

AAS Atomic absorption spectroscopy
AFM Atomic force microscope
AMS Atomic mass spectroscopy
APS Atomic fluorescence spectroscopy
CC-mode Constant current mode
CCCV Constant current constant voltage
CE Charging efficiency
CEI Cathode-electrolyte-interphase
CV-mode Constant voltage mode
CV Cyclic voltammetry
DEC Diethyl carbonate
DFT Density functional theory
DIS Diffusion-induced stress
DMC Dimethyl carbonate
EC Ethylene carbonate
EDAX Energy-dispersive X-ray spectroscopy
EDS Energy dispersive spectroscopy
EIS Electrochemical impedance spectroscopy
EMF Electromotive force
EQCM Electrochemical quartz crystal microbalance
ESCA Electron spectroscopy for chemical analysis
EV Electric vehicle
FIB Focused ion beam
FT-IR Fourier transform infrared spectroscopy
GC/MS Gas chromatography mass spectroscopy
GIC Li(solvent)xCy
GITT Galvanostatic intermittent titration technique
HEV Hybrid electric vehicle
I


HOMO Highest Occupied Molecular Orbital
HOPG Highly Ordered Pyrolytic Graphite
ICP Inductively coupled plasma
LFP Lithium iron phosphate
LCO Lithium cobalt oxide
LMO Lithium manganese spinel oxide
LUMO Lowest Unoccupied Molecular Orbital
MCMB Mesocarbon microbeads
NLS Nonlinear Least Squares
NMC Lithium nickel manganese cobalt oxide
NMC(111) LiNi1/3Mn1/3Co1/3O2
NMC(532) LiNi0.5Mn0.3Co0.2O2
NMR Nuclear magnetic resonance
ODE Ordinary differential equation
OES Optical emission spectroscopy
PC Propylene carbonate
PHEV Plug-in hybrid electric vehicle
PITT Potentiostatic intermittent titration technique
SEI Solid-electrolyte-interphase
SEM Scanning electron microscopy
SoC State-of-Charge
STEM Scanning transmission electron microscope
VC Vinylene carbonate
XAS X-ray absorption
XES X-ray emission
XPS X-ray photoelectron spectroscopy
II

List of Symbols

List of Symbols
Symbol Meaning Value Unit
Surface area of the graphite electrode m2
SEI covered surface areas of C6 electrode m2
Fresh surface areas of C6 electrode m2
Total surface areas of Fe particles on the graphite electrode m2
Total surface areas of LiFePO4 electrode m2
Charge efficiency at the anode -

Charge efficiency at the cathode -
Fe2+ concentration in the electrolyte at time mol·L-1
H+ concentration in the electrolyte at 0 mol·L-1
H+ concentration in the electrolyte at time mol·L-1

Surface concentration of LiFePO4 mol·m-2
/ Number of electrons reaching the graphite surface in unit time -

/ Differential voltage based on EMF curves V·Ah-1
Gravimetric energy density Wh·kg-1
Volumetric energy density Wh·L-1

Fermi level of LixC6 electrode eV
Fermi level of LiC6 electrode eV
Fermi level of the anode eV
Fermi level of the cathode eV
Energy throughout Wh
EMF curves at the initial state V
EMF curves at time V
Activation energy for cathode dissolution J·mol-1
Δ Energy barrier for electron tunneling eV
Δ Average energy barrier for electron tunneling eV
Charge of single electron 1.6·10-19 C
III

List of Symbols

Faraday constant 96500 C·mol-1

Δ Gibb’s free energy J
Current A
Charging current A
Current related to Li intercalation during charging A
Discharging current A
Current related to Li immobilization during charging A
Tunneling current A
SEI formation current during storage A
/ Integrated intensity ratio between D and G hand -
Lithium diffusion flux in the electrode materials mol·s-1
Total reduction flux of Fe2+ at the anode mol·s-1
Total dissolution flux of Fe2+ from the cathode mol·s-1
Current density of SEI formation A·m-2
Dissociation reaction constant of LiH2PO4 depends on parameter

Combined exchange reaction constant depends on parameter
Pre-exponential factor depends on parameter
Exchange reaction constant depends on parameter

Deposition reaction constant of Fe2+ on graphite depends on parameter
Thickness of the inner SEI layer on C6 m
Initial thickness of the inner SEI layer on C6 m

Δ Thickness increase of the inner SEI layer on C6 m
Δ Thickness increase of the inner SEI layer on Fe m
Initial thickness of the inner SEI on Fe m
Total thickness of the inner SEI layer on C6 m

Mole mass of graphite 72.06 g·mol-1
Mole mass of Fe 55.85 g·mol-1

Mole mass of Lithium 6.94 g·mol-1
Electron mass 9.11·10-28 g
Mass of anode materials g
IV

List of Symbols

Total mass of the battery g
Δ Li mass in the increased inner SEI layer g

Mass of all passive materials g
Δ Increased mass of the inner SEI layer g

Total numbers of Fe particles -
Avogadro number 6.02·1023 mol-1
Total amount of Fe reduced on graphite mol
Cycle numbers -
Δ Moles of Li consumed in the increased inner SEI layer mol

Tunneling probability -
Pre-exponential coefficient -
Battery capacity Ah
/ Specific capacity of the complete battery mAh·g-1
Charge capacity Ah
Charge capacity in cycle Ah

Total capacity of the graphite electrode Ah
, Width of the first plateau of the graphite electrode Ah
, Width of the second plateau of the graphite electrode Ah
, Width of the third plateau of the graphite electrode Ah
Discharge capacity Ah
Discharge capacity (pristine state) Ah
Discharge capacity at time Ah
Discharge capacity at cycle Ah
Discharge capacity in cycle Ah
, Capacity loss caused by Fe dissolution and deposition Ah
Total capacity of the LiFePO4 electrode Ah
Maximum LiFePO4 electrode capacity (pristine state) Ah
, Total amount of Li on the first plateau of the graphite electrode Ah
, Total amount of Li on the second plateau of the graphite electrode Ah
V

List of Symbols

, Total amount of Li on the third plateau of the graphite electrode Ah
Maximum battery discharge capacity Ah

Maximum battery discharge capacity at 0 Ah
Maximum charge capacity Ah
Maximum discharge capacity Ah
Maximum battery discharge capacity at time Ah
Maximum discharge capacity at time during cycling Ah
Maximum discharge capacity at time during storage Ah
Extracted amount of charge Ah
Irreversible capacity losses caused by SEI formation Ah
Capacity loss due to the SEI formation on Ah
Accumulated capacity loss due to the SEI formation on Ah
Capacity loss caused by SEI formation during cycling Ah
Capacity loss due to the SEI formation on in each cycle Ah
Capacity consumed by inner SEI growth Ah

Capacity loss caused by SEI formation during storage Ah
, Irreversible capacity losses caused by SEI formation on Fe Ah
Specific capacity of the anode mAh·g-1

Specific capacity of the cathode mAh·g-1
Specific capacity of LFP battery mAh·g-1
Δ Apparent capacity loss Ah
Δ Cathode capacity loss Ah

Δ The graphite electrode capacity loss Ah
Δ , Decline of the first plateau of the graphite electrode Ah
Δ , Decline of the second plateau of the graphite electrode Ah
Δ , Decline of the third plateau of the graphite electrode Ah
Δ , Li loss on the first plateau of the graphite electrode Ah
Δ , Li loss on the second plateau of the graphite electrode Ah
Δ , Li loss on the third plateau of the graphite electrode Ah
VI

List of Symbols

Δ Irreversible capacity loss Ah

Δ Irreversible capacity loss due to calendar aging Ah
Δ Total irreversible capacity loss induced by cycling aging Ah
Δ Total irreversible capacity loss during storage Ah

Δ LiFePO4 electrode capacity loss Ah
Resistance Ω
Charge transfer resistance Ω
Mass transfer resistance Ω
Ohmic resistance Ω
Ohmic resistance during charging Ω
Ohmic resistance during discharging Ω

Average radius of Fe particles m
Temperature K
Time s
Total rest time in each cycle s
Reference energy level eV
Energy level of free electrons eV
LUMO of the solvent eV
Fermi velocity of electrons in bulk of metallic Fe m·s-1
Δ Volume of the increased inner SEI layer m3
Volume of Fe particles m3
Initial discharge voltage V
End of discharge voltage V
Battery voltage V
Voltage discharge curve at the initial state V
Voltage discharge curve at time V
Battery voltage at the end of rest after charging but before discharging V
Battery voltage after commencing the discharging process V

Battery voltage at the end of rest after discharging but before charging V
EMF voltage of the battery V
VII

List of Symbols

Open circuit potential of battery V

Battery voltage at the end of resting period V
Initial battery voltage during charging V
Initial battery voltage during discharging V

Δ Electrode volume changes m3
Fermi velocity of electrons in bulk of graphite m·s-1
Average weight percentage of Li in inner SEI layer -

State-of-Charge of the graphite electrode -
Ratio of the first plateau to the total graphite electrode -
Ratio of the second plateau to the total graphite electrode -
Ratio of the third plateau to the total graphite electrode -
Critical concentration of Li-deficient phase -
Critical concentration of Li-rich phase -
Number of the electrons involved in the basis charge transfer reaction -

Li-deficient phase of LFP material -
Li-rich phase of LFP material -
Capacity fraction between inner and total SEI layer on C6 -
Capacity fraction between inner and total SEI layer on Fe -
Overpotential V
Electrode potential of LixC6 V
Electrode potential of LiC6 V
Capacity loss due to the SEI formation on in each cycle Ah

Density of Fe 7.86·106 g·m-3
Density of graphite 2.266·106 g·m-3
Density of the inner SEI layer g·m-3
VIII

Chapter 1
Introduction
1.1 Energy storage
In the past centuries the rapid development of our world was ensured by a vast amount of
fossil fuels. Nowadays it is commonly accepted that such dependence on minerals creates an
obstacle for sustainable development of our society. The limited resources which are irregularly
distributed between the countries and continents facilitate conflicts and social unrest. In addition
to that, excessive emissions of greenhouse gases pollutes the living environment and accelerates
global warming processes with potentially disastrous consequences.
Renewable energies such as solar, wind, tidal and geothermal are considered as an
alternative for fossil fuels. However, solar and wind are irregular in time and space, while tidal
and geothermal energies are location specific. All these sources require energy storage devices
to act as buffer to stabilize the energy output. Many storage methods have been proposed, such
as mechanical energy storage, electromagnetic energy storage, energy storage in organic fuels
and hydrogen and electrochemical energy storage [1, 2]. Among the various methods that can
be used for energy storage, the electrochemical methods, in particular batteries, draw
considerable attention in the last decades. Combining a high efficiency with portability and the
possibility for scaling, batteries develop quickly.
Table 1.1. Various battery systems including primary and secondary batteries
Aqueous electrolyte Non-aqueous electrolyte
Primary battery Manganese dry cell Metallic Li battery
Alkaline dry cell
Magnesium cell
Secondary battery Lead-acid battery Li-ion battery
Ni-Cd battery
Ni-MH battery
1

Chapter 1 Introduction

Batteries are nowadays the most popular devices for both stationary and portable energy
storage. Batteries are closed systems in which electrical energy is generated by conversion of
chemical energy via redox reactions at the negative and positive electrodes [3]. Currently
batteries are classified into primary batteries, which can be only discharged, and secondary
batteries, which permit multiple charging and discharging cycles. Table 1.1 summarizes the
main battery systems according to this classification.
1.1.1 Primary batteries
Primary batteries are produced in the charged state and provide only one continuous or
intermittent discharge. During the discharging process, the chemical energy stored inside the
battery is irreversibly converted into electrical energy. The working principles of a manganese
battery can be represented by a half-cell reaction at the negative electrode
Zn → Zn2+ + 2e-, [1.1]
that at the positive electrode can be represented by
2MnO2 + 2NH4+ +2e- → Mn2O3 + 2NH3 + H2O, [1.2]
and, consequently, the total reaction of the cell can be represented by
Zn + 2MnO2 + 2NH4+ → Zn2+ + Mn2O3 + 2NH3 + H2O. [1.3]
The battery voltage of a fresh manganese dry cell is 1.5V.

Metallic lithium batteries are the most energy-dense primary cells. Research of this lithium
battery has been started in 1912 [4]. The breakthrough came in 1958 when Harris revealed the
stability of metallic Li in a number of nonaqueous electrolytes. The passivation layer formed
on the Li surface prevents the direct chemical reaction between Li metal and the electrolyte but
still allows ionic transport [5]. Numerous inorganic compounds such as SO2, SOCl2, MnO2, CFx
have been used as cathode materials in Li batteries. Since the late 1960s a series of nonaqueous
Li batteries with various cathode materials have been developed and commercialized.
The advantage of this battery type was first demonstrated in the early 1970s. Owing to their
high capacity and variable discharge rate, Li batteries have been rapidly applied as power supply
for watches, calculators or implantable medical devices [4, 6]. Rechargeable Li batteries, using
Li-insertion compounds as positive electrode, were further developed since the mid 1970s.
Since these times extensive efforts have been spent to commercialize rechargeable Li batteries.
However, the dendrite formation on the Li electrode surface represents a challenge for the
development of rechargeable Li batteries. The Li dendrites formation frequently leads to short
2


circuits with catastrophic consequences. Li batteries are therefore preferably considered as

primary batteries only.
1.1.2 Secondary batteries
A secondary battery is a cell or group of cells where the chemical energy and electrical
energy can reversibly be stored and extracted. After being discharged the battery can be
recharged to its initial state by an electric current flowing in the direction opposite to the
discharge current [3]. Secondary batteries are usually assembled in the discharged state, and
have therefore to be charged first before these can be used as energy source.
a. Lead-acid batteries
The first rechargeable battery system, the lead-acid battery, was invented by Gaston Plante
in 1859. It survived tough competition with many other batteries and is still playing a dominant
role after more than 150 years of use in many applications. Lead-acid batteries have remained
one of the leading commercial systems mainly because of their low cost and the capability of
delivering large currents. It is also known to be a highly recyclable product. The anode of the
battery is composed of metallic lead and the cathode is PbO2. The electrolyte is H2SO4 solutions.
The working mechanisms of a lead-acid battery can be represented by a half-cell reaction at the
negative electrode

PbSO 2 SO Pb, [1.4]

the half-cell reaction at the positive electrode

PbSO 2H O 4H PbO SO 2 , [1.5]

and the total reaction of the cell can then be represented by

2PbSO 2H O PbO Pb 2H SO4 . [1.6]

b. NiCd batteries
The first nickel-based battery system was invented by Thomas Edison in 1898 and was
based on an iron anode, which was then further modified by Jungner by replacing iron with
cadmium to NiCd batteries. The working principles of NiCd batteries are represented by a half-
cell reaction at the negative electrode
3


Cd OH 2 Cd 2OH , [1.7]

at the positive electrode

2Ni OH 2OH 2NiOOH 2H2 O 2 , [1.8]

and the total reaction of the cell then can be represented by

Cd OH 2Ni OH 2NiOOH 2H2 O Cd. [1.9]

It is well known that NiCd batteries suffered from the “memory effect”, a phenomena of
sharp voltage drop during the initial discharging period. The reduction of the capacity due to
memory effect can be recovered by a deep discharge followed by a normal recharging. The
memory effect of NIcD should be checked.
c. NiMH batteries
The NiMH battery was made by replacing the cadmium anode by a hydride-forming
electrode (MH). During charging electrochemical hydriding of the MH electrode takes place,
while the reverse process occurs during discharging [7, 8]. In contrast to the NiCd battery, the
water content and the OH concentration inside the NiMH battery is constant. The main storage
reactions of the NiMH battery are represented by a half-cell reaction at the negative electrode

M H2 O MH OH , [1.10]

and the positive electrode

Ni OH OH NiOOH H2 O , [1.11]

with overall reaction

M Ni OH MH NiOOH. [1.12]

d. Li-ion batteries
The Li-ion technology has been developed on the basis of the existing primary Li batteries.
To circumvent the safety issues caused by Li-dendrite formation in metallic Li batteries, several
alternative approaches were pursued in which either the electrolyte or the negative electrode
was modified. Capitalized on earlier findings, carbonaceous materials were proposed as anode
4


material which finally led to the creation of the C/LiCoO2 rocking-chair cell commercialized
by Sony Corporation in June 1991 [6]. These batteries with a configuration of carbonaceous
material as negative electrode and lithiated metal oxides as positive electrode were called Li-
ion batteries.
Fig. 1.1. Comparison of the different battery technologies in terms of volumetric and gravimetric energy
density [6].
Li-ion batteries represent a breakthrough in the field of power sources for a variety of
applications. Fig. 1.1 shows the comparison of the various secondary batteries in terms of
volumetric and gravimetric energy density. Although the energy density of metallic Li batteries
is much higher than that of other battery systems, it suffers from serious safety issues. Obviously,
Li-ion batteries are the best option for mobile applications from both the energy density and
safety point of view. The advantage of Li-ion batteries can be summarized as
 High specific gravimetric and volumetric energy density
 Low self-discharge
 Long cycle life
 No maintenance
 No memory effect
 Fairly wide operating temperature range
 Fairly high rate capability
Details of this battery system will be discussed in the following section.
5


1.2 Li-ion batteries
1.2.1 Historical development
Li-ion batteries have been developed on the basis of metallic Li batteries. By 1972, the
concept of electrochemical intercalation and its potential application was clearly demonstrated
[6]. The concept of Li-ion batteries has been proposed in the late 1970s by Armand [9] who
suggested to use two different intercalation compounds as positive and negative electrodes in
the so-called rocking-chair battery. The Li+ ions can be transferred from one electrode to the
other during charging and discharging. Lazzari and Scrosati already assembled a cell with two
intercalation compounds in the early 1980s. However, this concept has drawn significant
attention only after the successful introduction of the Japanese industry. Sony and Sanyo in
1985 and 1988, respectively, have utilized new cathode materials following the fundamental
research of Goodenough in Oxford [10] and new anode materials following Armand and
Touzain in Grenoble [11], who evidenced the fast motion of lithium ions in layered host
structures. Table 1.2 shows the initial patents applied for Li-ion batteries. The first patent related
to the construction of a Li-ion batteries was issued on 5 October 1985 by Asahi Chemical Ind.
(Japan).
Table 1.2. Prior patents related to Li-ion batteries [4].
Inventor/company Patent title Patent number Application date
Goodenough JB, Mizushima K Fast ion conductors (AxMyO2) US 05-04-1979
(Kingdom Atomic Energy) 4,357,215 A
Goodenough JB, Mizushima K Electrochemical cell with new fast ion US 31-03-1980
(Kingdom Atomic Energy) conductors 4,302,518
Ikeda H, Narukawa K, Nakashima Graphite/Li in nonaqueous solvents Japan 18-06-1981
H (Sanyo) 1,769,661
Basu S (Bell labs) Graphite/Li in nonaqueous solvents US 13-09-1982
4,423,125
Yashino A, Jitsuchika K, Nakajima Li-ion battery based on carbonaceous Japan 05-10-1985
T (Asahi Chemical Ind.) material 1,989,293
Nishi N, Azuma H, Omaru A (Sony Nonaqueous electrolyte cell US 29-08-1989
Corp.) 4,959,281
Sony Energytec Inc. has commercialized the first type of Li-ion battery using LiCoO2 as
cathode and non-graphitic carbon as anode. The electrolyte they applied was propylene
carbonate (PC) / diethyl carbonate electrolyte (DEC) solution combined with LiPF6 salt. The
6


Li-ion battery was designed to be a 18650-type with an energy density of 253 Wh·L- 1. This
battery type was aimed to power mobile phones [4]. Lithium-ion chemistries represent a
significant step forward in battery technology. During the following decades investigation,
active R&D was directed towards alternative electrode materials resulting in small, light-weight
and high-energy-density batteries that have successfully enabling a wide range of new
applications [12].
1.2.2 Structure and working principles
Fig. 1.2 represents the layout of a Li-ion cell (C6/LiFePO4). Like all electrochemical cells,
a commercial Li-ion cell is composed of two electrodes, an anode and a cathode. A separator
membrane is placed between the two electrodes to prevent electrical contact. The cell is filled
with electrolyte composed of non-aqueous solvents and Li salts, e.g., EC (ethylene carbonate)
/ DEC (diethyl carbonate) / DMC (dimethyl carbonate) (1:1:1) VC (Vinylene carbonate) 2%,
LiPF6 (1M).
Fig. 1.2. Layout of Li-ion battery (C6/LiFePO4).
As shown in Fig. 1.2, during charging electrons are extracted from the LiFePO4 cathode and
flow into the graphite electrode through the outer circuit (red arrow). Simultaneously, Li+ ions
are delithiated from the LiFePO4 electrode and transported through the electrolyte to the
graphite electrode (red arrow) safeguarding charge neutrality in the electrodes. The electric
energy converts into chemical energy during charging. The transportation direction of electrons
and Li+ ions are reversed during discharging as indicated by black arrows in Fig. 1.2. The stored
chemical energy converts into electrical energy during discharging. The main electrochemical
storage reactions of this battery type can be represented by
7


C6 + xLi+ + xe- Li C , [1.13]

LiFePO4 Li FePO + xLi xe- , [1.14]

resulting in the following overall reaction

C6 LiFePO4 Li FePO Li C , 1.15

where x is the State-of-Charge (SoC), 0 ≤ x ≤ 1.

A battery consists of a group of interconnected electrochemical cells. Shapes and
components of various Li-ion battery configurations are shown in Fig. 1.3. Cylindrical batteries
in most products follow standard models in terms of size, for example, the 18650 and 26650
cell. Typical 18650 cells in commercial Li-ion battery products have a volumetric energy
density of 600–650 Wh·L−1, which are approximately 20% higher than those of their pouch
counterparts [13]. The main components of the Li-ion battery are the anode, cathode, separator,
electrolyte and current collectors. Generally, graphite is used as the anode electrode in most
commercial batteries while cathode materials can vary due to different requirements in the
various applications.
Fig. 1.3. Representation of the shape and components of various Li-ion battery configurations [6]:
cylindrical (a), prismatic (b), coin (c) and pouch cell (d).
The energy density and stability of Li-ion batteries are strongly determined by the electrode
materials and the electrolyte. The general requirements with respect to electrode materials and
8


electrolyte are schematically represented in Fig. 1.4. The relative electron energies in the
electrodes and the electrolyte of a thermodynamically stable cell with nonaqueous electrolyte
are shown in Fig. 1.4. The Fermi levels of the two electrodes are for the anode and for
the cathode. The energy difference of the Lowest Unoccupied Molecular Orbital (LUMO) and
the Highest Occupied Molecular Orbital (HOMO) of the electrolyte is the “operating window”
of the electrolyte [14, 15]. As illustrated in Fig. 1.4, in case of that LUMO* (red) is below ,
the electrolyte will be reduced at the anode; in case of that HOMO* (red) is above , the
electrolyte will be oxidized at the cathode. Therefore, LUMO (blue) should be higher than
and HOMO (blue) should be lower than , and consequently, the difference between and
should be within the stability window of the electrolyte. The open-circuit potential of the
cell is determined by
  E cf  E af 
VOCP  , [1.16]
e
where is the electronic charge.
Fig. 1.4. Schematic representation of the open-circuit energy diagram of a nonaqueous cell. and
are the Fermi level of the anode and cathode, respectively. LUMO and HOMO are the Lowest
Unoccupied Molecular Orbital and the Highest Occupied Molecular Orbital, respectively, of the
electrolyte. The red and blue lines show two cases of the LUMO and HOMO inside the batteries.
The performance of Li-ion batteries is strongly dependent on the thermal, mechanical and
physical stability of its materials. The combination of the anode, cathode, electrolyte and
separator materials will determine the energy and power densities, the storage and cycle life,
9


safety characteristics, etc. The electrodes have a composite structure comprising of active
materials, binders, and additives which are coated on the corresponding current collectors. The
electrolyte provides the path for the Li+ ions to shuttle between the electrodes during the
(dis)charging process. Several choices of materials for anode, cathode and electrolyte will be
discussed below.
a. Anode materials
An ideal active anode material for lithium-ion batteries should satisfy the following
requirements:
(i) High specific capacity. It should accommodate as much Li as possible to optimize
the gravimetric and volumetric capacity.
(ii) Robust structure and excellent cycling performance. The material should have
minimal volume expansion and stress accumulation during the (dis)charge process.
(iii) Low electrode potential. Its redox potential with respect to Li+/Li must be as low as
possible so that it can give a high battery voltage.
(iv) Excellent electronic and ionic conductivities.
(v) Excellent (electro)chemical stability. It should resist against corrosion in the
electrolyte.
(vi) Cheap and environmentally friendly.
Considering the above requirements, carbonaceous materials are preferred for anodes in current
commercial Li-ion batteries. There are many types of carbonaceous materials used as negative
electrode active material in Li-ion cells, including graphite, mesocarbon microbeads (MCMB),
carbon nanotubes, and carbon films [16]. The cycling performance of these materials strongly
depends on the solid-electrolyte-interphase (SEI) formation at the electrode surface. On the one
hand, the SEI layers prevent the graphite from exfoliation, on the other hand, the continuous
growth of the SEI layers leads to irreversible capacity loss. The theoretical capacity associated
to cycling between C6 and LiC6 is 372 mAh·g-1, which is relatively low compared with other
anode materials.
Among many higher-specific-capacity alternatives for graphite under investigation is Si,
one of the most promising anode materials because of its superior theoretical capacity (> 4000
mAh·g−1) and attractive operating voltage (~0.3 V versus Li+/Li) [13]. The huge volume change
of Si during (de)lithiation processes has been considered to be a big challenge for its
applications as anode material in Li-ion batteries. Some smart modifications of the electrode
structure [17-21] and binder designs [22-27] reported recently are believed to be solutions for
10


material pulverization and peeling-off. Another critical problem of Si electrodes is the

formation of unstable passivation layers. It is very difficult to develop a stable SEI layer to
protect the electrode surface when the electrode volume fluctuates drastically. A careful
selection of proper electrolyte composition is believed to be a solution for this interfacial issue
[28-31].
Other classes of materials reversibly forming alloys with lithium are Al (LiAl) and Sn
(Li22Sn5), etc. The theoretical capacity of these alloys is much higher than that of graphite
materials. Similar to the Si case, the drawback of these materials is also related to the large
volume change during (dis)charging process which finally leads to cracking and mechanical
disintegration of the anode structure.
The last group of materials that can be used as anode are certain lithium oxides such as TiO2,
Li4Ti5O12, etc., which have relatively high electrode potential vs Li+/Li. These materials have
smaller risk of lithium plating and electrolyte reduction, and therefore provide excellent cycling
performance. The main disadvantages of these materials are attributed to their relatively low
specific storage capacities as well as the higher electrode potentials vs Li+/Li, which finally lead
to a lower energy density compared with other materials (e.g. C6) [16].
b. Cathode materials
An ideal active cathode material for lithium-ion batteries should follow the following
requirements [12]:
(i) Should make use of readily reducible/oxidizable atoms, like transition metals.
(ii) Li+ ions can be (de)intercalated reversibly, preferably without changing the host
structure.
(iii) Having good electronic/ionic conductivity.
(iv) Being stable under wide voltage and temperature windows.
Apart from these intrinsic properties there are several physical properties that determine the
quality of cathode active materials and consequently the Li-ion battery characteristics. These
physical properties include particle size, particle shape, particle size distribution, water content,
tap density, specific surface area, impurity level, and so on [12].
The most common cathode materials used in commercial Li-ion batteries are layered oxides
LiMO2, spinel LiM2O4 and olivine LiMPO4 where M is a transition metal. Layered LiCoO2,
was introduced in 1991 as cathode material in the first commercially available Li-ion batteries.
The theoretical specific capacity of LiCoO2 is around 130 mAh·g-1 because only half Li can be
reversibly cycled without causing electrode degradation [12]. LiCoO2 was later replaced by
11


other layered oxides such as LiNi0.8Co0.15Al0.05O2, LiNi0.8Co0.2O2, Li1−xNi1−yCoyO2,

LiMn0.5Ni0.5O2, and LiNi1/3Co1/3Mn1/3O2 in commercial batteries. These materials have high
energy and power densities but still cannot be fully charged due to the structural collapse at
fully charged states. The structural stability of the spinel LiMn2O4 is better in comparison with
the layered oxides. Spinel LiMn2O4 also has a higher operating voltage and can be (dis)charged
at high rates but have a low energy density [16].
Since the introduction by Padhi et al. [32] of LiFePO4 as cathode material in Li-ion batteries,
LiFePO4-based (LFP) batteries are drawing a lot of attention due to the many favorable
characteristics, such as high safety, long life span, environmental friendliness, low cost and
wide-spread materials abundancy. Both the electronic and ionic conductivities of LiFePO4 are
low. However, the rate capability of LFP batteries has been highly improved by using nanosized
particles [33] and carbon coating technologies.
c. Electrolytes
Besides the electrodes the electrolyte constitutes the third key component of Li-ion batteries
[6]. The choice of electrolyte components is dictated by the electrode materials in use [4]. The
most common electrolytes used in commercial Li-ion batteries are composed of one or more
organic solvents and one salt. The preferred solvent for the electrolyte in lithium-ion batteries
is a combination of ethylene carbonate (EC) and dimethyl carbonate (DMC). The most common
salt is LiPF6 [16, 34].
The combination of electrolyte and the electrode materials is responsible for the formation
of the SEI that determines the cycling performance of Li-ion batteries. The properties of the
SEI layers depend on the electrolyte compositions, additives and impurities. For instance, EC
can provide a stable protective layer on the graphite electrode surface that prevents further
reaction (continuous electrolyte reduction and self-discharge) as well as solvent cointercalation
[6]. EC is therefore present in almost all commercial electrolytes, in combination with other
solvents [34]. The electrolyte additives (unstable electrolyte ingredients) can be reduced at
higher electrode potentials, the SEI formation would then be completed far before solvent
cointercalation occurs. Contaminants such as traces of water in the electrolyte can lead to
electrolyte decomposition. For example, HF is formed as by product, harmfully influencing the
battery performance.
12


1.3 Challenges of Li-ion batteries
The main challenges of Li-ion battery technology are related to energy and power densities,
battery life, safety and cost.
1.3.1 Energy and power densities
The battery energy is determined by the intrinsic properties of the individual electrode
materials and design characteristics. There are two types of energy density: gravimetric energy
density and volumetric energy density . The gravimetric energy density can be
calculated according to
1 V2
 
g
Ebat Qbat dVbat , [1.17]
mbat V1
Where Q bat is battery storage capacity, is the battery voltage, is the total mass of the
battery. The integration boundaries , represent the end voltages of the charged and
discharged states, respectively. Qbat mbat is known as the specific capacity of the complete
battery. Apparently, increasing the battery specific capacity and the voltage is an efficient way
to increase the battery energy density.
Qbat mbat is determined by the capacities of the individual electrodes, according to
Qbat qa qc
 , [1.18]
mbat  mo 
qa  1   qc
 ma 
where and are the specific capacities of the anode and cathode, respectively. is the
mass of anode active material, is the mass of all passive materials in the cell. The derivation
of Eq. 1.18 can be found in Appendix I.
Fig. 1.5 shows the contour plot of the specific battery capacity as a function of the specific
anode capacity and specific cathode capacity . For the purpose of clarity, only the mass of
the active materials are accounted for in the calculations, i.e. is assumed to be 0. The specific
capacities of the cathode materials are always smaller than those of the anode materials in all
of the current commercial batteries. For example, the theoretical capacity of graphite is 372
mAh·g-1, and that of LiFePO4 is 170 mAh·g-1. The specific capacity of the LFP battery, as
indicated in Fig. 1.5, is = 116 mAh·g-1. To increase the battery specific capacity from 116
mAh·g-1 to, for example, 125 mAh·g-1, two strategies can be applied. The specific capacities of
13


either the anode or the cathode can be enlarged. Apparently, the capacity-increment required
for the cathode is much smaller than that for the anode. Therefore, the cathode capacity is
considered to be the bottleneck to increase the energy density of Li-ion batteries.
500
225
400 200
175
-1
300 150
qc / mAh g
125
qc
200 100 qa
-1
75 qLFP = 116 (mAh g )
100 50
-1
25 (mAh g )
0
0 100 200 300 400 500
qa / mAh g -1
Fig. 1.5. Contour plot representing the specific battery capacities at various values for and . The
specific battery capacity has been calculated based on the electrodes materials only. The specific
capacity of LFP battery has been indicated by .
Extensive investigations have been carried out to search for high-capacity cathode materials.
Ternary systems such as LiNi0.8Co0.15Al0.05O2 and LiNi1/3Co1/3Mn1/3O2 were found to be
promising alternatives for LiFePO4 electrode because of their higher capacities and higher
voltage levels.
The power density of Li-ion batteries is strongly determined by the electronic and ionic
conductivities of the materials, which are intrinsic materials properties. The electronic
conductivity can be improved by, for example, carbon coating and the ionic conductivity can
be improved by doping and by reducing the particle size, applying state-of-the-art nano-
technologies [35].
1.3.2 Battery performance
Li-ion batteries require a cycle life of over 5000 cycles and a life span of 10-15 years or
more in order to make battery-powered vehicles competitive in a future sustainable economy
[36]. Factors contributing to the durability of large-format Li-ion batteries are complex and vary
widely with different electrode materials, battery manufacturing processes, cycling rate,
14


temperature and other operating conditions. The various battery components may undergo
different aging mechanisms, for example, electrolyte decomposition and SEI formation, current
collector corrosion, structural transformations and metal dissolution from the electrodes [37].
Side reactions and degradation processes in Li-ion batteries may cause a number of undesirable
effects leading to capacity losses [38].
Aprotic electrolytes can be thermodynamically unstable against reduction at negative
electrodes and against oxidation at positive electrodes [39]. As shown in Fig. 1.4, the difference
between the Fermi level of the electrode and the HOMO or LUMO levels of the electrolyte
determines the thermodynamic stability of the electrolyte against the electrodes.
In the presence of Li+ ions the calculated LUMO levels of most electrolyte components (e.g.
EC, VC etc.) decrease by ∼0.5 eV relative to pure solvents, rendering these electrolytes
thermodynamically unstable with respect to the intercalated Li in the graphite electrode [39].
The electrolyte will be reduced at the negative electrode and will form SEI layers comprising
of inorganic and organic Li salts [40]. The cyclable Li+ ions are partially immobilized in the
SEI products, leading to irreversible capacity loss. Moreover, the formation of a non-uniform
SEI layer can result in non-uniform lithium deposition and formation of lithium dendrites [41,
42], which can lead to battery internal short-circuits and battery failures.
High-voltage cathode materials such as Li1−xCoO2 (x < 0.7 above 4.5 vs Li+/Li) [43],
Li1−xMn1.5Ni0.5O4 (Ni4+/Ni2+ at 4.7 vs Li+/Li) [44], and Li1−xNiPO4 (Ni3+/2+ at 5.2 vs Li+/Li) [45]
have very low electrode Fermi levels, approaching the HOMO levels of aprotic electrolytes,
which can result in a thermodynamic driving force for electrolyte oxidation at the cathode [39].
Besides, other species in the electrolyte such as products resulting from reduction at the anode
can also be oxidized at the cathode surface. The solid products from these oxidation reactions
may form a surface film on the cathode, called cathode-electrolyte-interphase (CEI). Most
components of CEI films are electronically resistive, thus deteriorating the rate capability of the
full battery.
Electrode material degradation is also considered to be an important issue to address.
Graphite is commonly used in commercial Li-ion batteries. Graphite is generally considered to
be a stable material. However, structural degradations have been observed under severe aging
conditions, for example, at high temperatures and cycling at high C-rates [46]. The inter-layer
blockage caused by metal deposition [46] and structural deformations caused by stress
accumulation [47] are considered to be the main source of the graphite deterioration. Particle
isolation resulting from the continuous growth of SEI films is considered to be another factor
which eventually leads to a decrease of the graphite electrode capacity [48]. Cathode dissolution
15


is a common phenomenon at elevated temperatures in the presence of even small trace of

protons [49]. Metal ions dissolved from the cathode have a series of detrimental effects on the
battery performance. For example, metal ions can be transported to the anode and subsequently
be deposited on the graphite surface leading to battery capacity losses and graphite electrode
blockage [46]. In ternary cathode systems, such as LiNi1/3Co1/3Mn1/3O2, phase transition [50]
and/or Li-Ni site interchange [51] can be triggered during cycling.
1.3.3 Battery safety
Safety issues are extremely important in Li-ion battery applications. Different battery
chemistries have varying failure modes, but thermal runaway is one of the most common
abusive conditions. Thermal runaway is a catastrophe for Li-ion batteries since it may lead to
battery fire or even explosion. Various events [52-56], contributing to the battery safety hazards
are summarized in Table 1.3.
Table 1.3. Safety hazards of Li-ion batteries

Mechanical damage Electrical failure Thermal abuse
Crush Over (dis)charge Over heat
Impact External short circuit Fire
Shock Internal short circuit
Puncture
Shock
Vibration
Drop
1.3.4 Cost
The main sources of total battery cost are the battery materials and manufacturing processes
[57, 58]. The contributions to the materials cost include the following components: positive and
negative active electrode materials, separator, electrolyte, and current collector foils. Generally,
most Li-ion battery designs are based on the same current collectors and negative electrode
materials (graphite). The cost of the electrolyte and separator is more or less constant. Therefore,
the cost of various cathode materials determines the variations of the total battery cost [57].
The cost of positive electrode materials is driven to a large extent by the cost of the raw
materials from which it is made. Lithium cobalt oxide (LCO) is the original material
16


commercialized in Li-ion batteries for consumer electronics. LCO is expensive because the cost
of Co is high. Many other cathode materials are developed for Li-ion batteries such as lithium
nickel manganese cobalt oxide (NMC), lithium manganese spinel oxide (LMO) and lithium
iron phosphate (LFP). Reducing the quantity of Co in the positive electrode will reduce the total
battery price.
In general, to decrease the cost, earth-abundant transition metals should be dominatingly
present in cathode materials. Both iron and manganese are abundant and inexpensive transition
metals for intercalation materials. In addition, LMO is relatively easy to manufacture. In
contrast, LFP requires a reduced atmosphere [57] and a carbon coating step to reach the end
product. The increased complexity of the manufacturing process results in a higher the price of
LFP.
1.4 Scope of this thesis
The aging performance of Li-ion batteries is of vital importance, especially when applied in
an electric vehicle (EV), hybrid electric vehicle (HEV) and plug-in hybrid electric vehicle
(PHEV) where stable battery performance during a long period of time is required. Long
operation life of Li-ion batteries not only ensures good and reliable performance but also
reduces the usage cost for customers. The aim of this PhD thesis is to investigate the aging
mechanisms of Li-ion batteries under various operation conditions and facilitate the
understanding of battery degradation processes, and ultimately improve the battery life. Battery
ageing has been investigated both experimentally and by simulations. Several novel techniques,
such as electromotive force (EMF) determination, derivative voltage analysis, etc., will be
introduced in this thesis. Physical characterization methods, including X-ray photoelectron
spectroscopy (XPS), Raman spectroscopy, scanning electron microscopy (SEM), inductively
coupled plasma (ICP), etc., have been performed to boost the in-depth understanding of aging
mechanisms. The research work reported here also provides insight to improve the battery
design and material performance.
Chapter 2 gives an overview of degradation mechanisms of Li-ion batteries. Cyclable Li
loss and electrode material decay are the two most important processes in battery aging. Solid-
electrolyte-interphase (SEI) formation at the anode is generally accepted to be the main reason
responsible for cyclable Li losses. The SEI formation mechanisms, experimental
characterization methods and model developments will be systematically reviewed in this
chapter. Cathode electrode degradation which mainly occurs under more severe aging
17


conditions will also be discussed. Transition metal dissolution and structural transformation are
considered to be the main reasons of cathode degradation. Finally, degradation of the graphite
electrode will be discussed, which has been mainly attributed to structural deformation and
interspace blockage.
Chapter 3 introduces the details of the experimental work carried out in this thesis. Both
C6/LiFePO4 (LFP) and C6/LiNi1/3Co1/3Mn1/3O2 (NMC) batteries have been investigated and will
be described. The aging experiments include both storage and cycling performed with complete
batteries and have been carried out with automated cycling equipment. Material
characterization focused on dismantled electrodes and have been performed by XPS, Raman
spectroscopy, ICP spectrometry and SEM.
Chapter 4 discusses the most important methods used in this thesis, including EMF
determination and / analyses. The EMF curve is considered to be an important tool
to obtain an in-depth understanding of aging mechanisms inside Li-ion batteries. Various
parameters, such as maximum capacities ( ), irreversible capacity losses ( Δ ),
overpotentials ( ), etc. can be extracted from these EMF curves. The development of the second
depressions in the / curves is proven to be an interesting indicator for the graphite
electrode decay. A non-destructive approach will be proposed to quantitatively determine the
graphite inaccessibility after aging.
In Chapter 5 the capacity loss and material decay of LFP batteries have been investigated
under various storage conditions as a function of State-of-Charge (SoC) and temperature. The
irreversible capacity losses, which are accurately calculated on the basis of the maximum
storage capacity estimated from the EMF curves, increase as a function of temperature and SoC.
The loss of cyclable lithium during storage is considered to be the main cause of irreversible
capacity losses under all storage conditions. The graphite electrode decay at 60oC has been
quantitatively determined by non-destructive analyses on the basis of ⁄ curves.
Deposition of Fe on the graphite electrode has experimentally been confirmed by XPS and ICP
analyses. The increasing graphite inaccessibility will be shown to be the consequence of Fe
dissolution from the cathode with the subsequent deposition onto the anode.
Chapter 6 discusses the development of the irreversible capacity loss under various cycling
conditions and temperatures. The irreversible capacity loss as a function of cycle number and
time will be discussed. A new mathematical extrapolation method will be proposed to
distinguish between calendar ageing and cycling-induced ageing. The iron dissolution from the
cathode at 60oC and the subsequent deposition onto the anode have been characterized. The
influence of Fe deposition on the SEI formation at the graphite electrode will be investigated.
18


Simultaneously, the graphite electrode decay will be quantified in more detail by analyzing the
/ curves. The analyses show that the electrode decay can be related to the structural
deterioration and the inter-layer surface blockage of the graphite electrode, as will be
experimentally investigated by Raman and XPS spectroscopy.
Theoretical simulations of battery capacity losses upon both storage and cycling will be
described in Chapter 7. At moderate temperatures, the irreversible capacity losses can mainly
be attributed to the Li immobilization in the SEI layer. At elevated temperatures, iron
dissolution from the cathode and the metal deposition on the anode will influence the
irreversible capacity losses. The model predicts that the capacity losses during cycling are larger
than during storage due to the generation of cracks inside the SEI induced by the volumetric
changes of the graphite electrode during charging. The simulation results show that the capacity
loss during storage is dependent on the State-of-Charge (SoC), the storage time, and
temperature while during cycling it is dependent on the cycling current, the cycling time,
temperature and cycle number.
Finally, the degradation of NMC batteries will be discussed in Chapter 8. The irreversible
capacity losses have been systematically investigated as a function of time and cycle number at
various cycling currents and temperatures. / curves are used to determine the material
decay of the individual electrodes.
1.5 References
[1] R.A. Huggins, in, Springer, New York, 2010, pp. 424.
[2] J.M. Tarascon, Philosophical Transactions of the Royal Society a-Mathematical Physical
and Engineering Sciences, 368 (2010) 3227-3241.
[3] M. Winter, R.J. Brodd, Chemical Reviews, 104 (2004) 4245-4269.
[4] C. Julien, A. Mauger, A. Vijh, K. Zaghib, Lithium batteries, Springer, New York, 2016.
[5] R. Jasinski, High-energy batteries, Plenum, New York, 1967.
[6] J.M. Tarascon, M. Armand, Nature, 414 (2001) 359-367.
[7] P.H.L. Notten, M. Latroche, Metal hydride alloys, Elsevier, Amsterdam, 2009.
[8] H.J. Bergveld, W.S. Kruijt, P.H.L. Notten, Battery management system design by
modelling, Kluwer Academic Publischers, Dordrecht, The Netherlands, 2002.
[9] M. Armand, Intercalation electrodes, Plenum Press, New York, 1980.
[10] K. Mizushima, P.C. Jones, P.J. Wiseman, J.B. Goodenough, Materials Research Bulletin,
15 (1980) 783-789.
19


[11] M. Armand, P. Touzain, Materials Science and Engineering, 31 (1977) 319-329.

[12] R.J. Brodd, M.M. Doeff, J.B. Goodenough, K. Naoi, K. Bullock, Z. Zhang, G.D. Smith,
O. Borodin, P.J. Sideris, S.G. Greenbaum, B. Barnett, D. Ofer, S. Sriramulu, R.
Stringfellow, P. Ramadass, A. Yamada, in: R.J. Brodd (Ed.) Batteries for Sustainability,
Springer New York, USA, 2013.
[13] J.W. Choi, D. Aurbach, Nature Material, 1 (2016) 1-16.
[14] J.B. Goodenough, Y. Kim, Chemistry of Materials, 22 (2010) 587-603.
[15] J.B. Goodenough, K.S. Park, Journal of the American Chemical Society, 135 (2013)
1167-1176.
[16] S.C. Nagpure, B. Bhushan, S.S. Babu, Journal of the Electrochemical Society, 160 (2013)
A2111-A2154.
[17] T.H. Hwang, Y.M. Lee, B.S. Kong, J.S. Seo, J.W. Choi, Nano Letters, 12 (2012) 802-
807.
[18] N. Liu, H. Wu, M.T. McDowell, Y. Yao, C.M. Wang, Y. Cui, Nano Letters, 12 (2012)
3315-3321.
[19] D.S. Jung, T.H. Hwang, S.B. Park, J.W. Choi, Nano Letters, 13 (2013) 2092-2097.
[20] I.H. Son, J.H. Park, S. Kwon, S. Park, M.H. Rummeli, A. Bachmatiuk, H.J. Song, J. Ku,
J.W. Choi, J.M. Choi, S.G. Doo, H. Chang, Nature Communications, 6 (2015).
[21] Y. Yu, L. Gu, C.B. Zhu, S. Tsukimoto, P.A. van Aken, J. Maier, Advanced Materials, 22
(2010) 2247-2250.
[22] B. Koo, H. Kim, Y. Cho, K.T. Lee, N.S. Choi, J. Cho, Angewandte Chemie-International
Edition, 51 (2012) 8762-8767.
[23] T.W. Kwon, Y.K. Jeong, J.W. Choi, A. Coskun, Abstracts of Papers of the American
Chemical Society, 248 (2014).
[24] T.W. Kwon, Y.K. Jeong, I. Lee, T.S. Kim, J.W. Choi, A. Coskun, Advanced Materials,
26 (2014) 7979-7985.
[25] C. Wang, H. Wu, Z. Chen, M.T. McDowell, Y. Cui, Z.A. Bao, Nature Chemistry, 5 (2013)
1042-1048.
[26] Z. Chen, C. Wang, J. Lopez, Z.D. Lu, Y. Cui, Z.A. Bao, Advanced Energy Materials, 5
(2015) 1-8.
[27] Y.K. Jeong, T.W. Kwon, I. Lee, T.S. Kim, A. Coskun, J.W. Choi, Nano Letters, 14 (2014)
864-870.
20


[28] E.M. Erickson, E. Markevich, G. Salitra, D. Sharon, D. Hirshberg, E. de la Llave, I.

Shterenberg, A. Rozenman, A. Frimer, D. Aurbach, Journal of the Electrochemical
Society, 162 (2015) A2424-A2438.
[29] V. Etacheri, U. Geiger, Y. Gofer, G.A. Roberts, I.C. Stefan, R. Fasching, D. Aurbach,
Langmuir, 28 (2012) 6175-6184.
[30] E. Markevich, K. Fridman, R. Sharabi, R. Elazari, G. Salitra, H.E. Gottlieb, G.
Gershinsky, A. Garsuch, G. Semrau, M.A. Schmidt, D. Aurbach, Journal of the
Electrochemical Society, 160 (2013) A1824-A1833.
[31] E. Markevich, G. Salitra, A. Rosenman, Y. Talyosef, D. Aurbach, A. Garsuch, Rsc
Advances, 4 (2014) 48572-48575.
[32] A.K. Padhi, K.S. Nanjundaswamy, J.B. Goodenough, Journal of the Electrochemical
Society, 144 (1997) 1188-1194.
[33] F.M. Jiang, P. Peng, Y.Q. Sun, Journal of Power Sources, 243 (2013) 181-194.
[34] K. Xu, Chemical Reviews, 104 (2004) 4303-4417.
[35] S.Y. Chung, J.T. Bloking, Y.M. Chiang, Nature Materials, 1 (2002) 123-128.
[36] X.L. Wang, K. An, L. Cai, Z.L. Feng, S.E. Nagler, C. Daniel, K.J. Rhodes, A.D. Stoica,
H.D. Skorpenske, C.D. Liang, W. Zhang, J. Kim, Y. Qi, S.J. Harris, Scientific Reports,
2 (2012).
[37] V. Agubra, J. Fergus, Materials, 6 (2013) 1310-1325.
[38] P. Arora, R.E. White, M. Doyle, Journal of the Electrochemical Society, 145 (1998)
3647-3667.
[39] M. Gauthier, T.J. Carney, A. Grimaud, L. Giordano, N. Pour, H.H. Chang, D.P. Fenning,
S.F. Lux, O. Paschos, C. Bauer, F. Magia, S. Lupart, P. Lamp, Y. Shao-Horn, Journal of
Physical Chemistry Letters, 6 (2015) 4653-4672.
[40] P. Verma, P. Maire, P. Novák, Electrochimica Acta, 55 (2010) 6332-6341.
[41] O. Crowther, A.C. West, Journal of the Electrochemical Society, 155 (2008) A806-A811.
[42] R. Bhattacharyya, B. Key, H.L. Chen, A.S. Best, A.F. Hollenkamp, C.P. Grey, Nature
Materials, 9 (2010) 504-510.
[43] M. Menetrier, D. Carlier, M. Blangero, C. Delmas, Electrochemical and Solid State
Letters, 11 (2008) A179-A182.
[44] Q.M. Zhong, A. Bonakdarpour, M.J. Zhang, Y. Gao, J.R. Dahn, Journal of the
Electrochemical Society, 144 (1997) 205-213.
[45] J. Wolfenstine, J. Allen, Journal of Power Sources, 142 (2005) 389-390.
21


[46] D. Li, D. Danilov, J. Xie, L. Raijmakers, L. Gao, Y. Yang, P.H.L. Notten, Electrochimica
Acta, 190 (2016) 1124-1133.
[47] Y. Qi, H. Guo, L.G. Hector, Jr., A. Timmons, Journal of the Electrochemical Society,
157 (2010) A558-A566.
[48] C.S. Wang, X.W. Zhang, A.J. Appleby, X.L. Chen, F.E. Little, Journal of Power Sources,
112 (2002) 98-104.
[49] D. Aurbach, B. Markovsky, G. Salitra, E. Markevich, Y. Talyossef, M. Koltypin, L.
Nazar, B. Ellis, D. Kovacheva, Journal of Power Sources, 165 (2007) 491-499.
[50] S.C. Yin, Y.H. Rho, I. Swainson, L.F. Nazar, Chemistry of Materials, 18 (2006) 1901-
1910.
[51] S.K. Jung, H. Gwon, J. Hong, K.Y. Park, D.H. Seo, H. Kim, J. Hyun, W. Yang, K. Kang,
Advanced Energy Materials, 4 (2014).
[52] R. Spotnitz, J. Franklin, Journal of Power Sources, 113 (2003) 81-100.
[53] D. Doughty, E.P. Roth, The Electrochemical Society Interface, (2012) 37-44.
[54] F. Ren, T. Cox, H. Wang, Journal of Power Sources, 249 (2014) 156-162.
[55] C. Zhang, S. Santhanagopalan, M.A. Sprague, A.A. Pesaran, Journal of Power Sources,
290 (2015) 102-113.
[56] E. Sahraei, E. Bosco, B. Dixon, B. Lai, Journal of Power Source, 319 (2016) 56-65.
[57] P.A. Nelson, K.G. Gallagher, I. Bloom, D.W. Dees, in, http://www.osti.gov/bridge, Oak
Ridge, 2012.
[58] E.J. Berg, C. Villevieille, D. Streich, S. Trabesinger, P. Novak, Journal of the
22

Chapter 2
Overview of Degradation Mechanisms of
Li-ion Batteries
Aging is a critical issue of Li-ion batteries in many applications such as electric vehicle (EV), hybrid
electric vehicle (HEV), etc. Cyclable Li losses and electrode material decay are the two most
important processes in battery aging. Solid-electrolyte-interphase (SEI) formation on the anode is
generally believed to be the main reason responsible for the cyclable Li losses. The SEI formation
mechanisms, experimental characterizations and model developments will be systematically
reviewed in this chapter. Cathode electrode degradation, which mainly occurs under severe aging
conditions will also be discussed. Transition metal dissolution and structural transformation are
considered to be the main reason of cathode degradation. Finally, degradation of the graphite
electrode will be discussed, which is mainly caused by structural deformation and interspace
blockage.
23

Chapter 2 Overview of Degradation Mechanisms of Li-ion batteries

2.1 Capacity degradation
It is well known that Li-ion batteries suffer from capacity loss during both storage and
cycling [1-7]. Battery capacity loss can be classified as reversible (Δ ) and irreversible
capacity losses (Δ ) [7]. Δ is defined as losses that can be fully recovered in the subsequent
cycles under low-current conditions. The origin of Δ is still under debate. Some scientists
attributed Δ to current-dependent battery polarization [1]. Others proposed the formation of
a metastable “electron-ion-solvent” complex as the reason of Δ [8].
Extensive studies have been performed to investigate Δ [8-84]. It is generally understood
that the degradation mechanisms strongly depend on the battery chemistries, e.g., LFP [20] and
NMC [21-26]. Δ does not originate from a single aging process but rather combines various
processes and their interactions. The degree of influence of the individual processes on Δ
can vary and it is therefore a challenge to isolate the contributions of individual processes on
battery capacity loss.
Calendar-life studies revealed that Li-immobilization in SEI is the main source of Δ at
moderate storage temperatures [12-16]. However, at elevated temperatures the cathode
degradation may have a significant impact on Δ [18-26]. The influence of cathode
degradation on Δ is strongly dependent on the specific degradation mechanisms. For
example, metal ions dissolved from the cathode can be deposited on the graphite surface and,
consequently, accelerate the SEI formation [20, 21]. The structural transformations at the
cathode surface result in the kinetic decline which finally leads to battery capacity loss [22-26].
Apart from temperature, SoC is also considered to be an important factor influencing Δ [12-
25].
Similar to the case of storage, Δ has also been attributed to Li-immobilization in the SEI
layers under moderate cycling conditions (low temperatures, small currents). However, the
cycling-induced effect has an additional contribution in Δ [65-67]. Therefore, it is
commonly accepted that Δ is larger during cycling than under storage conditions [14]. The
influences of current [49-51], temperature [39-45] and cycling range [56-58] on Δ are found
to be more pronounced under cycling. The irreversible capacity losses are always accompanied
by the degradation of the electrodes, Δ and Δ for cathode and anode, respectively.
Therefore, electrode degradation under both storage and cycling conditions is important to
address.
24

Chapter 2 Overview of Degradation Mechanisms of Li-ion Batteries

Summarizing the discussions above, all aging processes related to Δ ,Δ and Δ can
be schematically presented in Fig. 2.1.
Fig. 2.1. Capacity degradation mechanisms inside commercial Li-ion batteries under both storage (blue
arrows) and cycling conditions (red arrows). represents current, ΔVol the electrode volume changes,
the Li diffusion flux, the SEI-covered surface area, the fresh surface area induced by cracks,
DIS the diffusion-induced stress, Δ is the battery irreversible capacity loss, Δ and Δ represent
the capacity losses of the anode and cathode, respectively. is the operation temperature.
In Fig. 2.1, the network constructed by the blue arrows represents all aging processes
happening under storage conditions and the red-arrow-connected network represents the
processes occurring upon cycling. Under storage both the current flow ( ) and the electrode-
volume changes (ΔVol) are zero. The graphite electrode surface area covered by the SEI layers
after the formation step has been considered to be constant ( ) [65]. The influence of the
cathode on Δ is negligible at moderate temperatures. Therefore, the continuous SEI growth
on becomes the only source for Δ .
However, at elevated temperatures (i.e. 60oC), the influence of the cathode electrode on
Δ becomes more significant. As can be seen in Fig. 2.1, the transition metal ions start to
25


dissolve from the cathode materials and can subsequently be transported to the graphite surface
where they are reduced into metallic clusters/dendrites. The cathode dissolution process will
directly lead to Δ , while the metal deposition on graphite surface has at least three
consequences, namely, blocking the graphite layers and leading to inaccessibility of the graphite
electrode (Δ ), consuming cyclable electrons at the anode, and accelerating the SEI formation.
If the size of the cluster/dendrite is large enough then it becomes a safety hazard because internal
short circuits may be induced. It should be noted that the structural transformation can also
occur in the case of NMC materials at elevated temperatures under storage, which will cause an
increase of both Δ and Δ .
The consequences caused by the cycling effect have been indicated by the red arrows shown
in Fig. 2.1. The capacity degradation mechanisms under cycling conditions are more complex
than under storage. Consider the case of moderate temperatures. Volume expansion and
contraction of the graphite electrode take place periodically during charging and discharging.
Cracks in the SEI layers will be induced by the volume expansion due to the fragile structure
of the SEI. The graphite surface areas under these cracks ( ) are exposed to solvent
immediately. Therefore, the total graphite surface area is divided in two parts, and
[85]. That leads to additional Δ in cycling conditions induced by the SEI formation on
in comparison with the case of storage conditions. Simultaneously, the Li (de)intercalation
processes during cycling can also induce a so-called diffusion-induced stress (DIS), which can
lead to fracture of the graphite particles. The fracture of the graphite particles has three
consequences: (i) it causes Δ [86], (ii) it generates more fresh surface areas, leading to more
SEI formation and Δ [85], (iii) part of the fractured graphite particles may be isolated from
the conductive matrix, generating “dead Li” which finally contributes an additional Δ .
When batteries are cycled at elevated temperatures, especially at high currents, the cathode
dissolution and structural transformation (i.e. NMC materials) will occur. The dissolution effect
is similar to that under storage. The structural transformation processes, however, can lead to
Δ and Δ due to the inactivation of the cathode particles.
The SEI formation as well as electrode degradation will be discussed in more details in the
following sections.
26


2.2 SEI formation
2.2.1 What is SEI?
SEI is an important concept in modern Li-ion battery technologies. It is well known that a
thin film covers the anode surface acting as a passivation layer between the electrode and the
electrolyte. Thus “interphase” has solid-electrolyte properties, namely, it allows ions to pass
through but prevents electron transport. It has been denoted by Peled as “Solid-Electrolyte-
Interphase (SEI)” [87]. The SEI is mainly formed by parasitic reactions on both cathode and
anode electrodes in Li-ion batteries. For example, solvent oxidation on the cathode leads to the
formation of cathode SEI while solvent reduction at the anode lead to anode SEI.
Conventionally, “SEI” is used to only refer to the anode SEI. The cathode SEI is therefore
denoted as CEI (Cathode-Electrolyte Interphase) in this thesis in order to distinguish from the
anode.
2.2.1.1 Properties of SEI on anodes
SEI layer studies of the graphite electrode have a long history. In earlier work, low
reversible capacity can be obtained from the graphite electrode using a propylene-carbonate
(PC)-based electrolyte [88]. It was found that PC decomposes on the graphite electrode at a
voltage of ~0.8 V, which is much higher than the potential for Li intercalation (< 0.2 V) [88].
The products of these decomposition reactions are deposited on the graphite electrode forming
a loose and porous film. The presence of these solvent decomposition reactions apparently
prohibits reversible Li intercalation [89]. After many efforts, researchers realized that the
structure of the thin films deposited on the anode surface plays a critical role in determining the
cycling performance of the graphite electrode. Dahn et al. [90] found that the reduction products
of ethylene carbonate (EC) form a much more dense layer which effectively inhibits the
irreversible reactions and prevents the graphite electrode from structural disintegration. The
difference in properties of the films built by the decomposition products of PC and EC were
attributed to methyl group in PC [91], known as the “alkyl loose tails” [92], which intervenes
with of good adhesion and cohesion of the formed film on the graphite surface. The utilization
of EC in the electrolyte finally overcomes the challenge of structural disintegration of graphite
and establishs reversible Li intercalation and de-intercalation processes. This finally led to the
application of graphite as anode material in various Li-ion battery chemistries [89].
27


The SEI layer on the graphite electrode plays a dual role in modern Li-ion batteries. On the
one hand, it protects the graphite electrode from exfoliation, induced by the solvent co-
intercalation, and prevents further solvent decomposition. On the other hand, the continuous
growth of SEI layers will lead to irreversible capacity loss [65]. Ideally, a good SEI should meet
the following requirements [89]: (i) electron isolation, (ii) high ionic conductivity, (iii) uniform
morphology and chemical composition for homogeneous current distribution, (iv) good
adhesion to the carbonaceous anode surface, (v) good mechanical strength and flexibility and
(vi) low solubility in the electrolytes.
The properties of the SEI layers have a strong impact on the cycle life of Li-ion batteries.
In order to improve the cycling performance of Li-ion batteries, extensive research has been
performed on the SEI formation mechanisms, both experimental and theoretical.
2.2.1.2 Properties of CEI on cathode electrode
Compared with the extensive research activities on the SEI at the anode electrode there are
only a few studies dedicated to the understanding of the interface formed between electrolytes
and cathode surfaces.
Goodenough and co-workers are considered to be the first authors who suggested that a film
exists at the cathode/electrolyte interface [89, 93]. In formulating new electrolyte compositions
that can withstand the high potentials of cathode materials, Guyomard and Tarascon also
realized that oxidative decomposition of electrolyte components occurred at the cathode
surfaces, leading to surface passivation and consequently preventing further electrolyte
decomposition [94, 95]. Aurbach et al. reported the existence of thin films on LiNiO2 and
LiMn2O4 electrodes. EDAX (energy-dispersive X-ray spectroscopy) and XPS analysis
confirmed that Li2CO3 and ROCO2Li are the main components of these films [96]. Dedryvere
et al. [97] observed LiF on the LiCoO2 electrode and revealed that the content of LiF is strongly
influenced by the battery voltage. Zhang et al. also studied the CEI on the LiNiO2 electrode and
argued that the formation of CEI is one of the main origins of the irreversible capacity losses in
the initial cycles [98]. Wang et al. [99] investigated surface films on NMC electrodes by XPS
analyses. They found Li2CO3 to be the main component of the surface film of pristine electrode
materials. However, -(CH2CH2O)n- , R-CH2OCO2Li and R-CH2OLi become more dominant in
aged cathode materials.
The CEI also plays a dual role in determining the cathode performance. The existence of
CEI can increase the cathode resistance thereby decreasing the battery rate capability, and lead
to capacity loss. On the other hand, the film can protect the cathode electrode from further
28


electrolyte oxidation. Therefore, a thin and robust film with high ionic conductivity is important
to safeguard the cathode performance.
2.2.2 SEI formation mechanisms
2.2.2.1 Electrolyte decomposition mechanisms
Fig. 2.2 schematically represents the difference of the electrodes Fermi level ( for the
anode, for the cathode) and the electrolyte redox stability windows in Li-ion batteries. The
open-circuit voltage ( ) of a cell is determined by the difference of the Fermi levels between
the cathode and anode, according to
  E cf  E af 
VOCP  . [2.1]
e
Fig. 2.2. Schematic representation of Fermi levels of anode and cathode and the HOMO/LUMO (blue)
levels of the electrolyte. HOMO*/LUMO* (red) represent the real energy levels of the electrolyte,
considering association of solvent molecules and Li ions.
The relationship between the energy levels of and the lowest unoccupied molecular orbital
(LUMO) levels of the electrolyte determine the thermodynamic stability of the electrolyte at
the anode. The relationship between and the highest occupied molecular orbital (HOMO)
levels of the electrolyte determine the thermodynamic stability of the electrolyte against the
29


cathode. Ideally, as schematically illustrated in Fig. 2.2 (blue case), LUMO is above the Fermi
level of the anode and HOMO is below the Fermi level of the cathode, consequently no
reduction and oxidation will occur inside the batteries [100]. However, in most practical cases,
solvent coordinated with Li ions is less stable than pure solvent molecules because of the
electron displacement to the carbonate end of the molecule via the induced dipole effect [9, 101,
102]. Therefore, the LUMO of the electrolyte in presence of the Li ions is lower than the Fermi
level of the anode and the energy difference between and LUMO* will drive the reduction
of the electrolyte at the anode surface (red case). Furthermore, Wang et al. [101] argued that
the presence of Li ions can stabilize the EC reduction intermediates. Likewise, on the positive
side, the HOMO* of the electrolyte becomes higher than due to the association of solvent
molecules with Li ions, which finally leads to oxidation of the electrolyte on the cathode [4].
Electrolyte reduction is an expected characteristic of all Li-ion batteries using carbon-based
insertion electrodes due to the instability of the electrolyte at the carbon electrode under
reducing conditions. Electrolyte reduction can jeopardize the battery capacity and cycle life
performance by consuming Li salt, solvent species and cyclable electrons and may lead to safety
problems by generating gaseous products. Minimization of the electrolyte reduction reactions
and capacity losses related to these processes require detailed knowledge about the mechanisms
of these reactions.
Dey et al. [88] studied the electrochemical decomposition of LiClO4 PC electrolytes on
graphite and proposed a two-electron decomposition mechanism based on gas chromatography
analyses
PC + 2Li+ + 2e- → CH3CH=CH2 + Li2CO3. [2.2]
Eq. 2.2 occurs during the first charge when the potential of the electrode is near 0.8 V vs. Li+/Li.
Propylene as a product was found to be highly soluble in PC.
Aurbach and co-workers [103-105] performed extensive studies of solvents reduction and
salt decomposition processes at the surface of carbon-based electrodes by a series of
spectroscopic analyses. On the basis of the functionalities detected in Fourier transform infrared
spectroscopy (FT-IR), X-ray microanalysis, Scanning electron microscopy (SEM), X-ray
photoelectron spectroscopy (XPS) and nuclear magnetic resonance (NMR) studies, they were
able to investigate the mechanisms involved in the reduction process of carbonate-based
solvents. Aurbach argued that the formation of (RCO3Li) species, resulting from the reduction
of PC, involved only one electron transfer step followed by radical termination reactions, rather
30


than the expected two electrons per PC molecule transfer. The proposed mechanism is
summarized in Fig. 2.3.
Fig. 2.3. Reaction scheme of one-electron reduction mechanism proposed by Aurbach [103].
The first electron transfer to PC will form a radical anion which can transfer to an
intermediate state by C-O bond cleavage. This relatively stable intermediate will further transfer
to various Lithium alkyl carbonates by radical termination reactions [103]. This reduction
mechanism was believed to be independent of the nature of the working electrode.
Wang et al. [101] investigated the decomposition mechanism of EC by using density
functional theory (DFT) calculation. They proposed that an additional electron can further
reduce the radical intermediate to form Li2CO3, as shown in the following reaction equations,
(EC)nLi+ + e ̶ → [(EC) ̶ ·Li+] + (EC)n-1
[(EC) ̶ ·Li+] + e ̶ → Li2CO3 + CH2=CH2 [2.3]
Matsumara et al. [106] investigated the irreversible capacity losses during the first cycle
and argued that the capacity losses were not only due to PC decomposition to Li2CO3, but also
because of additional side reactions. They concluded that there are two possible pathways for
the decomposition of PC during the first charge. In one branch, PC is directly reduced into
propylene and Li2CO3. In another branch, PC is firstly reduced to a radical anion, which is then
transferred into lithium alkyl carbonate(s) by radical termination reactions. Shu et al. [107]
studied the electrolyte decomposition during the first cycle of Li intercalation into graphite in 1
M LiCIO4 PC/EC (1:1) electrolyte. They suggested that at least two processes were involved,
namely, two-electron reduction of PC and EC to propylene and ethylene gases and one-electron
reduction to form lithium alkyl carbonates. The two-electron reduction has been further divided
into direct electrochemical and chemical reduction. The overall reaction scheme is shown in
Fig. 2.4.
31


Fig. 2.4. Reaction scheme of the solvent reduction mechanism proposed by Shu [107].
The chemical reduction pathway requires the formation of a complex [PCmLi+] followed by
the reduction of graphite at which PC/EC was further decomposed. The initial step for both
electrochemical reduction and solid electrolyte interphase (SEI) film formation involves the
formation of lithium carbonate complexes followed by one-electron reduction to radical anions.
These radical anions undergo further one-electron reduction, producing gaseous products or
radical termination to form an SEI film.
Naij et al. [108] cycled electrochemically two types of carbon electrodes in LiClO4-EC-
based electrolyte. They found that EC reduction takes place in two steps: firstly a two-electron
step process at 0.8V vs Li+/Li leading to the inorganic product Li2CO3 and secondly, a one-
electron process occuring at lower potentials, producing organic lithium alkylcarbonate
compounds.
2.2.2.2 Specific decomposition reactions
Based on the mechanisms discussed above, the reactions related to the SEI formation,
including solvent reduction and salt decomposition inside Li-ion batteries, can be summarized
as follows.
a. Solvent reduction [1]

(i) Reduction of propylene carbonate (PC), via a two-electron mechanism
PC + 2Li+ + 2e ̶ → CH3CH=CH2 (g) + Li2CO3 (s) [2.4]
via a one-electron mechanism
PC + e ̶ → PC ̶ (radical anion)
2PC ̶ → CH3CH=CH2 (g) + CH3CH(OCO2 ̶ )CH2(OCO2 ̶ )
CH3CH(OCO2 ̶ )CH2(OCO2 ̶ ) + 2Li+ → CH3CH(OCO2Li )CH2OCO2Li (s)
in total:
2PC + 2e ̶ + 2Li+ → CH3CH(OCO2Li )CH2OCO2Li (s) + CH3CH=CH2 (g) [2.5]
32


(ii) Reduction of ethylene carbonate (EC), similar to that for PC. Via a two-electron
mechanism
EC + 2Li+ + 2e ̶ → CH2=CH2 (g) + Li2CO3 (s) [2.6]
via a one-electron mechanism
EC + e ̶ → EC ̶ (radical anion)
2EC ̶ → CH2=CH2 (g) + CH2(OCO2 ̶ )CH2(OCO2 ̶ )
CH2(OCO2 ̶ )CH2(OCO2 ̶ ) + 2Li+ → CH2(OCO2Li )CH2OCO2Li (s)
in total:
2EC + 2e ̶ + 2Li+ → CH2(OCO2Li )CH2OCO2Li (s) + CH2=CH2 (g) [2.7]
(iii) Reduction of dimethyl carbonate (DMC) can be written as
CH3OCO2CH3 + Li+ + e ̶ → CH3OCO2Li (s) + CH3· [2.8a]
or
CH3OCO2CH3 + Li+ + e ̶ → CH3OLi (s) + CH3OCO· [2.8b]
The radicals CH3· and CH3OCO· can be re-assembled freely, forming C2H6 or
(CH3OCO)2 or CH3OCOCH3 .
(iv) Reduction of diethyl carbonate (DEC) can be written as
CH3CH2OCO2 CH2CH3 + Li+ + e ̶ → CH3CH2OCO2Li (s) + CH3CH2· [2.9a]
or
CH3CH2OCO2 CH2CH3 + Li+ + e ̶ → CH3CH2OCO· + CH3CH2OLi (s) [2.9a]
The radicals CH3CH2· and CH3CH2OCO· can be converted into C4H10 or
C2H5OCOC2H5 or (CH3CH2OCO)2 [96, 109, 110].
b. Salt decomposition [1, 96, 109, 110]

(i) Decomposition of LiPF6
LiPF6 ⇌ LiF + PF5 [2.10]
PF5 + 2 e ̶ + 2 Li+ → LiF + LixPF5-x [2.11]
PF6 ̶ + 2e ̶ + 3Li+ → 3LiF + PF3 [2.12]
Traces of water in the electrolyte can also be involved in the salt decomposition
reactions,
PF5 + H2O → 2HF + PF3O [2.13]
PF3O + 2 e ̶ + 2 Li+ → LiF + LixPF3-xO [2.14]
(ii) Decomposition of LiClO4

LiClO4 + 8e ̶ + 8Li+ → 4Li2O + LiCl [2.15a]
33


or
LiClO4 + 4e ̶ + 4Li+ → 4Li2O + LiClO2 [2.15b]
or
LiClO4 + 2e ̶ + 2Li+ → Li2O + LiClO3 [2.15c]
(iii) Decomposition of LiAsF6

LiAsF6 + 2e ̶ + 2Li+ → 3LiF + AsF3 [2.16]
AsF3 + 2 e ̶ + 2 Li+ → LiF + LixAsF3-x [2.17]
(iv) Decomposition of LiBF4 (similar to LiPF6)

LiBF4 + xe ̶ + xLi+ → LiF + LixBF4-x [2.18]
2.2.2.3 Factors influencing the SEI formation
Influence of graphite matrix
Yamaguchi et al. [111] used in-situ electrochemical Atomic Force Microscopy (AFM) to
study the SEI formation on the graphite surface in Li-ion batteries. From these results they
concluded that solvent decomposition can take place on both the edge and basal surfaces of the
graphite electrode. Chu et al. [112] studied the surface films formed on highly ordered pyrolytic
graphite (HOPG) electrodes during cathodic polarization in 1 M LiClO4 EC/DMC (1:1) and 1
M L1PF6 EC/DMC (1:1) electrolytes. They found the solvent reduction reactions to be
irreversible and suggested that these reactions initiate at a higher potential (1.6 and 2.0 V vs.
Li+/Li) on edge surfaces of HOPG while at a lower voltage (0.8 and 1.0 V) on the basal surface.
Peled et al. [113, 114] investigated the composition and morphology of the SEI formation
on both highly ordered and disordered carbon electrodes with XPS. They suggested that the
carbon matrix has more effect on the composition and SEI layer thickness than the nature of the
electrolyte. The experimental data show good evidence for compositional and morphological
distinctions between the SEI formed on the basal and edge surfaces of the graphite electrode.
Utsunomiya et al. [115] investigated the temperature dependence of the SEI formation on both
the edge and basal surfaces of the graphite electrode using Scanning Transmission Electron
Microscope (STEM) and Energy Dispersive X-ray Spectroscopy (EDS). They concluded that
the morphology and chemistry of the SEI formed on the edge surface is not influenced by the
storage temperature while the morphology and thickness of the SEI formed on the basal surface
34


depended strongly on storage temperature. The influence of the graphite particle size on the
electrolyte decomposition was investigated by Utsunomiya et al. [115].
Influence of solvent components
Haregewoin et al. [116] reported a systematic electrochemical and spectroscopic

comparison of the reduction of PC, EC, and diethyl carbonate (DEC) when used as single (PC),
binary (EC/PC, EC/DEC), and ternary (EC/PC/DEC) solvent systems by analyzing the products
with FTIR. FTIR analyses revealed that the reduction of EC and PC was not considerably
influenced by the presence of other alkyl carbonates. However, DEC exhibited a different
reduction product when used in EC/DEC and EC/PC/DEC solvent systems. The influence of
DEC was attributed to the additional reaction of its primary reduction intermediate with various
surrounding species. The reduction of EC occurred at higher potential than that of PC and DEC
and produced a passivating surface film that prevented carbon exfoliation caused by PC. The
EC/PC/DEC-based electrolyte exhibited favorable capacity retention, higher Li+ ion diffusivity,
and lower impedance compared with those of the EC/PC system.
Kim et al. [117] investigated the decomposition of the electrolyte with solvent mixtures of
EC and different linear carbonates dimethyl carbonate (DMC), DEC and ethyl methyl carbonate
(EMC) by using liquid Gas Chromatography-Mass Spectrometry (GC/MS) technique. They
concluded that the reduction voltage of DMC and EMC is higher than that of DEC. The
reduction products of DMC and EMC are soluble and has been detected by GC/MS.
Influence of Li salts in the electrolyte
The cycle life of rechargeable Li-ion batteries depends on the long-term reversibility of cell
chemistries, therefore the electrochemical stability of the electrolyte, especially with Li salts,
plays a crucial role. This section will discuss the electrochemical stability of various salts used
in state-of-the-art electrolytes and the influence of these salts on the SEI formation on the
graphite electrode.
Table 2.1 lists selected electrochemical stability data for a number of anions commonly used
in lithium-based electrolytes. Although it is known that the reduction of anions does occur,
sometimes even at high potentials, these processes are usually sluggish and standard redox
potentials for these reduction processes are usually hard to determine. The reduction of solvents,
occurring simultaneously with reduction of anions, further complicates the interpretation. For
35


this reason, only the anodic stability of salt anions is given in Table 2.1, while the cathodic limit
of the salt is in most cases set by the reduction of its cation (i.e. lithium deposition potential).
Table 2.1. Stability of anions on non-active electrodes [89]
concentration working
salt solvent (i / (mA·cm-2)) a ref
(mol /L) electrode
PC 0.65 GC b 6.1 (1.0) [118]
ClO4̶
PC Pt 4.6 [119]
PC 0.65 GC 6.6 (1.0) [118]
BF4̶
EC/DMC 1.0 Au 4.78
THF 0.001 GC 4.4 (0.1) [120]
PC 0.65 GC 6.8 (1.0) [118]
PF6̶ none GC 4.94 (1.0) [121]
none Pt 5.0 (1.0) [121]
EC/DMC 1.0 Au 4.55 [122]
PC 0.65 GC 6.8 (1.0) [118]
none GC 5.05 (1.0) [121]
AsF6̶ none Pt 5.1 (1.0) [121]
EC/DMC 1.0 Au 4.96 [122]
THF 1.0 GC 4.25 (0.1) [120]
THF 1.0 GC 4.1 (0.1) [120]
SbF6̶
PC 0.65 GC 7.1 (1.0) [118]
PC 0.65 GC 6.0 (1.0) [118]
̶
Tf PC 0.10 GC 5.0 (0.5)
EC/DMC 1.0 Pt 4.29 [122]
a
Anodic limit, potential referred to Li+/Li, cutoff current density in parentheses. b glassy carbon.
Coupling effects of solvent/salt on the electrolyte stability can be expected when solvent
mixtures are used. The stability of the electrolyte can be much more improved when proper
solvent/salt combination are selected. For example, the room-temperature breakdown voltage
of LiX/EC/DMC electrolytes is of the order
ClO4 ̶ ~ PF6 ̶ ~ BF4 ̶ > AsF6 ̶ > Tf ̶ .
Leroy et al. [123] investigated the formation of SEI layers at the surface of graphite
electrodes of full graphite/LiCoO2 batteries using LiBF4, LiPF6, LiN(SO2CF3)2, Li(SO2C2F5)2
in carbonate solvents. The results were analyzed by XPS. They have reported that the potential-
dependent character of the surface film and each salt. At 3.8 V, all salts produce identical
carbonated species. Beyond this potential, the specific behavior of LiPF6 was identified by the
high LiF content in the SEI, whereas for other salts, the formation process of the SEI appears
to be controlled by the solvent decomposition of the electrolyte. Andersson et al. [124] disclosed
the thermal stability of LiPF6, LiBF4, LiCF3SO3, and LiN(SO2CF3)2 in EC/DMC in the
following order: LiPF6 < LiBF4 < LiCF3SO3 < LiN(SO2CF3)2 . The main components of the
SEI in LiCF3SO3, and LiN(SO2CF3)2 salts containing electrolyte are related to lithium alkyl
36


carbonates and polymers while the main components of LiPF6 and LiBF4 salts are related to
lithium salts. Nie et al. [125] reported that SEI components are salt dependent, indicating that
Li salts are decomposed and involved in the formation of the SEI layers.
Influence of additives
It is generally accepted that SEI layers originate from the decomposition of the electrolyte
solvents and salts. SEI has a strong impact on the battery performance. The use of electrolyte
additives is believed to be an efficient method to improve the quality of the SEI layers.
Extensive studies have been carried out to investigate the influence of the electrolyte additives.
Vinylene carbonate (VC) is one of the most commonly used electrolyte additives in Li-ion
batteries.
Aurbach et al. [126] studied the impact of VC on the electrode cycling performance by
using cyclic voltammetry (CV), impedance spectroscopy, Electrochemical Quartz Crystal
Microbalance (EQCM) FTIR and XPS. They found that VC is reactive on graphite electrodes
and forms a flexible and cohesive polymeric surface species that suppresses both solvent and
salt anion reduction. The authors reported that the cyclability of the graphite electrode has been
improved by VC additives especially at elevated temperatures. Ota et al. examined the influence
of VC on the SEI formation at metallic Li [127] and graphite electrodes [128]. It was found that
cells containing VC have excellent performance at elevated temperatures while the performance
decreases at low temperatures. The presence of VC reduces the emission of reductive gases
such as C2H4, CH4, and CO from EC-containing electrolyte. Another conclusion from their
work is that VC improves the thermal stability of the SEI layers on graphite electrodes. The role
of VC on cathode electrodes has also been investigated. El Ouatani et al. [129] investigated the
influence of VC on the electrode/electrolyte interface in C/LiCoO2, C/LiFePO4 and
Li4Ti5O12/LiCoO2 systems and concluded that the same VC polymer was deposited on both
electrodes but with different mechanisms.
Some other less studied additives are trimethoxyboroxine (TMOBX) and LiN(CF3SO2)2.
Dahn et al. [130-134] studied TMOBX and LiN(CF3SO2)2 in LiCoO2 and NMC batteries and
concluded that these additives can increase the coulombic efficiency and thus extend the cycle
life of Li-ion batteries. Another interesting conclusion is that the additives also reduced the
charge end-point slippage, suggesting that the oxidation of the electrolyte at the positive
electrode has been depressed.
37


Influence of temperature
The influence of the temperature on the SEI formation and the SEI stability will be discussed
in this section. Generally, elevated temperatures enhance the kinetics of all (electro)chemical
processes inside Li-ion batteries, including side reactions [10]. Andersson et al. [135]
investigated the SEI formation at elevated temperatures and suggested that the thickness of the
SEI layer is much larger than at moderate temperatures, which is consistent with the trend of
capacity fade. Furthermore, the authors observed changes in morphology and compositions at
elevated temperatures. They disclosed that organic Li salts (ROCO2Li) decompose and dissolve
into the electrolyte at elevated temperatures. In contrast, the content of inorganic Li salts
increased [136]. Park et al. [137, 138] also observed the degradation of the SEI layers at elevated
temperatures, however, they argued that the damaged SEI can be repaired by electrolyte
decomposition and concomitant film deposition.
Wang et al. [139] revealed that the SEI film formed from PC solvent is stable at 0oC, but
becomes soluble when temperature increases to 30oC. However, in conventional electrolyte
systems, a moderate temperature (30oC) will promote the formation of a uniform SEI layer on
the graphite electrode [140] while at low temperature the lithium plating can occur [10].
2.2.3 Experimental characterization of SEI
The SEI is a mixture of several components. The stability of the SEI determines Li-ion
batteries longevity, electrochemical performance and safety. Therefore, extensive efforts have
been carried out to understand the chemical compositions and physical structures of the SEI
layers.
There are several challenges for analyzing the chemical compositions of the pristine SEI.
Firstly, the SEI is a mixture of various components with amorphous structures. Secondly, the
SEI is a very thin layer adhering to the active material surface, on nano-scale, which makes it
virtually impossible to establish the boundary between the SEI and the electrode (electrolyte).
Thirdly, the SEI layer has a loose outer layer structure which can be easily removed together
with the soluble components after rinsing. So, it is difficult and almost impossible to perform
precise compositional analyses. When analyzing the SEI along with the electrolyte there is
always an uncertainty which component actually belongs to the SEI and which one comes from
the electrolyte. Fourthly, most of the SEI components are highly sensitive to contamination, air,
and humidity. It is difficult to preserve its pristine nature during sample transfer. Therefore, for
38


ex situ analyses a transfer or encapsulating protection is required, which allows the sample to
be transported with an inert atmosphere without exposure to air and humidity.
2.2.3.1 Composition
A large variety of techniques have been used to analyze SEI compositions. These techniques
and corresponding results are summarized below.
Characterization techniques
1. FTIR (Fourier transform infrared) spectroscopy analysis

FTIR spectroscopy is a very conventional but sensitive tool for the analysis of the
chemistry of sample surfaces. It provides a beam with wavelength in the range of 1 μm
to 100 μm during the measurements. This makes the FTIR measurements harmless for
the properties of the materials. FTIR can identify various functional groups based on the
vibrational energy of bonds generated from dipole moments [141].
2. XPS (X-ray photoelectron spectroscopy) analysis
XPS is a highly efficient method for studying the surface chemistry of electrode
materials because of its high surface sensitivity (< 30Å). Furthermore, it can detect
information of species that are present in small amounts (>1%). Coupled with depth
profiling, XPS provides the possibility to get more insights into the bulk compositions
of the materials [142].
3. XES (X-ray emission) and XAS (X-ray absorption) spectroscopy analysis
Soft XES and XAS spectroscopies are traditionally used to study occupied and
unoccupied states of atoms, molecules, up to complex materials. These methods are
based on photon-in and photon-out and are therefore not limited to electrically
conducting systems, indicating that the charging problems often faced in XPS analysis
can be avoided. Both XES and XAS have atomic selectivity and chemical sensitivity
[143].
4. Other techniques such as Raman spectroscopy [45], NMR (Nuclear Magnetic
Resonance) [144] and EDS (Energy Dispersive Spectrometry) [145] etc.
Typical characterization results
FTIR and XPS are the most commonly used techniques for the SEI analyses. The results
reported in literature are summarized in Table 2.2 and 2.3.
39


Table 2.2. FTIR data of the SEI components from Aurbach et al. [109, 146, 147].
Component Functional group Vibration (cm-1)

Li2CO3 C̶O 1465
CO32- 869
CH3OCO2Li C̶H 3032 - 2863
C̶O 1009
C=O 1352
CO3 bend 830
ROCO2Li C̶H 2950 - 2820
C̶O 1077
C=O 1339, 1647
CH2 1416, 1458
CO3 bend 831
(CH2OCO2Li)2 C̶H 2957
C̶O 1083
C=O 1654, 1301
CH2 1400
CO3 bend 822
LiOCH3 C̶O 1077
C̶H 2930 - 2780
Table 2.3. XPS data of the SEI components.
Compound Binding energy / eV

C 1s O 1s F 1s P 1s Li 1s
284.3 [148]
Graphite
284.5 [149, 150]
686.5 [148] 55.5 [148]
LiF 684.9 [149, 150] 55.8 [149]
56.4 [150]
290 [148] 532.1 [149] 56.5 [148]
287 [142] 531.5 [148]
Li2CO3
289 [150] 532.0 [142]
533.6 [150]
LiOH 531.9 [148]
528.7 [148]
Li2O
528.3 [149]
R-CH2OLi 288.2 [148] 532.5 [148]
288.0 [148] 534.0 [148]
RCH2OCO2Li
285.1 [150]
CH2CH2 285.5 [148]
PEO 286.2 [148] 532.8 [148]
40


2.2.3.2 SEI thickness measurements
XPS depth profiling is a typical approach to measure the SEI thickness. The sample is etched
by Ar+ ion sputtering at the surface and then subsequently analyzed by XPS. By tuning the X-
ray energy, the escape depth of the photoelectrons can be changed and, hence, also the sampling
depth [142]. The SEI thickness can be determined by the sputtering time and sputtering rate
after approaching the bare graphite electrode surface [12].
Apart from XPS, Focused Ion Beam (FIB) and Atomic Force Microscopy (AFM) are also
used to analyze the thickness of the SEI layers. The values of the SEI thickness vary with respect
to various detecting approach, which of course also depend on the intrinsic properties of the
graphite electrode. Applying XPS, Niehoff et al. [149] reported the total SEI thickness
(including inorganic and organic layers) is ~ 2nm, while Edstrom et al. [142] reported that the
inner SEI layer is around 1.5-2 nm and the outer layer is approximately 90 nm. The value
estimated by Lee et al. [151] is even larger, around 140 ~ 220 nm. Wang et al. [145] estimated
the thickness of the SEI layers on highly oriented pyrolytic graphite (HOPG) to be 200 nm by
AFM. Zhang et al. [152] measured the SEI thickness on HOPG using FIB and reported the
thickness of the SEI layers to be in the range of 450 to 980 nm, which is much larger than other
researchers reported.
2.2.4 SEI formation models
Fig. 2.5 schematically shows a summary of the development of the knowledge and models
on SEI formation on the negative electrode during the last decades. The earliest investigation
of the passivated film on metallic Li dates back to the early 1970s, though the concept of “SEI”
was firstly proposed by Peled in 1979 [87] and used to describe the passivation film formed on
the metallic Li electrode. The decomposition of the PC solvent on the Li electrode was described
by Dey et al. [88] with a proposed two-electron solvent reduction mechanism. The first
mathematical simulations of passivation films associated with the metallic Li electrode in
aqueous electrolyte was reported in 1976 by Bennion et al. [153]. Afterwards an electron
tunneling-based mathematical model was proposed by Peled to describe the SEI formation on
the Li electrode in non-aqueous electrolyte [87]. Peled concluded that the SEI grows as a
function of t1/2. This conclusion has been widely accepted to estimate the battery capacity loss
caused by SEI formation. However, the model in that stage did not consider information about
the composition of the SEI layers, which indeed plays an important role in determining the SEI
formation mechanisms [4].
41


Fig. 2.5. Development of the SEI formation knowledge and SEI models on negative electrode during
the past five decades [4]. The left side representing the case of graphite and the right side representing
the case of metallic Li.
Nazri et al. [154] investigated SEI layers on the Li electrode by in-situ X-ray diffraction and
reported the existence of Li2CO3 and polymeric compounds in SEI. On the basis of FTIR and
XPS analyses, Aurbach et al. [103] demonstrated that lithium alkyl carbonates are the main SEI
components. Instead of the two-electron mechanism, Aurbach proposed a one-electron
mechanism where a meta-stable radical intermediate is assumed to be generated after the
solvent accepting one electron. The radical termination reaction leads to the formation of Li
alkyl oxides. Li2CO3 can be formed by further reducing the radical intermediate. A classical
SEI morphological model was proposed in 1997 by Peled et al. [155] after summarizing
42


previous findings. According to this model SEI is composed of multiple inorganic and organic
products resulting from electrolyte decomposition, including Li2O, Li2CO3, LiF, polyolefins
and semicarbonates. Based on the FTIR, XPS, EDS, XRD, SEM, AFM and EIS analysis,
Aurbach et al. [96] concluded that the SEI comprised of compact and porous layers, which can
be further identified as a couple of sublayers. These sublayers have been described by a group
of equivalent circuit elements.
After Dahn’s seminal work on graphite cycling investigation in EC-based electrolytes, the
graphite electrode has become the most popular anode material in commercial Li-ion batteries.
Consequently the SEI formation on the graphite electrode has attracted considerable attention
from then on. On the basis of the knowledge about graphite structure and the corresponding
intercalation properties, Besenhard et al. [156] proposed a SEI formation model that involves
the initial formation of a ternary GIC [Li(solvent)xCy] and its subsequent decomposition near
the edge sites of the graphite layers (see Fig. 2.5). Later Chu et al. [112] and Yamaguchi et al.
[111] revealed that the SEI can also be formed on the basal surface of the graphite electrode.
Vetter et al. [10] summarized the knowledge of the SEI formation mechanisms of the
graphite electrode and proposed a complete model, including all processes that may be involved
in the SEI formation such as graphite exfoliation and cracking, SEI growth and dissolution,
transition metal ions deposition and Li plating, etc. Based on this model researchers can have a
complete picture of the battery degradation related to the anode. After extensive XPS analyses,
Edström et al. [157] proposed a model based on the SEI compositions. The authors argued that
Li2O cannot be a component of the pristine inorganic SEI layer, Li2CO3 is not always observed
as well but only LiF is frequently observed during battery cycling.
The final purpose of the quantitative investigation of the SEI formation is that in-depth
understanding of battery aging has been obtained and that the battery performance has been
improved effectively. Many mathematical simulations have been carried out on the basis of
various proposed mechanisms. A number of these classical modeling papers will be presented
below.
Christensen and Newman [158] developed a continuum-scale mathematical model to
simulate the growth of the SEI and transport of Li+ ions and electrons through the surface films.
They attempted to combine the relevant mechanisms for the growth in a more general oxide
growth model that includes transport of cationic and anionic vacancies and interstitials as well
as electrons through the film but with explicit emphasis on Li-ion systems with graphic negative
electrode. Ploehn et al. [63] proposed a continuum mechanic model with the assumption that a
reactive solvent component diffuses through the SEI and undergoes two-electron reduction at
43


the graphite surface. In line with the conclusion obtained from Peled, this model also predicts
that the SEI thickness increases as the square root of time. Pinson and Bazant [67] developed a
single-particle model which was further extended to a porous electrode model to describe the
SEI formation and battery fading mechanisms. Li et al. [12, 65] further developed the electron
tunneling model proposed by Peled by considering the specific structure properties of the SEI
layers. The cycling-induced effects and the catalyst effects from the deposited metallic clusters
are discussed in detail.
2.3 Cathode electrode decay
2.3.1 LiFePO4 electrode
2.3.1.1 Structure of LiFePO4
Fig. 2.6. Crystal structure of olivine LiFePO4 [168].
LiFePO4 belongs to a general class of ‘polyanion’ compounds and has an olivine structure.
The structure of LiFePO4 as a cathode material has been firstly discussed in details by Padhi et
al. in 1997 [159]. The LiFePO4 olivine structure has metal atoms in half of the octahedral sites
and P atoms in one-eighth of the tetrahedral sites of a hexagonal close-packed oxygen array.
Fig. 2.6 shows the crystal structure of LiFePO4. The M1 (Li) sites have a ̅ symmetry, the M2
(Fe) octahedron has mirror symmetry with average M-O bond length, which is larger than that
in the M1 octahedron. The M1 (Li) sites form linear chains by sharing an edge of the octahedron,
running parallel to the c-axis in the alternate a-c planes. The M2 (Fe) sites form zigzag planes
44


by sharing a corner of the octahedron parallel to the c-axis in the other a-c planes. Since Li is
in the continuous chain of edge-shared octahedron on alternate a-c planes, a reversible
extraction/insertion of lithium from/into these chains can take place through a first-order phase
transition [160] between FePO4/LiFePO4 [62, 161-167].
Both LiFePO4 and FePO4 have the same space group, namely, Pmnb. The lattice parameters
of LiFePO4 and FePO4 are listed in Table 2.4. On electrochemical extraction of Li from
LiFePO4 as indicated by Eq. 1.14, there is a volume decrease of 6.81% and density increase of
2.59% [169]. This small volume change will not cause structural damage during charging and
discharging. The structure of LiFePO4 is highly robust due to the strong bonding between the P
and O atoms.
Table 2.4. Lattice parameters of LiFePO4 and delithiated FePO4 [169].

LiFePO4 FePO4
a (Å) 6.008(3) 5.792(1)
b (Å) 10.334(4) 9.821(1)
c (Å) 4.693(1) 4.788(1)
Volume (Å ) 291.392 272.357(1)
2.3.1.2 Electrochemistry of LiFePO4
Fig. 2.7. Schematic representation of Li (de)intercalation of a LiFePO4 particle during charging and
discharging. Reconstructed based on ref [163].
45


The theoretical storage capacity of a LiFePO4 electrode is calculated to be 169.7 mAh/g.

Unlike LiCoO2 material, all Li+ ions in the LiFePO4 electrode can be reversibly extracted. The
lithiation and delithiation processes are schematically represented in Fig. 2.7.
When discharge starts (indicated by the black arrows), Li+ ions are transported from the
anode to intercalate into the FePO4 material to form a Li-deficient phase . After a critical
concentration ( ) is reached, a shell of Li-rich phase with a Li concentration of is formed,
covering the Li-deficient phase. As discharge proceeds, more phase is converted into the
phase, resulting in the core shrinking. This process continues until the core is completely
converted and the material becomes a single Li-rich phase. The Li concentration increases from
to 1 when all available sites are filled with lithium when the material reaches the fully
discharged state.
When the completely discharged electrode is charged (indicated by the red arrows), the Li+
ions delithiate from LiFePO4 material. The Li-rich phase is firstly formed followed by the
formation of Li-deficient shell. The Li-rich core will be completely consumed when Li
concentration reaches . After the charging process a completed, all Li+ ions are extracted and
LiFePO4 has been completely converted into FePO4 again. It should be pointed out that the
processes discussed above are considered to be in equilibrium state.
Fig. 2.8. Charge and discharge voltage curves of a LiFePO4 electrode [163].
The electrode potential of a LiFePO4 electrode as a function of the Li concentration is almost

constant in the two-phase ( ) region. The voltage plateau in this region was reported to be
3.45V [159, 169]. However, there is quite some spread in the reported plateau width (
46


). Based on the shrinking-core model, Srinivasan and Newman [163] reported that the plateau
is in the range of 0.02 0.9525, while Gaberscek et al. demonstrated the two phase region
to be in the range of 0.05 0.89 [162]. The voltage curve of LiFePO4 electrode is shown
by Fig. 2.8.
From Fig. 2.8 it can be seen that the charge and discharge curves do not overlap even at
very low currents (see the solid curves). This hysteresis effect can be explained by the shrinking-
core model developed by Srinivasan and Newman [163]. The electrochemical charge transfer
reaction takes place at the Li-deficient phase during charging and at the Li-rich phase
during discharging. Therefore one can expect that kinetic parameters governing the charge-
transfer reactions during charging may be different from those on discharge.
It is well known that LiFePO4 is a low-electronic-conductive material [168, 170-172]. As
can be concluded from the crystal structure of LiFePO4, Li ions can only be transported in the
1D channels along the axis. However, the ions can be easily blocked by ionic disorder, foreign
phases or stacking faults, which also reduces the ionic conductivity [169]. Both electronic and
ionic conductivities can be improved by conductive coatings [164, 173-176], heterogeneous
doping [164, 168], nanocrystallization [177, 178] and antisite defects, etc. [179].
2.3.1.3 LiFePO4 electrode degradation

Fig. 2.9. Cyclic voltammetry (CV) of a LiFePO4 electrode at various temperatures [180].
Low temperatures will decrease the kinetics of both the charge transfer reaction at the
electrode/electrolyte interface and the rate of Li diffusion within the bulk of cathode materials
[169]. Fig. 2.9 shows the cyclic voltammetry (CV) measurements of a LiFePO4 electrode at
47


various temperatures [180]. The sluggishness of the Li (de)intercalation process at -20oC has
been attributed to the decrease of the Li diffusion capability in the bulk of LiFePO4 materials.
Obviously, the total capacity of the cathode material cannot be completely used due to the
kinetic limitations.
The capacity of the LiFePO4 electrode increases with increasing temperature since more
electrode materials can be utilized at high temperatures [181]. However, cathode dissolution,
as a side reaction, has also been observed at the elevated temperatures [20, 182-187].
Amine et al. [20] investigated the LFP batteries under both storage and cycling conditions
at high temperatures. At the end of the aging experiments, over 640 ppm and 535 ppm of Fe
ions have been detected in the LiFePO4 and C-LiFePO4 electrolyte solutions, respectively, by
ICP measurements. Furthermore, iron deposition on the graphite electrode has also been
confirmed by EDAX analysis. Similar investigation has been described in [184], where the Fe
deposition on graphite was confirmed by XPS. Interestingly, iron deposition on the separator
was also observed when batteries were aged at high temperatures [186].
Fig. 2.10. Layout of the dissolution mechanism of the LiFePO4 electrode.
It is generally accepted that the LiFePO4 electrode dissolution is related to the H+

contamination in the electrolyte [182, 188]. Residual water inside the electrolyte is considered
to be the origin of H+, according to
LiPF6 
 LiF
elevated T
H2O POF3 2H+ 2F ̶ . [2.19]
H+ ions can react with the LiFePO4 electrode by ion exchange reactions (2 ↔ ). The
produced LiH2PO4 will be further dissociated and release 2H+ ions. Therefore, the content of
H+ ions is constant. As schematically illustrated in Fig. 2.10, the dissolved iron ions can be
transported to the anode and be reduced at the graphite surface. Thereby, cathode dissolution
will not be influenced by the accumulation of Fe2+ ions in the electrolyte since the transition
48


metal ions are continuously removed from the solution due to their reaction on the anode side
[182].
2.3.2 LiCo1/3Ni1/3Mn1/3O2 (NMC)
Ohzuku et al. [189] was the first one who identified a novel Li insertion material with the
composition LiCo1/3Ni1/3Mn1/3O2, also commonly named stoichiometric NMC, NMC(111) or
simply NMC [190], by using a computational approach [191]. The NMC(111) cathode, similar
to LiCoO2, LiNiO2, LiCrO2, LiMnO2, etc. [192], has a layered Li MO2 type of structure, with
a space group of 3 and lattice parameters of a~2.8 Å and c~14 Å [193, 194]. As shown in
Fig. 2.11a the rhombohedral unit cell has two axes of equal length, and that each angle between
any two axes is the same. The transition metal ions Ni, Mn, Co (indicated in Fig. 2.11a as M in
random substitution) are located at the center of the oxygen octahedra. These MO6 octahedra
are edge-shared, forming a vertical slab parallel to the (001) planes. The Li+ ions are located
between the MO6 slabs.
Fig. 2.11. (a) The crystal structure of LixNiyCozMn(1-y-z)O2 (NMC) cathode material. The blue lines
indicate the unit cell. The Ni, Co, Mn atoms are randomly distributed on the M sites. (b) Change of
lattice parameters and the unit cell volume of the NMC cathode phase during charge and discharge [193].
2.3.2.1 Electrochemistry of NMC electrode
The de-insertion of 1 Li+ in NMC electrode provides a theoretical specific capacity of 276
mAh·g-1. The electrode capacity of NMC below 4.3 V is 160 mAh/g, even higher than that of
49


the LiCoO2 electrode (140 mAh/g). The oxidation states of Ni, Co, and Mn in the as-produced
NMC electrode have been identified to be +2, +3, and +4, respectively, on the basis of both
theoretical [195, 196] and experimental studies [196]. First-principle calculations suggest that
only Ni2+ and Co3+ are involved in the redox reactions [196]. Since the relative Fermi level of
the Ni2+/3+ and Ni3+/4+ is higher than that of Co3+/4+ [191, 195-197], it can be concluded that Ni
is oxidized first followed by the oxidation of Co at higher potentials. Ni2+ will be ultimately
oxidized to Ni4+ after extraction of the first two-thirds of the lithium during electrochemical
charging in lithium cells, and oxidation of Co3+ to Co4+ occurs only during extracting of the last
one-third [196]. As in LiNi0.5Mn0.5O2, Mn remains in its initial oxidation state of +4 throughout
normal operating cell voltages.
The most significant structural change in NMC during cycling is the lattice dilation. The
changes of lattice parameters for NMC has been detected in-situ by Wang et al. [193]. The
results are presented in Fig. 211b. It has been concluded that, during charging, c increases
whereas a decreases. These results can be qualitatively understood in terms of charge
compensation. Taking LixNiO2 as an example, and removing Li ions from the lattice, the ionic
radius of Ni decreases with increasing valence state of Ni ions (Ni2+ 0.69 Å, Ni3+ 0.56 Å, Ni4+
0.48 Å). As a result, the edge-shared NiO6 slabs shrink along the a-axis. Meanwhile, NiO6 slabs
become more positively charged and expel each other along the c-axis, resulting in an expansion
along the c-axis.
2.3.2.2 NMC electrode degradation
Transition metal dissolution
Metal dissolution is generally accepted to be a common phenomenon of transition-metal-

oxides-based cathode materials, such as LiMn2O4, NMC, etc., in acidic solutions (LiPF6-based
electrolyte) at high temperatures [21, 46, 198-203]. Several detrimental effects can be induced
by the metal dissolution process, including (i) capacity fading of the positive electrode, (ii)
metal reduction at the anode surface and the subsequent blockage of the graphite layers, (iii)
damage of the SEI layers leading to higher battery capacity losses. The details of the dissolution
mechanism in NMC positive electrode are still under debate. Some researchers attribute the
disproportion reaction of Mn ions to be the main reason of metal dissolution [204-206] while
others consider the acidic corrosion to be the origin [198-203].
Metal dissolution in various cathode materials has been reported by Choi and Manthiram
[206], see Table 2.5. The dissolution measurements have been performed by soaking the sample
50


powders in the electrolyte (EC/DEC 1:1, LiPF6) at 55oC for 7 days, followed by analyzing the
amount of metal ions in the electrolyte with atomic absorption spectroscopy (AAS). Although
NMC material is considered to be more stable than LiCoO2 and LiNi0.5Mn0.5O2, considerable
dissolution has still been observed. From Table 2.5 it can be concluded that NMC(111) material
is relatively more stable than other materials of NMC families.
Table 2.5. Comparison of transition metal dissolution from various cathode materials [206].
Metal ions dissolution % (based on sample weight)

Sample number Composition
Mn Ni Co Fe Total
1 LiCoO2 0.8 0.8
2 LiNi0.5Mn0.5O2 0.4 0.7 1.1
3 LiNi0.425Mn0.425Co0.15O2 0.3 0.8 0 1.1
4 LiNi0.33Mn0.33Co0.33O2 0.2 0.4 0.3 0.9
5 LiNi0.29Mn0.29Co0.42O2 0.4 1.1 0.3 1.8
6 LiNi0.25Mn0.25Co0.5O2 0.4 0.9 0.5 1.8
7 LiNi0.21Mn0.21Co0.58O2 0.3 0.8 0.5 1.6
8 LiMn0.8Cr0.2O2 2.6 2.6
9 LiMnO2 3.2 3.2
10 LiMn2O4 3.2 3.2
11 LiMn1.5Ni0.5O4 0.3 0.3 0.6
12 Li1.05Mn1.53Ni0.42O4 0.2 0.1 0.3
13 LiMn1.5Ni0.42Zn0.08O4 0.4 0.3 0.7
14 LiMn1.42Ni0.42Co0.16O4 0.3 0.3 0.6
15 LiFePO4 0.5 0.5
Structural transformation
It is known that the NMC(111) material experiences a phase transition from the
rhombohedral space group 3 (initial “O3” phase) to the monoclinic space group C2/m (“O1”
phase) beyond a charge voltage of 4.4 V vs Li+/Li [23, 26, 191, 207]. The “O1”
LiNi1/3Co1/3Mn1/3O2 phase has been clearly observed at 0.3 [194]. Cycling above this phase
transition point at higher potentials will lead to a faster capacity decay of the cathode.
Structural change induced by Li-Ni site interchange is considered to be another detrimental
effect on the electrode cycling performance [23-26, 207-209]. Due to the similar ionic radius of
Ni2+ (0.67 Å) and Li+ (0.76 Å), there is always a possibility that these two ions exchange their
crystallographic sites, which induces local disorder in the NMC(111) materials. The un-
removable Ni ions in the Li layer will then block Li diffusion pathways, leading to a decrease
51


of the cathode rate capability. High currents [25] and voltages [23] are considered as
unfavorable, leading distortion of the electrode surface.
It is worthwhile to point out that the degradation mechanism is composition dependent, for
example, Jung et al. [23] reported that the degradation of NMC(532) (LiNi0.5Mn0.3Co0.2O2)
material is attributed to the phase transformation from rhombohedral to spinel at the surface
while the degradation of NMC(111) material is due to the phase transformation from the O3 to
the O1 phase. Therefore the degradation mechanisms of NMC materials should be carefully
studied with respect to the specific compositions.
Particle isolation
Based on the fact that the impedance of cathode electrodes dramatically increase after aging,
some researchers proposed that particle isolation should be responsible for the capacity and
power fades of cathode materials [36, 45, 210]. The increase of the impedance has been
attributed to a loss of the conductive carbon and/or SEI layers on the cathode surface.
2.4 Graphite (C6) electrode decay
2.4.1 Physical properties of graphite
Graphite is the most important anode material nowadays applied in commercial Li-ion
batteries due to its excellent cycling performance, high safety performance and considerable
specific capacity. Furthermore, the electrode potential [211] of the lithiated graphite is lower
than 200 mV vs Li+/Li.
The ideal graphite structure is shown in Fig. 2.12a. The graphite consists of a sequence of
graphene layers parallel to the (001)-oriented plane of hexagonally linked carbon atoms [212].
The planes which are perpendicular to the -axis are defined as basal surface and those parallel
to the -axis are called edge surface. The interatomic distance within a layer plane is 1.42 Å,
the distance between the interlayers is 3.35 Å. Three valence electrons of carbon form regular
covalent bonds ( bonds) with adjacent carbon atoms while the remaining electron resonates
between the valence bonds, forming a conjugated bond in the whole graphene layer.
Therefore, the carbon bonding within the graphene layer is extremely strong, leading to stable
chemical properties of the graphite electrode. However, the binding between the adjacent
graphene layers is much weaker. The force combining the adjacent graphene layers is attributed
to the van der Waals force. Due to its intrinsic structural characteristic, the graphite conductivity
52


along the basal plane is 2.7 104 S·cm-1 which is comparable to the conventional conductors
(~105 S·cm-1), while only 6~10 S·cm-1 along the -axis [213, 214].
Fig. 2.12. Crystal structure of graphite (a) and types of carbon atoms in graphite (b). Drawing are based
on [212].
The graphene layers are stacked by an ABAB arrangement. The definition of A-type and
B-type carbon atoms are shown in Fig. 2.12b [215]. 17-22% of ABCABC stacking sequence
can be observed in natural graphite [216]. In the so-formed state of the “artificial” or “synthetic”
graphite, only a few percent at best of the ABCABC arrangement can be found. However,
deformation processes such as grinding substantially increase the percentage of this stacking
sequence [216].
2.4.2 Li intercalation into graphite electrode
Li-graphite intercalation compounds (Li-GIC) have originally been discussed in the mid
1950s [217]. The maximum stoichiometry of Li-GIC is LiC6, which has a similar chemical
reactivity compared with metallic Li. Fig. 2.13 shows the Li intercalation model proposed by
Daumas and Hérold [218] in 1960. Their model indicates that several “stages” are involved
during Li intercalation into the graphite electrode. For convenience, the Li-graphite structure of
these “stages” are designated as “stage n”, where the stage index n refers to a single Li-
intercalated layer for every n graphene sheets [219].
53


Fig. 2.13. (a) The classical model and (b) the model proposed by Daumas and Hérold for Li intercalation
into the graphite electrode [218].
However, Li-GIC as anode electrode for Li-ion batteries encountered tremendous

difficulties until 1980s. It was found that most solvent molecules can be decomposed on the
graphite surface at a potential of ~ 0.8 V vs Li+/Li which is much higher than where Li
intercalation is initiated [89]. Furthermore, solvent co-intercalation has been found to be another
problem for the graphite electrode since the subsequent decomposition products of the co-
intercalated solvent can lead to exfoliation of the graphite layered structure [88]. The situation
was improved only after the seminal work from Dahn and co-workers [90], revealing the role
of the SEI on the reversibility of carbonaceous electrodes and the effect of EC. The SEI layers
formed by EC not only prevent the physical disintegration of graphite that occurs at 0.8 V, but
also support the reversible Li intercalation and de-intercalation at low potentials (<0.2 V) with
an electrode capacity almost approaching the theoretical value of LiC6, 372 mAh·g-1.
Under the protection of the SEI layers, the reversible Li intercalation processes can be
described as follows. Starting from the pure graphite, 4-7% of Li will be intercalated into the
space of each graphene layer to form a “liquid-like” diluted stage 1 (also called stage 1L)
structure [220]. After increasing the Li concentration, the Li-graphite structure transfers from
stage 1L into diluted stage 4 (also called stage 4L) structure, which is accompanied by a phase
transition. Transitions between stage 4L, the dilute stage 3 (stage 3L) and the dilute stage 2
(stage 2L) are still under debate. Stage 3L (LiC18) can be observed at a temperature below 248
54


K, however, LiC18 forms a liquid-like stage 2L at room temperature. A phase transition from
stage 2L to dense 2L (with Li composition of LiC12) will take place with further increasing Li
concentration. The final phase transition occurs between the dense stage 2 and the dense stage
1 (LiC6). After the complete Li intercalation, the space between the graphene layers inside
graphite increases from 3.35 to 3.7 Å [221].

Fig. 2.14. The electrode voltage curve of graphite electrode during Li intercalation.
Fig. 2.14 shows the electrode voltage evolution during Li intercalation into the graphite
electrode. The potential of the graphite electrode is determined by in Li C , where 0
1. Several voltage stages have been observed during the Li (de)intercalation processes due to
the different structures of at various SoC values [12, 65, 146, 219, 220].
2.4.3 Graphite electrode degradation
Although the graphite electrode is considered to be a stable anode material, structural

degradation has still been observed [221-224]. Kostecki et al. [223] suggested that a non-
uniform current distribution across and within the anode can lead to co-intercalation of ion
aggregates, generating local graphite degradation or exfoliation. The inhomogeneity of the
current density within the graphite electrode is more significant at elevated temperatures [223].
Furthermore, the cycling range was found to have a significant influence on the graphite
electrode decay [224]. Typical Raman spectra of graphite electrodes reveal that the structural
damage is more pronounced when cycling at lower SoC ranges, for example, at 0 0.16.
After semi-quantitatively analysis from the Raman spectra it was concluded that the average
55


particle size of the graphite electrode decreased [222]. Particle cracks and flake exfoliation have
been visually observed from the aged graphite electrodes by SEM [222].
The stress evolution caused by the repeated volume expansion and contraction of the
graphite electrode during cycling is believed to be the origin of the particle fracture and
structural damage. Mathematical models based on the electrode volume changes have been
proposed to quantitatively describe graphite degradation upon cycling. Among these models
the diffusion-induced-stress (DIS) model [219, 225, 226] was well accepted and widely applied
to analyze the graphite deformation. The model suggests that particle fractures take place due
to severe tangential stresses developed during Li extraction. However, the model proposed by
Christensen and Newman suggests that the particle surface is likely to fracture at the end of
extraction while the center is most likely to fracture at the beginning of Li insertion [227, 228].
Apart from the mechanical stresses induced by Li (de)intercalation, metal dissolution from
the cathode and the subsequent deposition on graphite surface is considered to be another factor
responsible for graphite degradation. Transition metal ions on the graphite surface have been
detected by Energy Disperse X-ray Spectroscopy (EDS) [20, 83] and X-ray photoelectron
spectroscopy (XPS) [182, 184]. These metal clusters covering on the graphite surface hinder
the Li intercalation process, leading to inaccessibility of the graphite electrode [12].
2.5 References
3647- 3667.
[2] S.C. Nagpure, B. Bhushan, S.S. Babu, Journal of the Electrochemical Society, 160 (2013)
A2111-A2154.
[3] A. Barre, B. Deguilhem, S. Grolleau, M. Gerard, F. Suard, D. Riu, Journal of Power
Sources, 241 (2013) 680-689.
[4] M. Gauthier, T.J. Carney, A. Grimaud, L. Giordano, N. Pour, H.H. Chang, D.P. Fenning,
S.F. Lux, O. Paschos, C. Bauer, F. Magia, S. Lupart, P. Lamp, Y. Shao-Horn, Journal of
Physical Chemistry Letters, 6 (2015) 4653-4672.
[5] G. Sarre, P. Blanchard, M. Broussely, Journal of Power Sources, 127 (2004) 65-71.
[6] M. Broussely, P. Biensan, F. Bonhomme, P. Blanchard, S. Herreyre, K. Nechev, R.J.
Staniewicz, Journal of Power Sources, 146 (2005) 90-96.
[7] D. Guyomard, J.M. Tarascon, Solid State Ionics, 69 (1994) 222-237.
[8] R. Yazami, Y.F. Reynier, Electrochimica Acta, 47 (2002) 1217-1223.
56


112 (2002) 98-104.
[10] J. Vetter, P. Novak, M.R. Wagner, C. Veit, K.C. Moller, J.O. Besenhard, M. Winter, M.
Wohlfahrt-Mehrens, C. Vogler, A. Hammouche, Journal of Power Sources, 147 (2005)
269-281.
[11] M. Broussely, S. Herreyre, P. Biensan, P. Kasztejna, K. Nechev, R.J. Staniewicz, Journal
of Power Sources, 97-8 (2001) 13-21.
Acta, 190 (2016) 1124-1133.
[13] M. Kassem, C. Delacourt, Journal of Power Sources, 235 (2013) 159-171.
[14] M. Safari, C. Delacourt, Journal of the Electrochemical Society, 158 (2011) A1123-
A1135.
[15] S. Grolleau, A. Delaille, H. Gualous, P. Gyan, R. Revel, J. Bernard, E. Redondo-Iglesias,
J. Peter, S. Network, Journal of Power Sources, 255 (2014) 450-458.
[16] T. Yoshida, M. Takahashi, S. Morikawa, C. Ihara, H. Katsukawa, T. Shiratsuchi, J.
Yamaki, Journal of the Electrochemical Society, 153 (2006) A576-A582.
[17] G. Kwak, H.Q. Wang, K.H. Choi, K.H. Song, S.H. Kim, H.S. Kim, S.J. Lee, H.Y. Cho,
E.J. Yu, H.J. Lee, E.J. Park, L.S. Park, Macromolecular Rapid Communications, 28 (2007)
1317-1324.
[18] C.H. Huang, K.L. Huang, S.Q. Liu, Y.Q. Zeng, L.Q. Chen, Electrochimica Acta, 54
(2009) 4783-4788.
[19] J. Li, J. Zhang, X.G. Zhang, C.Z. Yang, N.X. Xu, B.J. Xia, Electrochimica Acta, 55
(2010) 927-934.
[20] K. Amine, J. Liu, I. Belharouak, Electrochemistry Communications, 7 (2005) 669-673.
[21] K. Amine, Z.H. Chen, Z. Zhang, J. Liu, W.Q. Lu, Y. Qin, J. Lu, L. Curtis, Y.K. Sun,
Journal of Materials Chemistry, 21 (2011) 17754-17759.
[22] N. Yabuuchi, K. Yoshii, S.T. Myung, I. Nakai, S. Komaba, Journal of the American
Chemical Society, 133 (2011) 4404-4419.
[24] D. Mohanty, H. Gabrisch, Journal of Power Sources, 220 (2012) 405-412.
[25] Y.B. He, F. Ning, Q.H. Yang, Q.S. Song, B.H. Li, F.Y. Su, H.D. Du, Z.Y. Tang, F.Y.
Kang, Journal of Power Sources, 196 (2011) 10322-10327.
[26] H. Gabrisch, R. Yazami, Electrochemical and Solid State Letters, 13 (2010) A88-A90.
57


[27] A.T. Stamps, C.E. Holland, R.E. White, E.P. Gatzke, Journal of Power Sources, 150
(2005) 229-239.
[28] A. Eddahech, O. Briat, H. Henry, J.Y. Deletage, E. Woirgard, J.M. Vinassa,
Microelectronics Reliability, 51 (2011) 1968-1971.
[29] Z. Li, L.G. Lu, M.G. Ouyang, Y.K. Xiao, Journal of Power Sources, 196 (2011)
9757- 9766.
[30] M. Dubarry, C. Truchot, B.Y. Liaw, Journal of Power Sources, 219 (2012) 204-216.
[31] D. Ansean, M. Gonzalez, J.C. Viera, V.M. Garcia, C. Blanco, M. Valledor, Journal of
Power Sources, 239 (2013) 9-15.
[32] T.G. Zavalis, M. Klett, M.H. Kjell, M. Behm, R.W. Lindstrom, G. Lindbergh,
Electrochimica Acta, 110 (2013) 335-348.
[33] J.C. Burns, A. Kassam, N.N. Sinha, L.E. Downie, L. Solnickova, B.M. Way, J.R. Dahn,
Journal of the Electrochemical Society, 160 (2013) A1451-A1456.
[34] A.J. Smith, J.C. Burns, J.R. Dahn, Electrochemical and Solid State Letters, 14 (2011)
A39 - A41.
[35] A.J. Smith, H.M. Dahn, J.C. Burns, J.R. Dahn, Journal of the Electrochemical Society,
159 (2012) A705-A710.
[36] M. Kerlau, J.A. Reimer, E.J. Cairns, Electrochemistry Communications, 7 (2005)
1249 - 1251.
[37] I. Bloom, B.W. Cole, J.J. Sohn, S.A. Jones, E.G. Polzin, V.S. Battaglia, G.L. Henriksen,
C. Motloch, R. Richardson, T. Unkelhaeuser, D. Ingersoll, H.L. Case, Journal of Power
Sources, 101 (2001) 238-247.
[38] M. Dubarry, B.Y. Liaw, Journal of Power Sources, 194 (2009) 541-549.
[39] D. Aurbach, B. Markovsky, A. Rodkin, E. Levi, Y.S. Cohen, H.J. Kim, M. Schmidt,
[40] A.J. Smith, J.C. Burns, D. Xiong, J.R. Dahn, Journal of the Electrochemical Society, 158
(2011) A1136-A1142.
[41] B. Markovsky, A. Rodkin, Y.S. Cohen, O. Palchik, E. Levi, D. Aurbach, H.J. Kim, M.
Schmidt, Journal of Power Sources, 119 (2003) 504-510.
[42] R.B. Wright, J.P. Christophersen, C.G. Motloch, J.R. Belt, C.D. Ho, V.S. Battaglia, J.A.
Barnes, T.Q. Duong, R.A. Sutula, Journal of Power Sources, 119 (2003) 865-869.
[43] Y.C. Zhang, C.Y. Wang, X.D. Tang, Journal of Power Sources, 196 (2011) 1513-1520.
[44] B. Ziv, V. Borgel, D. Aurbach, J.H. Kim, X.C. Xiao, B.R. Powell, Journal of the
58


[45] J. Shim, R. Kostecki, T. Richardson, X. Song, K.A. Striebel, Journal of Power Sources,
112 (2002) 222-230.
[46] Y. Li, M. Bettge, B. Polzin, Y. Zhu, M. Balasubramanian, D.P. Abraham, Journal of the
[47] J. Li, E. Murphy, J. Winnick, P.A. Kohl, Journal of Power Sources, 102 (2001) 294-301.
[48] K. Maher, R. Yazami, Journal of Power Sources, 247 (2014) 527-533.
[49] R. Premanand, A. Durairajan, B. Haran, R. White, B. Popov, Journal of the
[50] G. Ning, B. Haran, B.N. Popov, Journal of Power Sources, 117 (2003) 160-169.
[51] Y.H. Ye, Y.X. Shi, A.A.O. Tay, Journal of Power Sources, 217 (2012) 509-518.
[52] J.H. Kim, S.C. Woo, M.S. Park, K.J. Kim, T. Yim, J.S. Kim, Y.J. Kim, Journal of Power
Sources, 229 (2013) 190-197.
[53] A. Barre, F. Suard, M. Gerard, M. Montaru, D. Riu, Journal of Power Sources, 245 (2014)
846-856.
[54] W. Bogel, J.P. Buchel, H. Katz, Journal of Power Sources, 72 (1998) 37-42.
[55] K.C. Chiu, C.H. Lin, S.F. Yeh, Y.H. Lin, C.S. Huang, K.C. Chen, Journal of Power
Sources, 263 (2014) 75-84.
[56] G. Ning, R.E. White, B.N. Popov, Electrochimica Acta, 51 (2006) 2012-2022.
[57] P. Ramadass, B. Haran, P.M. Gomadam, R. White, B.N. Popov, Journal of the
[58] V.A. Agubra, J.W. Fergus, R.J. Fu, S.Y. Choe, Journal of Power Sources, 270 (2014)
213- 220.
[59] X. Lin, J. Park, L. Liu, Y. Lee, A.M. Sastry, W. Lu, Journal of the Electrochemical
Society, 160 (2013) A1701-A1710.
[60] J. Christensen, J. Newman, Journal of The Electrochemical Society, 152 (2005) A818.
[61] Y. Xie, J. Li, C. Yuan, Journal of Power Sources, 248 (2014) 172-179.
[62] M. Safari, C. Delacourt, Journal of the Electrochemical Society, 158 (2011) A63-A73.
[63] H.J. Ploehn, P. Ramadass, R.E. White, Journal of the Electrochemical Society, 151 (2004)
A456-A462.
[64] D. Li, D. Danilov, Z. Zhang, H. Chen, Y. Yang, P.H.L. Notten, ECS Transactions, 62
(2014) 8.
[65] D.J. Li, D. Danilov, Z.R. Zhang, H.X. Chen, Y. Yang, P.H.L. Notten, Journal of the
59


[66] H. Ekstrom, G. Lindbergh, Journal of the Electrochemical Society, 162 (2015)

A1003 - A1007.
[67] M.B. Pinson, M.Z. Bazant, Journal of the Electrochemical Society, 160 (2013) A243-
A250.
A1447.
[69] B.Y. Liaw, R.G. Jungst, G. Nagasubramanian, H.L. Case, D.H. Doughty, Journal of
Power Sources, 140 (2005) 157-161.
[70] M. Safari, M. Morcrette, A. Teyssot, C. Delacourt, Journal of the Electrochemical Society,
157 (2010) A892-A898.
[71] M. Safari, M. Morcrette, A. Teyssot, C. Delacourt, Journal of the Electrochemical Society,
157 (2010) A713-A720.
[72] W.J. Gu, Z.C. Sun, X.Z. Wei, H.F. Dai, Electrochimica Acta, 133 (2014) 107-116.
[73] K. Takei, K. Kumai, Y. Kobayashi, H. Miyashiro, N. Terada, T. Iwahori, T. Tanaka,
Journal of Power Sources, 97-8 (2001) 697-701.
[74] A.P. Schmidt, M. Bitzer, A.W. Imnre, L. Guzzella, Journal of Power Sources, 195 (2010)
7634-7638.
[75] L. Liu, J. Park, X.K. Lin, A.M. Sastry, W. Lu, Journal of Power Sources, 268 (2014)
482- 490.
[76] S. Ramesh, B. Krishnamurthy, Journal of the Electrochemical Society, 162 (2015)
A545- A552.
[77] J. Guo, Z.J. Li, M. Pecht, Journal of Power Sources, 281 (2015) 173-184.
[78] D. Aurbach, B. Markovsky, A. Rodkin, M. Cojocaru, E. Levi, H.J. Kim, Electrochimica
Acta, 47 (2002) 1899-1911.
[79] D.E. Demirocak, B. Bhushan, Journal of Power Sources, 280 (2015) 256-262.
[80] L. Castro, R. Dedryvere, J.B. Ledeuil, J. Breger, C. Tessier, D. Gonbeau, Journal of the
[81] R.J. Fu, S.Y. Choe, V. Agubra, J. Fergus, Journal of Power Sources, 261 (2014) 120-135.
[82] M. Klett, R. Eriksson, J. Groot, P. Svens, K.C. Hogstrom, R.W. Lindstrom, H. Berg, T.
Gustafson, G. Lindbergh, K. Edstrom, Journal of Power Sources, 257 (2014) 126-137.
[83] M. Klett, T.G. Zavalis, M.H. Kjell, R.W. Lindstrom, M. Behm, G. Lindbergh,
[84] V. Pop, H.J. Bergveld, P.P.L. Regtien, J.H.G.O.H. Veld, D. Danilov, P.H.L. Notten,
60


[85] D. Li, D. Danilov, L. Gao, Y. Yang, P.H.L. Notten, Electrochimica Acta, (2016) 11.
[86] D.J. Li, D.L. Danilov, J. Xie, L. Raijmakers, L. Gao, Y. Yang, P.H.L. Notten,
[87] E. Peled, Journal of the Electrochemical Society, 126 (1979) 2047-2051.
[88] A.N. Dey, B.P. Sullivan, Journal of the Electrochemical Society, 117 (1970) 222-&.
[89] K. Xu, Chemical Reviews, 104 (2004) 4303-4417.
[90] R. Fong, U. Vonsacken, J.R. Dahn, Journal of the Electrochemical Society, 137 (1990)
2009-2013.
[91] D. Aurbach, H. Teller, E. Levi, Journal of the Electrochemical Society, 149 (2002)
A1255 - A1266.
[92] Y. Ein-Eli, Electrochemical and Solid State Letters, 2 (1999) 212-214.
[93] M.G.S.R. Thomas, P.G. Bruce, J.B. Goodenough, Journal of the Electrochemical Society,
132 (1985) 1521-1528.
[94] D. Guyomard, J.M. Tarascon, Journal of the Electrochemical Society, 140 (1993) 3071-
3081.
[95] D. Guyomard, J.M. Tarascon, Journal of the Electrochemical Society, 139 (1992) 937-
948.
[96] D. Aurbach, Journal of Power Sources, 89 (2000) 206-218.
[97] R. Dedryvere, H. Martinez, S. Leroy, D. Lemordant, F. Bonhomme, P. Biensan, D.
Gonbeau, Journal of Power Sources, 174 (2007) 462-468.
[98] S.S. Zhang, K. Xu, T.R. Jow, Electrochemical and Solid State Letters, 5 (2002) A92-A94.
[99] Y.H. Wang, X. Yan, X.F. Bie, Q. Fu, F. Du, G. Chen, C.Z. Wang, Y.J. Wei,
[100] J.B. Goodenough, K.S. Park, Journal of the American Chemical Society, 135 (2013)
1167-1176.
[101] Y.X. Wang, S. Nakamura, M. Ue, P.B. Balbuena, Journal of the American Chemical
Society, 123 (2001) 11708-11718.
[102] Y. Wu, N. Hu, L.H. Wei, Y. Dai, Ecs Electrochemistry Letters, 3 (2014) A50-A53.
[103] D. Aurbach, M.L. Daroux, P.W. Faguy, E. Yeager, Journal of the Electrochemical
Society, 134 (1987) 1611-1620.
[104] D. Aurbach, Y. Gofer, Journal of the Electrochemical Society, 138 (1991) 3529-3536.
[105] D. Aurbach, Y. Gofer, M. Benzion, P. Aped, Journal of Electroanalytical Chemistry, 339
(1992) 451-471.
61


[106] Y. Matsumura, S. Wang, J. Mondori, Journal of the Electrochemical Society, 142 (1995)
2914-2918.
[107] Z.X. Shu, R.S. McMillan, J.J. Murray, Journal of the Electrochemical Society, 140 (1993)
922-927.
[108] A. Naji, J. Ghanbaja, B. Humbert, P. Willmann, D. Billaud, Journal of Power Sources,
63 (1996) 33-39.
[109] D. Aurbach, B. Markovsky, A. Shechter, Y. EinEli, H. Cohen, Journal of the
[110] D. Aurbach, E. Zinigrad, Y. Cohen, H. Teller, Solid State Ionics, 148 (2002) 405-416.
[111] S. Yamaguchi, H. Asahina, K.A. Hirasawa, T. Sato, S. Mori, Molecular Crystals and
Liquid Crystals Science and Technology Section a-Molecular Crystals and Liquid
Crystals, 322 (1998) 239-244.
[112] A.C. Chu, J.Y. Josefowicz, G.C. Farrington, Journal of the Electrochemical Society, 144
(1997) 4161-4169.
[113] E. Peled, D. Golodnitsky, A. Ulus, V. Yufit, Electrochimica Acta, 50 (2004) 391-395.
[114] V. Eshkenazi, E. Peled, L. Burstein, D. Golodnitsky, Solid State Ionics, 170 (2004) 83- 91.
[115] T. Utsunomiya, O. Hatozaki, N. Yoshimoto, M. Egashira, M. Morita, Journal of Power
Sources, 196 (2011) 8675-8682.
[116] A.M. Haregewoin, E.G. Leggesse, J.C. Jiang, F.M. Wang, B.J. Hwang, S.D. Lin,
[117] H. Kim, S. Grugeon, G. Gachot, M. Armand, L. Sannier, S. Laruelle, Electrochimica
Acta, 136 (2014) 157-165.
[118] M. Ue, M. Takeda, M. Takehara, S. Mori, Journal of the Electrochemical Society, 144
(1997) 2684-2688.
[119] G. Eggert, J. Heitbaum, Electrochimica Acta, 31 (1986) 1443-1448.
[120] C. Nanjundiah, J.L. Goldman, L.A. Dominey, V.R. Koch, Journal of the Electrochemical
Society, 135 (1988) 2914-2917.
[121] V.R. Koch, L.A. Dominey, C. Nanjundiah, M.J. Ondrechen, Journal of the
[122] K. Xu, M.S. Ding, T.R. Jow, Journal of the Electrochemical Society, 148 (2001)
A267 - A274.
[123] S. Leroy, H. Martinez, R. Dedryvere, D. Lemordant, D. Gonbeau, Applied Surface
Science, 253 (2007) 4895-4905.
62


[124] A.M. Andersson, M. Herstedt, A.G. Bishop, K. Edstrom, Electrochimica Acta, 47 (2002)
1885-1898.
[125] M. Nie, B.L. Lucht, Journal of the Electrochemical Society, 161 (2014) A1001-A1006.
[126] D. Aurbach, K. Gamolsky, B. Markovsky, Y. Gofer, M. Schmidt, U. Heider,
[127] H. Ota, K. Shima, M. Ue, J. Yamaki, Electrochimica Acta, 49 (2004) 565-572.
[128] H. Ota, Y. Sakata, A. Inoue, S. Yamaguchi, Journal of the Electrochemical Society, 151
(2004) A1659-A1669.
[129] L. El Ouatani, R. Dedryvere, C. Siret, P. Biensan, D. Gonbeau, Journal of the
[130] J.C. Burns, G. Jain, A.J. Smith, K.W. Eberman, E. Scott, J.P. Gardner, J.R. Dahn, Journal
of the Electrochemical Society, 158 (2011) A255-A261.
[131] J.C. Burns, N.N. Sinha, G. Jain, H. Ye, C.M. VanElzen, W.M. Lamanna, A. Xiao, E.
Scott, J. Choi, J.R. Dahn, Journal of the Electrochemical Society, 159 (2012)
A1095- A1104.
[132] J.C. Burns, R. Petibon, K.J. Nelson, N.N. Sinha, A. Kassam, B.M. Way, J.R. Dahn,
[133] R. Petibon, C.P. Aiken, N.N. Sinha, J.C. Burns, H. Ye, C.M. VanElzen, G. Jain, S.
Trussler, J.R. Dahn, Journal of the Electrochemical Society, 160 (2013) A117-A124.
[134] N.N. Sinha, J.C. Burns, J.R. Dahn, Journal of the Electrochemical Society, 160 (2013)
A709-A714.
[135] A.M. Andersson, K. Edstrom, N. Rao, A. Wendsjo, Journal of Power Sources, 81 (1999)
286-290.
[136] A.M. Andersson, K. Edstrom, Journal of the Electrochemical Society, 148 (2001)
A1100 - A1109.
[137] H. Park, T. Yoon, J. Mun, J.H. Ryu, J.J. Kim, S.M. Oh, Journal of the Electrochemical
Society, 160 (2013) A1539-A1543.
[138] H. Park, T. Yoon, Y. Kim, J.G. Lee, J. Kim, H.S. Kim, J.H. Ryu, J.J. Kim, S.M. Oh,
[139] Y.S. Wang, C.M. Huang, H.W. Hsieh, Y.F. Lin, C.Y. Lin, J.T. Lee, Electrochimica Acta,
142 (2014) 34-42.
[140] V.A. Agubra, J.W. Fergus, Journal of Power Sources, 268 (2014) 153-162.
63


[142] A.M. Andersson, A. Henningson, H. Siegbahn, U. Jansson, K. Edstrom, Journal of Power

Sources, 119 (2003) 522-527.
[143] A. Augustsson, M. Herstedt, J.H. Guo, K. Edstrom, G.V. Zhuang, P.N. Ross, J.E.
Rubensson, J. Nordgren, Physical Chemistry Chemical Physics, 6 (2004) 4185-4189.
[144] L. Gireaud, S. Grugeon, S. Laruelle, S. Pilard, J.M. Tarascon, Journal of the
[145] L.X. Wang, D. Deng, L.C. Lev, S. Ng, Journal of Power Sources, 265 (2014) 140-148.
[146] D. Aurbach, B. Markovsky, I. Weissman, E. Levi, Y. Ein-Eli, Electrochimica Acta, 45
(1999) 67-86.
[147] D. Aurbach, B. Markovsky, M.D. Levi, E. Levi, A. Schechter, M. Moshkovich, Y. Cohen,
Journal of Power Sources, 81 (1999) 95-111.
[148] M. Herstedt, A.M. Andersson, H. Rensmo, H. Siegbahn, K. Edstrom, Electrochimica
Acta, 49 (2004) 4939-4947.
[149] P. Niehoff, S. Passerini, M. Winter, Langmuir, 29 (2013) 5806-5816.
[150] S. Leroy, F. Blanchard, R. Dedryvere, H. Martinez, B. Carre, D. Lemordant, D. Gonbeau,
Surface and Interface Analysis, 37 (2005) 773-781.
[151] J.T. Lee, N. Nitta, J. Benson, A. Magasinski, T.F. Fuller, G. Yushin, Carbon, 52 (2013)
388-397.
[152] H.L. Zhang, F. Li, C. Liu, J. Tan, H.M. Cheng, Journal of Physical Chemistry B, 109
(2005) 22205-22211.
[153] D.N. Bennion, E.L. Littauer, Journal of the Electrochemical Society, 123 (1976)
1462 - 1469.
[154] G. Nazri, R.H. Muller, Journal of the Electrochemical Society, 132 (1985) 1385-1387.
[155] E. Peled, D. Golodnitsky, G. Ardel, Journal of the Electrochemical Society, 144 (1997)
L208-L210.
[156] J.O. Besenhard, M. Winter, J. Yang, W. Biberacher, Journal of Power Sources, 54 (1995)
228-231.
[157] K. Edstrom, M. Herstedt, D.P. Abraham, Journal of Power Sources, 153 (2006) 380-384.
[158] J. Christensen, J. Newman, Journal of the Electrochemical Society, 151 (2004)
A1977 - A1988.
[159] A.K. Padhi, K.S. Nanjundaswamy, J.B. Goodenough, Journal of the Electrochemical
Society, 144 (1997) 1188-1194.
[160] A. Van der Ven, J. Bhattacharya, A.A. Belak, Accounts of Chemical Research, 46 (2013)
1216-1225.
64


[161] W.J. Zhang, Journal of Power Sources, 196 (2011) 2962-2970.

[162] M.A. Roscher, J. Vetter, D.U. Sauer, Journal of Power Sources, 195 (2010) 3922-3927.
[163] V. Srinivasan, J. Newman, Electrochemical and Solid State Letters, 9 (2006) A110-A114.
[164] S.H. Yu, C. Park, H. Jang, C.B. Shin, W. Cho, Bulletin of the Korean Chemical Society,
32 (2011) 852-856.
[165] Y.R. Zhu, Y. Xie, R.S. Zhu, J. Shu, L.J. Jiang, H.B. Qiao, T.F. Yi, Ionics, 17 (2011)
437 - 441.
[166] C.T. Hsieh, C.T. Pai, Y.F. Chen, P.Y. Yu, R.S. Juang, Electrochimica Acta, 115 (2014)
96-102.
[167] K. Shimoda, H. Sugaya, M. Murakami, H. Arai, Y. Uchimoto, Z. Ogumi, Journal of the
[168] S.Y. Chung, J.T. Bloking, Y.M. Chiang, Nature Materials, 1 (2002) 123-128.
[169] L.X. Yuan, Z.H. Wang, W.X. Zhang, X.L. Hu, J.T. Chen, Y.H. Huang, J.B. Goodenough,
Energy & Environmental Science, 4 (2011) 269-284.
[170] H. Huang, S.C. Yin, L.F. Nazar, Electrochemical and Solid State Letters, 4 (2001)
A170 - A172.
[171] P.P. Prosini, D. Zane, M. Pasquali, Electrochimica Acta, 46 (2001) 3517-3523.
[172] S.F. Yang, Y.N. Song, P.Y. Zavalij, M.S. Whittingham, Electrochemistry
Communications, 4 (2002) 239-244.
[173] D. Lepage, F. Sobha, C. Kuss, G. Liang, S.B. Schougaard, Journal of Power Sources, 256
(2014) 61-65.
[174] H.C. Shin, W.I. Cho, H. Jang, Journal of Power Sources, 159 (2006) 1383-1388.
[175] J. Nanda, S.K. Martha, W.D. Porter, H. Wang, N.J. Dudney, M.D. Radin, D.J. Siegel,
[176] J.J. Wang, J.L. Yang, Y.J. Tang, R.Y. Li, G.X. Liang, T.K. Sham, X.L. Sun, Journal of
Materials Chemistry A, 1 (2013) 1579-1586.
[177] B. Kang, G. Ceder, Nature, 458 (2009) 190-193.
[178] A.S. Arico, P. Bruce, B. Scrosati, J.M. Tarascon, W. Van Schalkwijk, Nature Materials,
4 (2005) 366-377.
[179] S.Y. Chung, S.Y. Choi, T. Yamamoto, Y. Ikuhara, Angewandte Chemie-International
Edition, 48 (2009) 543-546.
[180] X.Z. Liao, Z.F. Ma, Q. Gong, Y.S. He, L. Pei, L.J. Zeng, Electrochemistry
Communications, 10 (2008) 691-694.
65


[181] M. Takahashi, S. Tobishima, K. Takei, Y. Sakurai, Solid State Ionics, 148 (2002)
283 - 289.
[182] M. Koltypin, D. Aurbach, L. Nazar, B. Ellis, Electrochemical and Solid State Letters, 10
(2007) A40-A44.
[183] M. Koltypin, D. Aurbach, L. Nazar, B. Ellis, Journal of Power Sources, 174 (2007)
1241 - 1250.
[184] H.C. Wu, C.Y. Su, D.T. Shieh, M.H. Yang, N.L. Wu, Electrochemical and Solid State
Letters, 9 (2006) A537-A541.
[185] N. Dupre, J.F. Martin, J. Degryse, V. Fernandez, P. Soudan, D. Guyomard, Journal of
Power Sources, 195 (2010) 7415-7425.
[186] H.F. Jin, Z. Liu, Y.M. Teng, J.K. Gao, Y. Zhao, Journal of Power Sources, 189 (2009)
445-448.
[187] X.D. Wang, K. Cheng, J.W. Zhang, L.G. Yu, J.J. Yang, Advanced Powder Technology,
24 (2013) 593-598.
[188] J.J. Wang, Y.J. Tang, J.L. Yang, R.Y. Li, G.X. Liang, X.L. Sun, Journal of Power Sources,
238 (2013) 454-463.
[189] T. Ohzuku, Y. Makimura, Chemistry Letters, (2001) 642-643.
[191] R. Robert, C. Villevieille, P. Novak, Journal of Materials Chemistry A, 2 (2014)
8589 - 8598.
[192] N. Nitta, F.X. Wu, J.T. Lee, G. Yushin, Materials Today, 18 (2015) 252-264.
2 (2012).
[194] S.C. Yin, Y.H. Rho, I. Swainson, L.F. Nazar, Chemistry of Materials, 18 (2006)
1901 - 1910.
[195] Y. Koyama, I. Tanaka, H. Adachi, Y. Makimura, T. Ohzuku, Journal of Power Sources,
119 (2003) 644-648.
[196] B.J. Hwang, Y.W. Tsai, D. Carlier, G. Ceder, Chemistry of Materials, 15 (2003)
3676 - 3682.
[197] C.M. Julien, A. Mauger, K. Zaghib, H. Groult, Inorganics, 2 (2014) 23.
66



[199] M. Ochida, Y. Domi, T. Doi, S. Tsubouchi, H. Nakagawa, T. Yamanaka, T. Abe, Z.
Ogumi, Journal of the Electrochemical Society, 159 (2012) A961-A966.
[200] J.L. Esbenshade, A.A. Gewirth, Journal of the Electrochemical Society, 161 (2014)
A513 - A518.
[201] T. Joshi, K. Eom, G. Yushin, T.F. Fuller, Journal of the Electrochemical Society, 161
(2014) A1915-A1921.
[202] H. Takahara, Y. Kobayashi, K. Shono, H. Kobayashi, M. Shikano, T. Nakamura, Journal
[203] D.R. Gallus, R. Schmitz, R. Wagner, B. Hoffmann, S. Nowak, I. Cekic-Laskovic, R.W.
Schmitz, M. Winter, Electrochimica Acta, 134 (2014) 393-398.
[204] D.H. Jang, Y.J. Shin, S.M. Oh, Journal of the Electrochemical Society, 143 (1996)
2204 - 2211.
[205] Y.Y. Xia, M. Yoshio, Journal of the Electrochemical Society, 143 (1996) 825-833.
[206] W. Choi, A. Manthiram, Journal of the Electrochemical Society, 153 (2006)
A1760 - A1764.
[207] H. Gabrisch, T.H. Yi, R. Yazami, Electrochemical and Solid State Letters, 11 (2008)
A119-A124.
[208] B. Xu, D.N. Qian, Z.Y. Wang, Y.S.L. Meng, Materials Science & Engineering R-Reports,
73 (2012) 51-65.
[209] J. Shu, R. Ma, L.Y. Shao, M. Shui, K.G. Wu, M.M. Lao, D.J. Wang, N.B. Long, Y.L.
Ren, Journal of Power Sources, 245 (2014) 7-18.
[210] M. Kerlau, M. Marcinek, V. Srinivasan, R.M. Kostecki, Electrochimica Acta, 53 (2007)
1385-1392.
[211] K. Xu, Energies, 3 (2010) 135-154.
[212] J.C. Tan, A.K. Cheetham, Chemical Society Reviews, 40 (2011) 1059-1080.
[213] A.K. Dutta, Physical Review, 90 (1953) 187-192.
[214] D.Z. Tsang, M.S. Dresselhaus, Carbon, 14 (1976) 43-46.
[215] A.S. Milev, N.H. Tran, G.S.K. Kannangara, M.A. Wilson, Physical Chemistry Chemical
Physics, 12 (2010) 6685-6691.
[216] Properties and characteristic of graphite, Entegris, 2013.
[217] A. Herold, Bulletin De La Societe Chimique De France, (1955) 999-1012.
[218] N. Daumas, A. Herold, C.R.Acad.Sc.Paris, serie C (1969).
67


157 (2010) A558-A566.
[220] M. Hess, P. Novak, Electrochimica Acta, 106 (2013) 149-158.
[221] L.J. Hardwick, M. Marcinek, L. Beer, J.B. Kerr, R. Kostecki, Journal of the
[222] E. Markervich, G. Salitra, M.D. Levi, D. Aurbach, Journal of Power Sources, 146 (2005)
146-150.
[223] R. Kostecki, F. McLarnon, Journal of Power Sources, 119 (2003) 550-554.
[224] V.A. Sethuraman, L.J. Hardwick, V. Srinivasan, R. Kostecki, Journal of Power Sources,
195 (2010) 3655-3660.
[225] Y.T. Cheng, M.W. Verbrugge, Journal of Applied Physics, 104 (2008).
[226] R. Deshpande, Y.T. Cheng, M.W. Verbrugge, Journal of Power Sources, 195 (2010)
5081-5088.
[227] J. Christensen, J. Newman, Journal of Solid State Electrochemistry, 10 (2006) 293-319.
[228] J. Christensen, J. Newman, Journal of the Electrochemical Society, 153 (2006) A1019-
A1030.
68

Chapter 3
Experimental
Both C6/LiFePO4 (LFP) and C6/LiNi1/3Co1/3Mn1/3O2 (NMC(111)) batteries have been investigated
in this thesis. The aging experiments include both storage and cycling experiments performed with
complete batteries and have been carried out with automated cycling equipment (Maccor). Material
characterization focused on dismantled electrodes and have been performed by X-ray Photoelectron
Spectroscopy (XPS), Raman spectroscopy, Inductively Coupled Plasma (ICP) spectrometry and
Scanning Electron Microscopy (SEM).

69

Chapter 3 Experimental

3.1 Batteries selected
Two different chemistries of Li-ion batteries (i) C6/LiFePO4 (LFP) and (ii)
C6/LiNi1/3Co1/3Mn1/3O2 (NMC(111)) have been selected for the aging experiments. Fig. 3.1
illustrates the shapes and sizes of these batteries. The C6/LiFePO4 chemistry is represented by
a prismatic 50 Ah battery (Fig. 3.1a) and a cylindrical (ANR26650) 2.3 Ah battery (Fig. 3.1b).
The C6/LiNi1/3Co1/3Mn1/3O2 battery (Fig. 3.1c) is a cylindrical 18650 type of battery with a
nominal capacity of 2.0 Ah. All these batteries are commercially available.
Fig. 3.1. Photographs of C /LiFePO batteries from Huali (a) and A123 (b) and a
C /LiNi ⁄ Co ⁄ Mn ⁄ O battery from B&K (c).
3.2 Storage experiments
Prismatic C6/LiFePO4 50 Ah batteries from Huali Company (Fig. 3.1a) and cylindrical
ANR26650 batteries from A123 (Fig. 3.1b) have been selected for the storage experiments
under various conditions. The electrochemical experiments were performed with automatic
cycling equipment (Maccor). The storage temperature of Huali batteries was at room
temperature (RT) while storage of the A123 batteries was carried out in climate chambers to
control the temperature at 20, 40 and 60oC. A F12-ED refrigerator/heating circulator (Julabo)
with glycol fluid was employed to control the temperature of the climate chambers. The detailed
storage conditions are summarized in Table 3.1 and the experimental details of these two types
of batteries are summarized in Fig. 3.2.
70


Table 3.1. Experimental conditions of the storage experiments for two types of C6/LiFePO4 batteries.
SoC Temperature
Huali 30% 70% 100% 20oC
10% 50% 100% 20oC
A123 10% 50% 100% 40oC
10% 50% 100% 60oC
3.2.1 Storage experiments of prismatic 50 Ah batteries
The batteries were activated with a current of 5 A (0.1 C) for 4 cycles, then fully charged
and discharged under a constant-current constant-voltage (CCCV) regime in order to find the
maximum storage capacities and . The cut-off conditions were 3.65 V until 0.1 A
(1/500 C) for deep-charging and 1.6 V until 0.1 A for deep-discharging. Subsequently, three
batteries were charged to 0.3 0.7 and and then stored
for 20 days at room temperature. The batteries were fully discharged after each 20 days and the
corresponding capacities were denoted as where superscript represents the storage
period. The batteries were charged to the previous storage capacity after fully
discharging in order to make sure that all batteries will continue the previous storage cycle.
3.2.2 Storage experiments of A123 batteries
Before conducting these storage experiments, the A123 batteries were activated for 5 cycles
at room temperature. In order to obtain EMF curves before the storage experiments were started,
characterization cycles were measured at 20, 40 and 60oC, corresponding to the subsequent
storage temperatures. To determine the EMF from the characterization cycles all batteries were
charged in a constant-current constant-voltage (CCCV) mode. A 1C charging rate was used in
the CC-mode and the batteries were allowed to continue charging in the CV-mode at 3.6 V for
2 hours. Then the batteries were discharged at various constant currents (0.1, 0.2, 0.3, 0.5, 0.75,
1, 1.5 and 2C-rate) in the subsequent cycles, using a cut-off voltage of 1.6 V. The EMF curves
have been regularly determined by mathematical extrapolation of these measured voltage
discharge curves. These extrapolation methods will be described in Chapter 4.
71


Fig. 3.2. Scheme of the storage experiments.
After completing the characterization process all batteries were charged at 1C rate to various
predetermined levels of SoC (10, 50 and 100%). The corresponding charge capacities are
denoted as , (Fig. 3.2). The batteries were discharged after one month of storage with the
same current and cut-off voltage, which resulted in the corresponding discharge capacities ( .
Re-characterization was regularly performed before a new storage period was initiated. After
completing these characterization cycles the batteries were charged to the previous discharging
capacity, i.e. , in order to make sure that all batteries will continue the previous
storage periods.
3.3 Cycling experiments
For the cycling experiments all three types of batteries, shown in Fig. 3.1, have been selected.
The cycling regime was divided into two sections (i) full cycling between SoC = 0-100% and
(ii) partial cycling between SoC = 0-30%, 35-65%, 70-100%. The experimental details will be
discussed below.
3.3.1 Full cycling measurements
3.3.1.1 Prismatic 50 Ah batteries (Huali)
The electrochemical experiments were carried out with Maccor automatic cycling
equipment. The cycling current was selected to be 0.1 C (5A). The environment temperature
72


was at room temperature. The batteries were activated using the same method as for the storage
experiments described in Section 3.2.1. Subsequently, the batteries were cycled in the CC mode
with 0.1 C. The cut-off voltage was 3.65 V for charging and 1.6 V for discharging.
3.3.1.2 Cylindrical 2.3 Ah batteries (A123)
Before conducting the cycling experiments all batteries were activated for 5 cycles at room
temperature. The characterization processes were subsequently carried out at 20, 40 and 60oC,
corresponding to the cycling temperatures, to obtain the EMF curves. The details of the
characterization procedures are the same with those in the storage experiments as discussed in
Section 3.2.2. After the characterization process has been completed all batteries were cycled
under various conditions summarized in Table 3.2. Various cycling currents are given in the
second column. The last column shows the duration of each cycle. Since the cycling current is
different, the duration of each cycle is also different. All batteries were regularly re-
characterized after approximately every 20 days. Note that the actual duration of each cycle is
variable due to the decreasing of the battery capacity after cycling.
Table 3.2. Experimental conditions of the cycling experiments of A123 batteries.
Temperature Duration
o
20 C 40oC 60oC of each cycle (hours)
Current 0.1 0.1 0.1 22.6~19.3
(C-rate) 0.5 0.5 0.5 5.5~4.8
1.0 1.0 1.0 2.3~2.2
2.0 2.0 2.0 2.1~1.8
3.3.1.3 Cylindrical NMC batteries (B&K)
Before conducting the cycling experiments all batteries were activated for 5 cycles at 40
and 60oC, corresponding to the cycling temperatures. The characterization processes were
subsequently carried out to obtain the EMF curves. During characterization all batteries were
charged in the CCCV mode. A 1C charging rate was used in the CC-mode followed by CV
charging at 4.2 V during 1 hours. The batteries were then discharged at various constant currents
(0.1, 0.2, 0.3, 0.5, 0.75, 1.0, 1.5 and 2.0 C-rate) in the subsequent cycles using a cut-off voltage
of 2.7 V. On the basis of these sets of discharge curves, the EMF was extracted by extrapolation
at either constant SoC (vertical extrapolation) or constant voltage (horizontal extrapolation) at
the end of the discharge process. Details of these extrapolation methods can be found in Chapter
73


4. After the characterization process has been completed all batteries were cycled under various
conditions shown in Table 3.3. Since the cycling current is different, the cycling time in each
cycle is also different. All batteries were regularly re-characterized after approximately every
20 days. Note that the actual duration of each cycle is variable due to the decreasing of the
battery capacity after cycling.
Table 3.3. Experimental conditions of cycling NMC batteries.
Temperature Duration
in each cycle (hours)
40oC 60oC
Current 0.1 0.1 25 ~22
(C-rate) 0.3 0.3 8.2~7.1
0.5 0.5 5.4~4.0
1.0 1.0 3.3~2.9
2.0 2.0 2.6~2.2
3.3.2 Partial cycling measurements
In order to investigate the influence of cycling State-of-Charge on the battery performance,

the A123 batteries have been selected to perform these partial cycling measurements. The
cycling conditions are summarized in Table 3.4. The activation and characterization processes
of each type of battery are the same to those discussed in Section 3.3.1. After the
characterization process has been completed all batteries were cycled at 1C rate in various
predetermined SoC ranges and at various temperatures, as indicated in Table 3.4. Re-
characterization was regularly performed before a new cycling period was initiated.
Table 3.4. Partial cycling conditions of A123 batteries.
Battery Temperature SoC

20oC 40oC 60oC 0-30%
A123 o o o
20 C 40 C 60 C 35-65%
(LFP)
o o o
20 C 40 C 60 C 70-100%
74


3.4 X-ray photoelectron spectroscopy (XPS)
X-ray Photoelectron Spectroscopy (XPS) also known as Electron Spectroscopy for

Chemical Analysis (ESCA) is the one of the most widely used surface analytical techniques
because it can be used to analyze a broad range of materials. XPS can provide valuable
quantitative information of the chemical state of the investigated material surface. The average
depth analysis of XPS is approximately 5 nm. Spatial distribution information can be obtained
by scanning the sample surface with a micro-focused x-ray beam. Depth distribution
information can be obtained by sputtering the sample surface with Argon cluster ions.
In order to investigate the SEI formation and the deposition of Fe on the graphite electrode
upon aging, XPS measurements have been carried out on graphite electrodes dismantled from
the batteries aged under various conditions. The batteries after storage and cycling were fully
discharged at 1C-rate before opening in an Argon-filled glove box. Small pieces of the graphite
electrodes were cut from different locations and rinsed with pure solvent (Diethyl Carbonate).
The collected samples were dried under vacuum for ~3 days before transferring to the XPS
equipment. XPS analyses were carried out on a K-Alpha system (Thermo Scientific) with a
resolution of 0.2 eV, using an Al monochromatic irradiation (1486.6 eV) at a working
pressure smaller than 7 10-8 bar. Depth profiling was carried out, using Ar ion-beam sputtering
with 500 eV. The sputtering rate was equivalent to 0.26 nm/s on Ta2O5.
3.5 Raman spectroscopy
Confocal Raman spectroscopy is a non-destructive, fast, and high-resolution tool for the
characterization of battery materials. Raman spectroscopy is generally used in chemistry to
provide a fingerprint by which molecules can be identified. Carbonaceous materials with
conjugated sp2 or close to sp2 carbons such as graphite, carbon nanotubes, and graphene etc.,
have been widely studied by Raman spectroscopy in recent years [1-8]. Raman spectroscopy
provides spatially resolved information about the vibrational spectrum and the electronic band
structure via the mechanism of double-resonant Raman scattering [5]. Moreover, Raman
spectroscopy can be efficiently used to monitor number of layers, quality of layers, doping level
and confinement of those graphitic materials.
Graphite is commonly used as the anode material in most commercial Li-ion batteries.
Single-crystal graphite belongs to the 6 / space group [3, 4]. Its isogonal point
group is . The vibration modes of graphite can be assigned to the types of , ,
and as shown in Fig. 3.3. The two in-plane vibrations ( modes) are Raman active and
75


have been identified with a band at 1575 [1] ~ 1580 [8] cm-1 and a low-frequency neutron
scattering feature at 47 cm-1. The earliest Raman spectrum on graphite has been experimentally
analyzed by Tuinstra and Koenig [1]. The only Raman line called G band at 1580 cm-1 has been
observed in single-crystal graphite and is therefore considered to be characteristic for graphite
materials. In most commercial graphite materials a so-called D band at ~1360 cm-1 is usually
also observed [1].

Fig. 3.3. Vibration modes of single crystal graphite and the corresponding spectroscopy actives [3].
The origin of D band is still debatable. Some researchers attributed the D band to a decrease
in symmetry near microcrystallite edges, where the symmetry reduced from to or even
[3]. The selection rules for Raman activity has been changed at these edges. Some certain
phonons which were inactive in the infinite lattice became active at edges. Tuinstra and Koenig
reported that the intensity of the D band was proportional to the percentage of the boundaries
76


in the samples [1]. Obviously, the effective crystallite size in the direction of the graphite
plane determines the percentage of the edges which was considered to be the origin of the D
band. When the laser spot is probing these edges, even when the bulk structure is perfect, the D
peak will appear in the spectra. This edge effect was validated by Pimenta et al. [2]. Apart from
the edge effect, structural defect / disorder was considered to be another origin of the presence
of the D band. Defects include bond length and angle disorder at the atomic scale. In principle,
the sample edges can also be considered as defects [6]. Therefore, it can be concluded that the
smaller and higher numbers of defects will lead to a higher D peak intensity. Apart from the
D band, another band at 1620 cm-1 (D’ band) is also considered to be related to defects [8].
However, the D’ band is much more difficult to distinguish as it strongly coincides with the
large G band.
Raman spectroscopy was used to analyze the structural degradation of the graphite electrode
after the cycling experiments have been completed. The dismantled electrodes were rinsed by
Dimethyl Carbonate (DMC) and dried in a vacuum chamber for 1 week before the
measurements were carried out. The excitation wavelength was supplied by an internal He-Ne
632 nm laser. The size of the laser beam at the sample was ~ 0.2 . Raman mapping of the
anode surface was carried out across a typical 40 40 area at 0.8 lateral resolution,
using a software-controlled motorized XY stage.
3.6 Scanning electron microscopy (SEM)
The morphology of the electrode was characterized by scanning electron microscope (SEM)
from Philips/FEI XL 40 EFG. The samples were prepared by dismantling the batteries aged
under various conditions. Pieces of electrode were cut down and rinsed with Dimethyl
Carbonate (DMC) and then dried under vacuum for 3 days. The SEM images were taken at 10
kV acceleration voltage at a beam current of 0.54 nA. The particle size distribution has been
analyzed based on the SEM images.
3.7 Inductively Coupled Plasma-Optical Emission Spectrometry (ICP-OES)
The concentration of iron ions in the electrolyte is very small [9]. Conventional analytical
technologies are not sufficient to obtain quantitative information of trace elements in the
electrolyte. The most commonly used technique for the determination of small concentrations
of elements is based on atomic spectrometry. Each element has its own characteristic set of
energy levels and thus its own unique set of emission and/or absorption wavelengths. Atomic
77


Absorption Spectrometry (AAS), Optical Emission Spectrometry (OES), Atomic Fluorescence

Spectrometry (AFS) and Atomic Mass Spectrometry (AMS) are the four most popular atomic
spectroscopic techniques. An important advantage of OES is that it is flexible to select specific
emission wavelengths characteristic for individual elements. [10].
In OES, the samples have to be subjected to considerably high temperatures in order to
reach the excited (ionized) states of the atoms. These excited atoms can decay to lower states
through thermal or radiative energy transitions. The intensity of the light emitted at specific
wavelengths is measured and used to determine the concentrations of the elements of interest.
Argon supported inductively coupled plasma (ICP) is the state-of-the-art in plasma sources as
it generates extremely hot plasma to excite and/or ionize the atoms for atomic and ionic
emission. The combination of an ICP and OES is called ICP-OES [10].
In order to quantitatively determine the Fe deposition on graphite electrodes, three samples
have been prepared based on the graphite electrode dismantled from (i) pristine, (ii) cycled and
(iii) stored batteries. The corresponding aging conditions of the cycling and stored batteries are
2 C-rate at SoC = 10% at 60oC, respectively. The graphite electrodes have been cut into small
pieces of 2.9 2cm (both sides are coated by active materials), and immersed into the sulfuric
acid solution with a concentration of 1 mol/L (prepared from H2SO4, 99.999%, Aldrich). The
samples have been stored at room temperature for 10 days in order to dissolve the Fe sufficiently.
The ICP measurements are carried out on these samples, using standard calibration solutions.
3.8 References
[1] F. Tuinstra, J.L. Koenig, Journal of Chemical Physics, 53 (1970) 1126-&.

[2] M.A. Pimenta, G. Dresselhaus, M.S. Dresselhaus, L.G. Cancado, A. Jorio, R. Saito,
Physical Chemistry Chemical Physics, 9 (2007) 1276-1291.
[3] Y. Wang, D.C. Alsmeyer, R.L. Mccreery, Chemistry of Materials, 2 (1990) 557-563.
[4] S. Reich, C. Thomsen, Philosophical Transactions of the Royal Society a-Mathematical
Physical and Engineering Sciences, 362 (2004) 2271-2288.
[5] D. Graf, F. Molitor, K. Ensslin, C. Stampfer, A. Jungen, C. Hierold, L. Wirtz, Solid
State Communications, 143 (2007) 44-46.
[6] A.C. Ferrari, Solid State Communications, 143 (2007) 47-57.
[7] K.N. Kudin, B. Ozbas, H.C. Schniepp, R.K. Prud'homme, I.A. Aksay, R. Car, Nano
Letters, 8 (2008) 36-41.
78


[8] V. Zolyomi, J. Koltai, J. Kurti, Physica Status Solidi B-Basic Solid State Physics, 248
(2011) 2435-2444.
[10] B.B. Charles, J.F. Kenneth, in, PerkinElmer, Inc., USA, 2004.
79

80

Chapter 4
Methodologies and Terminologies
This chapter introduces the most important methods used in this work, including Electromotive
Force (EMF) determination and / analysis. The EMF curve is considered to be an
important tool to obtain an in-depth understanding of aging mechanisms inside Li-ion batteries.
Various parameters, such as maximum capacities ( ), irreversible capacity losses (Δ ),
overpotentials ( ), etc., can be extracted from these EMF curves. The development of the second
depressions in the / curves is proven to be an interesting indicator for the graphite
electrode decay. A non-destructive approach has been proposed to quantitatively determine the
graphite inaccessibility after aging.

EMF
-dVEMF / dQ
Voltage / V
(a) (b)
Voltage / V
Qout / Ah
C-rate C-rate
Qout / Ah Qout / Ah

81

Chapter 4 Methodologies and Terminologies
4.1 EMF determination
The Electromotive force (EMF) is the battery voltage at its equilibrium state. The
relationship between the EMF and the thermodynamic properties has been described by
G   zFVEMF , [4.1]
where Δ is the change of the Gibb’s free energy, is the number of the electrons involved in
the basis charge-transfer reaction (Eqs. 1.13-15), the Faraday constant and is the EMF
voltage of the battery. The EMF curve describes the equilibrium voltage as a function of SoC
[1]. The maximum storage capacity of the battery can then be easily obtained from the EMF
curve. Several methods have been used to measure the battery EMF, such as Galvanostatic
Intermittent Titration Technique (GITT), Potentiostatic Intermittent Titration Technique (PITT),
etc. In this work we adopted a more classical and convenient approach, which is based on the
regression extrapolation of voltage discharge curves obtained under various loading conditions
[1-5].
3.5
3.34
3.0 (a)
0.5Ah
Voltage / V
Voltage / V
3.30 0.7Ah 2.60 (b)

2.4V
2.5 0.9Ah
2.8V
Qout / Ah
1.5Ah
EMF 3.26 2.52
0.1 0.3 0.5 2.9V
0.1C
C-rate
0.2C
2.0 0.3C 3.0V
0.5C
2.44
0.75C 0.1 0.3 0.5
1.0C C-rate
1.5C t
Qmax
2.0C
1.5
0 0.5 1 1.5 2 2.5
Qout / Ah
Fig. 4.1. Set of voltage discharge curves (solid lines) obtained during the characterization process of a
LFP battery (A123) at 60°C with various indicated discharge currents. The extrapolated EMF curve is
represented by the dotted curve. The insets show an example of the vertical (a) and horizontal
extrapolation (b) procedure.
82


Fig. 4.1 shows, as an example, a set of voltage discharge curves for a C6/LiFePO4 battery at
60oC and, extracted from these results, the extrapolated EMF-curve (dotted line). Dependent on
the SoC, the extrapolation is performed at either constant SoC (vertical extrapolation) at the
beginning of discharging or at constant voltage (horizontal extrapolation) at the end of the
discharge process. The insets show an example of such a vertical (a) and horizontal
extrapolation (b) procedure. Linear relationships are observed between the current and voltage
in the vertical direction and between the current and extracted amount of charge ( ) in the
horizontal direction, indicating that the extrapolation is highly accurate.
The maximum capacities of the batteries ( ) can then be determined from these
extrapolated EMF curves as indicated in Fig. 4.1. The obtained values for represents the
total amount of the cyclable Li ions inside the batteries.
4.2 Parameters and definitions
Parameters and definitions involved in the aging experiments will be introduced in this
section. Fig. 4.2 shows the voltage curves of a LFP battery during charging and discharging.
The corresponding charge and discharge capacities are represented by , and ,
for the and 1 cycle, respectively.
Fig. 4.2. Voltage curves of a LFP battery where , (charging capacity) and ,
(discharging capacity) are defined at the and 1 cycle.
83

The charge capacity ( ) represents the total amount of charge transferred from the cathode
to the anode during the charging while the discharge capacity ( ) represents the total amount
of charge that can be re-extracted from the anode and transported to the cathode during
discharging. The subsequent charging capacity ( ) denotes the amount of charge that can
be re-extracted from cathode and transported to the anode during the subsequent charging
process. Ideally, . However, under real conditions there are always cyclable
Li losses due to parasitic side-reactions, such as the SEI formation at the anode, and electrode
material decay, resulting from, for example, structural degradation of the cathode during
(dis)charging, implying that the above capacities may be different from each other. CE a is
introduced to describe the charge efficiency at the anode, according to

Qdn
CEa  . [4.2]
Qchn
Traditionally, CE a is denoted as the coulombic efficiency. Similarly, the charge efficiency at
the cathode  CE c  can be defined as
Qchn 1
CE c  . [4.3]
Qdn
The values of CE a and CE c are determined by the side-reactions at the anode and cathode,
respectively.
Fig. 4.3a shows the methods and parameters used in this thesis. The black thick line in Fig.
4.3a represents the extrapolated EMF curve at the initial state ( ) after full battery
activation but before aging, including storage and cycling. The red thick line shows the
extrapolated EMF curve after long-time aging ( ), where the superscript denotes the
aging time. The corresponding maximum capacities of the EMF curves are denoted by
and where is considered to be the total storage capacity at any aging time . The
irreversible capacity loss (Δ can then be defined as
Qir  Qmax
0
 Qmax
t
. [4.4]
The voltage discharge curves in the initial and aged state are represented by (thin black
curve) and (thin red curve), respectively. The corresponding discharge capacities are
denoted by and . The apparent discharge capacity loss (Δ ) in Fig. 4.3a can then be
defined by
Qapp  Qd0  Qdt . [4.5]
84


The voltage difference between the discharging voltage curve and the EMF curve is defined
as the total overvoltage ).

Fig. 4.3. Schematic representation of the methods and defined parameters used in this thesis. (a) The
EMF curves ( for pristine and for aged batteries) and the voltage discharge curves (
and ) define the various storage capacities ( , and , ), the apparent capacity loss
(Δ ) and irreversible capacity loss (Δ ), and the overpotential . (b) Schematic representation
of the voltage curves of the LiFePO4 electrode, the graphite electrode and the complete LFP battery. The
various voltage regions are denoted by , , , , , for the graphite electrode and , , , ,
, for complete batteries. The corresponding ratios of the various graphite electrode stages are
represented by , and . The total capacity is represented by for the graphite electrode,
for the complete battery and by for the cathode.
Fig. 4.3b schematically shows the EMF curves of the two individual electrodes and added,
the EMF of the complete battery. It is well known that several voltage plateau regions of the
graphite electrode can be identified during Li+ (de)intercalation. These voltage regions are
denoted by , , , and , . Plateau I has been related to the (de)intercalation processes
85

at the so-called dense stage 1 and stage 1 - 2, plateau II to stage 2 and stage 2 - 2L, and plateau
III corresponds to the remaining intercalation processes, including stages 3L and 4L and the
dilute stage 1L [6-7]. The width of these voltage plateaus ( , and ) can be related to in
Li C (see Fig. 4.3b) where 1 . The storage capacities of the individual
plateaus ( , , , and , ) can then be related to the total graphite electrode capacity
( , according to
QC6 ,I  xIQC6 , [4.6]
QC6 ,II  x IIQC6 , [4.7]
QC 6 ,III  x IIIQC6 , [4.8]
where
QC6  QC6 ,I  QC6 ,II  QC6 ,III . [4.9]
Generally, as schematically shown in Fig. 4.3b the anode electrode capacity ( ) is always
designed larger than the cathode electrode capacity ( ) in order to avoid metallic Li-
plating on the anode surface. Therefore the LiFePO4 electrode is the battery capacity-limiting
electrode in LFP batteries. Since the voltage curve of the LiFePO4 electrode has a wide and flat
plateau, the various identified battery voltage plateau regions (red curve in Fig. 4.3b) must be
attributed to the graphite electrode. The storage capacities of the various battery plateau regions
can be represented by Li,I , Li,II and Li,III . The total battery capacity is therefore a summation
of these three plateaus
Qmax  QLi,I  QLi,II  QLi,III . [4.10]
4.3 Overpotential and resistance determinations
The battery overpotential (also called overvoltage) established at a certain current as a

function of SoC can be written as
  SoC   V EM F  SoC   V bat  SoC  , [4.11]
where is the EMF at various SoC and is the battery voltage at a given
discharge current as a function of SoC. Fig. 4.4 shows an example of the calculated
overpotentials, according to Eq. 4.11, as a function of SoC and discharge currents.
86


0.5
0.10 C
0.20 C
0.4 0.30 C
0.50 C
0.75 C
Overpotential / V
1.00 C
0.3 1.50 C
2.00 C
0.2
0.1
0
0 0.2 0.4 0.6 0.8 1
SoC
Fig. 4.4. Development of the overpotentials at various indicated discharging currents as a function of
SoC.
0.5
0.10 C
0.20 C
0.30 C
0.4
0.50 C
0.06 0.75 C
1.00 C
R (SoC) / 
0.3
1.50 C
2.00 C
0.04
0.2
0.1 0.02
0 0.2 0.4 0.6 0.8 1
0
0 0.2 0.4 0.6 0.8 1
SoC
Fig. 4.5. Development of the total battery resistance at various indicated discharging currents as a
function of SoC.
The total battery resistance at various discharging currents as a function of State-of-Charge

( ) can be calculated on the basis of the overpotentials, according to
  SoC 
R  SoC   , [4.12]
Id
where represents the discharge current. Fig. 4.5 shows the corresponding values
calculated from the data shown in Fig. 4.4, using Eq. 4.12.
87

3.5
3.0
Voltage / V
3.48 Vend
r
2.5
3.45
Vini
d
 t = 15 ms
2.0
3.42
0 5 10 15 20 25
Time / ms
1.5
0 0.4 0.8 1.2
Time / h
Fig. 4.6. Development of the battery voltage initiated by a discharging current application after resting.
The inset shows a zoom-in of the instantaneous voltage drop after current applying.
Fig. 4.6 shows the development of the voltage profile when the battery discharge process is
initiated. The battery voltage at the end of the resting period is denoted as and that after
commencing the discharge process is denoted by . The response time of the cycling device
(Maccor) is about 15 ms. Since Δ is sufficiently small, the voltage drop in such a period can
be considered as ohmic, representing the ohmic resistance of the battery where can be
written as
Vrend  Vdini
R  . [4.13]
Id
4.4 Non-destructive quantification of QC6
Using a reference electrode is considered to be an effective way to determine the graphite

electrode capacity. However, batteries have to be opened in order to position reference
electrodes inside the electrode package, which will influence the cycling performance. In this
section, a non-destructive approach is proposed, which is based on the analyses of the plateaus
on the battery EMF curves to quantitatively determine the graphite electrode capacity .
has been represented by , , , and , by dividing the corresponding ratios ,
and according to Eqs. 4.6-4.8. Although , , , and , are unknown parameters,
88


some connections to the battery capacities can, however, be found. Details will be discussed
below.
The observed plateaus in the battery EMF curves are attributed to the graphite electrode
since the cathode voltage curve has a wide and flat plateau region (Fig. 4.3). It is known that
during charging a graphite electrode, it is thermodynamically more favorable to preferentially
occupy the available sites in the graphite electrode starting from the plateau III host sites,
followed by the plateau II sites to finally occupy the plateau I sites at the end of the charging
process [6]. Obviously, the reverse holds for the discharging processes. Interestingly, as
schematically shown in Fig. 4.3b, the identifiable storage capacities of the various battery
voltage plateaus are quantitatively related to those of the graphite electrode by
Q Li,II  QC6 ,II  x IIQC 6 , [4.14]
Q Li,III  QC6 ,III  x IIIQC6 . [4.15]
However,
QLi,I  QC6 ,I , [4.16]
because the amount of Li-ions to be delivered by the LiFePO4 electrode is not sufficient to
occupy all available host sites in the graphite electrode. can in principle be calculated on
the basis of , and , , according to Eqs. 4.14 and 4.15. From an experimental of view,
, and are more convenient to be accurately determined. Therefore, Eq. 4.14 will be
adopted to determine the graphite electrode capacity in this thesis. In order to facilitate a more
accurate determination of the width of the second plateau , , differential voltage analyses
has been carried out from the determined EMF curves and the details will be discussed in the
following section.
4.5 Analyses of dV/dQEMF curves
Differential voltage analysis is an effective method to investigate various battery systems,

including Li-ion batteries [8-14]. / curves are highly useful to examine, for example,
phase transitions of electrode materials and may offer a detailed understanding of ionic
intercalation mechanisms in host materials.
Li-ion intercalation and consequent phase transitions inside the electrode materials are one
of the most important processes occurring during battery operation. However, it is very difficult
to extract information of the individual electrodes from the battery voltage curves as the total
89

battery voltage is the difference between the cathode ( ) and the anode ( ) electrode
voltages, i.e.
, [4.17]
Differential voltage analysis provides, however, a possibility to get more insight into the
intercalation processes and phase transitions of the individual electrodes [10-12, 14]. The
amount of charge exchanged between the positive and negative electrode are always exactly
balanced as the two electrodes are obviously operating in series. Therefore the following
expression applies
dVbat dVc dVa

  . [4.18]
dQ dQ dQ
Eq. 4.18 implies that the information of the individual electrodes can, in principle, be
identified from the battery voltage curves by making use of the differential voltage technique.
The main drawback of the dependence of / is that includes the overpotential
contributions of both the electrodes and electrolyte during battery operation, i.e. during
(dis)charging. The overpotentials obscure the signals of the electrode materials in the /
curves, making it difficult to draw conclusions with respect to the operation of the electrode
materials. Instead of analyzing the / curves, / curves can provide more
underlying information about intercalation and phase transitions for the individual electrode
[10-12] as
dVEMF dVLiFePO4 dVC6

  , [4.19]
dQ dQ dQ
where , and refer to the battery voltage and to the individual electrode
voltages, respectively.
Fig. 4.7a shows an example of a / curve determined for a LFP (A123) battery at
60oC. The depressions on / curve are corresponding to the plateaus in the EMF curves
while the peaks on / curve correspond to the slopes on EMF curve. As shown in Fig.
4.3b, the voltage plateau of the LiFePO4 electrode is very flat. / in Eq. 4.19 can be
considered zero at almost all SoC, implying that the depressions and peaks on the /
curve must be related to the graphite electrode voltage, especially in the plateau II region. The
width of the region II on / curve is used to determine the graphite electrode capacity,
according to Eq. 4.14.
90


Fig. 4.7. (a) Illustration of / curve of a LFP (A123) battery at 60oC. (b) The / curves
define the storage capacities of the various battery voltage regions denoted as , , , , , and
, , , , , for pristine and aged batteries, respectively.
The storage capacity of the graphite electrode may decrease upon aging. Eqs. 4.14 and 4.15
imply that a decrease of both , and , will lead to the same decrease of the battery
voltage characteristics , and , . Storage capacity loss of the graphite electrode has, for
example, been attributed to particle isolation and layer blockage [6]. As a consequence of these
deterioration mechanisms all voltage plateau regions of the graphite electrode are expected to
be reduced proportionally. The reduction of the second battery voltage plateaus can then be
represented by
QLi,II  QC6 ,II  xII QC6 . [4.20]
QLi,III  QC6 ,III  xIII QC6 . [4.21]
91

As the width of plateau II in the / curves is exactly the same to that of the graphite
electrode, the decrease of Δ , obtained from / curves can be considered as an
indicator for the graphite electrode degradation.
Fig. 4.7b schematically shows an example of the corresponding ⁄ curves for a
pristine (black) and aged (red) LFP battery. The amount of charge related to the various battery
voltage plateau regions are represented by , , , and , at pristine state, and by , ,
, , , in the aged state. Obviously,
Δ Li,I Li,I Li,I , [4.22]
Δ Li,II Li,II Li,II , [4.23]
Δ Li,III Li,III Li,III , [4.24]
and
Δ Δ Li,I Δ Li,II Δ Li,III . [4.25]
I II III
-dVEMF / dQ
1200
900
600
300
0 1 2
Qout / Ah
Fig. 4.8. Development of / curves as a function of cycle number during cycling at 2 C and
o
60 C.
Fig. 4.8 illustrates, as an example, the development of the / curves as a function

of cycle number at 2 C and 60oC. The conventional / plots, as shown in Fig. 4.8, are
usually normalized at the charged state. The advantage of this plot is that the total irreversible
capacity loss can be easily identified as indicated by the red arrow. However, the changes of
the individual plateaus are difficult to monitor. In order to emphasize the changes in width of
92


regions I and II, all curves are aligned with respect to the first peak at about 0.8 Ah. As shown
in Fig. 4.9, the blue vertical lines indicate the position of the first and second peak in the pristine
state. A red sloping line, connecting the second peak positions in the / curves is added
to emphasize the decline of the plateau II regions. The cycling-induced changes of regions I and
II, indicated by red arrows, now become more visible.
I II III
-dVEMF / dQ
1200
900
600
300
0 1 2
Qout / Ah
Fig. 4.9. Development of / curves as a function of cycle number during cycling at 2 C and
60oC.
The technique of differential of the extrapolated EMF voltage curves analysis provides a
non-destructive approach to accurately determine the graphite electrode decay, which allows to
continuously perform cycling tests without interruption.
4.6 References
[1] V. Pop, H.J. Bergveld, J.H.G. Op het Veld, P.P.L. Regtien, D. Danilov, P.H.L. Notten,
[2] D. Danilov, R.A.H. Niessen, P.H.L. Notten, Journal of the Electrochemical Society, 158
(2011) A215-A222.
[3] M.S. Rad, D.L. Danilov, M. Baghalha, M. Kazemeini, P.H.L. Notten, Electrochimica
Acta, 102 (2013) 183-195.
[4] H.J. Bergveld, W.S. Kruijt, P.H.L. Notten, Battery Management Systems, Design by
Modeling, Kluwer Academic Publishers, Boston 2002.
93

[5] V. Pop, H.J. Bergveld, P.P.L. Regtien, J.H.G.O.H. Veld, D. Danilov, P.H.L. Notten,
Acta, 190 (2016) 1124-1133.
[8] M. Dubarry, B.Y. Liaw, M.S. Chen, S.S. Chyan, K.C. Han, W.T. Sie, S.H. Wu, Journal
of Power Sources, 196 (2011) 3420-3425.
[10] I. Bloom, J. Christophersen, K. Gering, Journal of Power Sources, 139 (2005) 304-313.
[11] I. Bloom, A.N. Jansen, D.P. Abraham, J. Knuth, S.A. Jones, V.S. Battaglia, G.L.
Henriksen, Journal of Power Sources, 139 (2005) 295-303.
[12] I. Bloom, J.P. Christophersen, D.P. Abraham, K.L. Gering, Journal of Power Sources,
157 (2006) 537-542.
Sources, 101 (2001) 238-247.
[14] I. Bloom, L.K. Walker, J.K. Basco, D.P. Abraham, J.P. Christophersen, C.D. Ho, Journal
of Power Sources, 195 (2010) 877-882.
94

Chapter 5
Degradation Mechanisms of LFP Batteries
Experimental Analyses of Calendar Aging
The capacity loss and material decay of LFP batteries have been investigated under various storage
conditions as a function of State-of-Charge (SoC) and temperature. The electromotive force (EMF)
curves, determined by mathematical extrapolation of the measured voltage discharge curves, are
used to investigate the aging mechanisms during storage. The irreversible capacity loss, which is
accurately calculated on the basis of the maximum storage capacity estimated from the EMF curves,
increases as a function of temperature and SoC. The loss of cyclable lithium during storage is
considered to be the main source of the irreversible capacity loss under all storage conditions.
Strikingly, during storage at 60oC another important degradation process was discovered: the
inaccessibility of graphite. The graphite electrode decay has been quantitatively determined by non-
destructive analyses on the basis of ⁄ curves. Deposition of Fe on the graphite electrode
has experimentally been confirmed by XPS and ICP analysis. The increasing graphite inaccessibility
is shown to be the consequence of Fe dissolution from the cathode with the subsequent deposition
onto the anode.

95

Chapter 5 Degradation mechanisms of LFP batteries: Experimental Analysis of Calendar Aging
5.1 Introduction
Although LFP batteries have many advantages, the capacity loss and the electrode material
decay are still important issues to address. Capacity losses generally include both the as-denoted
irreversible and reversible losses [1]. Irreversible losses are due to the immobilization of Li ions
inside the batteries and reversible capacity losses are caused by kinetic limitations resulting
from higher internal resistances. Extensive studies have been performed to investigate these
degradation mechanisms [2-29].
It is well known that electrons extracted from the cathode during charging are partially
consumed by parasitic reactions occurring at the anode, leading to low coulombic efficiencies
in the first (dis)charge cycles. The reaction products form a protective layered structure on the
graphite surface, known as Solid-Electrolyte-Interphase (SEI). Continuous growth of the SEI
layer during aging lead to irreversible capacity losses [30-33]. Various SEI formation
mechanisms have been proposed in the last decades. Some researchers consider electron
tunneling through the SEI to be rate determining [34, 35] while others believe solvent diffusion
through the SEI [36, 37] or charge transfer to be rate limiting [38]. The structure of the SEI
layer has been unraveled by experimental studies. A thin and dense inorganic layer was found
to be present on the graphite surface followed by a more porous organic layer [39-41]. A new
tunneling model based on this specific SEI morphology will be proposed in this thesis [2, 3].
Besides Li-immobilization in the SEI layer, both cathode and anode electrode material
decay has been reported [6]. Dissolution of the LiFePO4 electrode has been discussed
extensively [10-14]. Fe-ion dissolution into the electrolyte has been examined by Inductively
Coupled Plasma (ICP) spectroscopy [12, 13] and was found to be affected by many factors,
such as the impurity levels in the cathode material [42], water contamination in the electrolyte
[10] and operating conditions [13]. Obviously, Fe dissolution leads to a change in surface
morphology, which may suppress the cathode electrode kinetics. Several strategies have been
reported to improve the LiFePO4 electrode stability, such as carbon coatings [43], electrolyte
additives [44], etc.
The decay of the graphite electrode has also been investigated by many researchers [5, 6,
45-49]. The disparity of the Li ions within graphite particles during (de)lithiation is considered
to be the main reason for the structural decay. A diffusion-induced stress (DIS) model was
proposed to explain the degradation during (dis)charging [47]. However, to the best of our
knowledge no reports are addressing the mechanisms of the graphite electrode decay during
long-term storage.
96


The calendar life of Li-ion batteries is of major importance, especially in applications such
as EVs and HEVs which reside most of the life time in the so-called parking mode [7]. Moreover,
in-depth investigation of aging mechanisms during storage is helpful to understand the battery
fading phenomena under other operating conditions. Prismatic 50 Ah and cylindrical 26650
LFP batteries were used to carry out the present storage experiments. The capacity loss, material
decay and resistance development up to 1 year storage has been analyzed and will be discussed.
The deposition of Fe on the graphite electrodes has been analyzed by XPS and ICP. Based on
these results a new graphite decay mechanism during storage is proposed in the present chapter.
5.2 Results
5.2.1 Aging of prismatic batteries
The experimentally observed capacity degradation data of prismatic batteries (50 Ah, Huali
company) are shown in Fig. 5.1. The results were obtained after storage at 30% (black curve),
70% (blue curve) and 100% (red curve) at 20oC. After storage at 30% SoC for more than 3000
hours the capacity decreased 0.935 Ah, approximately 2% of the initial value, while storage at
70% and 100% induced a two times higher degradation with losses of 1.844 Ah and 1.841 Ah,
respectively. Under the present storage conditions the electrodes decay are considered to be
negligible, therefore, the capacity losses are mainly attributed to the SEI formation [2, 3].
56
55 30%
Qmax / Ah
54
70%
t
53
100%
52
51
0 1000 2000 3000
Time / h
Fig. 5.1. Measured discharge capacity of prismatic batteries (50Ah, Huali company) under various
indicated SoC and room temperature as a function of storage time.
97

It is well known that the graphite surface has already been covered by the SEI layers after
the activation cycle(s) performed by the manufacture. The SEI layers consist of a dense inner
layer and a porous outer layer [30, 39, 40, 50]. The porous structure of the outer layer can
facilitate the solvent molecular passing through. However, the compact inner layer is tightly
adhered to the graphite surface and will prevent solvent molecules from penetrating. The
continuous growth of the SEI layer can be explained by electron tunneling. Electrons can tunnel
continuously through the inner SEI layer and reduce the solvent molecules. Consequently, the
SEI layer will slowly grow at the interphase between the inner and outer SEI layer during
storage. The electrons for this reduction process are delivered by the oxidation of Li stored in
the graphite electrode.
The electron tunneling probability is largely determined by the energy barrier which is
influenced by the graphite electrode potential [2, 3]. The lower the graphite electrode potential,
the higher the tunneling probability; consequently, the capacity losses will be faster. Therefore,
the resulting capacity losses depend on the SoC and total storage time. A more detailed
discussion about the capacity losses under various storage conditions will be given in Chapter
7.
5.2.2 Aging of A123 Batteries
Charge and discharge curves of LFP batteries stored at SoC = 100% at different
temperatures are shown in Figs. 5.2a-c. A decline of the discharge capacity as a function of
storage time is found in all cases. It is evident that the capacity declines faster at 60oC than at
20 and 40oC. Moreover, an overvoltage increase is observed after storage at 60oC (Fig. 5.2c).
Figs. 5.2d-f show the corresponding (discharge) EMF curves after storage under various
conditions. Similarly, the capacity losses calculated from the EMF curves increase with
increasing storage temperatures; the irreversible losses are most pronounced at 60oC. The total
capacities obtained from the discharge curves are in all cases smaller than those determined
from the corresponding EMF curves due to kinetic limitations. As the EMF curves represent
the voltage profiles in the equilibrium state, the values for exclude the influence of
polarization caused by the charge transfer reactions and mass transport processes.
98


(a) 20 C
o (d) 20oC
3.5 3.5
Voltage / V
Voltage / V
0 month 0 month
2.5 1 months 2.5 1 month
2 months 2 months
5 months 5 months
10 months 10 months
1.5 1.5
(b) 40 C o (e) 40oC

3.5 3.5
Voltage / V
Voltage / V
0 month
0 month 2.5
2.5 1 months
1 months
2 months
2 months
5 months
5 months
10 months
10 months
1.5 1.5
(c) 60oC
(f) 60oC
3.5 3.5
Voltage / V
Voltage / V
0 month 0 month
2.5 2.5
1 month 1 month
2 months 2 months
5 months 5 months
10 months 10 months
1.5 1.5
0 1 2 0 1 2
Capacity / Ah Capacity / Ah
Fig. 5.2. Charge (1C) and discharge (1C) voltage curves after storage A123 batteries at SoC=100% and
20oC (a), 40oC (b) and 60oC (c). (d)-(f) are the corresponding extrapolated EMF curves. The different
voltage curves correspond to various indicated storage times.
5.2.2.1 Apparent capacity loss
20 C
o
40oC o
60 C
0.9 (a) (b) (c) 100%
ΔQapp / Ah
0.6 100% 100%

50%
0.3 50%
50%
10%
10%
0 10%
0 4000 8000 0 4000 8000 0 4000 8000

Time / h Time / h Time / h
Fig. 5.3. The apparent discharge capacity loss (Δ ) of A123 batteries as a function of storage time
at 20oC (a), 40oC (b) and 60oC (c) at 10% (black curves), 50% (blue curves) and 100% (red curves) SoC.
Note that the axes are kept the same for all temperatures.
99

The values for Δ are calculated from the discharge capacities under various storage
conditions shown in Figs. 5.3a-c, according to Eq. 4.5. Fig. 5.3 reveals that Δ increases in
most case as a function of SoC at all temperatures. Remarkably, Δ decreases as a function
of storage time at SoC=10% at 20oC as has also been reported by Kassem, et al. [7]. Δ is
found to be significantly larger at 60oC than at 20oC and 40oC, again clearly indicating
accelerated degradation at elevated temperatures.
5.2.2.2 Irreversible capacity losses
Δ has been calculated according to Eq. 4.4 from the EMF curves shown in Figs. 5.2d-f.
Fig. 5.4 shows the development of Δ as a function of storage time at various SoC and
temperatures. In line with the trend found for Δ in Fig. 5.3, Δ increases with increasing
SoC. The influence of the temperature on Δ is, however, more significant than the
dependence on SoC. The irreversible capacity losses at 60oC are significantly accelerated,
indicating that severe decay occurs at elevated temperatures.
20 C
o o
40 C 60oC
0.6
(a) (b) (c)
100%
0.4
50%
ΔQir / Ah
10%
100%
0.2 50%
100% 10%
10% 50%
0 4000 8000 0 4000 8000 0 4000 8000

Fig. 5.4. The irreversible capacity loss (Δ ) of LFP batteries stored at the indicated SoC at 20oC (a),
40oC (b) and 60oC (c). Black symbols represent Δ measured at SoC = 10%, blue symbols SoC = 50%
and red symbols SoC = 100%. Note that the axes are kept the same for all temperatures.
Δ represents the immobilization of cyclable Li ions and is mainly attributed to the SEI
formation on the graphite electrode. The SEI layers are well known for their dual role during
Li-ion battery operation. On the one hand, the SEI layers protect the electrode from solvent co-
intercalation, thereby preventing exfoliation of the graphite layers. On the other hand, SEI
continuously grows due to solvent reduction, initiated by the electron tunneling process through
100


the inner SEI layer. The electron tunneling represents a rate-determining step of the SEI growth
during storage [2, 3]. The electron tunneling probability increases with decreasing graphite
electrode potential. Therefore the SEI formation rate is indeed expected to increase at higher
SoC when the graphite electrode potential is lower.
ln(dQ ir / dt) / A -9.0
-10
100%
-11 50%
10%
-12
2.9 3.1 3.3 3.5
1 / T  1000 / K -1
Fig. 5.5. Temperature dependence of the rate of the irreversible capacity loss ( ⁄ ) at 10%, 50%
and 100% SoC.
Fig. 5.5 shows the rate of the irreversible capacity loss (ln ⁄ as a function of
temperature at different SoC after 4000 h storage. It is generally accepted that reaction rates
follow an Arrhenius-type dependence as a function of reciprocal temperature [35]. Although
the number of investigated temperatures for these elaborate experiments had to be limited, Fig.
5.5 clearly shows that accelerated degradation takes place at higher temperatures, suggesting
that an additional aging mechanism is responsible for the increased immobilization of lithium
under this condition.
The influence of temperature on the irreversible capacity loss of LFP batteries has been
extensively discussed [6, 13, 19, 22, 35, 51]. The iron dissolution from the LiFePO4 cathode is
considered as the most likely reason for faster degradation at elevated temperatures.
Subsequently, the metal ions migrate to the anode, pass the SEI layers and can be reduced at
the graphite/SEI interface. The deposition of metallic clusters at the graphite electrode has at
least three consequences:
(i) Iron reduction at the graphite electrode consumes electrons, which leads to
irreversible capacity losses;
101

(ii) The deposited iron clusters may influence the SEI morphology and may, being a
good electronic conductor, facilitate electron transport across the SEI layer, thereby
accelerating the SEI formation and Li immobilization;
(iii) The iron clusters may block Li intercalation into the graphite electrode, leading to
inaccessibility of the graphite layers.
The irreversible capacity losses related to mechanisms (i) and (ii) are accelerated at elevated
temperatures (see Fig. 5.5) as more metal ions can be dissolved from cathode and deposited on
the graphite surface. The inaccessibility of graphite particles related to mechanism (iii) occurs
only at 60°C, see detailed discussion in Section 5.4.
10% 50% 100%

0.6 (a) (b) (c)
-dV/dQ
0.3 III III III

I II I II I II
o 0
20 C
0.6 (d) (e) (f)
-dV/dQ
0.3 III III III

I II I II I II
o 0
40 C
0.6 (g) (h) (i)
-dV/dQ
0.3 III III III

I I II I II
II
o 0
60 C
0 1 2 0 1 2 0 1 2
Capacity / Ah Capacity / Ah Capacity / Ah
Fig. 5.6. ⁄ curves obtained from the discharge EMF curves before (black lines) and after
storage (red lines) at various SoC and temperature. The different stages, characteristic for the graphite
electrode, are indicated by I, II and III.
5.2.2.3 Analysis of / curves
Fig. 5.6 shows the development of the / curves before (black) and after (red)
storage at various indicated SoC and temperatures. The depressions are corresponding to the
plateaus and the peaks are corresponding to the slopes in the voltage curve of the graphite
102


electrode (see Figs. 5.2d-f). The total shrinkage of the red curves compared with the black
curves indicates the total irreversible capacity losses Δ . The capacity losses of the red curves
are increasing with SoC. It is also found that the capacity losses increase significantly when the
temperature increases as shown in Figs. 5.6a, d and g. The influence of the temperature is even
larger than the SoC.
10% 50% 100%

0.6 (a) (b) (c)
−dV/dQ
0.3 III III III

I II I II I II
o
20 C 0
0.6 (d) (e) (f)

−dV/dQ
0.3 III III III

I II I II I II
40oC 0
0.6 (i)
(g) (h)
−dV/dQ
0.3 III III III

I II I II I II
o
60 C 0
0 1 2 0 1 2 0 1 2
Capacity / Ah Capacity / Ah Capacity / Ah
Fig. 5.7. ⁄ curves for A123 batteries obtained from the discharge EMF curves before (black
lines) and after storage (red lines) at various SoC and temperatures. All curves are aligned with respect
to the first peak at approximately 0.8Ah. The different stages, characteristic for the graphite electrode,
are indicated by I, II and III.
Apart from the total capacity losses, the decreases in the width of the various regions can
also be observed from the red curves. The decline of the individual regions, for example, region
I and II, reflects different aging processes inside the battery. It has been identified that during
charging a pristine LFP battery it is thermodynamically more favorable to occupy the as-
denoted stage III sites in the graphite electrode followed by the occupation of stage II sites.
Finally, at the end of the charging process all stage I sites will be occupied. Therefore the
apparent reduction of voltage plateau I in Fig. 5.6 must be attributed to the Li-immobilization
process [52, 53]. The decrease of region II (Δ , ) is believed to be related to the graphite
materials blockage as discussed in section 4.4.
103

10% 50% 100%

0.2
(a) (b) (c)
ΔQir
ΔQ / Ah
0.1
ΔQ
ΔQ ir
ΔQ
ir ΔQ Li,I
ΔQ Li,I ΔQLi,II ΔQ
o Li,I Li,III
20 C 0 ΔQ ΔQ ΔQ
Li,II ΔQ
Li,III
Li,II Li,III
0.2
(d) (e) ΔQir (f)
ΔQ ΔQ
ir ΔQ ir
ΔQ / Ah
ΔQ Li,I
0.1 Li,I
ΔQ
Li,I
o ΔQLi,II ΔQ ΔQLi,II ΔQ ΔQ ΔQLi,III

40 C 0 Li,III Li,III Li,II
0.6
(g) (h) ΔQ
ir
(i) ΔQir
ΔQir
ΔQ
ΔQ / Ah
0.3 Li,I ΔQ
Li,I
ΔQ
Li,I
ΔQ ΔQ ΔQ ΔQ ΔQ ΔQ
o Li,II Li,III Li,II Li,III Li,II Li,III
60 C 0
0 4000 8000 0 4000 8000 0 4000 8000
Fig. 5.8. Δ (black curves), Δ , (red curves), Δ , (blue curves) and Δ , (cyan curves) as
a function of storage time at 20oC, SoC = 10% (a), 50% (b) and 100% (c); at 40oC, SoC = 10% (d), 50%
(e) and 100% (f); and at 60oC, SoC = 10% (g), 50% (h) and 100% (i).
In order to properly compare the changes in magnitude of region I and II, all curves are
aligned with respect to the first peak at approximately 0.8 Ah. Fig. 5.7 shows the /
curves after alignment. The width of region I clearly decreases as a function of SoC and
temperature which is mainly attributed to lithium immobilization in the SEI layer. The influence
of the temperature is more dominant than the influence of SoC. In contrast, a decrease in width
of region II can only be observed at 60oC. The as-obtained values for Δ , (red curves), Δ ,
(blue curves) and Δ , (cyan curves) as well as the total irreversible capacity losses (Δ
presented by the black curves) are shown in Fig. 5.8. It can indeed be concluded that Δ , is
very close to the values of Δ and that Δ , and Δ , are much smaller than Δ , in
most cases.
5.2.2.4 Resistance
Fig. 5.9 shows the development of the battery resistance during discharging after 6000 h of
storage under various conditions. The black curves show the pristine battery characteristics and
the red curves correspond to those after storage. The differences between the two curves are
almost negligible when batteries are stored at 20oC (Figs. 5.9a-c) and 40oC (Figs. 5.9d-f) at all
104


SoC. However, a significant resistance increase is found for batteries stored at 60oC, in
particular at SoC=50% (Fig. 5.9h) and 100% (Fig. 5.9i). The maxima found in the resistance
curves can be related to different (de)lithiation stages of the graphite electrode.
10% 50% 100%

60
(a) (b) (c)
40
R / m
20
pristine
o aged
20 C 0
60
(d) (e) (f)
40
R / m
20
o
40 C 0
60
(g) (h) (i)
40
R / m
20
o
60 C 0
0 1 2 0 1 2 0 1 2
Qout / Ah Qout / Ah Qout / Ah
Fig. 5.9. The development of the total battery resistance (A123) as a function of during constant
current discharging of pristine batteries (black curves) and batteries stored for 1 year at various indicated
conditions (red curves) at 20°C (a)-(c), 40°C (d)-(f) and 60°C (g)-(i). Storage experiments has been
carried out at 10% SoC ((a), (d) and (g)), 50% SoC ((b), (e) and (h)) and 100% SoC ((c), (f) and (i)).
The total battery resistance is generally considered to consist of three terms
[5.1]
where the ohmic resistance not only combines contributions of the current collectors,
internal (particle-particle) connections and electrode materials but also the migration
component of the electrolyte-related overvoltage [54], relates to the charge-transfer
resistances of the two battery electrodes and relates to the mass-transfer resistances,
including all Li+ diffusion processes taking place in LFP batteries.
105

60
20oC
10%
(a) 50%
R / m 
100%
30
0
60
R / m  (b) 40oC
30
0
60
(c) 60oC
R / m 
30
0
0 3000 6000
Time / h
Fig. 5.10. The development of the ohmic resistance ( ) for A123 batteries stored at various
temperatures and SoC = 10% (black symbols), SoC = 50% (blue symbols) and SoC = 100% (red
symbols).
Fig. 5.10 shows the development of the ohmic resistance measured during the initial stages
of discharging under different storage conditions. Details of the measurements can be found in
Chapter 3. is found to be essentially independent on the storage time and SoC at 20oC (Fig.
5.10a) and 40oC (Fig. 5.10b). However, the SoC has a significant influence on storage at 60oC.
increases significantly after about 3000 hours (Fig. 5.10c), which must be attributed to a
decrease of the electronic conductivity of the active battery materials [14, 17].
Fig. 5.11 shows the development of the calculated values for ( ).
Obviously, is smaller at higher temperatures than at lower temperatures. The

differences between the pristine (black curves) and stored batteries (red curves) are only
marginal, except at 60oC at SoC =100%. Obviously, the maxima in Fig. 5.11 correspond to
those found in Fig. 5.9. High temperatures will both increase the charge-transfer kinetics and
diffusion processes, which is in line with the decrease in [55]. The values for
(Fig. 5.11) are similar in magnitude to those for at 20oC and 40oC (Fig. 5.10), while
106


is much smaller than at 60oC, indicating that the ohmic resistances are more dominant at
higher temperatures.
10% 50% 100%

30 (a) (b) (c)
R - R / m
20
10
pristine
o aged
20 C 0
30 (d) (e) (f)

R - R / m
20
10
o
40 C
0
30 (g) (h) (i)
R - R / m
20
10
o
60 C 0
0 1 2 0 1 2 0 1 2
Fig. 5.11. Development of the sum of the charge-transfer and mass-transfer resistances on after
storage at 10%, 20oC (a), 50% SoC, 20oC (b), 100% SoC, 20oC (c), 10% SoC, 40oC (d), 50% SoC, 40oC
(e), 100% SoC, 40oC (f), 10% SoC, 60oC (g), 50% SoC, 60oC (h), 100% SoC, 60oC (i). The black lines
refer to pristine batteries, the red lines to the aged batteries after 6000 h storage.
5.2.3 Graphite electrode characterization
5.2.3.1 Scanning electron microscope (SEM)
In order to visually observe the changes of the graphite electrode after storage, the
morphologies of the graphite electrodes have been investigated by SEM. Fig. 5.12 shows an
example of the SEM images of both the pristine (a) and aged graphite electrodes (b). Quite
some changes of the graphite electrode can be seen after storage. The particle surface of the
pristine graphite electrode is smooth and clean (a) while it becomes rough and vague after
storage (b). The changes of the graphite electrode morphology are considered to result from the
SEI formation. The thickness of the SEI layers can be detected by XPS analysis which will be
discussed in the following section.
107

Fig. 5.12. SEM images of the graphite electrode before (a) and after (b) storage under 60oC for 6000
hours.
5.2.3.2 XPS analysis of SEI
(a) C 1s (b) C 1s
Intensity (ARB)
Intensity (ARB)
960s 960s
420s 420s
180s 180s
120s 120s
60s 60s
0s 0s
298 293 288 283 298 293 288 283

Binding Energy Binding Energy
Fig. 5.13. C 1s spectra obtained from the graphite electrode after storage for about 6000 hours at 20oC
(a) and 60oC (b).
Fig. 5.13 shows the evolution of C1s spectra of a graphite electrode after storage for about
6000 hours at 20oC (a) and 60oC (b) as a function of sputtering time in the XPS equipment. The
graphite signal could not be detected on the electrode surface at t = 0 at all temperatures,
indicating that the graphite surface is covered by SEI layers. Interestingly, it is found that the
intensity of C1s spectra of C6 (indicated by the arrow) at sputtering time of 60 s decreases as a
function of temperature. The intensity of the graphite C1s spectra at 20oC is quite high as
indicated in Fig. 5.13a, and becomes much weaker at 60oC as shown in Fig. 5.13b. It can be
concluded that the thickness of SEI layers on the graphite electrode increases as a function of
108


temperature. After a longer sputtering time the C6 signal becomes more and more significant as
the bare graphite surface has been reached.
The thickness of the SEI layer is proportional to the total irreversible capacity losses Δ .
The results observed in Fig. 5.13 show that the irreversible capacity losses are larger at elevated
temperatures than at lower temperatures. It should be noted that after rinsing some of the SEI
layers can be dissolved or detached. Therefore the real thickness of the SEI layers might be
thicker than what is detected by XPS.
5.2.3.3 XPS analysis of Fe deposition
The iron deposition on the graphite electrode has also been analyzed by XPS. Fig. 5.14
shows Fe 2p spectra obtained after 60 s sputtering on the dismantled graphite electrodes, which
have been stored for 6000 hours at different temperatures and at SoC = 10%. The peak at 707.6
eV is observed for the electrode stored at 60oC, and has been assigned to the 2p3/2 peak of
metallic iron.
Fe2p1/2 Fe2p3/2
60oC
Intensity (ARB)
40oC
20oC
740 730 720 710 700

Binding Energy / eV
Fig. 5.14. Fe 2p spectra obtained from the dismantled graphite electrode after storage A123 batteries at
different temperatures.
Fig. 5.15a shows the evolution of the Fe 2p spectra of a graphite electrode after storage for
6000 hours at 60oC, SoC =10% as a function of sputtering time in the XPS equipment. No
metallic iron was detected on the electrode surface at t = 0. However, the iron intensity quickly
increases with sputtering time to decrease again after longer sputtering times. The Fe 2p3/2 core
level peak is at 707.6 eV compared to a theoretical value of 706.8 eV for metallic Fe and 710
109

eV for Fe2+ ions [56]. The slight shift towards higher binding energies as compared to metallic
Fe suggests that the Fe atoms have some interactions with the SEI material, donating charge to
the atoms, having high affinity energies, such as O and F [57]. Fig. 5.15b shows the C 1s spectra
(284.6 eV) at the same etching times.
(a) Fe 2p (b) C 1s C1s
Fe2p Fe2p
960s 1/2 3/2
Intensity (ARB)
Intensity (ARB)
420s
960s
180s
420s
120s
180s
60s 120s
60s
0s
0s
740 730 720 710 700 298 293 288 283

Binding Energy / eV Binding Energy / eV
Fig. 5.15. Fe 2p spectra (a) and C 1s spectra (b) for a dismantled graphite electrode after various
sputtering times. The battery has been stored, in this example, for 6000 hours at temperature of 60oC at
SoC = 10%.
Graphite cannot be detected without sputtering at 0, indicating the existence of the SEI
layers. After sputtering the C6 signal slowly increases and saturates when the bare graphite
surface has been reached. Combining the results in Fig. 5.15a and b it can be concluded that the
iron clusters are embedded in the SEI layers covering the graphite surface when stored at high
temperatures. Contrastingly, no iron deposits could be identified by XPS inside the SEI layers
at low temperatures.
In order to quantitatively determine the total amount of the iron deposited on the graphite
electrode, ICP measurements have been carried out. Details will be discussed below.
5.2.3.4 ICP-OES measurements
The Fe deposition on the graphite electrode which has been confirmed by XPS analysis,
has been quantitatively determined by ICP measurements. Graphite electrodes dismantled from
3 different batteries, (i) pristine, (ii) cycled at 2 C-rate and 60oC for 1700 cycles, (iii) storage at
60oC and 10% for 10 months have been dissolved into 1M H2SO4 solutions for 10 days. The
total amount of iron deposited on the graphite is summarized in Table 5.1.
110


No iron has been detected in pristine graphite. However, considerable iron is found in the
aged graphite electrodes as shown in Table 5.1. In order to obtain the total amount of iron
deposited on the aged graphite electrode inside LFP batteries, the iron concentration measured
by ICP can be recalculated on the basis of the sample geometries. The parameters of the samples
as well as the complete battery are summarized in Table 5.1 (columns 2-4).
Table 5.1. Dimensions of dismantled and cut graphite electrodes and the amount of iron deposited
V/L / mg·L-1 / mg
W / cm L / cm A / cm2
(H2SO4) (ICP) (calculated)
(sample), pristine 2 2.9 11.6 0.015 - -
(sample), cycled 2 2.9 11.6 0.015 2.26 0.0339
(sample), stored 2 2.9 11.6 0.015 2.22 0.0333
(battery), cycled 5.5 175.5 1930.5 - - 5.64
(battery), stored 5.5 175.5 1930.5 - - 5.54
The surface A is 2 as the electrode is coated double-sided. The total amount of iron
( ) deposited on the graphite electrode of both samples and batteries is calculated and given
in the last column in Table 1. The total amount of iron deposited on the graphite electrode inside
the LFP batteries is calculated to be 5.64 mg when cycled 1700 cycles at 2 C and 60oC. This
amount is about the same (5.54 mg) after storage at 10% and 60oC for 10 months. The deposition
of Fe calculated above causes blockage of the graphite layered structure leading to the graphite
electrode capacity loss. Furthermore, the battery capacity fading will be accelerated due to the
so-called catalyst effect of these Fe clusters. The influence of Fe deposition on the battery
capacity loss will be quantitatively discussed in details in Chapter 7.
5.2.4 Quantification of the inaccessibility of the graphite electrode
The stability of the graphite electrode is highly dependent on the quality of the SEI layers.
Although extensive efforts have been carried out to improve the quality of the SEI layers,
graphite degradation is still observed during storage and cycling, especially at the elevated
temperatures.
The degradation of the graphite electrode has at least two distinctive features:
(i) Structural deformation of the graphite electrode can lead to an increase of the
electrode impedance.
111

(ii) The electrode decay will cause inaccessibility for Li intercalation leading to a
decline of the electrode capacity. Li plating will take place when the graphite
electrode capacity becomes smaller than the battery capacity.
In-depth understanding of the graphite degradation mechanisms and the quantitative
determination of the graphite electrode decay is therefore essential for the investigation of the
irreversible capacity losses inside the LFP batteries. Several mechanisms have been proposed
to explain the decay of graphite electrodes [46, 47] under cycling conditions. Strain generated
during the (de)lithiation processes was considered to be the main origin for graphite structure
decay. However, in case of the present storage experiments no current has been applied to the
batteries and no (de)lithiation is involved. The graphite decay mechanism can therefore not be
related to strain but must be attributed to the iron dissolution-precipitation mechanism. The
blockage of the graphite layered structure is considered to be the main reason responsible for
the graphite capacity loss [58]. The quantification of the graphite capacity loss will be discussed
below.
0.06
o
60 C
0.04
/ Ah
Li,II
ΔQ
0.02 o
40 C
o
0 20 C
0 2000 4000 6000 8000
Time / h
Fig. 5.16. Decrease of the second battery voltage plateau (Δ , as a function of storage time at SoC
= 10% and various indicated temperatures.
Fig. 5.16 shows, in more detail, the development of Δ ,II at different temperatures and at
SoC=10%. At lower temperatures (20oC and 40oC) Δ ,II is negligible but it becomes
considerable at 60oC. As can be concluded from the presence of plateau I voltage region in Fig.
5.7g there are still “stage I sites” available and the decrease of Δ ,II upon storage can therefore
112


only be explained by the fact that some sites in the graphite electrode have become inaccessible
at 60°C.
As discussed in the Chapter 4.5, the total graphite capacity loss can be calculated on the
basis of Δ , according to Eq. 4.20
QLi,II
QC6  . [5.2]
xII
It is generally assumed that 0.5 1 for , , 0.25 0.5 for , and 0 0.25
for , as is schematically indicated in Fig. 4.2b. Therefore, 0.5, 0.25 and
0.25. Fig. 5.16 indicates that no graphite material losses occur during storage at 20 and 40oC,
which is in line with the conclusions of Safari, et al. [5] who showed that no graphite electrode
decay takes place during storage at moderate temperatures. Δ ,II increases to 0.04 Ah after
6000 h storage at 60oC corresponding, according to Eq. 5.2, to a decrease of graphite electrode
capacity Δ of 0.16 Ah.
5.3 Calendar ageing model
From the above experiments it can be concluded that storage at higher temperatures has
quite some impact on the stability of LFP batteries. The various results show that iron
dissolution from the cathode and precipitation on the anode plays an essential role in the
graphite electrode degradation process. In order to understand the mechanistic details the
following battery model is proposed where both the low and high temperature degradation are
considered in Figs. 5.17 and 5.18, respectively.
Fig. 5.17a shows the low-temperature case where the SEI formation and the accompanying
lithium immobilization is considered to be the only process responsible for ageing. Electrons
are assumed to tunnel through the inner SEI layer, reducing solvent molecules at the inner/outer
SEI interphase. The development of the storage capacities are schematically shown in Fig. 5.18a.
The maximum storage capacity of the battery is considered to be equal to that of the
cathode ( as the capacity of the anode ( is oversized with respect to that of the
cathode. is considered to be constant at moderate temperatures because of its excellent
stability. It is well-known that during the activation process, commenced after the LFP batteries
have been manufactured, the SEI layers grow onto the graphite electrode. Consequently, lithium
ions are immobilized in the SEI. This process continues during the initial activation procedure
in the present experiments. The lithium immobilization process results in a reduction of the
113

reversible storage capacity indicated by Δ in Fig. 5.18a. The total reversible capacity
is indicated by the red area in Fig. 5.18a. The total number of host sites inside the graphite
electrode, as represented by , are still available and can be considered constant. According
to the XPS analyses no iron precipitation is found on the graphite electrode under low-
temperature conditions.
Fig. 5.17. Calendar ageing models at low (a) and elevated temperatures (b).
Apart from the SEI formation, other deterioration processes are clearly involved during
storage at elevated temperatures. Fig. 5.17b includes the following processes: accelerated SEI
formation, iron dissolution from the cathode and metal deposition onto the graphite electrode,
making parts of the graphite electrode inaccessible for Li intercalation. Iron ions can dissolve
from the cathode by a chemical exchange reaction with protons from the electrolyte [10].
Subsequently, these ions diffuse/migrate to the anode where they are electrochemically
converted into metallic iron, which is in good agreement with the present XPS results. These
metal clusters have two clearly visible detrimental effects:
(i) It accelerates the SEI formation by facilitating electronic transport;
114


(ii) It simultaneously blocks the graphene layers, making the various stages of the
graphite electrode partly inaccessible for lithium intercalation, as schematically
shown in Fig. 5.17b.
Fig. 5.18. Schematic representation of the development of the storage capacity of LFP batteries at low
(a) and elevated temperatures (b).
Fig. 5.18b illustrates the storage capacity losses upon storage at elevated temperatures. Δ
is found to be significantly larger at 60oC than at lower temperatures. A decline of , caused
by electrode blockage, is also indicated in Fig. 5.18b. The capacity losses of the cathode,
Δ , caused by the iron dissolution are considered to remain negligibly low during
storage [5] and is therefore not considered in Fig. 5.18b. Interestingly, both degradation
mechanisms have a distinctive appearance in the development of the ⁄ curves:
immobilization of lithium ions in the SEI layers are visualized by the reduction of the stage I
voltage plateau region only whereas, on the other hand, graphite electrode blocking induced by
the iron particles influence all three voltage regions, which can also be identified in the
⁄ curves.
115

5.4 Conclusions
Capacity degradations of the prismatic C6/LiFePO4 (50Ah) batteries have been investigated
under various storage SoC at room temperature. The capacity loss, which has been attributed to
the SEI formation on the graphite electrode, is found to be larger at higher SoC than that at
lower SoC.
The capacity loss and material decay of cylindrical C6/LiFePO4 batteries (A123 system)
during storage have been systematically investigated as a function of both SoC and temperature.
The EMF curves, which are regularly determined by mathematical extrapolation of the
measured voltage discharge curves, are used to investigate the aging mechanisms during storage.
The irreversible capacity loss, which is accurately calculated on the basis of the maximum
storage capacity estimated from the EMF curves, increases as a function of temperature and
SoC. It is concluded that the origin of Δ is mainly related to the lithium immobilization in
the SEI layers formed on the graphite electrode. The thickness of the SEI layers has been
determined by XPS analysis by sputtering. It is shown that the thickness of the SEI layers at
high temperatures is much thicker than at lower temperatures which is in line with the
conclusions of Δ .
Based on the EMF curves, / curves have been applied to investigate the aging
mechanisms. Detailed analyses of the / curves provide a non-destructive approach to
quantitatively determine the graphite inaccessibility. The accessibility of the graphite electrode
is found to be reduced at higher temperatures. The iron deposition on the graphite surface is
considered to be the origin for the graphite decay under storage conditions since the graphite
electrode will be blocked and the Li intercalation will be hindered. The presence of iron at
higher storage temperatures has been experimentally proven by XPS and ICP. The deposited
metal clusters on the graphite electrode facilitate electron transport through the SEI layer and,
consequently, accelerate the SEI formation and growth. Iron deposition was found to be
negligible at moderate temperatures.
5.5 References
112 (2002) 98-104.
(2014) 1-8.
116


A1447.
A1135.
[7] M. Kassem, J. Bernard, R. Revel, S. Pelissier, F. Duclaud, C. Delacourt, Journal of Power
Sources, 208 (2012) 296-305.
[9] Y.H. Ye, Y.X. Shi, A.A.O. Tay, Journal of Power Sources, 217 (2012) 509-518.
(2007) A40-A44.
[11] M. Koltypin, D. Aurbach, L. Nazar, B. Ellis, Journal of Power Sources, 174 (2007) 1241-
1250.
Power Sources, 195 (2010) 7415-7425.
[15] K. Edstrom, T. Gustafsson, J.O. Thomas, Electrochimica Acta, 50 (2004) 397-403.
[17] M. Klett, T.G. Zavalis, M.H. Kjell, R.W. Lindstrom, M. Behm, G. Lindbergh,
[18] M.H. Kjell, S. Malmgren, K. Ciosek, M. Behm, K. Edstrom, G. Lindbergh, Journal of
Power Sources, 243 (2013) 290-298.
[19] S. Grolleau, A. Delaille, H. Gualous, P. Gyan, R. Revel, J. Bernard, E. Redondo-Iglesias,
J. Peter, S. Network, Journal of Power Sources, 255 (2014) 450-458.
[21] H. Kim, S. Grugeon, G. Gachot, M. Armand, L. Sannier, S. Laruelle, Electrochimica
Acta, 136 (2014) 157-165.
117

[22] E. Sarasketa-Zabala, I. Gandiaga, L.M. Rodriguez-Martinez, I. Villarreal, Journal of

Power Sources, 272 (2014) 45-57.
445-448.
[24] K. Shimoda, H. Sugaya, M. Murakami, H. Arai, Y. Uchimoto, Z. Ogumi, Journal of the
[27] P. Liu, J. Wang, J. Hicks-Garner, E. Sherman, S. Soukiazian, M. Verbrugge, H. Tataria,
J. Musser, P. Finamore, Journal of the Electrochemical Society, 157 (2010) A499-A507.
[28] J. Wang, P. Liu, J. Hicks-Garner, E. Sherman, S. Soukiazian, M. Verbrugge, H. Tataria,
J. Musser, P. Finamore, Journal of Power Sources, 196 (2011) 3942-3948.
[29] C.R. Birkl, E. McTurk, M.R. Roberts, P.G. Bruce, D.A. Howey, Journal of The
Electrochemical Society, 162 (2015) 10.
[30] D. Aurbach, B. Markovsky, M.D. Levi, E. Levi, A. Schechter, M. Moshkovich, Y. Cohen,
[31] A.M. Andersson, K. Edstrom, N. Rao, A. Wendsjo, Journal of Power Sources, 81 (1999)
286-290.
A1988.
[35] M. Broussely, S. Herreyre, P. Biensan, P. Kasztejna, K. Nechev, R.J. Staniewicz, Journal
of Power Sources, 97-8 (2001) 13-21.
A456-A462.
A250.
[38] X.K. Lin, J. Park, L. Liu, Y. Lee, A.M. Sastry, W. Lu, Journal of the Electrochemical
Society, 160 (2013) A1701-A1710.
118


[40] D. Aurbach, B. Markovsky, A. Rodkin, M. Cojocaru, E. Levi, H.J. Kim, Electrochimica

Acta, 47 (2002) 1899-1911.
[41] K. Edström, M. Herstedt, D.P. Abraham, Journal of Power Sources, 153 (2006) 380-384.
238 (2013) 454-463.
[43] H.C. Shin, W.I. Cho, H. Jang, Journal of Power Sources, 159 (2006) 1383-1388.
Letters, 9 (2006) A537-A541.
195 (2010) 3655-3660.
[47] Y. Qi, S.J. Harris, Journal of the Electrochemical Society, 157 (2010) A741-A747.
Sources, 101 (2001) 238-247.
[53] A.J. Smith, J.C. Burns, D. Xiong, J.R. Dahn, Journal of the Electrochemical Society, 158
(2011) A1136-A1142.
[54] D. Danilov, P.H.L. Notten, Electrochimica Acta, 53 (2008) 5569-5578.
[55] C.T. Hsieh, C.T. Pai, Y.F. Chen, P.Y. Yu, R.S. Juang, Electrochimica Acta, 115 (2014)
96-102.
[56] P.C.J. Graat, M.A.J. Somers, Applied Surface Science, 100 (1996) 36-40.
[57] M. Fahlman, H. Guan, J.A.O. Smallfield, A.J. Epstein, in: Annual Technical Conference
- ANTEC, Soc Plast Eng, Atlanda, Georgia, United States, 1998, pp. 1238.
Acta, 190 (2016) 1124-1133.
119

120

Chapter 6
Degradation Mechanisms of LFP Batteries
Experimental Analyses of Cycling-induced Aging
Electromotive force (EMF) voltage curves are regularly determined to facilitate in-depth
understanding of aging mechanisms of LFP batteries during cycling. The irreversible capacity losses
under various cycling conditions and temperatures are accurately obtained from the extrapolated
EMF curves and are found to increase with cycle number and time. A new mathematical
extrapolation method is proposed to distinguish between calendar ageing and cycling-induced
ageing. The capacity losses due to calendar aging are obtained by extrapolating the total irreversible
capacity losses to zero cycle number. It is found that calendar ageing increases logarithmically in
time. On the other hand, cycling-induced ageing is accurately determined by extrapolating the
capacity losses to zero time. In this case the capacity losses are found to increase linearly with cycle
number. It is furthermore found that iron dissolution from the cathode at 60oC and the subsequent
deposition onto the anode enhances significantly the SEI formation on the graphite electrode and,
consequently, battery ageing. Interestingly, the graphite electrode decay has been quantified in much
more detail, by analyzing the / curves. The analyses show that the electrode decay can be
related to both the structural deterioration and the inter-layer surface blockage of the graphite
electrode, as has also been experimentally confirmed by Raman and XPS spectroscopy.
121

Chapter 6 Degradation mechanisms of LFP batteries: Experimental Analysis of Cycling-induced Aging
6.1 Introduction
Aging of Li-ion batteries is one of the most important issues in the development of EV and
HEV. Many studies have been carried out to investigate the cycling performance of Li-ion
batteries in order to understand the underlying aging mechanisms [1-22]. For example, Li et al.
[1-3] have investigated the battery capacity degradation by both experiments and simulations.
Dubarry et al. [4-6] studied battery aging on the basis of / analysis and Safari et al. [7-9]
have investigated the capacity fade of LFP batteries at different cycling temperatures.
Christensen et al. [20] discussed the battery capacity loss by mathematical simulations and
Smith et al. have investigated battery aging at low currents and elevated temperatures [21].
Cyclable Li-ion losses and electrode materials decay are considered to be the most important
degradation mechanisms [23-27]. The irreversible capacity losses due to Li immobilization are
mainly attributed to the SEI formation onto the graphite electrode. Volume changes induced by
cycling [28-33] and active materials blockage [3] are considered to be the main reasons, causing
the graphite electrode decay. In addition, the cathode dissolution has also been reported at
elevated temperatures [14]. Metal dissolution from the cathode at elevated temperatures and the
subsequent deposition onto the anode leads to accelerated battery capacity losses. Cathode
dissolution is considered to result from water contamination in the electrolyte [34-40]. Material
decay reduces the electrode capability to reversibly accommodate Li-ions but, simultaneously,
also increase the electrodes resistance [13, 14].
C6/LiFePO4 (LFP) batteries are excellent candidates to investigate battery aging as the
cathode has outstanding stability properties at moderate temperatures and the focus can
therefore be put on the anode degradation processes. In this chapter the degradation of LFP
batteries have been studied under various cycling conditions, such as different currents,
temperatures, as well as cycle ranges. The influence of the cycle number and time on the
irreversible capacity losses will be discussed in detail. The anode capacity losses caused by the
structural degradation of the graphite electrode will be analyzed on the basis of the /
curves. The anode degradation has been experimentally confirmed by Raman spectroscopy.
Iron dissolution and precipitation, induced at elevated temperatures, is confirmed by XPS
spectroscopy.
122


6.2 Results
6.2.1 Cycling prismatic batteries (50Ah)
Prismatic batteries have been cycled at 0.1C and at room temperature. Fig. 6.1 shows the
development of the normalized capacity as a function of total cycling time (black curve). As
expected the battery capacity decreases as a function of cycle numbers. For convenience the
capacity loss during cycling is compared with those unraveled under storage at various SoC
reported in chapter 5 (see colored curves in Fig. 6.1). It can be seen that the capacity decay
during cycling is faster than that under storage. The capacity losses under both cycling and
storage are plotted in Fig. 6.2. It can be concluded that the capacity losses resulting from cycling
are higher compared to those found upon storage, which is in agreement with the experimental
data reported before [1]. This is caused by the volumetric changes of the graphite electrode
during charging and discharging.
101
100
30%
Normalized capacity %
99
98
100%
97
70%
96
95 0.1C
94
0 500 1000 1500 2000 2500 3000 3500
Time [ h ]
Fig. 6.1. Normalized capacity decay of 50 Ah batteries upon storage at various SoC: 30% (red curve),
70% (blue curve), 100% (magenta curve) and upon cycling at 0.1 C-rate (black curve).
The aging performance of C /LiFePO batteries under various cycling conditions will be
investigated in more detail with smaller size batteries, namely the A123 (2.3Ah) system. The
aging results of this battery type will be discussed in the following sections.
123

3.0
2.5
0.1C
2.0
Capacity loss [ Ah ]
100%
1.5
70%
1.0
0.5 30%
0.0
-0.5
0 500 1000 1500 2000 2500 3000 3500 4000
Time [ h ]
Fig. 6.2. The capacity losses under storage at various SoC: 30% (red curve), 70% (blue curve), 100%
(magenta curve) and upon cycling at 0.1 C-rate (black curve).
6.2.2 Cycling cylindrical batteries (2.3Ah) in the full SoC range
3.6 (a) 0.1C (b) 0.5C

3.0
Voltage / V
2.5 0cycle 0cycle

30 100
60 200
2.0 100 300
150 400
200 500
1.5
3.6 (c) 1C (d) 2C

3.0
Voltage / V
2.5
0cycle 0cycle
200 300
2.0 500 800
800 1200
1050 1500
1.5
0 0.5 1.0 1.5 2.0 2.5 0 0.5 1.0 1.5 2.0 2.5
Qout / Ah Qout / Ah
Fig. 6.3. Development of the extrapolated EMF curves as a function of cycle number at 0.1C (a), 0.5C
(b), 1C (c) and 2C (d) and at 60oC.
124


Figs. 6.3a-d show the development of the extrapolated EMF curves upon cycling A123
batteries at various discharge-rates at 60oC. Each discharge voltage curve corresponds to
specific cycle number indicated in the legend. It is clearly visible that the maximum storage
capacity declines upon cycling significantly faster at higher current than at low current.
Figs. 6.4a-c illustrate the development of the normalized maximum capacities ( / )
as a function of cycle number at various temperatures and cycling currents. It can be seen that
/ decreases faster at higher temperatures than at lower temperatures. Fig. 6.4a
reveals that influence of the current is minor at low temperature, but becomes significant at
elevated temperatures as shown in Figs. 6.4b-c. From Figs. 6.4a-c it follows that /
decreases faster at lower currents. Figs. 6.4d-f illustrate / as a function of time at
various temperatures and cycling currents. The influence of current is clearly observed at low
temperature (see Fig. 6.4d). In contrast with the results shown in Figs. 6.4a-c, the decrease of
/ is found to be faster at higher currents than at lower currents.
20oC 40oC 60oC

100
Qtmax / Q0max / %
90
80 0.1C
0.5C
1C
70
(a) 2C
(b) (c)
0 500 1000 1500 0 500 1000 1500 0 500 1000 1500
Cycle number Cycle number Cycle number
100
/ %
90
max
/ Q0
max
80
Qt
0.1C
0.5C
1C
70
(d) 2C
(e) (f)
0 2500 5000 0 2500 5000 0 2500 5000
Fig. 6.4. Normalized maximum capacities ( / ) as a function of cycle number at various

cycling currents and at 20oC (a), 40oC (b) and 60oC (c). (d)-(f) show the development of / at
o o o
various cycling currents as a function of cycle time at 20 C (d), 40 C (e) and 60 C (f).
125

6.2.2.1 Irreversible capacity losses ( )
As it has been defined in Chapter 4, the irreversible capacity loss (Δ ) is the difference
between and . Figs. 6.5a-c illustrate the development of Δ as a function of cycle
number at various cycling currents and temperatures. In line with the results observed in Figs.
6.4a-c, Δ is found to be larger at the elevated temperatures than at lower temperatures.
Furthermore, it can be concluded that Δ is larger for lower currents (see Figs. 6.5a-c) when
plotted as a function of cycle number. However, as shown in Figs. 6.5d-f, Δ is larger at
higher currents when plotted as a function of cycle time. Summarizing Figs. 6.5a-f it can be
concluded that Δ is both cycle number and cycling time dependent. In order to have an in-
depth understanding of the influence of the cycle number and time on Δ , a new mathematical
extrapolation method is proposed to distinguish between calendar aging and cycling-induced
aging.
0.8
20oC 40oC 60 C
o
0.1C
(a) 0.5C (b) (c)
1C
0.6 2C
ΔQ / Ah
0.4
ir
0.2
0
0 500 1000 1500 0 500 1000 1500 0 500 1000 1500
0.8
(d) 0.1C
0.5C (e) (f)
1C
0.6 2C
ΔQ / Ah
0.4
ir
0.2
0
0 2500 5000 0 2500 5000 0 2500 5000
Fig. 6.5. Irreversible capacity losses (Δ ) as a function of cycle number at various cycling currents at
20oC (a), 40oC (b) and 60oC (c). (d)-(f) show the development of Δ under various cycling currents as
a function of cycle time at 20oC (a), 40oC (b) and 60oC (c).
126


Fig. 6.6a shows the development of Δ as a function of both cycle number and cycling
time at 60oC. Δ clearly increases with both cycle number and cycling time, as indicated by
the red and blue arrows, respectively. For a given cycle time Δ increases significantly with
increasing discharge current because the cycle number is, obviously, much larger at higher
currents. In order to identify the influence of the calendar time on the total irreversible capacity
loss and to exclude the cycling effect, a mathematical extrapolation towards zero current was
applied. The irreversible capacity loss obtained by extrapolating towards zero current is defined
as calendar ageing (Δ ). Obviously Δ only depends, in a logarithmic way, on time and
not on current and cycle number (see red region in Fig. 6.6a).
Fig. 6.6. (a) Three-dimensional representation of Δ measured upon cycling at 0.1, 0.5, 1 and 2C and
plotted as a function of cycle number and time. The capacity loss due to calendar ageing (red region)
has been obtained by extrapolation to I = 0C. (b) Ageing induced by cycling (red region) is obtained by
extrapolation to t = 0.
Fig. 6.6b shows the development of Δ as a function of cycle number up to 300 cycles,
which is the maximum cycle number reached under low-current cycling conditions. Again,
Δ increases with both cycle number and time. Strikingly, for a given cycle number, Δ
increases with decreasing current, as is indicated by the red arrow. The reason for this behavior
is related to the much longer duration of the cycling experiments at lower currents, as indicated
in Table 3.2. Similarly, in order to eliminate the influence of calendar ageing and to study only
the influence of cycling, the total irreversible capacity losses have been mathematically
extrapolated to zero time, as is indicated by the red region in Fig. 6.6b. The irreversible capacity
loss due to cycling has been attributed to the SEI formation on the graphite electrode and the
volumetric changes of this electrode upon (dis)charging [1]. It has been shown that under these
127

conditions the existing SEI layer is partly peeled off, exposing new uncovered graphite surface
where new SEI will start to grow [1]. The capacity loss related to this peeling-off process has
been denoted as . Remarkably, a linear relationship between and cycle number is
found in Fig. 6.6b, in line with the model prediction reported before [1]. A detailed discussion
of will be given in Section 6.3.
Fig. 6.7. Development of Δ and corresponding values for Δ and at 20oC, 40oC and 60oC as
a function of time (a-c) and cycle number (d-f).
Fig. 6.7 shows an overview of the development of Δ as a function of cycling time (a-c)
and cycle number (d-f) at various indicated cycling currents and temperatures. The extrapolated
‘red areas’ in Figs. 6.7a-c represent the development of calendar ageing as a function of time.
It can be concluded that calendar ageing increases logarithmically with time and is significantly
higher at elevated temperatures. The extrapolated red areas for Δ as a function of cycle
number, shown in Figs. 6.7d-f, unravel the influence of cycling only. The capacity losses related
to cycling-induced ageing can be related to the accelerated SEI formation due to the volumetric
changes of the graphite electrode and are found to increase linearly with cycle number. Also in
this case a strong temperature dependence is found.
128


The mathematical extrapolation technique, adopted in Figs. 6.6 and 6.7, offers a new but
simple method to distinguish between the individual contribution of calendar ageing and
cycling-induced ageing during long-term cycling experiments. To the best of our knowledge,
this is the first time that the influence of cycling time and cycle number has been independently
extracted from conventional battery cycling experiments. Based on these analyses the various
ageing mechanisms can be analyzed and modeled in detail.
6.2.2.2 / analyses
/ curves can easily be obtained from the extrapolated EMF curves. It has been
reported that analysis of / curves is helpful to understand the aging mechanisms
during cycling [3-5, 41, 42]. For clarity reasons, only the curves at 0.1 and 2C-rate at 20oC (a
and b) and 60oC (c and d) are shown in Fig. 6.8 at various cycle numbers. The three indicated
regions I, II and III have been discussed in detail in Chapter 4 (Fig. 4.2). The peaks in the
/ curves correspond to the sloping regions in the EMF curves and the depressions
correspond to the plateaus.
1.0
(a) 0.1C (b) 2C
Qir
Qir
I II III
-dV/dQ
0.5
150 1600
100 1200
75 800
50 550
25 300
0 0
o 0
20 C
1.0
(c) 0.1C (d) 2C
Qir Qir
-dV/dQ
0.5
150 1700
110 1300
80 800
50 550
25 300
0 0
0
o
60 C 0 1 2 0 1 2
Qout / Ah Qout / Ah
Fig. 6.8. The development of dV /dQ curves without peak alignment at 0.1C (a and c) and 2C-rate
o o
(b and d) at 20 C and 60 C as a function of the various indicated cycle numbers.
129

/ curves shown in Fig. 6.8 are normalized at the charged state. Therefore, the
shrinkage of the various / curves indicated by the red arrow represents the total
irreversible capacity loss (Δ ). As discussed in Fig. 6.5, Δ increases as a function of both
current and temperature. Apart from the shrinkage of the whole / curves, the decrease
in the width of individual regions can also be observed, especially at the elevated temperatures.
The decline of regions I and II provides insight into the understanding of the aging mechanisms.
In order to facilitate a proper analysis of the width of regions I and II, a more logic alignment
has been adopted to replot these / curves. The details will be discussed in Fig. 6.9.
1.0
(a) 0.1C (b) 2C
I II III
−dV/dQ
0.5
150 1600
100 1200
75 800
50 550
25 300
0 0
20oC 0
1.0
(c) 0.1C (d) 2C
−dV/dQ
0.5
150 1700
110 1300
80 800
50 550
25 300
0 0
0
o
60 C 0 1 2 0 1 2
Q / Ah Q / Ah
out out
Fig. 6.9. The development of dV /dQ curves with peak alignment of the first peak at about 0.8 Ah
at 0.1C (a and c) and 2C-rate (b and d) at 20oC and 60oC as a function of the various indicated cycle
numbers.
Fig. 6.9 shows the replotted / curves with alignment on the basis of the first peak
at approximately 0.8 Ah. A small decrease of region I is observed in Fig. 6.9a. When the
discharge current increases to 2C-rate (Fig. 6.9b), the decrease of region I becomes somewhat
larger, indicating higher capacity losses at higher currents. The change of region II at 20oC is
130


negligible in both cases. However, the decrease of region I during cycling becomes significant
at 60oC, even at small current (Fig. 6.9c). In addition, the decrease of region II becomes very
pronounced at this high temperature, especially at 2C (Fig. 6.9d). This becomes even more
evident with the help of the red sloping lines, connecting the peaks of the various /
curves.
0.4
20 oC 40 oC 60 oC
(a) 0.1C (b) (c)
0.5C
QLi,I / Ah
1C
0.2
2C
0.4
(d) 0.1C (e) (f)
0.5C
QLi,II / Ah
1C
0.2
2C
0.4
(g) 0.1C (h) (i)
0.5C
QLi,III / Ah
0.2 1C
2C
0 600 1200 1800 0 600 1200 1800 0 600 1200 1800

Fig. 6.10. The development of Δ , (a-c), Δ , (d-f) and Δ , (g-i) as a function of cycle number
at various currents (0.1, 0.5, 1, 2C-rate) and temperatures (20, 40 and 60oC). For comparative reasons
the values of all axes are kept the same.
The capacity losses of the individual regions, denoted by Δ , ,Δ , and Δ , , are

obtained from the / curves under all cycling conditions. The specific values are
summarized in Fig. 6.10 as a function of cycle number for the various investigated temperatures
and discharge currents. It is found that Δ , increases strongly with increasing temperature
(Figs. 6.10a-c). Strikingly, Δ , decreases to even negative values during the initial stages of
cycling at 1C and 2C at 20oC (Fig. 6.10a). Both Δ , (d-f) and Δ , (g-i) also increase as a
function of cycle number.
131

6.2.2.3 Quantitatively determination of graphite capacity decay ( )
As described in Chapter 4.5, the graphite degradation can be quantified on the basis of
Δ , (Figs. 6.10d-f), according to Eq. 4.20
QLi,II
QC6  .
xII
where , as discussed in Chapter 5.2.4, is assumed to be 0.25. Fig. 6.11 shows the
development of the graphite electrode capacity decay under various cycling conditions as a
function of cycle number. Obviously, the development of Δ is in line with the trend of
Δ , as shown in Figs. 6.10d-f. It is found that Δ is small at 20oC but becomes much more
significant at 60oC, indicating a strong temperature dependence.
o o o
20 C 40 C 60 C
1.2
0.1C
(a) 0.5C
(b) (c)
1C
2C
0.8
QC / Ah
6
0.4
0
0 900 1800 0 900 1800 0 900 1800
Fig. 6.11. The capacity decline of the graphite electrode under various cycling currents (0.1, 0.5, 1 and
2 C) at 20oC (a), 40oC (b) and 60oC (c) as a function of cycle number.
Graphite degradation is considered to be induced by (i) structural deformation of the

graphite electrode and (ii) iron deposition onto the graphite electrode. Structural deformations
are generally attributed to the volume changes in the graphite electrode during cycling. These
volume changes induce stress, which leads to a poor connection between the graphite particles
as well as cracking of the graphite particles. The isolated/deformed graphite particles therefore
become inaccessible and this will result in an increase of Δ . Particle isolation and/or
structural deformation of graphite electrodes can be investigated by SEM and Raman analyses.
The iron deposition will be shown to occur by XPS measurements. Details will be discussed in
the following sections.
132


6.2.2.4 Raman spectroscopy
The integrated intensity ratio / between the D and G bands is widely used to
qualitatively characterize the amount of defects in graphite [43]. An increasing / ratio
implies an increasing number of materials defects. For example, Sethuraman et al. [33] have
reported the degradation of graphite electrodes after cycling by analyzing the evolutions of the
/ .
20oC G
o
40 C
o
60 C
Intensity
D
′
D
1200 1300 1400 1500 1600 1700

−1
Wavenumber / cm
Fig. 6.12. Raman spectra of the cycled graphite electrodes cycled at 0.1C at various temperatures. The
Gauss function was used to fit the peaks (bold lines). The intensities are normalized with respect to the
G band.
The Raman spectra of the graphite electrodes measured after cycling at various temperatures
are shown in Fig. 6.12. The intensities of the G bands in the various spectra are normalized with
respect to each other. The presence of the D and D’ bands clearly indicates the existence of
defects in all graphite electrodes. The peak height of the D band increases, however,
significantly at higher temperatures, indicating that the structural degradation of graphite
electrodes at high temperature cycling is more severe.
Fig. 6.13 shows the Raman mapping plots of the / ratio of the uncycled graphite
electrode (Fig. 6.13a) as well as the graphite electrodes after cycling at various temperatures
(Figs. 6.13b-d). The low values of the / ratio represented by dark blue color indicate a
better crystalline structure of graphite. It can be seen that the dark blue areas decrease with
133

increasing temperature, implying increased graphite disorder. The orange and red spots in the
maps indicate the inhomogeneity of graphite during cycling, especially at higher temperatures.
Fig. 6.13. Surface Raman maps of the / ratio of a 40 40 graphite electrode area at pristine
o o o
state (a) and after cycling at 2C-rate at 20 C (b), 40 C (c) and 60 C (d).
6.2.2.5 XPS analysis of Fe deposition
Precipitation of the iron ions at the graphite electrode under various aging conditions has
been investigated by XPS. The Fe 2p spectra for the graphite electrodes after cycling at various
temperatures are shown in Fig. 6.14. It is confirmed that no iron precipitation occurred on the
graphite electrode at 20 and 40oC. However, the Fe 2p signal located at 707.6 eV is clearly
detected when the graphite electrode was cycled at 60oC, indicating that severe Fe dissolution
and deposition onto the anode has taken place at elevated temperatures. The slight shift of this
Fe 2p3/2 peak with respect to the theoretical value of 706.8 eV is attributed to the interaction
between Fe and the atoms, such as O and F, having high affinity energies [44].
134

Fe2p
Intensity (ARB)
60oC
40oC
20oC
740 720 700

Binding Energy / eV
Fig. 6.14. Fe 2p spectra obtained from the dismantled graphite electrodes after cycling A123 batteries
at different temperatures.
C1s
(a) Fe 2p Fe2p1/2 Fe2p3/2
(b) C 1s
360s
Intensity (ARB)
Intensity (ARB)
240s
360s
180s
240s
120s
180s
60s 120s
60s
0s
0s
740 730 720 710 700 298 293 288 283

Binding Energy Binding Energy
Fig. 6.15. Fe2p spectra (a) and C 1s spectra (b) collected for a dismantled graphite electrode after cycling
at 60oC with a current of 0.5C at various sputtering time.
Fig. 6.15a shows the development of the Fe 2p spectra (707.6 eV for Fe 2p3/2) of the graphite
electrode cycled at 60oC at different sputtering time. Hardly any iron is detected at t = 0 s, while
the intensity becomes significant after t = 60 s. Fig. 6.15b shows the C1s peak (284.6 eV) for
C6 at the same etching time. No graphite can be detected at t = 0 s, indicating the existence of
the SEI layers. However, after sputtering the C6 signal slowly increases and saturates when
approaching the bare graphite surface. In contrast to the development of the C1s peak for C6,
135

the intensity of Fe 2p spectra decreases after longer sputtering time. Combining the results
shown in Figs. 6.16a and Fig. 6.15b it can be concluded that the iron clusters are embedded
inside the SEI layers, covering the graphite electrode surface. Precipitated iron partially blocks
the graphene layers leading to inaccessibility of graphite particles.
(a) C1s (b) (c)
240s
Intensity (ARB)
180s
120s
60s
0s
295 290 285 280 295 290 285 280 295 290 285 280
Binding Energy / eV
Fig. 6.16. The evolution of the C1s spectra of the dismantled graphite electrodes cycled at 20 oC (a), 40
o
C (b) and 60oC (c) at 0.1C as a function of the indicated sputtering time.
6.2.2.6 XPS analysis of SEI
The thickness of the SEI is also investigated by XPS. Figs. 6.16a-c show the evolution of
the C1s spectra of the graphite electrodes dismantled from the batteries after 300 cycles at 20,
40 and 60oC, respectively, as a function of the indicated sputtering time. For clarity reasons,
only the C1s spectra at t = 60 s are fitted with respect to the peaks of C6 and the various SEI
components. The C1s peak of C6 is located at 284.3-284.5 eV [45, 46]. Peaks located at higher
binding energies are assigned to various components of the SEI layers. For example, the peak
at 286.5 eV can be assigned to (CH2CH2O)n , that at 287.6 eV is corresponding to CH3OLi,
and the peak at 291 eV has been attributed to Li2CO3 or R-CH2OCO2Li [47].
The graphite signal cannot be detected on the electrode surface at t = 0 s at all temperatures,
indicating that the graphite electrode surface is covered by SEI layers. After longer sputtering
136


time the C6 signal becomes more and more significant when the bare graphite surface has been
reached. Interestingly, it is found that the intensity of C1s spectra of C6 (indicated by the arrows)
at a sputtering time of 60 s decreases as a function of temperature. The intensity of the graphite
C1s spectra at 20oC is quite high, as indicated in Fig. 6.16a. It becomes less significant at 40oC
and becomes weak at 60oC as shown in Figs. 6.16b and c, respectively. It can be concluded that
the thickness of SEI layers on the graphite electrode increases as a function of temperature.
(a)C1s (b) (c) (d)
320s
Intensity (ARB)
240s
180s
120s
60s
0s
300 290 280 300 290 280 300 290 280 300 290 280
Binding Energy / eV
Fig. 6.17. The evolution of the C1s spectra of the dismantled graphite electrodes cycled at 60 oC at 0.1C
(a), 0.5C (b), 1C (c) and 2C (d) at the same cycling time (t 3000 h) as a function of the indicated
sputtering time.
Fig. 6.17 shows the influence of the cycling current on the thickness of the SEI layers. The
graphite electrodes dismantled from the batteries have been cycled for the same time. It is
known from the discussions in the previous sections that the irreversible capacity losses at high
currents are larger than at low currents. However, it can be concluded from Fig. 6.17 that the
thickness of the SEI layers is very similar at various cycling currents, indicating a time
dependence and cycle number independence of the SEI thickness. This results will be explained
in detail in the discussion section of this chapter.
137

6.2.3 Cycling cylindrical batteries in various SoC ranges
Figs. 6.18a-c show an example of the cycling voltage curves in various indicated cycling
windows at 60oC. The various SoC ranges are controlled by the corresponding cut-off voltages.
The capacities released in various cycling windows vary as a function of cycle number. The
decline of the cycling capacities is attributed to the increase of the battery impedance since the
SoC is quite sensitive to the electrode potential in the flat voltage regions. Figs. 6.18d-f show
the development of the corresponding EMF curves as a function of cycle number. An obvious
decline of the maximum capacities can be observed at all cycling windows. However, the
decline at higher SoC range is faster than at lower SoC.
3.5 (a) (b) (c)

70-100% 35-65%
Voltage / V
0-30%
2.5
1.5
0 0.4 0.8 0 0.4 0.8 0 0.4 0.8
3.5
I II (d) I II
(e) I II (f)
III III III
Voltage / V
70-100% 35-65% 0-30%
2.5 0cycle
150
300
450
600
750
900
1.5
0 1 2 3 0 1 2 3 0 1 2 3

Fig. 6.18. Cycling voltage curves in various SoC-windows: (a) 70-100%, (b) 35-65%, (c) 0-30% at 60oC.
(d)-(f) illustrate the corresponding EMF curves as a function of cycle number.
The maximum capacities at various SoC-windows and temperatures are obtained

from the corresponding EMF curves. Fig. 6.19 shows the development of / as a
function of cycle number. It is evident that / declines faster at 60oC than at 20 and
40oC. Apart from the temperature dependence, / is also found to be strongly
depended on the SoC-windows. / declines faster at higher SoC (red curves in Fig.
138


6.19). / deceases fastest when cycling in the full SoC range (light blue curves),
indicating that the wider SoC-windows leads to faster capacity decay.
o o o
20 C 40 C 60 C
100
95
Qtmax / Q0max / %
90
85 0−30%
35−65%
70−100%
(a) 0−100% (b) (c)
80
0 500 1000 1500 2000 0 500 1000 1500 2000 0 500 1000 1500 2000

Fig. 6.19. Development of / in various SoC-windows (0-30%, 35-65%, 70-100% and 0-

100%) at 20oC (a), 40oC (b) and 60oC (c) as a function of cycle number.
o o o
20 C 40 C 60 C
100
95
Qtmax / Q0max / %
90
85 0−30%
35−65%
70−100%
(a) 0−100% (b) (c)
80
0 2000 4000 6000 0 2000 4000 6000 0 2000 4000 6000
Et / Wh Et / Wh Et / Wh

Fig. 6.20. Development of / in various SoC-windows (0-30%, 35-65%, 70-100% and 0-

o o o
100%) at 20 C (a), 40 C (b) and 60 C (c) as a function of energy-throughput ( ).
Battery capacity losses are often plotted as a function of cycle number or cycling time (Fig.
6.19). However, from an application point of view, the energy-throughout ( ) upon aging is
also highly relevant. has been defined as the accumulation of energies extracted from the
batteries during the entire cycling period. The energy-throughput is essential for practical
battery applications.
The battery capacity decline as a function of at various cycling windows and
temperatures are shown in Fig. 6.20. Again, the elevated temperature is a leading factor in
139

accelerating the capacity decay. / decreases faster at high SoC-windows, in line with
the trend observed in Fig. 6.19. However, the decrease rate of / at full cycling SoC-
window at various temperatures in Fig. 6.20 deviates from the trend visible in Fig. 6.19. The
influence of cycling windows on / is found to be minor at 20oC (Fig. 6.20a), but
becomes pronounced at elevated temperatures (Figs. 6.20b and c). From Figs. 6.20b and c it
follows that the decrease rate of / for SoC = 0-100% is smaller than cycling in the
range of SoC = 70-100% at 40oC (Fig. 6.20b) and is small compared to other partial cycling
SoC-windows at 60oC (Fig. 6.20c).
6.2.3.1 Irreversible capacity losses ( )
The irreversible capacity losses Δ have been calculated according to Eq. 4.4 on the basis
of maximum capacities shown in Fig. 6.19. The development of Δ as a function of
cycle number under various cycling windows at 20, 40 and 60oC is shown in Fig. 6.21. Δ
obtained in the high SoC-window is larger than that in the low SoC-window. The influence of
temperature on Δ is more pronounced than the influence of the SoC-windows. The
irreversible capacity losses at 60oC are significantly accelerated indicating that severe decay
occurs at elevated temperatures.
o o
20oC 40 C 60 C
0.5
(a) 0−30%
35−65% (b) (c)
70−100%
0.4
ΔQ / Ah
0.3
ir
0.2
0.1
0
0 500 1000 1500 2000 0 500 1000 1500 2000 0 500 1000 1500 2000
Fig. 6.21. The irreversible capacity loss Δ in various cycling windows (0-30%, 35-65% and 70-100%)
at 20oC (a), 40oC (b) and 60oC (c) as a function of cycle number.
6.2.3.2 / analyses
Fig. 6.22 shows the / curves obtained from the corresponding EMF curves.
Different curves are corresponding to the various indicated cycle numbers. In order to
discriminate the changes in width of Region I and II, all curves are aligned with respect to the
140


first peak at about 0.8 Ah. The cycling-induced changes in width of region I and II are indicated
by red arrows. Figs. 6.22c-i clearly show that Region I decreases as a function of SoC and
temperature. The reduction of region I has been related to Li+ immobilization in the SEI layer
[3]. Strikingly, region I increases during the initial stages of cycling between SoC = 0-30% at
20 and 40oC, as indicated in Figs. 6.22a and b, respectively. This phenomenon was also
observed in the storage experiments at 20oC, SoC = 10% reported before [3] (see also Chapter
5).
20oC 40oC 60oC

(a) (b) (c)
III
dVEMF / dQ
I II
1200
900
600
300
0
0-30%
(d) (e) (f)

dVEMF / dQ
35-65%
(g) (h) (i)

dVEMF / dQ
70-100%
0 1 2 0 1 2 0 1 2
Fig. 6.22. The development of the dV /dQ curves obtained from the discharge EMF curves after
cycling at various temperatures and SoC. The different stages, characteristic for the graphite electrode,
are indicated by I, II and III. The evolutions of region I and II are indicated by red arrows.
The blue vertical lines indicate the position of the first and second peaks at cycle number is
0. A red sloping line connected the second peak of various / curves in each figure is
added to make the changing of Region II more visible. The decrease of Region II is clearly
observed in Figs. 6.22a-c, f and i, which correspond to SoC window 0-30% at all temperatures
and at 60oC at all cycling windows.
141

20oC 40oC 60oC

0.4
0-30%
(a) 35-65% (b) (c)
QLi,I / Ah 70-100%
0.2
0.0
0.4
(d) (e) (f)
QLi,II / Ah
0.2
0.0
0.4
(g) (h) (i)
QLi,III / Ah
0.2
0.0
0 1000 2000 0 1000 2000 0 1000 2000

Fig. 6.23. The evolution of Δ , (a-c), Δ , (d-f) and Δ , (g-i) as a function of cycle numbers at
various SoC (0-30%, 35-65%, 70-100%) and temperatures (20, 40 and 60oC).
The capacity losses of the individual regions, denoted by Δ , ,Δ , and Δ , , are

obtained from the / curves under all cycling conditions. The specific values are
summarized in Fig. 6.23 as a function of cycle number for the various investigated temperatures
and SoC-windows. It is found that Δ , is larger in the high SoC-windows. A strong
temperature dependence of Δ , is also observed as can be seen in Figs. 6.23a-c. Strikingly,
as shown in Figs. 6.23a and b, Δ , decreases to even negative values during cycling at 0-30%
at 20 and 40oC. From Figs. 6.23d-e it can be seen that Δ , is hardly affected during cycling
at 35-65% and 70-100% at both 20 and 40oC but is much more significant at SoC = 0-30% for
all temperatures. Fig. 6.23f shows the development of Δ , as a function of cycle number at
various SoC-windows and 60oC. It can be seen from Fig. 6.23f that Δ , becomes
considerable at all SoC-windows but is more significant at 0-30%. The development of Δ ,
is very similar to that of Δ , , as shown in Figs. 6.23g-i.
142


6.2.3.3 Quantitatively determination of graphite capacity decay ( )
As described in chapter 4.5, the graphite degradation can be quantified on the basis of Δ ,
(Figs. 6.23d-f), according to Eq. 4.20 assuming that 0.25 . Fig. 6.24 shows the
development of the graphite capacity decay under various cycling conditions as a function of
cycle number. Obviously, the development of Δ is in line with the trend of Δ , shown in
Figs. 6.23d-f. It can be concluded that the graphite electrode decay is more significant at lower
SoC (0-30%) and becomes substantial at elevated temperatures.
o
20oC 40oC 60 C
1.0
(a) 0−30%
35−65%
(b) (c)
70−100%
ΔQC6 / Ah
0.5
0
0 1000 2000 0 1000 2000 0 1000 2000
Fig. 6.24. The development of Δ under various cycling windows at 20oC, (a), 40oC (b) and 60oC (c)
as a function of cycle number.
6.2.3.4 Raman analysis
Fig. 6.25 shows the Raman mapping plots, revealing the ratio of the / Raman
reflections [33] of the graphite electrodes after cycling at various temperatures and cycling
ranges. The Raman map of an uncycled electrode is shown in Fig. 6.13a. In line with Fig. 6.13,
the low values of the / ratio represented by the dark blue color indicate a better crystalline
structure of graphite. It can be seen that the dark blue areas clearly decrease at lower SoC,
indicating increased graphite disorder. This becomes even more pronounced at higher
temperatures. The orange- and red-colored spots indicate that more inhomogeneity is induced
in the graphite electrode upon cycling, especially at higher temperatures. Based on the results
from Fig. 6.25 we can conclude that both the SoC window and temperature have a significant
influence on the graphite electrode degradation.
143

Fig. 6.25. Surface Raman maps of the / ratio of a 40 40 graphite electrode area after
cycling at 20oC in SoC-windows of 70-100% (a), 0-30% (b) and at 60oC in SoC-windows of 70-100%
(c), 0-30% (d).
6.2.3.5 XPS analysis
In order to investigate the influence of the SoC-window and the temperature on the SEI
growth, the C1s spectra of the dismantled graphite electrodes cycled at various SoC-windows
and temperatures have been analyzed as a function of sputtering time. Fig. 6.26 illustrates the
influence of the SoC-windows on the growth of the SEI layers at 40oC. No graphite signal was
detected on the electrode surface at t = 0 s, indicating the existence of SEI layers. The graphite
signal appears after sputtering 20 s on the graphite electrode cycled at 0-30% (Fig. 6.26a). The
location of the graphite peak is indicated by the blue arrow. The C1s spectra with the same
intensity of the graphite signal at various cycling SoC-windows are plotted in red (Figs. 6.26a-
c). It can be seen that to obtain the same intensity of the graphite peak, the sputtering time is t
= 40 s in Fig. 6.26a, 60 s in Fig. 6.26b and 80 s in Fig. 6.26c. The increase of the sputtering
time indicates that the SEI layers grow faster at high SoC-window (70-100%). It can be
concluded that the evolution of the SEI layers in the various cycling SoC-windows is in line
with the increasing trend of the irreversible capacity losses observed in Fig. 6.24.
144

(a) C1s (b) (c)

200s
Intensity (ARB)
180s
160s
140s
120s
100s
80s 80s
60s 60s
40s
20s
0s
298 290 280 298 290 280 298 290 280

Binding Energy / eV
Fig. 6.26. The evolution of the C1s spectra of dismantled graphite electrodes cycled at 40oC in SoC-
windows of 0-30% (a), 35-65% (b) and 70-100% (c) as a function of indicated sputtering time. The blue
arrows indicate the signal of graphite.
(a) C1s (b) (c)

200s
Intensity (ARB)
180s
160s
140s
120s 120s
100s
80s
60s 60s
40s
20s
0s
298 290 280 298 290 280 298 290 280

Binding Energy / eV
Fig. 6.27. The evolution of the C1s spectra of the dismantled graphite electrodes cycled at SoC = 70-
100% and 20oC (a), 40oC (b) and 60oC (c) as a function of the indicated sputtering time. The blue arrows
indicate the signal of graphite.
145

Fig. 6.27 shows the C1s spectra of the graphite electrode cycled at SoC = 70-100% at
various temperatures as a function of the indicated sputtering time. Again, no graphite signal
was detected on the electrode surface at t = 0, indicating the presence of SEI layers. The graphite
signal appears at t = 20 s (blue arrow) when the electrode was cycled at 20oC (Fig. 6.27a). The
C1s spectra for the electrode cycled at 40 and 60oC with the same intensity of the graphite peak
as at t = 20 s in Fig. 6.27a has been plotted by the red curve in Figs. 6.27b (t = 60 s) and
c (t = 120 s). The significant time increase shows that the thickness of the SEI layers increases
with temperature. The dependence of the SEI layers on the cycling SoC-windows and
temperatures as concluded from Fig. 6.26 and 6.27 are in line with the development of the
irreversible capacity losses shown in Fig. 6.24.
6.3 Discussion
6.3.1 SEI formation model
The SEI formation onto the graphite electrode is generally considered to be the origin of the
irreversible capacity losses in LFP batteries [48]. The SEI formation during storage at various
temperatures has been systematically discussed in Chapter 5. The details of the SEI formation
under cycling conditions will be discussed here. Various results show that iron dissolution at
elevated temperatures from the cathode and the subsequent precipitation on the anode plays a
significant role in the SEI formation [3]. Figs. 6.28a and b illustrate battery degradation models
for the low- and high-temperature situation, accordingly.
Fig. 6.28a shows the low-temperature case where the SEI formation is considered to be the
only process responsible for aging. Dense and stable SEI layers play an important role in
protecting the layered structure of the graphite electrodes, preventing solvent co-intercalation
and, consequently, improving the cycle life performance of Li-ion batteries. It has been argued
that the continuous growth of the SEI layers is controlled by the electron tunneling process
across the inner SEI layer [49]. The tunneling current is proportional to the surface area of the
graphite electrode [1, 2].
146

Fig. 6.28. SEI formation model under cycling conditions at low (a) and at elevated temperatures (b). Li
It is well-known that the surface area of the graphite electrode is changing during cycling
due to the volume expansion and/or fracturing of the electrode materials. The total surface area
of the graphite electrode can therefore be divided into two parts [1]: the SEI-covered surface
and the freshly formed surface area induced by the cracking process, as
schematically indicated in Fig. 6.28a. is assumed to be uniformly covered by the SEI
layers and is considered constant during cycling. , on the other hand, is generated by SEI
cracking due to the graphite volume expansion and/or particle fracturing during cycling. The
electron tunneling probability is strongly determined by the thickness of the inner SEI layer.
The SEI formation on is slow [1, 2] as the thickness of inner SEI layer has already been
well developed. The SEI formation on , on the other hand, is much faster as the fresh
electrode surface area is directly exposed to solvent molecules and the SEI layers will quickly
be formed on the uncovered graphite electrode surface.
The continuous growth of the SEI layers at lead to capacity losses, which are denoted
by . The SEI layers formed at during charging and the peeling off during the
147

cr
subsequent discharging lead to capacity losses, denoted by QSEI [1]. The total irreversible
capacity loss Qir can then be written as the summation of these two components
Qir  QSEI
cov
 QSEI
cr
. [6.1]
cov
QSEI is determined by , the total aging time ( and the electrode potential [1]. If the
cov
battery is in the equilibrium state at current = 0, then QSEI obviously corresponds to calendar
ageing losses ( Qirca ). These values have been extracted from the present experiments by
extrapolating the total irreversible capacity loss to the zero current, as shown in Figs. 6.6a and
6.7a-c. Qirca is found to increase logarithmically in time (see e.g. Figs. 6.7a-c), which is in line
with the mathematical model proposed before [2] (see also Chapter 7).
cr
QSEI represents the accumulated capacity losses due to the SEI-formation on fresh surfaces
induced by cycling. As the volume changes in each cycle can be considered constant under the
permanent cycling conditions, the surface area generated by these cracks can be considered
constant in each cycle. Furthermore, when the duration of each cycle is also approximately
cr
constant for a fixed (dis)charge current, QSEI can be represented by
cr
QSEI  n , [6.2]
cr
where is the capacity loss per cycle and is the cycle number. QSEI has been extracted from
the cycling experiments shown in Figs. 6.7d-f and was indeed found to increase linearly with
cycle number in accordance with Eq. 6.2.
The total irreversible capacity loss is, in principle, proportional to the total thickness of the
SEI layers. However, the graphite electrodes used for the determination of the SEI thickness
have been rinsed before carrying out the XPS analysis. The detached SEI layers ( ) may be
partly removed after rinsing. Therefore the thickness of the SEI layer determined by the XPS
analysis might be attributed to . As discussed above, is only time dependent (if the
overpotential is negligible). The SEI thickness shown in Fig. 6.17 at various currents is very
similar as the cycling time is the same.
In general, elevated temperatures accelerate the kinetics of all battery processes, including
the charge-transfer kinetics and transport-related processes. The SEI formation rate at higher
temperatures is, therefore, also expected to be larger than that at lower temperatures. Apart from
the SEI formation processes discussed in Fig. 6.28a, other degradation processes may be
148


involved during cycling at elevated temperatures. Fig. 6.28b schematically shows these
processes, which will be discussed below.
It has been reported that under high-temperature conditions iron ions dissolve from the
LiFePO4 cathode into the electrolyte by a chemical exchange reaction with residual H+ in the
electrolyte [34]. Subsequently, these metal ions diffuse/migrate to the anode where they are
electrochemically reduced into metallic iron. The precipitation of these iron ions has several
disadvantages with regard to the battery performance:
(i) Reduction of these ions will consume part of the cyclable electrons leading to
irreversible capacity losses.
(ii) The metal clusters can accelerate the SEI formation by facilitating electronic
transport (see Fig. 6.28b).
(iii) Iron can block the graphene layers, leading to partial inaccessibility of the graphite
electrode for Li intercalation (see Fig. 6.28b).
Δ is, therefore, expected to be larger at elevated temperatures, according to mechanism
(i) and (ii). Fig. 6.16 shows that this is the case, as the SEI layers are thicker at elevated
temperatures, according to the XPS analyses. The inaccessibility of the graphite electrode
related to mechanism (iii) has been quantified in Section 6.2.2.3.
6.3.2 Evolution of the individual graphite electrode plateaus
When the cyclable Li ions are immobilized in the SEI layers, the host sites are in principle
still available inside the graphite electrode. Therefore the reduction of region I (Δ , ) must be
attributed to the Li-immobilization process. When the graphite electrode deteriorates, all
plateaus reduce proportionally and can no longer accommodate the same amount of Li-ions as
before. The Li-ions which were initially located at the plateau II and III sites can, however, shift
to the plateau I sites, as the graphite electrode storage capacity is designed to be significantly
larger than that of the cathode. Δ , is therefore a consequence of cyclable Li-losses (Δ )
and the graphite electrode capacity decay (Δ ). Δ is a summation of the decrease of all
three plateaus, i.e.
 Qir   Q Li,I   Q Li,II   Q Li,III , [6.3]
Eliminating  Q Li,III in Eq. 6.3 by using Eqs. 4.20 and 4.21,  Q Li,I can be represented by
149

 x 
 Q Li,I   Qir   1  III   Q Li,II , [6.4]
 x II 
From Eq. 6.4 several cases can now be distinguished, which are related to different aging
mechanisms:
(i) Plateau II is constant

This refers to cases that Δ , 0 (Figs. 6.9a and b and Figs. 6.22d, e, g and h), indicating
that no graphite electrode decay takes place upon cycling (Eq. 4.20). However, Δ , (=
Δ ) is significant in these cases. The decrease of Δ , can be attributed to the Li-
immobilization process, trapping Li ions irreversible inside the SEI layer at the graphite
electrode surface [48].
(ii) Plateau II decreases

This refers to cases that Δ , 0. Referring to Eq. 6.4, three distinct cases can
additionally be distinguished:
 x 
a. When  Q ir   1  III   Q Li,II , then Δ , 0 and plateau I is expected to decrease, as
 x II 
experimentally found in Figs. 6.9c and d and Figs. 6.22c, f and i.

 x III 
b. When  Q ir   1    Q Li,II , then Δ , 0 and plateau I is expected to increase as
 x II 
found in Figs. 6.22a and b. The increase of plateau I will finally lead to Li plating on the
graphite electrode surface when the graphite electrode becomes the capacity limiting
electrode in LFP batteries.
 x 
c. When  Q ir   1  III   Q Li,II , plateau I remains constant ( Δ , 0 ). Only the
 x II 
decrease of plateau II and III can be observed.
6.3.3 Influence of graphite degradation on battery capacity loss
Figs. 6.29a-c show the development of and the calculated values for as a function
of cycle number obtained at various cycling currents (0.1, 0.5, 1.0 and 2.0 C-rate) and
temperatures (20, 40 and 60oC). The development found for for the high current (2C)
cases are schematically indicated by the shaded areas. The development of Δ and Δ at
2C at various temperatures are summarized in Figs. 6.29d-f. It is found that both Δ and Δ
150


are small at 20oC but become much more significant at 60oC. The graphite electrode degradation
rate is always higher under these cycling conditions. It can be seen in Fig. 6.29c that crosses
the initial battery storage capacity ( ) after about 1000 cycles but as is still lower than
the graphite electrode this has no influence on the battery performance. However, on the basis
of the present results it is to be expected that and will cross with each other after long-
term cycling. In that case it is predicted that a second slope will occur in the cycle life plots,
which is often seen in the cycle life characteristics of Li-ion batteries.
1.2
ΔQ
(d) 20oC ir
3 Q
C6
ΔQ ΔQ
C6 C6
0.8
Qmax / Ah
ΔQ / Ah
Qt ΔQir
2 max
0.4
0.1C
1
0.5C
(a) 20oC 1C
2C
0 0
1.2
ΔQC6 (e) 40oC
3 Q
C6
ΔQir 0.8
Qmax / Ah
ΔQ / Ah
2 Qt
max
0.4
1
(b) 40oC
0 0
1.2
(f) 60oC
3 ΔQ
C6
QC6
0.8
/ Ah
ΔQ / Ah
2
Qt ΔQ
max
max ir
Q
0.4
1
(c) 60oC
0 0
0 500 1000 1500 0 500 1000 1500
Cycle number Cycle number
Fig. 6.29. Maximum capacity of LFP batteries and graphite electrode capacity as a function of cycle
number under various current loading conditions (0.1, 0.5, 1.0, 2.0 C-rate) at 20 oC (a), 40 oC (b) and
60oC (c). The shaded areas indicate the irreversible capacity loss of the battery (Δ ) and the capacity
loss of graphite electrode (Δ ), accordingly. (d-f) show the development of Δ and Δ at 2C-rate
at 20, 40 and 60oC, respectively.
151

6.4 Conclusions
In this chapter, two types of commercial C6/LiFePO4 batteries have been investigated,
including prismatic 50 Ah batteries and cylindrical 2.3Ah batteries. The prismatic battery has
been cycled at low C-rate only. The capacity losses induced by cycling have been compared
with the corresponding storage conditions. A faster capacity decay is found during cycling than
during storage.
The cylindrical batteries have been investigated in more detail at various temperatures (20,
40 and 60oC), cycling currents (0.1 C, 0.5 C, 1 C and 2 C) and cycling SoC-windows (0-30%,
35-65% and 70-100%). The maximum capacities have been calculated on the basis of EMF
curves, which have been regularly determined during cycling. The irreversible capacity losses
Δ are accurately calculated on the basis of these maximum storage capacities.
It is concluded that the origin of Δ during cycling is mainly attributed to the Li-ion
immobilization in the SEI layers formed on the graphite surface. In addition, the reduction of
metal ions at the graphite electrode accelerates this process. Δ increases as a function of both
cycle number and time. The individual contributions of the cycle number and time have been
discriminated by a newly proposed mathematical extrapolation method. The capacity losses
induced by calendar aging ( Δ ) at the equilibrium state have been determined by
extrapolation Δ to zero current. Δ is found to increase logarithmically with time. The
capacity losses related to crack formation ( ), induced by electrode volume changes, have
been obtained by extrapolation from Δ to zero time. The growth of is found to be linear
with cycle number. Both Δ and are temperature dependent. The total irreversible
capacity loss during cycling is a summation of and . can be obtained on the basis
of Δ by considering the influence of overpotentials developed at the various cycling currents.
The thickness of the SEI layers has been investigated by XPS analyses. It has been
concluded that the SEI thickness determined by XPS is mainly related to the SEI formation on
the SEI covered surface areas ( ). Therefore the SEI thickness is very similar at various
cycling currents when the cycling time is kept the same. Both temperature and cycling range
can influence the SEI thickness. It is found that the SEI layers are thicker at higher temperatures
and higher SoC windows.
In order to facilitate the understanding of the aging mechanisms, the / curves are
obtained on the basis of the corresponding EMF curves. The depressions in the /
curves are related to the three plateau regions in the graphite electrode. The changes of the
second plateau on the / curves (Δ , ) are found to be an indicator of the graphite
152


degradation. The inaccessibility of graphite electrode has been quantitatively calculated on the
basis of Δ , . Temperature, current and the cycling SoC-ranges are found to have an influence
on the inaccessibility of the graphite electrode. The graphite electrode capacity loss (Δ ) is
found to be minor at low temperatures but becomes significant at elevated temperatures when
the batteries are cycled in the full SoC-range. However, Δ becomes significant when battery
is cycled at low SoC-windows even at low temperatures. Furthermore, Δ is found more
pronounced at high currents than at low currents.
Generally, the graphite electrode capacity decreases faster than battery capacity, especially
at higher temperatures. The second slope on the battery capacity degradation can be related to
the case that the graphite electrode becomes the capacity-limited electrode. Graphite
degradation is concluded to be a consequence of graphite structural deterioration as well as
metal deposition on the anode, which have been confirmed by Raman and XPS analyses,
respectively. Results from Raman analyses are in line with the results quantified from the
/ curves. The cathode dissolution which is mainly temperature related is also
confirmed by the Fe analyses on the graphite electrode by XPS.
6.5 References
(2014) 1-8.
Acta, 190 (2016) 1124-1133.
[6] M. Dubarry, B.Y. Liaw, M.S. Chen, S.S. Chyan, K.C. Han, W.T. Sie, S.H. Wu, Journal
of Power Sources, 196 (2011) 3420-3425.
A1135.
A1447.
[9] M. Safari, C. Delacourt, Journal of the Electrochemical Society, 158 (2011) A63-A73.
153

[10] J. Wang, P. Liu, J. Hicks-Garner, E. Sherman, S. Soukiazian, M. Verbrugge, H. Tataria,

J. Musser, P. Finamore, Journal of Power Sources, 196 (2011) 3942-3948.
Power Sources, 195 (2010) 7415-7425.
[15] Z. Li, L.G. Lu, M.G. Ouyang, Y.K. Xiao, Journal of Power Sources, 196 (2011) 9757-
9766.
[16] J.H. Kim, S.C. Woo, M.S. Park, K.J. Kim, T. Yim, J.S. Kim, Y.J. Kim, Journal of Power
Sources, 229 (2013) 190-197.
[18] M.H. Kjell, S. Malmgren, K. Ciosek, M. Behm, K. Edstrom, G. Lindbergh, Journal of
Power Sources, 243 (2013) 290-298.
A829.
[21] A.J. Smith, H.M. Dahn, J.C. Burns, J.R. Dahn, Journal of the Electrochemical Society,
159 (2012) A705-A710.
[22] R. Castaing, Y. Reynier, N. Dupre, D. Schleich, S.J.S. Larbi, D. Guyomard, P. Moreau,
3647-3667.
[24] R.P. Ramasamy, J.W. Lee, B.N. Popov, Journal of Power Sources, 166 (2007) 266-272.
Society, 160 (2013) A1701-A1710.
[26] B. Ziv, V. Borgel, D. Aurbach, J.H. Kim, X.C. Xiao, B.R. Powell, Journal of the
[27] R. Yazami, Y.F. Reynier, Electrochimica Acta, 47 (2002) 1217-1223.
[28] J. Christensen, J. Newman, Journal of Solid State Electrochemistry, 10 (2006) 293-319.
154


[30] Y. Qi, S.J. Harris, Journal of the Electrochemical Society, 157 (2010) A741-A747.
195 (2010) 3655-3660.
(2007) A40-A44.
[36] M. Koltypin, D. Aurbach, L. Nazar, B. Ellis, Journal of Power Sources, 174 (2007) 1241-
1250.
238 (2013) 454-463.
445-448.
Letters, 9 (2006) A537-A541.
[41] I. Bloom, J. Christophersen, K. Gering, Journal of Power Sources, 139 (2005) 304-313.
[42] I. Bloom, A.N. Jansen, D.P. Abraham, J. Knuth, S.A. Jones, V.S. Battaglia, G.L.
Henriksen, Journal of Power Sources, 139 (2005) 295-303.
[43] F. Tuinstra, J.L. Koenig, Journal of Chemical Physics, 53 (1970) 1126-&.
[44] M. Fahlman, H. Guan, J.A.O. Smallfield, A.J. Epstein, in: Annual Technical Conference
- ANTEC, Soc Plast Eng, Atlanda, Georgia, United States, 1998, pp. 1238.
[45] K. Edstrom, M. Herstedt, D.P. Abraham, Journal of Power Sources, 153 (2006) 380-384.
Science, 253 (2007) 4895-4905.
1885-1898.
[48] M. Kassem, C. Delacourt, Journal of Power Sources, 235 (2013) 159-171.
155

156

Chapter 7
Degradation Mechanisms of LFP Batteries:
Modeling Calendar and Cycling-induced Aging
An advanced model is proposed, describing the capacity losses of C6/LiFePO4 batteries under
various storage and cycling conditions. At moderate temperatures the capacity losses are mainly
attributed to Li immobilization in Solid Electrolyte Interface (SEI) layers at the surface of graphite
particles in the negative electrode. The SEI formation model assumes the existence of an inner and
outer SEI layer. Electron tunneling through the inner SEI layer is considered to be the rate-
determining step. The inner SEI layer grows much slower than the outer SEI layer. At elevated
temperatures, iron dissolution from the cathode and subsequent metal deposition on the anode
surface will influence the capacity losses. The SEI formation on the metal-covered surface is faster
than the conventional SEI formation. The model predicts that capacity losses during cycling are
higher than that during storage. This is caused by the generation of cracks in the SEI layers induced
by the volumetric changes during (dis)charging. These cracks generated at the graphite surface will
expose free graphite areas to the electrolyte and facilitate new SEI formation. The model has been
validated by storage and cycling experiments. The simulation results show that the capacity loss
during storage depends on the State-of-Charge (SoC), the storage time, and temperature. The
capacity losses during cycling depend on the cycling currents, the cycling time, temperature and
cycle number. The simulations are in good agreement with all experimental results.
157

Chapter 7 Degradation mechanisms of LFP batteries: modeling calendar and cycling-induced aging
7.1 Introduction
Capacity degradation has become an important topic of research since batteries are widely
applied in portable electronic devices, electric vehicles and smart grid energy storage, etc. Long
calendar and cycle life are critical for Li-ion batteries in these applications [1, 2]. Understanding
the aging mechanisms is essential for improving the life-span of Li-ion batteries.
The SEI formation and electrode material degradation are considered to be the main issues
in battery aging. It has been unraveled that the SEI formation on the anode is a major process
causing battery capacity losses [3, 4]. Considerable efforts have been made to study the
structural and chemical composition of SEI layers [5-18]. It has been found that the SEI plays
a dual-role in the performance of Li-ion batteries. On the one hand, it protects the negative
electrode from solvent co-intercalation, preventing exfoliation of the graphene layers. On the
other hand, it consumes cyclable lithium inside the battery, which is therefore no longer
available for the energy storage processes and hence leads to irreversible capacity losses.
Experimental studies demonstrate that the SEI consists of a compact inner SEI layer and a more
porous outer SEI layer [6, 7, 9]. The inner SEI-layer is dense and prevents the surface of the
graphite electrode from direct contact with the electrolyte, thereby preventing solvent co-
intercalation inside the graphite electrode. At the same time, the inner SEI layer is highly
conductive for unsolvated Li+ ions. The outer SEI layer is highly porous, easily allowing
solvated Li+ ions to pass through.
Despite the fact that a lot of studies were made to investigate the SEI by experimental
methods, the understanding of the SEI formation process is still limited due to the complexity
of the SEI formation reaction, which was found to be highly dependent on the composition of
the electrolyte, the electrode voltage and electrode surface morphology. Modeling is an efficient
way to investigate the SEI formation process, however, only a few studies are related to the SEI
growth mechanism, which are still under discussion [19-30]. Some researchers assumed the
electron diffusion process to be rate determining [22-24] while others considered solvent
diffusion through the SEI layer to be rate limiting [27-30].
Apart from the SEI formation, cathode dissolution at elevated temperatures is believed to
be another important process during battery aging [31-36]. The dissolved transition metal ions
can be transported to the anode and subsequently deposited on the graphite surface [32, 33].
Both the metal dissolution and the subsequent reduction can directly lead to a decrease of the
158


battery capacity [33]. Furthermore, the metal clusters on the graphite electrode surface can
accelerate the SEI formation by facilitating the electron transport [31-33].
Based on the experimental results, discussed in Chapters 5 and 6, an advanced electron-
tunneling-based SEI formation model and temperature-dependent cathode dissolution model
are proposed in this chapter. The models have been validated by the experimental data. Good
agreement between the simulated and experimental results are found.
7.2 Model development
Cyclable Li+ ions and electrons are originally stored in the LiFePO electrode in
C /LiFePO batteries after manufacturing. The maximum battery capacity ( ) therefore
equals to the total amount of cyclable Li+ ions in the cathode materials ( ),
0
Qmax  QLiFePO
0
4
. [7.1]
During charging (ch) electrons are extracted from the LiFePO electrode and flow into
graphite electrode via the outer circuit. In the meanwhile, Li+ ions are transported from the
LiFePO electrode via the electrolyte to the graphite electrode to safeguard the system electro-
neutrality. The reverse reactions take place during discharging (d). The main electrochemical
storage reactions in this type of battery can be represented by [22]

ch
C 6  Li +  e    LiC 6 ,
 [7.2]
d

ch
LiFePO 4   FePO 4  Li +  e  .
 [7.3]
d
Electrons extracted from the cathode can be partially consumed by many processes such as
parasitic side reactions taking place at the graphite electrode during cycling and storage, leading
to irreversible capacity losses. The detailed mechanisms of battery capacity losses during
cycling and storage will be discussed in the following sections.
7.2.1 Aging mechanisms of LFP batteries
7.2.1.1 SEI formation
The formation mechanisms and composition of SEI layers have been widely investigated
by many researchers. Although there are still some debates, it has been generally accepted that
the SEI layers are composed of inorganic and organic salts. These components constitute the
159

dense inner layer and porous outer layer, respectively, as schematically represented in Fig. 7.1.
The inner SEI layer is considered to be an insulator for electrons. It also prevents solvents from
passing through and co-intercalating into the graphene layers. Solvent reduction is therefore
assumed to take place at the interface of the inner and outer SEI layer.
Electrons at the graphite electrode surface tunnel through the inner SEI layer to the Lowest
Unoccupied Molecular Orbital (LUMO) of the solvent. Modern literature reports several values
for the Fermi level of metallic Li and LiC6 [37-39]. In the present work we adopt the results
reported by Wertheim et al. [39], namely -2.36 eV for Li and -2.80 eV for LiC6 vs vacuum. The
LUMO of ethylene carbonate (EC), one of the most important constituents in LFP battery
electrolytes, was reported to be -2.99 eV [40] in the presence of Li+ ions. This value is more
negative than the Fermi level of the lithiated graphite electrode and, consequently, EC will be
reduced, assuming that electrons can cross the energy barrier by tunneling.
The SEI formation is initiated when the voltage of the graphite electrode drops below
approximately 1.0 V vs Li+/Li [9, 12, 13, 15, 24]. The SEI is formed during the activation
procedure after the battery manufacturing process has been completed. The quality of the SEI
formed in this period determines the battery cycling performance in their forthcoming usage.
An ideal SEI layer formed during the activation procedure can dramatically decrease the new
SEI formation rate and, therefore, maintain a high battery coulombic efficiency. However, the
SEI layers normally continues to grow during battery usage, leading to continuous irreversible
capacity losses.
During (dis)charging, Li+ ions can easily pass both the inner and outer SEI layers. Solvent
molecules, present in the electrolyte, can also easily pass the porous outer SEI layer but cannot
penetrate the inner SEI layer [27, 30]. The inner SEI layer is a good insulator but electrons can
tunnel through it when its thickness is sufficiently small (< 3 nm). The tunneling process is
assumed to be the rate-determining step. The products of these reduction reactions increase the
thickness of both SEI layers but the individual formation rates might be significantly different
as will be shown below.
Fig. 7.1a schematically shows the SEI formation process in the case of storage. No current
flows through the outer circuit under open-circuit conditions, therefore no volume changes of
the graphite electrode occur during storage. When a battery is stored at a certain State-of-Charge
(SoC 0), electrons from the graphite electrode surface can freely tunnel through the inner
layer and reduce the solvent molecules at the inner/outer SEI interface. Obviously, this
reduction reaction should be counter-balanced under open-circuit conditions by an oxidation
reaction as no external current is flowing. Oxidation of Lithium stored inside the graphite
160


electrode will simultaneously take place at the graphite/inner SEI layer interface, resulting in
Li+ ion diffusion through the inner SEI layer. This electron tunneling process combined with
Li+ ion leaving the electrode ultimately results in irreversible capacity loss even under open-
circuit conditions. The SEI formation during storage of LFP batteries is considered to be the
main cause of irreversible capacity loss.
Fig. 7.1. Schematic representation of SEI formation on the graphite electrode inside a C6/LiFePO4
battery under conditions of storage (a) and upon cycling (b). (c) The influence of elevated temperatures
on aging processes inside LFP batteries. Li
When a charging current ( ) is applied to the battery, this current is used to drive two
electrochemical processes at the graphite electrode, namely the main electrochemical storage
reaction (Eq. 7.2) and the solvent reduction reactions, according to
161

I ch  I ct  I ir , [7.4]
where corresponds to the charge transfer current used to facilitate Li+ ion intercalation into
the graphite electrode and represents the current related to the Li immobilization process.
Obviously, the major part of the current is used to drive the main electrochemical storage
reaction. The fraction of with respect to the total charging current is defined as the
charging efficiency (CE) [22],
I ct
CE   100% . [7.5]
I ch
The charging efficiency is approximately equal to the coulombic efficiency when the capacity
loss is negligible during discharging.
Fig. 7.1b schematically shows the SEI formation process in case of cycling. It is well known
that the volumetric expansion/shrinkage of graphite electrodes during (dis)charging can
significantly influence the SEI formation process [22, 41-43]. Due to the mechanical stress,
resulting from these volumetric changes, cracks are formed in the SEI layer, inducing the
formation of uncovered graphite ( ). As a result, the solvent is directly brought in contact
with pristine graphite and will immediately be reduced, leading to new SEI products. Therefore,
these volumetric changes have a considerable influence on the increased capacity losses upon
cycling.
7.2.1.2 Cathode dissolution
At elevated temperatures other degradation processes can take place, such as cathode
dissolution, electrolyte degradation, graphite deformation, etc. Among these processes, cathode
dissolution has the most significant impact.
Transition metal ions are believed to be exchanged by protons in the electrolyte [35].
Protons originate from the water contamination, according to
Li +  PF6  
elevated T
 LiF + PF5 , [7.6]
 2H   2F   POF3 .
PF5  H 2 O  [7.7]
It can be concluded from Eqs. 7.6 and 7.7 that the temperature and water content are essential
conditions to produce protons in the electrolyte.
Fig. 7.1c schematically shows the cathode dissolution process and the subsequent influence
on the SEI formation. Fe2+ ions are assumed to be dissolved from the cathode by proton
162


exchange with the electrolyte (see section 7.2.3.1). These Fe2+ ions are then subsequently
transported to the anode and reduced at the graphite surface (see section 7.2.3.2). XPS analyses
have confirmed that the deposited metal clusters are embedded inside the SEI layers [32, 44].
These metal particles strongly accelerate the SEI formation since electron transport will be
highly facilitated. Therefore, more severe capacity degradation may be expected at the elevated
temperatures.
7.2.2 SEI formation model
7.2.2.1 Electron tunneling
When electron tunneling is considered to be rate limiting, as discussed in Ref [22, 23], the
tunneling current  I tl  is proportional to the number of electrons reaching the surface of the
graphite electrode per unit of time ( dN dt , [s-1]) and the probability ( , dimensionless) of
electrons to tunnel through the inner SEI layer. The electron tunneling current can then be
written as
dN
I tl  P e, [7.8]
dt
where is the electron charge. In the present work, the electrical double layer (dis)charging
currents are excluded and the only focus is on the description of the faradic tunneling currents.
Fig. 7.2. One-dimensional rectangular barrier model for electron tunneling through the inner SEI layer.
Region I corresponds to the graphite electrode LixC6, region II is the inner SEI layer, region III
corresponds to the solvent side.
163

A simple rectangular energy barrier model has been adopted to describe the electron
tunneling process (Fig. 7.2). is the Fermi level of the LixC6 electrode, is the energy
level of free electrons and Δ represents the energy barrier, is the SEI inner
layer thickness, is a reference energy level, is the LUMO of the solvents. The Fermi
level of the LixC6 electrode can be described by

E f  x   E f  LiC6   e  Li xC6   LiC6 ,  [7.9]
where LiC is the Fermi level of the fully charged LiC6, the electrode potential of
LixC6, the electrode potential of LiC6, LiC can be found from ref [45] and can
be obtained from the EMF curve of the graphite electrode.
The tunneling probability has been derived in Appendix II as Eq. A2.17, i.e.
in
16k1k2 2 e 2 l
P , [7.10]
 
 1    k1k2  e  
2 2 2
  k1  k2  e
2 2 l in 2 l in
2 2
1
where
1
k1  2m  E f  x   U I  , [7.11a]

1
k2  2 m  E f  x   U III  , [7.11b]

1
 2 m U II  E f  x   , [7.11c]

is the reduced Planck constant [J·s], and m is the electronic mass [g]. With some mild
assumptions (see the more detailed derivation in the Appendix II) can be simplified to
 2l in 2mE 
P  P0 exp   . [7.12]
  
P0 is a pre-exponential coefficient, which is given by
16k1k2 2
P0  , [7.13]
 2  k1  k2    2  k1k2 
2 2
and is determined by , , and , according to Eqs. 7.11a-c.
164


The electron flux from the bulk of the graphite to the surface can be derived from refs. [46,
47], which leads to
dN  6  x  N A C6 ve
e Ae , [7.14]
dt 4 M C6
where 0 1, C denotes the graphite density [g·m-3], ve is the Fermi velocity of electrons
6
in the bulk of graphite [m·s-1], the surface area of the graphite electrode available for electron
tunneling [m2], M C6 the molar mass of graphite [g·mol-1], N A the Avogadro number [mol-1],
multiplier 6 in the numerator of Eq. 7.14 corresponds to the amount of free electrons of each
graphite molecule and multiplier 4 in the denominator stems from the assumption that the
electronic velocity vector within the graphene layers can only proceed along one of the four
orthogonal directions with equal probabilities 1 / 4  .
Combining Eqs. 7.8, 7.12 and 7.14, the tunneling current Itl passing the inner SEI layer can
be represented by
 6  x  N AC ve  2l in 2mE 
I tl  6
AeP0 exp   . [7.15]
4 M C6   
Since
, [7.16]
Eq. 7.15 can be rewritten as
 6  x  F C ve  2l in 2mE 
I tl  6
AP0 exp   . [7.17]
4 M C6   
The Li consumption by the SEI layer can then be expressed by
dQSEI  t   6  x  F  C6 ve  2l in  t  2mE 
 AP0 exp    , [7.18]
dt 4 M C6  
 
where represents the amount of Li charge captured in both SEI layers from time 0
up to time and is the total thickness of the inner SEI layer at any time 0. At 0,
corresponds to the thickness of initial inner SEI layer after the activation process in the
manufacturing process. is an adjustable parameter dependent on the activation conditions
165

and is of the order of ~2.0 nm [11, 13]. The total inner SEI layer thickness can then, at any
moment of time , be written as
l in  t   l0in  l in  t  , [7.19]
where Δ is the increase of the inner layer thickness during storage. Using Eq. 7.19, Eq.
7.18 can be rearranged into
dQ SEI  t   6  x  F  C6 v e  2  l0in   l in  t   2 m  E 
 AP0 exp   . [7.20]
dt 4 M C6   
 
Assuming a homogeneous SEI layer thickness, the volume of the increased inner SEI layer
 V  t   can be represented by
in
V in  t   Al in  t  , [7.21]
and the corresponding mass  mSEI

in
 t   by
mSEI
in
 t    in Al in  t  , [7.22]
where  in is the (average) gravimetric density of the inner SEI layer. It is convenient for the
optimization process to be discussed later to consider the weight fraction of Li in the inner SEI
layers instead of the total mass of the inner layer. Denoting wLiin as the average weight fraction
of Li in the inner SEI layer, this leads to
 m Liin  t 
wLiin  , [7.23]
 m SEI
in
t 
Where  mLiin  t   represents the Li mass in the increased inner SEI layer. Eqs. 7.22 and 23 can
be rearranged to
l in  t  
mSEI
in
t  . [7.24]
 AwLiin
in
The mass of consumed lithium is connected with the number of moles of consumed lithium
 n  t   , according to
in
Li
mSEI
in
 t   M Li nLiin  t  , [7.25]
166


where is the molar mass of Li. The molar amount of consumed Li is related to the amount
of Li consumed in the inner SEI layer QSEI
in
 t   by
nLiin 
in
QSEI t  , [7.26]
F
Combining Eqs. 7.25 and 7.26, the following expression for the mass of consumed Li is
obtained
in
M LiQSEI t  ,
m in
SEI t   [7.27]
F
Replacing Δ in Eq. 7.24 using Eq. 7.27, the inner SEI layer thickness can be expressed
by
l in  t  
in
M LiQSEI t  , [7.28]
 AwLi F
in in
and the amount of charge involved to form the inner SEI layer
 in AwLiin F l in  t 
in
QSEI t   . [7.29]
M Li
can be defined as the fraction of the capacity loss related to the inner SEI layer with respect
to that of the total SEI layers . is dependent on the aging conditions. This leads to
in
QSEI   QSEI  t  . [7.30]
The increase of the inner SEI layer thickness as a function of time is then given by
M Li QSEI  t 
l in  t   . [7.31]
 in AwLiin F
Finally, the Li consumption rate to form the SEI layer satisfies the following Ordinary
Differential Equation (ODE)
  in M Li QSEI  t   
 2  l0   2mE 
dQSEI  t   6  x  F C6 ve A   in
Awin
F 
 P0 exp   Li . [7.32]
dt 4 M C6   
 
 
167

7.2.2.2 SEI formation during storage
In the case of storage the SEI formation rate can then be expressed by
  in M Li QSEI

st
 t   2mE 
 2  l  
 t    6  x  F C6 ve AC6 P exp     AC6 wLi F 
st 0 in in
dQSEI 
dt 4 M C6
0   . [7.33]
 
 
 
where is the amount of Li+ charge captured [C] in both SEI layers during storage from
0 up to . The State-of-Charge and Δ are more or less constant in case of storage, therefore,
Eq. 7.33 is now simple and can be solved analytically, leading to
 in wLiin AC F  6  x   C ve M Li P0 2mE  2l0in 2mE 

st
QSEI t   6
ln  1  6
t exp    . [7.34]
2 M Li 2mE  2 M C6  in wLiin    
The complete derivation of Eq. 7.34 from Eq. 7.33 can be found in Appendix III. Considering
Eq. 7.34, the SEI formation current during storage is given by
 2l0in 2mE 
 6  x  C6 e Li C6 0  
F   in
v w in
A  P exp 
st
dQSEI   
st
I SEI   . [7.35]
dt  2l0 2mE 
in
4 M C6  wLi   2  6  x   C6 ve M Li P0 2mE exp  
in in
t
  
7.2.2.3 SEI formation during cycling
In the case of cycling, cracks in the SEI layers will be periodically generated when the
volume of the graphite particles expands during charging. Therefore, the SEI formation will
continue on the SEI covered surface which leads to a capacity loss and, will occur
at the freshly formed cracks ( ), which leads to a capacity loss . It is worthwhile to note
that corresponds to the capacity loss in each cycle. The total capacity loss caused by the
SEI formation at cracks upon cycling ( ) is a summation of at various cycles. It has

been determined that 0.93 and 0.07 [48], where [m2] is the total area of the
graphite.
168


The SEI formation rate on is defined by the general expression given by Eq. 7.32. To
simplify this equation a 50% State-of-Charge is assumed and Δ Δ , which is an unknown
constant. The following expression can then be derived
2 l 
cov
 t   2mE 
 0  
dQ cov
SEI  t    6  0.5 F C ve Acov

P0 exp    A wLi F 
in cov in
,
6
[7.36]
dt 4 M C6   
 
 
where is the capacity loss due to the SEI formation at the covered surface. Similar to
the case of storage, the capacity loss can be calculated by integrating Eq. 7.36, which leads to
 in wLiin Acov F   6  0.5  C ve M Li P0 2mE  2l0in 2mE 

cov
QSEI t   ln  1  6
t exp    . [7.37]
2 M Li 2mE  2 M C  in wLiin 
6   
When SEI is formed on the fresh surface, the inner SEI layer develops from 0 to at the
cycle. The SEI formation rate can be written as
fr
dQSEI  ti    6  0.5 F C6 ve A fr P exp   2 M Li QSEI
fr
 ti  2mE 
  . [7.38]
dti 4 M C6
0
  A wLi F 
in fr in

Note that time in Eq. 7.38 refers to the duration of the cycle. This differential equation is
the same for each cycle, therefore can be considered as a constant when does not
change too much, i.e.
. [7.39]
The total capacity loss caused by the SEI formation on cracks during the whole cycling process
can be written as
n
cr
QSEI  QSEI
fr
i   n . [7.40]
i 1
The development of and upon cycling is schematically shown in Fig. 7.3. It can
be seen that develops from 0 to in each cycle while increases continuously as a

function of cycle number.
169

Fig. 7.3. Schematic representation of the development of (red) at each cycle and (black)
integrated upon cycling.
7.2.3 Cathode dissolution model
7.2.3.1 Iron dissolution at the cathode
The iron ions dissolved from the cathode are assumed to be Fe2+. The dissolution of Fe2+
can be represented by the following protons exchange reaction

2H+ LiFePO4 
ke
 Fe2+ LiH2PO4. [7.41]
The formed LiH PO can be further dissociated and dissolved into the electrolyte, according to

LiH2PO4 
kd
 2H+ Li+ PO43-, [7.42]
where and represent the rate constants of the exchange and dissociation reactions,
respectively. is expected to be much larger than , therefore the exchange reaction is
considered to be the rate-determining step. is temperature dependent, according to
 Ee 
ke  ke0 exp   a , [7.43]
 RT 
where is the pre-exponential factor and is the activation energy [J·mol-1].
170


It is worthwhile to note that in principle Eq. 7.42 can be a reversible reaction. However, in
present work the dissociation reaction is considered to be an irreversible process. This
assumption is reasonable when the total concentration of LiH2PO4 is very small. Combining
Eqs. 7.41 and 7.42, it can be concluded that the H+ ions are continuously regenerated by the
exchange and dissociation process. Consequently, the total concentration of H+ ions is not
influenced by the above processes, [mol·m-3] can therefore be considered constant.
Cyclable Li+ ions are released into the electrolyte from the cathode material during these
dissolution and dissociation processes. The dissolution flux of Fe2+ from the cathode ,
[mol·s-1·m-2], is obtained from Eq. 7.41, according to
c
J Fe  2
2   ALiFePO k e c + cLiFePO , [7.44]
4 H 4
where is the surface area of the cathode [m2]. Assuming that is constant
during cycling (storage), ke  ALiFePO4 ke can be combined to rate constant. When furthermore
the activity of LiFePO4 is considered to be 1 ( cLiFePO4  1 ), Eq. 7.44 can be written as
2  ke c + .
c 2
J Fe H
[7.45]
7.2.3.2 Iron reduction at the anode
Fe2+ ions will be transported to the graphite electrode and, subsequently, be deposited on
the graphite surface. The reduction reaction of Fe2+ ions can be represented by

Fe2+ + 2 Fe, [7.46]
where is the Fe2+ charge-transfer constant. To maintain charge neutrality of the graphite
electrode, two cyclable Li+ ions must be transported from the electrode to the electrolyte,
according to
 2C6  2e   2Li + .
2LiC6  [7.47]
Fe2+ reduction at the graphite electrode is an electrochemical process which can in principle
be described by the Butler-Volmer equation. However, the standard redox potential of Fe2+/Fe
is approximately 2.6 V vs Li+/Li , which is much higher than the electrode potential of Li xC6 ,
indicating that the overpotential of the Fe2+ reduction reaction (Eq. 7.46) is very high.
Considering the low Fe2+ concentration, it can therefore be assumed that the reduction of Fe2+at
the graphite electrode is diffusion-controlled.
171

 c  a
Considering steady state  Fe  0  , the reduction flux of Fe2+ ions at the anode J Fe2
2
 
 t 
must be equal to the dissolution flux given in Eq. 7.45
2  J  ke cH2 + .
a c
J Fe Fe2 
[7.48]
The total molar amount of Fe reduced on graphite can be obtained on the basis of Eq. 7.48,
according to
t
N Fe   J Fe
a
2  dt
0
 ke cH2 + t . [7.49]
7.2.3.3 Cyclable Li loss caused by iron dissolution and reduction
Cyclable Li+ ions are simultaneously released into the electrolyte from the cathode material
during the iron dissolution, according to Eqs. 7.41 and 7.42. The flux of Li+ ions released at the
cathode is equal to the flux of Fe dissolution,
J Lic   J Fe
c
2  kec + .
2
H
[7.50]
During the reduction of Fe2+ ions at the anode, cyclable Li+ ions must be transported from
the electrode to the electrolyte in order to maintain charge neutrality of the graphite electrode.
It can be concluded from Eqs. 7.46 and 7.47 that after the reduction of each Fe2+ ion, two
cyclable Li+ ions will be released. Therefore, the flux of Li+ ions oxidized at the anode is twice
the flux of Fe reduction,
J Lia   2 J Fe
a
2   2ke c + .
2
H
[7.51]
The battery capacity loss caused by Fe dissolution and reduction can therefore be calculated by
QLi ,Fe  F   J Lia   J Lic  dt

t
 3Fke cH2 + t . [7.52]
172


7.2.3.4 Fe-induced SEI formation
The iron clusters deposited on the graphite surface can accelerate the SEI formation by
facilitating electron transport. In order to make the discussion more general, the iron particles
deposited on the graphite surface are assumed to have a semi-spherical morphology. For the
purpose of simplification, the semi-spherical iron clusters are assumed to have the same size
with mean radius . The total volume of iron particles VFe [m3] can then be written as
2 Nr 3
VFe  . [7.53]
3
where is a constant, representing the total numbers of the Fe clusters. The volume of the Fe
particles can be represented by
N Fe M Fe
VFe  , [7.54]
 Fe
where  Fe is the density of Fe [g·m-3] and is the molar mass of Fe [g·mol-1]. can be
obtained by combining Eqs. 7.53 and 7.54

13
 3N M 
r   Fe Fe  . [7.55]
 2 N  Fe 
The surface area available for enhanced SEI formation at Fe is

23
 3N M 
AFe  2 N  Fe Fe  . [7.56]
 2 N  Fe 
Considering Eq. 7.49, the dependence of as a function of time can be represented by

23
 3kecH2 + M Fet 
AFe  2 N  . [7.57]
 2 N  
 Fe 
Similar to the SEI formation on the graphite electrode, the SEI formation on Fe can also be
described by an electron tunneling model. The rate of the SEI formation on Fe surface can be
written as
dQ SEI , Fe  t  F  Fe ue AFe  2  l Fe
0
 l Fe
in
 t   2m E 
 P0 exp   , [7.58]
dt M Fe   
 
173

where Δ is the thickness increase of the inner SEI layer on the Fe surface [m], is the
initial thickness of the inner SEI layer on Fe surface, is the Fermi velocity of electrons in
metallic Fe [m·s-1]. Assuming a homogeneous SEI layer thickness on the Fe particles and the
radius of Fe to be much larger than the total thickness of inner SEI layer (see Fig. 7.1c), Δ
can be represented, according to Eq. 7.31, by
M Li QSEI , Fe  t 
l Fe
in
 , [7.59]
 in AFe wLiin F
where is the fraction of the capacity loss related to the inner SEI layer with respect to that of
the total SEI layers on the Fe surface. Substituting Δ given by Eq. 7.59 in Eq. 7.58, this
leads to
  0 M Li QSEI ,Fe  t   

 2  lFe   2mE 
dQSEI ,Fe  t  F  Feue AFe    in AFe wLiin F  .
 P0 exp  [7.60]
dt M Fe   
 
 
 
Replacing in Eq. 7.60 using Eq. 7.57, yields
   
2 3    
dQSEI ,Fe  t  2 F  Fe ue N  3k e cH2 + M Fe    0 M Li QSEI ,Fe  t   2 m E 
 P0 exp  2  lFe 
dt M Fe  2 N   23 
 3k e cH2 + M Fe t   

 Fe    
  2 N  in in
w F    
 2 N  Fe  
Li
  
[7.61]
7.2.4 Summary of the aging model
The total capacity losses during storage Δ at any time can be written as
Qirst  t   QSEI
st
 t   QLi ,Fe  t   QSEI ,Fe  t  . [7.62]
The maximum capacity of the battery at any storage time can therefore be
represented by
st
Qmax  t   Qmax
0
 Qirst  t 
 Qmax
0
 QSEI
st
 t   QLi ,Fe  t   QSEI ,Fe  t  . [7.63]
174


The total capacity losses during cycling Δ at any time can be written as
Qircy  t   QSEI
cov
 t   nQSEI
fr
 QLi , Fe  t   QSEI ,Fe  t  . [7.64]
The maximum capacity of the battery at any cycling time can then be represented
by
cy
Qmax  t   Qmax
0
 Qircy  t 
 Qmax
0
 QSEI
cov
 t   nQSEI
fr
 QLi ,Fe  t   QSEI ,Fe  t  . [7.65]
7.3 Results and discussion
The maximum capacities upon storage and cycling have been

theoretically simulated by the presented model. The Nonlinear Least Squares (NLS) method
was used to determine the unknown parameters in the simulations and the corresponding
software was implemented in Matlab. All the parameters in the model under various aging
conditions are optimized simultaneously under a combined condition. The optimized
parameters under storage and cycling conditions are listed in Tables 7.1 and 7.2, respectively.
The experimental results shown in this chapter have been presented and discussed in Chapter 5
and 6, respectively.
100
/ Q0max / %
90
max
Qst
80
10%
50%
(a) 20oC 100% (b) 40oC (c) 60oC
70
0 2500 5000 7500 10000 0 2500 5000 7500 10000 0 2500 5000 7500 10000
Fig. 7.4. Experimental (symbols) and simulated (lines) capacity development of cylindrical 2.3Ah LFP
batteries upon storage at various SoC (10%, 50%, 100%) at 20oC (a), 40oC (b) and 60oC (c).
175

176
Table 7.1. Model parameters of LFP batteries stored under various SoC and temperature conditions.
Calendar aging
20oC 40oC 60oC
10% 50% 100% 10% 50% 100% 10% 50% 100%
/nm 2.63 2.63 2.63 2.63 2.63 2.63 2.63 2.63 2.63
/m 23.86 23.86 23.86 23.86 23.86 23.86 23.86 23.86 23.86
Δ /eV 2.90 2.84 2.80 2.90 2.84 2.80 2.90 2.84 2.80
/Ah 2.652 2.583 2.614 2.618 2.649 2.646 2.527 2.566 2.614
2.6·10-2 2.6·10-2 2.6·10-2 9.4·10-3 9.4·10-3 9.4·10-3 2.7·10-3 2.7·10-3 2.7·10-3
/m6 mol∙s 1.1·10-20 1.1·10-20 1.1·10-20 3.3·10-17 3.3·10-17 3.3·10-17 4.0·10-14 4.0·10-14 4.0·10-14
/m6 mol∙s 6.18·1034 6.18·1034 6.18·1034 6.18·1034 6.18·1034 6.18·1034 6.18·1034 6.18·1034 6.18·1034
/ ∙ mol 3.07·105 3.07·105 3.07·105 3.07·105 3.07·105 3.07·105 3.07·105 3.07·105 3.07·105
- - - - - - 1.03·106 1.03·106 1.03·106
- - - - - - 1.0·10-12 1.0·10-12 1.0·10-12
/nm - - - - - - 2.44 2.44 2.44
/m - - - - - - 2.82·10-5 2.82·10-5 2.82·10-5

Table 7.2. Model parameters of LFP batteries during cycling at various currents and temperatures.
Cycling aging
20oC 40oC 60oC
0.1C 0.5C 1C 2C 0.1C 0.5C 1C 2C 0.1C 0.5C 1C 2C
/nm 2.63 2.63 2.63 2.63 2.63 2.63 2.63 2.63 2.63 2.63 2.63 2.63
/m 23.86 23.86 23.86 23.86 23.86 23.86 23.86 23.86 23.86 23.86 23.86 23.86
Δ /eV 2.83 2.81 2.78 2.74 2.83 2.81 2.78 2.74 2.83 2.81 2.78 2.74
/Ah 2.616 2.676 2.640 2.630 2.666 2.671 2.692 2.662 2.583 2.584 2.605 2.660
/Ah 5.3·10-5 5.3·10-5 5.3·10-5 5.3·10-5 8.3·10-5 8.3·10-5 8.3·10-5 8.3·10-5 1.4·10-4 1.4·10-4 1.4·10-4 1.4·10-4
2.6·10-2 2.6·10-2 2.6·10-2 2.6·10-2 9.4·10-3 9.4·10-3 9.4·10-3 9.4·10-3 2.7·10-3 2.7·10-3 2.7·10-3 2.7·10-3
/m6 mol∙s 1.1·10-20 1.1·10-20 1.1·10-20 3.3·10-17 3.3·10-17 3.3·10-17 3.3·10-17 3.3·10-17 4.0·10-14 4.0·10-14 4.0·10-14 4.0·10-14
/m6 mol∙s 6.18·1034 6.18·1034 6.18·1034 6.18·1034 6.18·1034 6.18·1034 6.18·1034 6.18·1034 6.18·1034 6.18·1034 6.18·1034 6.18·1034
/ ∙ mol 3.07·105 3.07·105 3.07·105 3.07·105 3.07·105 3.07·105 3.07·105 3.07·105 3.07·105 3.07·105 3.07·105 3.07·105
- - - - - - - - 1.03·106 1.03·106 1.03·106 1.03·106
- - - - - - - - 1.0·10-12 1.0·10-12 1.0·10-12 1.0·10-12
/nm - - - - - - - - 2.44 2.44 2.44 2.44
/m - - - - - - - - 2.82·10-5 2.82·10-5 2.82·10-5 2.82·10-5
177

Fig. 7.4 shows the development of the experimental (symbols) and simulated (solid lines)
/ values as a function of storage time at various indicated SoC and temperatures.
The simulation results are in all cases in good agreement with the experimental results.
Obviously, the battery capacity degradation at low temperatures is much smaller than that at
high temperatures, indicating a strong temperature dependence of the capacity degradation. The
influence of SoC on capacity degradation is also observed. It can be clearly seen that the
capacity loss at low SoC is smaller than that at high SoC at all temperatures; the difference is
larger at high temperatures (Fig. 7.4c).
100
/ Q0max / %
90
max
Qcy
80
0.1C
0.5C
1C
(a) 20oC 2C (b) 40oC (c) 60oC
70
0 2000 4000 6000 0 2000 4000 6000 0 2000 4000 6000
Fig. 7.5. Experimental (symbols) and simulated (lines) capacity development of cylindrical 2.3Ah LFP
batteries upon cycling at various currents (0.1, 0.5, 1 and 2 C-rate) at 20oC (a), 40oC (b) and 60oC (c).
Fig. 7.5 shows the development of / as a function of time at various indicated

cycling currents and temperatures. Symbols again correspond to experimental results and solid
lines represent the simulation results. A good agreement between the experimental and
simulated results are also observed in Fig. 7.5. The influence of temperature on / is
in line with what has been observed under storage conditions. Apart from the temperature, the
(dis)charge current also plays an important role in the capacity degradation rate. It can be seen
that / decays faster at higher currents. This is due to the different energy barriers at
different currents as shown in Table 7.1.
Comparing Fig. 7.5 with Fig. 7.4 it is obvious that / decreases faster during
cycling than during storage for all comparable conditions. The volumetric changes of the
graphite particles during cycling is the main reason causing faster capacity degradation. Cracks
generated in the SEI layers due to graphite volume changes can accelerate new SEI formation,
leading to more capacity losses during cycling compared to storage. Furthermore, as can be
seen from Table 7.1 and 7.2, the tunneling barrier during cycling is always smaller than during
storage. Again, this will lead to more capacity losses during cycling.
178


7.3.1 Irreversible capacity losses under storage
Fig. 7.6 shows the development of the experimental (symbols) and simulated (lines)
irreversible capacity losses as a function of storage time at various storage SoC and
temperatures. A good agreement between the simulated and experimental results can be found
here. In line with the trend of / observed in Fig. 7.4, both experimental and
simulated results under storage show that Δ increases with increasing SoC. The influence of
temperature on Δ is, however, more significant than the dependence on SoC. The individual
contributions of SoC and temperatures on Δ will be discussed in detail below.
0.6
(a) 20oC 10% (b) 40oC (c) 60oC
50%
100%
0.4
ΔQst / Ah
ir
0.2
0
0 2500 5000 7500 10000 0 2500 5000 7500 10000 0 2500 5000 7500 10000
Fig. 7.6. The experimental (symbols) and simulated (lines) irreversible capacity loss (Δ ) of LFP
batteries as a function of storage time at various indicated SoC at 20oC (a), 40 oC (b) and 60 oC (c).
7.3.1.1 The influence of SoC on
Fig. 7.7a shows the development of the experimental and simulated Δ of LFP batteries
when stored at various SoC at 20oC. Δ is mainly attributed to the SEI formation ( ) at
20oC. The SEI formation rate is strongly dependent on the electron tunneling barrier Δ as can
be found in Eq. 7.35. It is concluded that the lower the value for Δ is, the faster the SEI
formation rate will be. Δ can be calculated on the basis of the graphite electrode Fermi level
( ) according to Δ . Furthermore, is a function of the graphite Electro-Motive
Force (EMF) as described by Eq. 7.9. The lower the graphite electrode potential, the higher ,
and the smaller Δ will be.
Fig. 7.7b shows the (the graphite electrode state-of-charge) dependence of the graphite
EMF. It clearly shows the presence of several plateaus as a function of . The electrode voltage
179


decreases with increasing . Since the anode is designed to be larger than the cathode in most
commercial batteries, the battery SoC shifts with respect to . The slippage of the battery SoC
is schematically shown in Fig. 7.7b. It can be concluded from that the electrode potential at
various SoC follows % % %, indicating that the electrode Fermi level at
various SoC follows 100% 50% 10% and Δ 100% Δ 50%
Δ 10% (see Eq. 7.9). Therefore, the capacity loss at high SoC is larger than that at low SoC.
0.10 0.4
(a) (b)
0.08
0.3
10%
Voltage / V
0.06
/ Ah
0.2
ΔQst
ir
0.04 50%
100%
0.1
0.02
10%
50%
100%
0 0
0 2000 4000 6000 8000 0 0.2 0.4 0.6 0.8 1
Time / h x
Fig. 7.7. (a) The experimental (symbols) and simulated (lines) irreversible capacity loss (Δ ) of LFP
o
batteries stored at various indicated SoC at 20 C. (b) the Electro-Motive Force (EMF) of the graphite
electrode as a function of (the graphite electrode state-of-charge). The corresponding battery SoC has
been indicated on the graphite EMF curve.
7.3.1.2 The influence of temperature
At low/moderate temperatures the irreversible capacity loss is mainly attributed to the SEI
formation. However, the contribution of cathode dissolution on the capacity loss becomes
significant at elevated temperatures.
Cathode dissolution
The origin of the cathode electrode dissolution was related to the proton exchange reaction
described by Eqs. 7.41 and 7.42. The concentration of protons in the electrolyte is essential for
Fe2+ ions dissolution. Fe2+ ions can be reduced at the anode. The Fe dissolution flux and the
subsequent reduction flux are described by Eqs. 7.45 and 7.48, respectively. The optimized
180


exchange constant can be found in Table 7.1 and is clearly found to be dependent on the
temperature.
0.6 0.06
(a) +
[H ] = 20ppm (b)
[H+] = 20ppm
0.4 0.04
NFe / mmol
QLi,Fe / Ah
0.2 +
0.02
[H ] = 10ppm +
[H ] = 10ppm
[H+] = 5ppm +
[H ] = 5ppm
0 0
0 2500 5000 7500 10000 0 2500 5000 7500 10000
Time / h Time / h
Fig. 7.8. (a) The total amount of Fe ions ( ) deposited on the graphite electrode as a function of
storage time in case of various indicated proton concentrations in the electrolyte at 60oC. The black dot
is the experimental result obtained from the ICP measurements. (b) The corresponding irreversible
capacity loss ( , ) caused by the Fe dissolution and the subsequent reduction on the graphite electrode
as a function of storage time.
The total amount of Fe deposited on the graphite electrode ( ) has been calculated by Eq.
7.49. Fig. 7.8a shows the development of as a function of storage time in case of various
indicated proton concentrations in the electrolyte at 60oC. It can be seen that clearly
depends on the proton concentration in the electrolyte. ICP measurement shows that the Fe
deposited on the graphite electrode is around 0.1 mmol after 7000 hours storage (black symbol),
in line with the prediction under the condition of 10 ppm H+ ions. The corresponding capacity
losses , caused by Fe dissolution and deposition under various indicated conditions are
shown in Fig. 7.8b. , is around 0.01 Ah after 7000 hours of storage when the concentration
of H+ is 10 ppm.
Fe-induced SEI formation
Fig. 7.9a shows the development of the surface area of the precipitated Fe particles )
on the graphite electrode. It can be seen from Eq. 7.57 that is determined by , which is
temperature dependent, and by , which is determined by the electrolyte system. Therefore,
is only determined by the environment temperature and the H+ concentration in the
electrolyte, independent on the testing conditions, such as SoC and cycling current. The SEI
181


formation on these Fe particles is also assumed to be determined by electron tunneling.

However, the thickness of the inner SEI layer on the Fe surface is found to be thinner than the
corresponding layer on the graphite electrode (see Table 7.2). Therefore the SEI formation rate
on the Fe surface is faster than that at the graphite electrode. Fig. 7.9b shows the development
of , as a function of storage time at various SoC at 60oC. , increases with increasing
SoC since the electron tunneling barrier on the Fe surface is smaller at higher SoC.
4 0.10
(a) (b)
0.08 100%
3
QSEI,Fe / Ah
/ cm2
0.06
50%
2
Fe
A
0.04
10%
1
0.02
0 0
0 2500 5000 7500 10000 0 2500 5000 7500 10000
Time / h Time / h
Fig. 7.9. The development of the surface area of the precipitated Fe particles on the graphite electrode
(a) and the capacity loss ( , ) caused by the SEI formation at the Fe surface at the various indicated
SoC upon storage (b).
0.6
(a) 10% (b) 50% (c) 100% ΔQst
ir
ΔQst Qst
ir
SEI
0.4
Qst
ΔQst SEI
ΔQ / Ah
ir
Qst
SEI
0.2
QSEI,Fe
QSEI,Fe
QSEI,Fe
QLi,Fe QLi,Fe QLi,Fe
0
0 2500 5000 7500 10000 0 2500 5000 7500 10000 0 2500 5000 7500 10000
Fig. 7.10. Δ (black curves), (red curves), , (pink curves) and , (blue curves) as a
function of storage time at 60oC with SoC = 10% (a), 50% (b) and 100% (c).
Fig. 7.10 shows the development of Δ (black curves), (red curves), , (pink
curves) and , (blue curves) as a function of storage time at 60oC with SoC = 10% (a), 50%
(b) and 100% (c). , is independent on SoC while both , and increase with
increasing SoC. Therefore, the total irreversible capacity loss Δ increases with increasing
182


SoC. Although the contribution of , and , on the total capacity loss becomes
considerable at 60oC, is still dominant in determining Δ .
Temperature dependence of
−3.0
0.0260
−4.0
0.0094
ln δ
δ
−5.0
0.0027
−6.0
2.9 3.0 3.1 3.2 3.3 3.4 3.5

1/T×103 / K −1
Fig. 7.11. Temperature dependence of .
Parameter is defined as the capacity ratio between the inner and the total SEI layer (Eq.
7.30). This parameter reflects the quality of the SEI layers, i.e. the larger , the better the quality
of the total SEI layer will be. The optimized at various storage conditions are listed in Table
7.1. For the purpose of simplification, is assumed to be SoC-independent in the model, it is,
however, influenced by the temperature. Fig. 7.11 shows the temperature dependence of . The
left axis is ln and the right axis shows the real value of . A linear relationship between
ln and can be observed, indicating that follows an Arrhenius-type of temperature
dependence.
7.3.1.3 Inner-SEI layer and outer-SEI layer growth under storage
Fig. 7.12 shows the simulated growth of the inner SEI layer upon storage at various
indicated SoC and temperatures. It has been well documented that the SEI layer is composed of
the inner and outer SEI layer [49]. The structure of inner layer is considered to be dense and
compact. The thickness of inner layer was measured by Edström [10], using X-ray
photoelectron spectroscopy (XPS). According to this study, the inner SEI layer thickness is
around 20 Å. The thickness of this simulated layer, using the present model, is in a good
agreement with these experimental results. Obviously, the simulated inner SEI-layer thickness
183


in Fig. 7.12 should be considered as an average thickness. Although the factors affecting the
formation of the inner SEI layer are still under discussion it has been identified that the electrode
potential has a significant influence on the products of the SEI formation [12]. Fig. 7.12 shows
that the inner SEI layer grows faster at high SoC at all temperatures. Moreover, the inner SEI
layer grows faster at lower temperatures than that at higher temperatures at any SoC. According
to the definition of , the growth of inner SEI layer is proportional to the product of and the
total irreversible capacity losses. The influence of temperature on the inner SEI growth can be
explained by the temperature dependence of as shown in Fig. 7.11 while the different growth
rates at various SoC are attributed to the different tunneling barriers.
3
(a) 20oC (b) 40oC (c) 60oC
Inner layer thickness / nm
2.9
2.8
2.7
10%
50%
100%
2.6
0 2500 5000 7500 10000 0 2500 5000 7500 10000 0 2500 5000 7500 10000
Fig. 7.12. The development of the inner SEI layers upon storage at various indicated SoC at 20oC (a),
40oC (b) and 60oC (c).
75
(a) 20oC (b) 40oC (c) 60oC
Increment of outer layer / nm
60
45
30
15 10%
50%
100%
0
0 2500 5000 7500 10000 0 2500 5000 7500 10000 0 2500 5000 7500 10000
Fig. 7.13. The development of the outer SEI layer upon storage at various indicated SoC at 20oC (a),
184


The growth of the outer SEI layer is shown in Fig. 7.13. It reveals a similar growth trend as
the inner SEI layer but at significantly higher rates. At 20oC, the outer SEI layer increased with
about 15 nm after 9000 hours storage (Fig. 7.13a), while this amounts to, for example, ~75 nm
at 60oC at 100% SoC (Fig. 7.13c). The influence of SoC on the outer layer growth is minor at
20oC, it becomes however more significant at 60oC. The increase of the outer SEI layer is about
45 nm when stored at 10% SoC while is 75 nm at 100% SoC at 60oC.
Experimental detection of the outer SEI layer thickness is difficult due to its porous and
fragile structure. Before conducting SEI characterization, the graphite electrode is usually
rinsed to remove the Li salt crystalized from the electrolyte. However, it has also been reported
that the outer SEI layers can be partially dissolved in the electrolyte/solvent, which makes the
measurements inaccurate. The simulations based on SEI models provides an efficient way to
accurately estimate the growth of both the inner and outer SEI layers.
According to Eq. 7.35 the SEI formation rate depends on the energy barrier Δ , the inner
layer thickness, the State-of-Charge and the surface area. When batteries are produced in the
same batch, it may be expected that these batteries all have a similar initial SEI inner layer
thickness and surface area. So the energy barrier Δ and State-of-Charge are the most important
factors, determining the SEI formation rate for that particular batch. Table 7.1 gives a summary
of all discussed parameters.
7.3.2 Irreversible capacity losses during cycling
0.8
(a) 20oC 0.1C
0.5C
(b) 40oC (c) 60oC
1C
0.6 2C
/ Ah
0.4
ΔQcy
ir
0.2
0
0 1000 2000 3000 4000 5000 0 1000 2000 3000 4000 5000 0 1000 2000 3000 4000 5000

Fig. 7.14. The irreversible capacity loss (Δ ) of LFP batteries as a function of cycling time at various
indicated cycling currents at 20 C (a), 40 C (b) and 60 oC (c).
o o
185


Both experimental (symbols) and simulated (lines) irreversible capacity losses under
various cycling currents and temperatures are shown in Fig. 7.14. The simulation results are in
all cases in good agreement with the experimental results under all cycling conditions. In line
with the trend of / observed in Fig. 7.5, both experimental and simulated results
show that Δ increases with increasing cycling currents. The influence of temperature on
Δ is, however, more significant than the dependence on the current. The individual
contributions of currents and temperatures on Δ will be discussed in detail below.
7.3.2.1 Cycling effects
Influence of and
The cycle number is an important parameter, describing the capacity loss during cycling.
The cycle number is a function of total cycling time , the (dis)charging current , ,
rest time in each cycle and the (dis)charging capacity and , according to
t
n . [7.67]
Qd Qch
  tr
Id I ch
If a constant charging regime is applied and the charging and discharging currents are the
same, , ≅ . Therefore, Eq. 7.67 can be rewritten as
It
n . [7.68]
2Qbat  Itr
3000
2C
2000
Cycle number
1C
1000 0.5C
0.1C
0
0 1000 2000 3000 4000 5000
Time / h
Fig. 7.15. Relationship between the total cycling time and cycle number at various cycling currents.
186


The relationship between the total cycling time and at various cycling currents is shown in
Fig. 7.15. It can be seen that at a fixed , increases with .
0.8
(a) 2C (b)
0.6 1C
/ Ah
0.5C
0.4
ΔQcy
ir
0.1C
0.2
0
0 2000 4000 6000 0 500 1000 1500 2000
Time / h Cycle number
Fig. 7.16. Irreversible capacity losses Qir

cy
  as a function of cycle time (a) and cycle number (b) at
various indicated cycling currents at 60oC.
Fig. 7.16 shows the capacity losses Qircy as a function of time (a) and cycle number (b).
Fig. 7.16a demonstrates that Qircy increases with increasing current. This is because the cycle
number at higher currents is much larger than that at lower currents at the same . Fig. 7.16b
shows that Qircy increases with current decreasing at the same . This result is attributed to a
shorter at higher currents when the cycle number is the same. These results observed from
these two figures imply that Qircy is both and dependent. The individual contributions of
and will be explained as follows.

Due to the volumetric changes of the graphite electrode, the SEI formation during cycling
has been classified into two cases. As shown in Fig. 7.1b, one process is the SEI formation on
cov
the covered surface areas Acov , denoted as QSEI , which is dependent and has been described
by Eq. 7.37. The other process is the SEI formation on the fresh surface areas A fr caused by
cr
cracks, denoted as QSEI , which is dependent and has been described by Eq. 7.40. It can be
cov cr
seen that QSEI increases logarithmically with (see Eq. 7.37), while QSEI increases linearly
cov cr
with cycle number (see Eq. 7.40). The development of the calculated QSEI (blue line), QSEI
(red line) and Qircy (black line) as well as the experimental results (black symbols) at 40oC and
187


1C-rate is shown in Fig. 7.17. Considering the relationship between and described by Eq.
cr
7.68, QSEI is also linear with .
0.3
cy
Qir
0.2
/ Ah
Qcov
SEI
cy
Qir
Qcr
0.1 SEI
0
0 1000 2000 3000 4000 5000
Time / h
Fig. 7.17. Experimental (symbols) and simulated total capacity losses  Q ircy (black line) and the
cov cr
contribution of the covered SEI Q SEI (blue line) and the SEI due to cracks Q SEI (red line) as a function
of cycle time at 1C-rate at 40oC.
The influence of current on and
0.20
0.15
/ Ah
0.10
Qcov
SEI
0.05 0.1C
0.5C
1C
2C
0
0 2000 4000 6000
Time / h
cov
Fig. 7.18. Development of Q SEI as a function time at various cycling currents at 40oC.
188


cov
The development of QSEI as a function of time at various cycling currents at 40oC is shown
cov
in Fig. 7.18. It can be concluded that QSEI increases logarithmically with time and is higher at
larger currents. The corresponding current densities related to the SEI formation on Acov
during cycling at various currents are plotted in Fig. 7.19. Like in the case of storage (Eq. 7.35),
a reciprocal relationship between and time can be observed. In line with the observation
in Fig. 7.18, the SEI formation rate at higher currents is larger than at lower currents. The
variation of at various currents is attributed to the overpotential development. The higher
the current, the larger the overpotential. The electron tunneling probability in case of a larger
graphite electrode overpotential is larger since the energy barrier is smaller. However, as can
be seen from Fig. 7.19, the difference of becomes smaller after long-term cycling since the
thickness of the inner SEI layer becomes the dominating factor in determining the electron
tunneling probability.
16
0.1C
0.5C
1C
2C
12
−2
μA⋅m
15
8
/
10
jSEI
5
4
0 200 400
0
0 1000 2000 3000 4000 5000 6000
Time / h
Fig. 7.19. Calculated current density used for SEI formation  jSEI  at various cycling currents at
40oC.
The dependence of on the energy barrier Δ is discussed in Fig. 7.20. The optimized
values for Δ at various cycling currents can be found in Table 7.2. The values of at 4
different cycling time: 1000 h (black symbols), 2000 h (blue symbols), 3000 h (pink symbols)
and 4000 h (red symbols) at 0.1, 0.5, 1 and 2 C-rate are extracted from Fig. 7.19. The calculated
ln jSEI are plotted as a function of E in Fig. 7.20. Obviously, a linear relationship between
189


ln jSEI and E can be observed at various cycling times, indeed indicating that follows
an exponential-type dependence on E .
−12.0
2C
1C
0.5C t1 = 1000h
0.1C
−12.5
ln jSEI / A⋅m−2
t2 = 2000h
−13.0
t = 3000h
3
t4 = 4000h
−13.5
−1.68 −1.67 −1.66 −1.65

1/2 1/2
(ΔE) / (eV)
Fig. 7.20. From Fig. 7.19 extracted values for at 4 different cycling times: 1000 h (black symbols),
2000 h (blue symbols), 3000 h (pink symbols) and 4000 h (red symbols) at various indicated cycling
currents. The corresponding solid lines indicate the linear dependence of ln jSEI on the tunneling
barriers E .
7.3.2.2 Influence of temperature
3 0.04
(a) (b) 2C
0.03 1C
2
/ Ah
2
0.5C
AFe / cm
0.02
SEI,Fe
0.1C
Q
1
0.01
0 0
0 1000 2000 3000 4000 0 1000 2000 3000 4000
Time / h Time / h
Fig. 7.21. Development of the surface area of precipitated Fe particles at the graphite electrode during
cycling at 60oC (a) and the capacity loss caused by the SEI formation on the Fe surface , (b) at
the various indicated cycling currents.
190


When batteries are cycled at the elevated temperatures, the cathode dissolution and the
subsequent metal deposition on the anode will influence the total irreversible capacity losses.
Fig. 7.21a shows the development of the surface area of precipitated Fe particles at the graphite
electrode during cycling at 60oC. As discussed in section 7.3.1.2b, the cathode dissolution and
metal deposition are only influenced by the and . Therefore, when the temperature and
the electrolyte are fixed, the cathode dissolution and metal deposition rates inside the batteries
cycled at various currents will be the same. Indeed, the cathode dissolution and metal deposition
rates under cycling are the same as those obtained during storage at the same temperature. The
growth of in Fig. 7.21a and Fig. 7.9a is exactly the same.
Fig. 7.21b shows the capacity losses caused by the SEI formation on the Fe particle surface
, at various cycling conditions. Different from the growth rate on the graphite electrode,
the SEI formation rate on the Fe surface becomes higher as a function of time. This is because
the surface area also increases with time. Moreover, the growth rate of , is higher at
higher currents due to the smaller electron tunneling barriers.
0.8
(a) 20oC (b) 40oC (c) 60oC
0.6
ΔQcy
ir
/ Ah
0.4 Qco
SEI
ΔQcy
ir
ΔQcy
ir
Qcr
SEI
0.2 ΔQcy
ir Qco
SEI
Qco
SEI Qcr
SEI QFe,SEI
Qcr
SEI
QLi,Fe
0
0 1000 2000 3000 4000 5000 0 1000 2000 3000 4000 5000 0 1000 2000 3000 4000 5000
Fig. 7.22. Various components of the total irreversible capacity loss (Δ ) during cycling at 1C-rate at
o o o
20 C (a), 40 C (b) and 60 C (c).
Fig. 7.22 shows the various components of the total irreversible capacity loss (Δ ) during
cycling at 1C-rate at 20oC (a), 40oC (b) and 60oC (c). Since the cathode dissolution at 20 and
40oC is negligible, Δ only comprises and .Δ consists of , , , ,
and at 60oC. It can be observed that both and are temperature dependent.
The development of as a function of cycling time at 20, 40 and 60oC at 1 C-rate is
shown in Fig. 7.23. It can be seen that increases strongly with increasing temperature. The
current density related to the SEI formation on has been calculated on the basis of
and is shown in Fig. 7.24. It clearly shows that decreases strongly as a function of
191


time but increases as a function of temperature. The dependence of on temperature is

further illustrated in Fig. 7.25. The values of at 4 different cycling time: 1000 h (black
symbols), 2000 h (blue symbols), 3000 h (pink symbols) and 4000 h (red symbols) at 20, 40
and 60oC are obtained from Fig. 7.24. The calculated value for are then plotted as a function
of in Fig. 7.25. Obviously, a linear relationship between ln jSEI and can be observed
at various cycling time , indicating that follows an Arrhenius-type of dependence on .
0.4
60oC
0.3
/ Ah
0.2
40oC
SEI
Qcov
0.1
o
20 C
0
0 2000 4000 6000
Time / h
Fig. 7.23. The development of as a function of cycling time at 20, 40 and 60oC at 1C-rate.
12
20oC
40oC
60oC
9
−2
μA⋅m
6
/
SEI
j
0
0 1000 2000 3000 4000 5000 6000
Time / h
Fig. 7.24. The development of jSEI as a function of cycling time at various indicated temperatures at
2 C-rate.
192


−12 60oC
o
−13 40 C
/ A⋅m−2
o
−14 20 C
SEI
ln j
t1 = 1000h
−15 t = 2000h
2
t3 = 3000h
t4 = 4000h
−16
2.9 3.0 3.1 3.2 3.3 3.4 3.5
3 −1
1/T×10 / K
Fig. 7.25. Calculated value for ln jSEI from Fig. 7.24 at 4 different cycling time: 1000 h (black symbols),
2000 h (blue symbols), 3000 h (pink symbols) and 4000 h (red symbols) at various cycling currents. The
corresponding solid lines represent the linear dependence of ln jSEI on the reciprocal of temperatures
( ).
−8.5
14
−9.0
×10−5 / Ah
ln QfrSEI / Ah
8.3
−9.5
SEI
Qfr
5.3
−10
2.8 3.0 3.2 3.4 3.6

1/T×103 / K −1
Fig. 7.26. The temperature dependence of .
193


Fig. 7.26 shows ln as a function of . The left axis is ln and the right axis
shows the real value of . A linear relationship between ln and is observed,

indicating that also follows an Arrhenius-type of dependence on temperature.
7.3.2.3 Growth of inner and outer SEI layers upon cycling
3
(a) 20oC (b) 40oC (c) 60oC
Inner layer thickness / nm
2.9
2.8
2.7 0.1C
0.5C
1C
2C
2.6
0 2000 4000 6000 0 2000 4000 6000 0 2000 4000 6000
Fig. 7.27. The development of the inner SEI layer upon cycling at various indicated currents at 20oC (a),
Fig. 7.27 shows the simulated growth of the inner SEI layer thickness upon cycling at 0.1,
0.5, 1 and 2 C at 20oC (a), 40oC (b) and 60oC (c). Similar to the case of Fig. 7.12, the simulated
the inner SEI-layer thickness must be considered as an average thickness here. Fig. 7.27 shows
that the inner SEI layer grows faster at higher currents at all temperatures. Moreover, the inner
SEI layer grows faster at lower temperatures. The growth of the inner SEI layer is determined
by both  and . The influence of temperature on the inner SEI growth rate can be
explained by the temperature dependence of  as shown in Fig. 7.11 while the different growth
rates at various currents are attributed to the differences in as shown in Fig. 7.18.
The growth of the outer SEI layer is shown in Fig. 7.28. It reveals a similar growth trend as
shown in Fig. 7.13. At 20oC, the outer SEI layer increased with about 15 nm after 5500 hours
cycling (Fig. 7.28a), while this amounts to about 75 nm when cycled at 60oC at 2 C (Fig. 7.28c).
The influence of the current on the outer SEI layer growth is minor at 20oC, however it becomes
significant at 60oC. Comparing with Fig. 7.13 it can be concluded that the growth rate of the
outer SEI layer is higher during cycling than during storage at a given temperature.
194


75
o o o
(a) 20 C (b) 40 C (c) 60 C
Increment of outer layer / nm
60
45
30
0.1C
15 0.5C
1C
2C
0
0 2000 4000 6000 0 2000 4000 6000 0 2000 4000 6000
Fig. 7.28. The development of the outer SEI layers upon cycling at various indicated currents at 20oC
(a), 40oC (b) and 60oC (c).
7.3.2.4 Charging efficiency (CE)
0.1C 0.5C
600 o 100
20 C
o
40 C
450 60oC 99.98
CE / %
Iir / μA
300 99.96
150 99.94
(a) (b)
0 99.92
1C 2C
600 100
450 99.98
CE / %
Iir / μA
300 99.96
150 99.94
(c) (d)
0 99.92
0 100 200 300 0 100 200 300
Fig. 7.29. Current of Li immobilization (solid line, corresponding to left y-axis) and charging efficiency
(dashed line, corresponding to right y-axis) of LFP batteries as a function of cycle number under 20oC
(black), 40oC (blue), 60oC (red), at 0.1 C (a), 0.5 C (b), 1 C (c) and 2 C (d).
195


The battery charging efficiencies (CE) are calculated according to Eq. 7.5. CE represents
the percentage of the charge-transfer current reversibly stored in the electrode during charging.
It can be influenced by the cycling conditions, i.e. temperature and currents. Fig. 7.29 shows
the development of (solid lines) and CE (dashed lines) as a function of cycle number at 20oC
(black lines), 40oC (blue lines), 60oC (red lines) and at 0.1 C (a), 0.5 C (b), 1 C (c) and 2 C (d).
The left-hand -axis represents and the right-hand -axis shows CE. As expected, is
larger at higher temperatures. Obviously, CE is smaller at higher temperatures. It can also be
observed that increases with increasing current, and, CE increases with increasing current.
7.3.3 Influence of graphite parameters on the irreversible capacity losses
Fig. 7.30 shows the dependence of the storage capacity loss as a function of surface area of
the graphite electrode. In line with Eq. 7.34, a linear relationship is found. The three lines
correspond to different storage SoC. Extrapolating the surface area to 0, the corresponding
capacity loss is found to be 0 also, indicating that no capacity loss is indeed to be expected when
no active surface area is exposed to the electrolyte.
0.20
0.15
/ Ah
0.10
SEI
Qst
0.05
10%
50%
100%
0
0 20 40 60 80 100
2
A / m
Fig. 7.30. Calculated capacity loss as a function of surface area after 1000 hours storage at 20oC at SoC
= 10%, 50% and 100%.
196


0.25
10%
50%
100%
0.20
/ Ah
0.15
SEI
Qst
0.10
0.05
0
0 5 10 15 20
Particle size / μm
Fig. 7.31. Calculated capacity loss as a function of the graphite particle size after 1000 hours storage at
20oC at SoC = 10%, 50% and 100%.
Fig. 7.31 shows the storage capacity loss as a function of particle size from which the
graphite electrode is composed. The capacity loss strongly decreases with increasing particle
size. A reciprocal relationship is found between the capacity loss and particle size in all cases.
From the model description it follows that the larger the particle size, the better the capacity
retention is. However, it is well known that a larger particle size can negatively influence the
mass transport properties for Li+ ions in the active materials. Larger particles increase the
diffusion distance of Li+ ions and may decrease the rate capability of the battery. So the
optimized particle size is a trade-off between the capacity loss and rate capability of Li-ion
batteries.
7.4 Conclusions
An advanced electron-tunneling-based SEI formation model and a temperature-dependent

cathode dissolution model are proposed, describing the capacity losses of C6/LiFePO4 batteries
at various storage and cycling conditions, which can be used to predict the calendar life and
cycle life performance of LFP batteries, respectively.
The SEI formation model assumes the existence of a compact inner and porous outer SEI
layer. The rate-determining step is considered to be electron tunneling through the inner SEI
197


layer. Both SEI layers are growing at the interface of the inner and outer SEI layer. The inner
SEI layer grows much slower than the outer layer. The initial thickness of the inner layer
developed after activation will largely determine the future degradation rate.
The capacity losses at moderate aging temperatures are all attributed to the SEI formation.
It has been concluded that the capacity losses are strongly dependent on the storage SoC and on
the cycling currents. The capacity losses during cycling are larger than during storage at the
same operation time due to the cracks formation upon charging and discharging. These cracks
generate free graphite surface areas exposed to the electrolyte, where new SEI will be easily
formed. The capacity losses due to the SEI formation on these fresh graphite surfaces are a
function of both the charging time and cycling number. A logarithmical relationship between
the capacity loss and aging time is found.
A cathode dissolution model is used to describe the transition-metal dissolution process at
the elevated temperatures under both storage and cycling conditions. Cathode dissolution is
assumed to be initiated by a proton-exchange reaction. The concentration of protons in the
electrolyte ultimately determines the cathode dissolution rate. The dissolved metal ions can be
transported to the anode and reduced at the graphite surface. Both cathode dissolution and metal
deposition will induce battery capacity loss. The SEI formation on these metal-cluster surfaces
has been simulated, which also cause capacity loss.
Both the particle size and particle surface area of graphite have a large influence on the
capacity losses. A linear relationship between the surface area and capacity loss is observed
while a reciprocal relationship between the particle size and the capacity loss is found in
accordance with the presented model.
7.5 References
[1] P.H.L. Notten, D.L. Danilov, Advances in Chemical Engineering and Science, 4 (2014)
62-72.
Acta, 102 (2013) 183-195.
3647-3667.
[4] M. Kassem, J. Bernard, R. Revel, S. Pelissier, F. Duclaud, C. Delacourt, Journal of Power
Sources, 208 (2012) 296-305.
[5] M. Lu, H. Cheng, Y. Yang, Electrochimica Acta, 53 (2008) 3539-3546.
198


[6] E. Peled, D. Golodnitsky, A. Ulus, V. Yufit, Electrochimica Acta, 50 (2004) 391-395.

[8] K. Xu, Energies, 3 (2010) 135-154.
1885-1898.
[11] S.K. Jeong, M. Inaba, Y. Iriyama, T. Abe, Z. Ogumi, Journal of Power Sources, 119
(2003) 555-560.
[13] J.T. Lee, N. Nitta, J. Benson, A. Magasinski, T.F. Fuller, G. Yushin, Carbon, 52 (2013)
388-397.
Science, 253 (2007) 4895-4905.
[16] K. Xu, A. von Cresce, Journal of Materials Chemistry, 21 (2011) 9849.
[17] Y. Xie, J. Li, C. Yuan, Journal of Power Sources, 248 (2014) 172-179.
[18] F.M. Wang, M.H. Yu, Y.J. Hsiao, Y. Tsai, B.J. Hwang, Y.Y. Wang, C.C. Wan, Int J
Electrochem Sc, 6 (2011) 1014-1026.
A1988.
[20] P. Lu, S.J. Harris, Electrochemistry Communications, 13 (2011) 1035-1037.
[21] E. Peled, D. Golodnitsky, G. Ardel, Journal of the Electrochemical Society, 144 (1997)
L208-L210.
(2014) 8.
A829.
[26] M. Onuki, S. Kinoshita, Y. Sakata, M. Yanagidate, Y. Otake, M. Ue, M. Deguchi, Journal
of The Electrochemical Society, 155 (2008) A794.
199


[27] A.J. Smith, J.C. Burns, J.R. Dahn, Electrochemical and Solid-State Letters, 13 (2010)
A177.
A513-A518.
Society, 160 (2013) A1701-A1710.
A456-A462.
[33] D. Li, D.L. Danilov, J. Xie, L. Raijmakers, L. Gao, Y. Yang, P.H.L. Notten,
Letters, 9 (2006) A537-A541.
(2007) A40-A44.
238 (2013) 454-463.
[37] N.A.W. Holzwarth, S. Rabii, L.A. Girifalco, Physical Review B, 18 (1978) 5190-5205.
[38] N.A.W. Holzwarth, S. Rabii, Materials Science and Engineering, 31 (1977) 195-200.
[39] G.K. Wertheim, P. Vanattekum, S. Basu, Solid State Communications, 33 (1980) 1127-
1130.
[40] B. Markovsky, A. Rodkin, Y.S. Cohen, O. Palchik, E. Levi, D. Aurbach, H.J. Kim, M.
Schmidt, Journal of Power Sources, 119 (2003) 504-510.
A250.
[42] R. Deshpande, M. Verbrugge, Y.-T. Cheng, J. Wang, P. Liu, Journal of the
[43] D. Li, D. Danilov, G. Lu, Y. Yang, P.H.L. Notten, Electrochimica Acta, 210 (2016) 11.
Acta, 190 (2016) 1124-1133.
[45] D. Cintora-Juarez, C. Perez-Vicente, S. Ahmad, J.L. Tirado, Physical Chemistry
Chemical Physics, 16 (2014) 20724-20730.
200


[46] H.S. Taylor, S. Glasstone, A treatise on physical chemistry: a cooperative effort by a

group of physical chemists, Van Nostrand, 1947.
[47] G.W. Castellan, Physical Chemistry, 3rd ed., Addison-Wesley Pub. Co., London, 1983.
157 (2010) A558-A566.
[49] D. Aurbach, M.D. Levi, E. Levi, H. Teller, B. Markovsky, G. Salitra, U. Heider, L. Heider,
Journal of the Electrochemical Society, 145 (1998) 3024-3034.
201

202

Chapter 8
Degradation Mechanisms of NMC Batteries:
Experimental Analysis of Cycling-induced Aging
The EMF curves of NMC(111) batteries have been regularly determined by mathematical
extrapolation of the measured voltage discharge curves. The irreversible capacity losses, which have
been accurately determined from the EMF curves, have been systematically investigated as a
function of time and cycle number and were found to increase with cycling current and temperature.
The charge efficiency of the individual electrodes has also been investigated. The charge efficiency
is always found to be larger than 100% for the cathode and lower than 100% for the anode. Parasitic
side reactions, occurring at the cathode and anode are considered to be responsible for the deviation
of the charge efficiency. The ohmic resistances of charging ( ) and discharging ( ) at 40 and
o
60 C are calculated on the basis of the initial voltage changes during (dis)charging. is found to
be larger than . Both and increase as a function of cycle number. Finally, /
curves are calculated from the determined EMF curves. Changes of the peak position in the
/ curves can be used to determine the electrode material decay and voltage slippage of the
individual electrodes.

203

Chapter 8 Degradation mechanisms of NMC batteries: Experimental analysis of Cycling-induced Aging
8.1 Introduction
The demand for large-scale rechargeable batteries in the application of electric vehicles and
smart grids has been growing rapidly in the past few years [1]. Higher energy density combined
with long cycle life and high safety is one of the key requirements in these applications. Lithium
nickel-manganese-cobalt oxide (NMC) cathode material is considered to be a promising
candidate for high-energy-density batteries, due to their larger theoretical specific capacity
( 260 mAh·g-1) compared with olivine (LiFePO4, 160 mAh·g-1) or spinel materials (LiMn2O4,
150 mAh·g-1) [2-4].
The ternary NMC material has a layered structure, which is similar to that of LiCoO2 [1].
Electronic structure studies have shown that NMC consists of Ni2+, Mn4+ and Co3+ in the as-
made materials [5-8]. Ni2+ will be oxidized to Ni4+ during the initial stages of charging, while
Co3+ will be oxidized to Co4+ in the higher voltage range. Mn4+ remains inactive throughout
normal charging [7, 9] and provides structural stability [10]. The NMC electrode has a similar
or higher achievable specific capacity compared to LiCoO2 when cycled in the potential
window of 2.5-4.3 V [7, 11]. Advantageously, the cyclability of the NMC electrode is better
than LiCoO2 due to its higher thermal stability [12, 13].
A lot of work has been carried out to investigate the degradation mechanisms of NMC
batteries [2-4, 10, 12, 14-31]. Generally, Li immobilization in the SEI layers on the graphite
electrode is considered to be the main origin of the battery capacity loss [14-19]. The cathode
material decay becomes significant under severe aging conditions, e.g. at high temperature,
using high (dis)charge currents and upon overcharging [4, 12, 27-30]. The degradation
mechanism of NMC is still under discussion. It is well known that the NMC material
experiences a phase transition from the rhombohedral space group 3 (initial “O3” phase) to
the monoclinic space group C2/m (“O1” phase) when the charge voltage is beyond 4.4 V vs
Li+/Li [2, 7, 26, 27]. The “O1” LiyNi1/3Co1/3Mn1/3O2 phase has been clearly observed at 0.3
[32]. Cycling above this phase transition voltage (> 4.4 V) will lead to a faster capacity decay
of the cathode. Structural changes induced by Li-Ni site interchange, is considered to be another
detrimental effect on the electrode cycling performance [2, 10, 12, 26-28, 33]. High currents
[12] and voltages [2] are believed to be more detrimental to cause such distortion at the surface
of these materials. Moreover, metal dissolution from the NMC electrode in acidic solutions
(LiPF6 based electrolyte) at high temperatures [4, 7, 31, 34-38] has also been reported.
The ternary NMC electrode system contains a large group of family members including
LiNi1/3Co1/3Mn1/3O2 (NMC(111)), LiNi0.5Co0.2Mn0.3O2 (NMC(532)), LiNi0.425Co0.15Mn0.425O2,
204


LiNi0.25Co0.5Mn0 25O2, etc. [2, 25]. However, at present only NMC(111) and NMC(532) have
been successfully introduced as cathode materials in commercial Li-ion batteries [2]. It is
worthwhile to point out that the degradation mechanisms of NMC materials depend on their
specific compositions [2]. In this chapter, the aging mechanisms of NMC(111) batteries will be
investigated and discussed.
8.2 Results and discussion
Cylindrical 18650 NMC batteries with a nominal capacity of 2.2 Ah have been cycled at
various currents at 40 and 60oC. Fig. 8.1a shows an example of the development of the voltage
discharge curves upon cycling at 60oC and 2 C-rate. The corresponding cycle numbers of the
discharging curves are indicated in the figure. It can be seen that the discharge curves contract
upon cycling number, indicating a decline of the discharge capacity. The corresponding EMF
curves are shown in Fig. 8.1b. Similar to the discharge curves, a contraction of the EMF curves
upon cycling is observed. Obviously, the capacities extracted from the EMF curves are larger
than those extracted from the discharge curves in Fig. 8.1a. The capacity loss found in the EMF
curves is smaller than that of the discharge curves. Moreover, the voltage plateaus of EMF
curves in Fig. 8.1b are higher than those of the discharge curves in Fig. 8.1a since overpotentials
are developed when a current flows through the battery.

4.2 4.2
3.7 3.7
Voltage / V
Voltage / V
400 200 1 400 200 1
3.2 3.2
(a) (b)
2.7 2.7
0 0.5 1.0 1.5 2.0 2.5 0 0.5 1.0 1.5 2.0 2.5
Q / Ah Qout / Ah
out

Fig. 8.1. Development of the voltage discharge curves (a) and corresponding EMF curves (b) of a
NMC(111) battery at 60oC and 2C-rate. The cycle number is indicated.

205

2.4
2.0
Qtmax
Capacity / Ah
1.6
Qt
d
1.2
0 100 200 300 400 500
Cycle number

Fig. 8.2. Maximum storage capacity and discharge capacity for a NMC(111) battery as a
function of cycle number at 2 C-rate at 60oC.
The extracted discharge capacity (black symbols) and the maximum capacity (red
symbols) are shown in Fig. 8.2 as a function of cycle number. In line with the trend observed
in Fig. 8.1, is higher than but the decline rate of is smaller than that of . Fig.
8.2 implies that the polarization (overpotential) of the battery is clearly influenced by the cycle
number. More and more cyclable Li+ ions cannot be extracted when the overpotentials increase.
The apparent capacity loss Δ calculated from deviates from the calculated
irreversible capacity loss Δ . Therefore, only the irreversible capacity loss
Δ is considered in the present work.
Fig. 8.3 shows the development of the normalized maximum capacity ( / ) at
various currents as a function of cycle number ((a) and (c)) and time ((b) and (d)) at 40 and
60oC. It can be seen that / decreases faster at higher temperatures than at lower
temperatures when the cycling current is low, i.e. at 0.1-0.5 C-rate. However, when the cycling
current increases to 1 C, / decreases very fast at both 40 and 60oC. Figs. 8.3a and
8.3c illustrate that the degradation rate of / decreases with increasing current (apart
from 1 C at 40oC) while Figs. 8.3b and 8.3d show that the degradation rate of /
increases with increasing current. The above results indicate that both the cycle number and
time influence the battery capacity degradation, which has been discussed in detail in Chapter
6 for LFP batteries.
206


100
0.1C
0.3C
0.5C
1C
Qtmax / Q0max / %
90
80
o (b)
40 C (a)
70
100
Qtmax / Q0max / %
90
80
o
60 C (c) (d)
70
0 400 800 1200 0 2000 4000 6000 8000
Cycle number Cycle time / h

Fig. 8.3. The development of / at various cycling currents as a function of cycle number ((a)
and (c)) and time ((b) and (d)) at 40 and 60oC.

100
/ %
90
max
/ Q0
max
80
Qt
0.1C
0.5C
o
40 C (a) LFP 1C (b) NMC
70
100
90
80
o
60 C
(c) LFP (d) NMC
70
0 2000 4000 6000 0 2000 4000 6000
Cycle time / h Cycle time / h

Fig. 8.4. Comparison of / for LFP (A123) and NMC(111) batteries at various cycling
currents as a function of cycle time at 40 oC ((a) and (b)) and 60oC ((c) and (d)).
207

Figs. 8.4a and 8.4b show the comparison of / for LFP and NMC(111) batteries
at various cycling currents as a function of cycle time at 40oC. It can be seen that the capacity
degradation of LFP batteries is similar to that of NMC(111) batteries at 0.1 C at 40oC. However,
/ of NMC(111) batteries decreases much faster than that of LFP batteries at 1 C,
indicating that the cycling stability of LFP (A123) batteries at higher currents is better than that
of NMC(111) batteries. Figs. 8.4c and 8.4d show the comparison of / for LFP and
NMC(111) batteries upon cycling at 60oC. It is clearly visible that the capacity loss of NMC(111)
is similar to that of LFP batteries at 0.5 and 1 C-rate at 60oC but is even better at 0.1 C-rate,
indicating that the cycling performance of NMC(111) batteries at higher temperatures is better
than that of LFP batteries.
100
0.5/1C 0.5/1C
1/1C 1/1C
Q0max / Qtmax / %
90
80
(a) (b)
70
0 400 800 1200 0 2000 4000 6000

Fig. 8.5. Influence of the charging current of a NMC battery on / as a function of cycle
number (a) and time (b) at 40oC.
Fig. 8.5a shows the development of / as a function of cycle number with

charging current of 0.5 C and 1 C but with the same discharging current of 1 C. It can be
concluded that battery degradation is larger when the charging current is higher. Fig. 8.5b shows
the same results now plotted as a function of time. The trend is similar to Fig. 8.5a, however,
the difference between the two curves in Fig. 8.5b becomes larger since the charging time at
0.5 C is longer than at 1 C.
Fig. 8.6 shows the irreversible capacity loss Δ of NMC(111) batteries cycled at 0.1, 0.3
and 0.5 C-rates at 40 and 60oC as a function of cycle number and time. Obviously, Δ is larger
at 60oC than at 40oC. In line with the results observed in Fig. 6.5, it is found that Δ is larger
at lower currents (see Figs. 8.6a and 8.6c). However, as shown in Figs. 8.6b and 8.6d, when
208


plotted as a function of time Δ is larger at higher currents. It can be concluded that the battery
capacity loss is both cycle number and time dependent.
0.8
(a) 0.1C (b)
0.3C
0.6 0.5C
ΔQ / Ah
0.4
ir
0.2
o
40 C
0
0.8
(c) (d)
0.6
ΔQ / Ah
0.4
ir
0.2
o
60 C
0
0 400 800 1200 0 2000 4000 6000 8000

Fig. 8.6. Irreversible capacity loss Δ at various cycling currents as a function of cycle number ((a)
and (c)) and time ((b) and (d)) at 40 and 60oC.
Fig. 8.7 shows the development of the charging capacity ( ) and the discharging capacity
( ) as a function of cycle number at 0.1 C-rate and 40oC. It can be seen that is always
smaller than while is fluctuating around . represents the total amount of
cyclable Li+ ions extracted from the cathode. However, a small part of these Li+ ions are
immobilized by the SEI formation process during intercalation into the graphite electrode.
Therefore, the amount of cyclable Li+ ions extracted from the anode during the subsequent
discharging is always smaller than . Ideally, under the same charging cut-off
conditions the charging capacity in the subsequent cycle should be equal to if there
are no side-reactions at the cathode. Apparently this is not the case as Fig. 8.7 shows. There are
two possibilities to explain the fluctuation of around . One is that the internal resistance
of the battery system is not stable, leading to fluctuating overpotentials, which influences the
209

final charging cut-off current in the constant-voltage charging period. Another reason might be
that there are some parasitic side reactions at the cathode, such as electrolyte oxidation, which
generate electrons at the cathode.

2.255
Qnch
Qn+1
ch
2.254
Capacity / Ah
Qn+2
ch Qn+3
Qnd Qn+1
ch
d
2.253
Qn+3
d
Qn+2
d
2.252
n n+1 n+2 n+3
Cycle number

Fig. 8.7. The development of charging capacity ( ) and discharging capacity ( ) as a function of
cycle number at 0.1C-rate at 40oC.
100.10
CE
c
CEc / %
100.05
100.00
CEa / %
99.95
CEa
99.90
0 100 200 300 400 500
Cycle number

Fig. 8.8. The development of the charge efficiency for the anode  CEa  and cathode  CEc  as a
function of cycle number at 0.5 C-rate and 60oC.
210


In order to investigate the charge efficiency for the individual electrodes, the development
of CE a (the charge efficiency at the anode, see definition Eq. 4.2) and CE c (the charge activity
ratio for the cathode, see definition Eq. 4.3) as a function of cycle number at 0.5 C-rate at 60oC
are shown in Fig. 8.8. Obviously, CE a is always smaller than 100%, indicating that the parasitic
side reactions, occurring at the anode, consumes cyclable Li+ ions during charging. In contrast,
CE c in most cases is larger than 100%, indicating that the processes occurring at the cathode
can release more charge during the charging processes. Both CE a and CE c are approaching
100% after long-term cycling, implying that the system becomes stable and the parasitic
reactions at both electrodes are depressed.

Fig. 8.9. Definitions of the battery voltage at the end of the resting period after charging and at
the initial stage of discharging . represents the battery voltage at the end of the resting
period after discharging and is the initial battery voltage during charging.

Fig. 8.9 represents the definitions of the voltage-related terminologies used in this chapter.
represents the battery voltage at the end of the resting period after charging but just before
the discharging step is commenced (the last point of the resting period), is the battery
voltage at the moment of initiating the discharging step (the first point of discharging) after
resting, denotes the battery voltage at the end of the resting period after discharging but
just before the charging step and is the initial battery voltage during charging (the first
point of charging). and indicate the equilibrium state of the battery after charging
211

and discharging, respectively. and reflect the overpotential of discharging and

charging processes, respectively.
(a) 40oC, 0.5C (b) 60oC, 0.5C

4.4
Vend Vend
C−R
C−R
4.0 Vini
R−D
Vini
R−D
Voltage / V
3.6
Vini
R−C
Vini
R−C
3.2
Vend
D−R Vend
D−R
2.8
0 300 600 900 0 300 600 900

Fig. 8.10. The development of , , and as a function of cycle number at 0.5 C-rate
at 40oC (a) and 60oC (b).

The development of , , and as a function of cycle number at 0.5 C-

rate at 40oC (a) and 60oC (b) are shown in Fig. 8.10. and are almost constant upon
cycling at both temperatures, indicating that the equilibrium state of the battery after charging
is the same. Therefore the extra capacity flowing into the battery during charging CEc  100%
cannot be attributed to the variation of the charging state. On the other hand, there is no clear
trend of and at 40oC observed in Fig. 8.10a. However, it can clearly be seen that
and increase upon cycling at 60oC in Fig. 8.10b. The increase of is attributed
to the cyclable Li immobilization in the anode.
The difference between and , and determines the overpotentials of
the charging and discharging process, respectively. The corresponding ohmic resistance during
charging ( ) and discharging ( ) can be calculated on the basis of these overpotentials and
the (dis)charge current and is shown in Fig. 8.11. Fig. 8.11a shows the development of the
charging ( ) and discharging ( ) ohmic resistance upon cycling at 0.5 C-rate at 40oC. It can
be seen that both and increase as a function of cycle number. Strikingly, is found
to be larger than in call cases. Fig. 8.11b shows the development of and upon cycle
numbers at 0.5 C-rate at 60oC. Similar to the case in Fig. 8.11a, and also increase with
cycle number.
212


80
(a) 40oC (b) 60oC
70 Rch Rch
Ω Ω
RΩ / mΩ
60
50
Rd Rd
Ω Ω
40
0 300 600 900 0 300 600 900

Fig. 8.11. The development of charging ( ) and discharging ( ) ohmic resistance as a function of
o o
cycle number at 0.5 C-rate at 40 C (a) and 60 C (b).

Fig. 8.12. The development of – / versus curves as a function of cycle number at various
cycling currents at 40oC (a-c) and 60oC (d-f). Three distinct peaks are indicated as (gray), (blue) and
(red).
213

Fig. 8.12 shows the development of – / versus curves as a function of cycle

number at 0.1, 0.3 and 0.5C at 40oC and 60oC. Three distinct peaks are observed which are
indicated as (gray), (blue) and (red). The slopes and peaks in / curves are
corresponding to the slopes in EMF curves while depressions in / curves correspond
to the plateaus in the EMF curves. In order to facilitate the analysis of the shift of these peaks,
all curves are aligned with respect to peak at approximately 1.6 Ah. The blue vertical lines
indicate the position of and peak at cycle number is 0. The red sloping lines connect the
peak of the various / curves in order to make the shift of the peak more visible. It
can be seen that the shift of the peak is negligible when the current is 0.1C at both 40 and
60oC. This shift at 60oC becomes more pronounced than at 40oC when the current increases to
0.5C. The shift of the peak indicates the material degradation during cycling [39, 40].
Summarizing the above results it can be concluded that the electrode material degradation at
60oC is more severe than at 40oC especially at higher currents.
Obvious changes of the peak upon cycling appear in Figs. 8.12c and 8.12f. It can be seen
that the peak becomes indistinct upon cycling to almost disappear after 500 cycles. The
changes of the peak can be attributed to the voltage slippage [41] of the individual electrodes.
8.3 Conclusions
extrapolation of the measured voltage discharge curves. The irreversible capacity losses, which
have been accurately determined from the EMF curves, have been systematically investigated
as a function of time and cycle number under various cycling currents and temperatures. Similar
to the conclusions obtained from LFP batteries, it is found that the capacity loss increases with
temperature and current. The irreversible capacity losses of LFP and NMC(111) batteries under
the same cycling conditions have been compared and it was found that the capacity degradation
of NMC(111) batteries is smaller than that of LFP batteries at low current but becomes
significantly larger at high current.
The charge efficiency of the individual electrodes has been investigated. The charge
efficiency at the cathode was found always larger than 100% while was lower than 100% at the
anode. The parasitic side reactions occurring at the cathode and anode are considered to be
responsible for the deviation of the charge efficiency.
214


The ohmic resistances of the charging and discharging processes at 40 and 60oC were
calculated on the basis of the initial voltage changes during charging and discharging. was
found to be larger than . Both and increased as a function of cycle number.
Finally, the / curves calculated from the corresponding EMF curves have been
investigated. The changes of the peaks observed from the / curves can be used to
determine the electrode material decay and voltage slippage of the individual electrodes.
8.4 References
2 (2012).
[3] N. Yabuuchi, K. Yoshii, S.T. Myung, I. Nakai, S. Komaba, Journal of the American
Chemical Society, 133 (2011) 4404-4419.
[4] K. Amine, Z.H. Chen, Z. Zhang, J. Liu, W.Q. Lu, Y. Qin, J. Lu, L. Curtis, Y.K. Sun,
Journal of Materials Chemistry, 21 (2011) 17754-17759.
[5] B.J. Hwang, Y.W. Tsai, D. Carlier, G. Ceder, Chemistry of Materials, 15 (2003) 3676-
3682.
[6] Y. Koyama, I. Tanaka, H. Adachi, Y. Makimura, T. Ohzuku, Journal of Power Sources,
119 (2003) 644-648.
[7] R. Robert, C. Villevieille, P. Novak, Journal of Materials Chemistry A, 2 (2014) 8589-
8598.
[8] J. Choi, A. Manthiram, Journal of the Electrochemical Society, 152 (2005) A1714-A1718.
[10] D. Mohanty, H. Gabrisch, Journal of Power Sources, 220 (2012) 405-412.
[11] N. Nitta, F.X. Wu, J.T. Lee, G. Yushin, Materials Today, 18 (2015) 252-264.
[12] Y.B. He, F. Ning, Q.H. Yang, Q.S. Song, B.H. Li, F.Y. Su, H.D. Du, Z.Y. Tang, F.Y.
Kang, Journal of Power Sources, 196 (2011) 10322-10327.
215

[13] N. Kiziltas-Yavuz, M. Herklotz, A.M. Hashem, H.M. Abuzeid, B. Schwarz, H.

Ehrenberg, A. Mauger, C.M. Julien, Electrochimica Acta, 113 (2013) 313-321.
[14] B. Stiaszny, J.C. Ziegler, E.E. Krauss, J.P. Schmidt, E. Ivers-Tiffee, Journal of Power
Sources, 251 (2014) 439-450.
[15] B. Stiaszny, J.C. Ziegler, E.E. Krauss, M.J. Zhang, J.P. Schmidt, E. Ivers-Tiffee, Journal
of Power Sources, 258 (2014) 61-75.
[16] D.A. Stevens, R.Y. Ying, R. Fathi, J.N. Reimers, J.E. Harlow, J.R. Dahn, Journal of the
[17] Y. Li, M. Bettge, B. Polzin, Y. Zhu, M. Balasubramanian, D.P. Abraham, Journal of the
[18] Y. Kobayashi, T. Kobayashi, K. Shono, Y. Ohno, Y. Mita, H. Miyashiro, Journal of the
[19] Y. Kobayashi, T. Kobayashi, K. Shono, Y. Ohno, Y. Mita, H. Miyashiro, Journal of the
[20] S. Bourlot, P. Blanchard, S. Robert, Journal of Power Sources, 196 (2011) 6841-6846.
[21] Y.H. Wang, X. Yan, X.F. Bie, Q. Fu, F. Du, G. Chen, C.Z. Wang, Y.J. Wei,
[22] M. Kerlau, M. Marcinek, V. Srinivasan, R.M. Kostecki, Electrochimica Acta, 53 (2007)
1385-1392.
[23] S.H. Kang, D.P. Abraham, W.S. Yoon, K.W. Nam, X.Q. Yang, Electrochimica Acta, 54
(2008) 684-689.
[24] M. Kerlau, J.A. Reimer, E.J. Cairns, Electrochemistry Communications, 7 (2005) 1249-
1251.
[25] W. Choi, A. Manthiram, Journal of the Electrochemical Society, 153 (2006) A1760-
A1764.
[26] H. Gabrisch, R. Yazami, Electrochemical and Solid State Letters, 13 (2010) A88-A90.
[27] H. Gabrisch, T.H. Yi, R. Yazami, Electrochemical and Solid State Letters, 11 (2008)
A119-A124.
[28] J. Shu, R. Ma, L.Y. Shao, M. Shui, K.G. Wu, M.M. Lao, D.J. Wang, N.B. Long, Y.L.
Ren, Journal of Power Sources, 245 (2014) 7-18.
[29] J. Shim, R. Kostecki, T. Richardson, X. Song, K.A. Striebel, Journal of Power Sources,
112 (2002) 222-230.
[30] A. Vu, L.K. Walker, J. Bareno, A.K. Burrell, I. Bloom, Journal of Power Sources, 280
(2015) 155-158.
216


[31] D.R. Gallus, R. Schmitz, R. Wagner, B. Hoffmann, S. Nowak, I. Cekic-Laskovic, R.W.

Schmitz, M. Winter, Electrochimica Acta, 134 (2014) 393-398.
[32] S.C. Yin, Y.H. Rho, I. Swainson, L.F. Nazar, Chemistry of Materials, 18 (2006) 1901-
1910.
[33] B. Xu, D.N. Qian, Z.Y. Wang, Y.S.L. Meng, Materials Science & Engineering R-Reports,
73 (2012) 51-65.
[35] M. Ochida, Y. Domi, T. Doi, S. Tsubouchi, H. Nakagawa, T. Yamanaka, T. Abe, Z.
Ogumi, Journal of the Electrochemical Society, 159 (2012) A961-A966.
A513-A518.
[37] T. Joshi, K. Eom, G. Yushin, T.F. Fuller, Journal of the Electrochemical Society, 161
(2014) A1915-A1921.
[38] H. Takahara, Y. Kobayashi, K. Shono, H. Kobayashi, M. Shikano, T. Nakamura, Journal
[40] D.J. Li, D.L. Danilov, J. Xie, L. Raijmakers, L. Gao, Y. Yang, P.H.L. Notten,
[41] A.J. Smith, J.C. Burns, J.R. Dahn, Electrochemical and Solid State Letters, 14 (2011)
A39-A41.
217

218

Chapter 9
Summary
In the past centuries the rapid development of our world was ensured by a vast amount of
fossil fuels. Nowadays it is commonly accepted that such dependence on fossil fuels creates an
obstacle for sustainable development of our society. The limited resources which are irregularly
distributed between the countries and continents induce conflicts and social unrest. Furthermore,
excessive emissions of greenhouse gases pollutes the living environment and accelerates the
global warming processes with potentially disastrous consequences [6-8, 282, 302]. The
application of Li-ion batteries in electric vehicles (EV) and hybrid electric vehicles (HEV) has
recently drawn major attention. The replacement of internal-combustion-engine-driven vehicles
by EV is considered to be an effective way to mitigate the air pollution problems in large cities.
The aging performance has a significant consequence on the reliability and safety of Li-ion
battery systems. Cyclable Li loss and electrode material decay are the two most important
processes responsible for battery aging. Solid-electrolyte-interphase (SEI) formation at the
anode is generally believed to be the main reason responsible for cyclable Li losses. The SEI
formation mechanisms, experimental characterization and model development have been
systematically reviewed in Chapter 2. Cathode degradation which mainly occurred under
severe aging conditions has also been discussed. Transition metal dissolution and structural
transformation are considered to be the main cause of cathode degradation. Finally, degradation
of the graphite electrode has been discussed, which is mainly caused by structural deformation
and inter-layer blockage.
Proper battery tests are essential to understand the battery degradation mechanisms. Testing
the lifetime of Li-ion batteries under real EV application conditions may take several years.
Therefore, accelerated testing is usually carried out under various conditions. Testing details
are described in Chapter 3. Both LFP and NMC(111) batteries have been investigated in this
thesis. The aging experiments include both storage and cycling experiments performed with
complete batteries and have been carried out with automated cycling equipment. The storage
experiments were conducted under various SoC and temperature conditions. The cycling
219

Chapter 9 Summary

experiments were performed with various cycling currents, SoC ranges and temperatures.
Material characterization focused on dismantled electrodes and have been performed by X-ray
Photoelectron Spectroscopy (XPS), Raman spectroscopy, Inductively Coupled Plasma (ICP)
spectrometry and Scanning Electron Microscopy (SEM).
To get more insight into the aging mechanisms from conventional experiments, novel and
efficient analysis methods have been adopted. The details of the methodologies adopted in this
thesis have been addressed in Chapter 4. The Electromotive Force (EMF) curves, which have
been regularly determined on the basis of a set of discharge voltage curves, are considered to
be an important tool to obtain in-depth understanding of aging mechanisms inside Li-ion
batteries. Various parameters, such as maximum capacities ( ), irreversible capacity losses
(Δ ), overpotentials ( ), etc., can be extracted from these EMF curves. The development of
the second depression in the / curves of LFP batteries is proven to be an interesting
indicator for the graphite electrode decay (Δ ).
The capacity loss and material decay of LFP batteries during storage have been
systematically investigated and discussed in Chapter 5. Δ is found to increase as a function
of temperature and SoC. It is concluded that the origin of Δ is mainly related to the lithium
immobilization in the SEI layers formed on the graphite electrode. The thickness of the SEI
layers has been determined by XPS analysis by sputtering. It was found that the thickness of
the SEI layers at high temperatures is much thicker than at lower temperatures which is in line
with the conclusions of Δ . Based on the EMF curves, / curves have been applied
to investigate the aging mechanisms. Detailed analyses of the / curves provide a non-
destructive approach to quantitatively determine the graphite inaccessibility. The accessibility
of the graphite electrode is found to be reduced at higher temperatures. The iron deposition on
the graphite surface is considered to be the origin for the graphite decay under storage
conditions since the graphite electrode will be blocked and Li intercalation will be hindered.
The presence of iron at higher storage temperatures has been experimentally proven by XPS
and ICP. The deposited metal clusters on the graphite electrode facilitate electron transport
through the SEI layer and, consequently, accelerate the SEI formation and growth. Iron
deposition was found to be negligible at moderate temperatures.
The capacity loss and the graphite electrode material decay of LFP batteries during cycling
have been discussed in detail in Chapter 6. In this chapter, two types of commercial LFP
batteries have been investigated, including prismatic 50 Ah batteries and cylindrical 2.3Ah
batteries. The prismatic battery has been cycled at low C-rate only. The capacity losses induced
220

Chapter 9 Summary

by cycling have been compared with the corresponding storage conditions. A faster capacity
decay was found during cycling than during storage.
The cylindrical batteries have been investigated in more detail at various temperatures (20,
40 and 60oC), cycling currents (0.1C, 0.5C, 1C and 2C) and cycling SoC-windows (0-30%, 35-
65% and 70-100%). The maximum storage capacities have been calculated on the basis of the
EMF curves, which have been regularly determined during cycling. The irreversible capacity
losses Δ are accurately calculated on the basis of these maximum storage capacities. It is
concluded that the origin of Δ during cycling is mainly attributed to the Li-ion
immobilization in the SEI layers formed on the graphite surface. In addition, the reduction of
metal ions at the graphite electrode accelerates this process. Δ increases as a function of both
cycle number and time. The individual contributions of the cycle number and time have been
discriminated by a newly proposed mathematical extrapolation method. The capacity losses
induced by calendar aging ( Δ ) at the equilibrium state have been determined by
extrapolation Δ to zero current. Δ is found to increase logarithmically with time. The
capacity losses related to crack formation (Δ ), induced by the electrode volume changes,
have been obtained by extrapolation from Δ to zero time. The growth of Δ is found to be
linear with cycle number. Both Δ and are temperature dependent.
The thickness of the SEI layers has been investigated by XPS analyses. It has been
concluded that the SEI thickness determined by XPS is mainly related to the SEI formation on
the SEI covered surface areas (Δ ). Therefore the SEI thickness is very similar at various
cycling currents when the cycling time is kept the same. Both temperature and cycling range
can influence the SEI thickness. It is found that the SEI layers are thicker at higher temperatures
and at higher SoC windows.
The inaccessibility of the graphite electrode during cycling has been quantitatively
calculated on the basis of the / curves. Temperature, current and the cycling SoC-
ranges are found to have an influence on the inaccessibility of the graphite electrode. The
graphite electrode capacity loss (Δ ) was found to be minor at low temperatures but became
significant at elevated temperatures when the batteries were cycled in the full SoC-range.
However, Δ becomes significant when the battery was cycled at low SoC-windows even at
low temperatures. Furthermore, Δ was found to be more pronounced at high currents than
at low currents. Graphite degradation during cycling was concluded to be a consequence of
structural deterioration as well as metal deposition on the anode, which have been confirmed
by Raman and XPS analyses, respectively. It has been concluded from the LFP battery
221

Chapter 9 Summary

measurements that the graphite electrode capacity decreases faster than the battery capacity,
especially at higher temperatures. The second slope on the battery capacity degradation can be
related to the case that the graphite electrode becomes the capacity-limiting electrode.
Modeling is an efficient way to investigate battery degradation. The developed models are
generally classified as empirical and theoretical models. Empirical models are widely used in
Battery Management System (BMS) of electronic portable devices and electric vehicles. Such
models are typically developed using data that is collected under tightly controlled experimental
conditions to predict its future state. The advantage of empirical models is that the computation
time is short. However, the battery characteristics cannot be updated along with battery aging
processes since the physics-based parameters cannot be determined by the empirical model. In
contrast with empirical models, theoretical models have been developed on the basis of physical
parameters which can accurately describe the battery state under various operating conditions.
Only a few models are related to degradation mechanisms. In Chapter 7 an advanced model is
proposed considering both the SEI formation and cathode dissolution processes.
The capacity losses at moderate aging temperatures are all attributed to the SEI formation.
The SEI formation model assumes the existence of a compact inner and porous outer SEI layer.
The rate-determining step is considered to be electron tunneling through the inner SEI layer.
Both SEI layers are growing at the interface of the inner and outer SEI layer. The inner SEI
layer grows much slower than the outer layer. The initial thickness of the inner layer, developed
after activation, will largely determine the future degradation rate. It has been concluded that
the capacity losses are strongly dependent on the storage SoC, i.e. by the electrode potential and
tunneling probability, and on the cycling currents (overpotential development and tunneling
barriers). The capacity losses during cycling are larger than during storage for the same
operating time due to the crack formation in the SEI during (dis)charging. These cracks generate
free graphite surface areas exposed to the electrolyte, where the new SEI will be formed easily.
The capacity losses due to the SEI formation on these fresh graphite surface areas are a function
of charging time and cycling number. A relationship between the capacity loss and aging time
( ln t ) is found when the capacity losses are attributed to the SEI formation only.
A cathode dissolution model has been proposed to predict the transition-metal dissolution
processes at elevated temperatures under storage and cycling conditions. The cathode
dissolution is assumed to be driven by the proton exchange reaction. The concentration of
protons in the electrolyte determines the cathode dissolution rate. The dissolved metal ions can
be transported to the anode and deposited on the graphite surface. Both cathode dissolution and
222

Chapter 9 Summary

metal deposition can cause battery capacity losses. Moreover, the SEI formation on these metal-
cluster surfaces has been simulated, confirming considerable capacity losses.
Finally, the influence of both the particle size and particle surface area of graphite on the
capacity losses has been discussed. A linear relationship between the surface area and capacity
loss was observed while a reciprocal relationship between the particle size and the capacity loss
was found in accordance with the proposed model. The model has been applied to predict the
calendar life and cycle life performance of LFP batteries.
The demand for rechargeable batteries for application in electric vehicles and energy storage
systems connected to the smart grid has been growing rapidly in the past few years. The ternary
NMC electrode material has drawn more and more attention due to their high energy density,
long cycling life and high safety performance. The ternary NMC system contains a large group
of family members including LiNi1/3Co1/3Mn1/3O2 (NMC(111)), LiNi0.5Co0.2Mn0.3O2
(NMC(532)), LiNi0.425Co0.15Mn0.425O2, LiNi0.25Co0.5Mn0.25O2, etc. However, at present only
NMC(111) and NMC(532) have been successfully commercialized as cathode materials in Li-
ion batteries. It is worthwhile to point out that the degradation mechanism of NMC materials
depends on their specific composition. In Chapter 8, aging mechanisms of NMC(111) batteries
have been discussed.
extrapolation from the measured voltage discharge curves. The irreversible capacity losses,
which have been accurately obtained from the EMF curves, have been systematically
investigated as a function of time and cycle number under various cycling currents and
temperatures. Similar to the conclusions obtained from the LFP batteries, it was found that the
capacity losses increase with temperature and current. The irreversible capacity losses of LFP
and NMC(111) batteries have been compared under the same cycling conditions and it was
found that the degradation of NMC(111) batteries is smaller than that of LFP batteries at low
currents but becomes much larger at high currents.
The charge efficiency of the individual electrodes has been investigated and was always
found to be larger than 100% for the cathode and smaller than 100% for the anode. Parasitic
side-reactions occurring at both the cathode and anode are considered to be responsible for the
deviation of the charge efficiency. Interestingly, / curves calculated from the
corresponding EMF curves have been analyzed. Changes in the observed peaks gave
information about the electrode material decay and the voltage slippage of individual electrodes.

223

Chapter 9 Summary

Fig. 9.1. Chart of the battery aging investigation strategy

The research strategy adopted in this thesis to investigate Li-ion battery aging processes is
summarized in Fig. 9.1. The study is composed of three steps: (i) battery test, (ii) data analyses
and (iii) the simulations.
(i) Battery testing
The testing step constitutes storage and cycling measurements. Different battery SoC and
ambient temperatures have been controlled during storage. The currents, cycling ranges and
temperatures are used as variables in the cycling experiments. Battery characterization is
based on a set of voltage discharge curves with various C-rates, and has been regularly
performed to determine the EMF curves upon aging. After the aging experiments, the
batteries were dismantled for further characterizing the electrode materials, by Raman
analysis and XPS.
(ii) Data analyses
Experimental data collected from the measurements has been systematically analyzed. The
EMF curves determined during characterization are essential in the investigation of aging
mechanisms. The maximum battery storage capacities obtained from the EMF curves
are used to calculate the irreversible capacity losses Δ . Furthermore, / plots
can be obtained from the EMF curves. In LFP systems the changes of the second depressions
in these / curves are considered to be a useful indicator for the graphite electrode
decay Δ .
(iii) Simulations
The irreversible capacity losses Δ are simulated on the basis of a complete aging model
which is composed of an SEI formation model and a cathode dissolution model. Electron
tunneling is assumed to be rate determining in the SEI formation model. The cathode
dissolution is assumed to be controlled by proton exchange with the transition metal.
224

Chapter 9 Summary

References
Acta, 102 (2013) 183-195.
[2] P.H.L. Notten, D.L. Danilov, Advances in Chemical Engineering and Science, 4 (2014)
62-72.
[3] P.H.L. Notten, M. Latroche, Metal hydride alloys, Elsevier, Amsterdam, 2009.
[4] H.J. Bergveld, W.S. Kruijt, P.H.L. Notten, Battery management system design by
modelling, Kluwer Academic Publischers, Dordrecht, The Netherlands, 2002.
[5] J.M. Tarascon, M. Armand, Nature, 414 (2001) 359-367.

225

226

Appendix I. Derivation of Eq. 1.18

Appendix I
Derivation of Eq. 1.18
Under ideal conditions, the anode electrode capacity is equal to the cathode electrode
capacity and also equal to the battery capacity
. [A1.1]
The electrode capacities can be calculated according to
, [A1.2]
and
, [A1.3]
where and are the specific capacities of the anode and cathode electrode materials,
respectively. and are the mass of the corresponding active materials. The total mass of
the battery is given by
, [A1.4]
where represents the mass of other species, including carbon black, binders, electrolyte,
separator, battery can, etc. The specific capacity of battery is, therefore, given by
Qbat Qbat
 . [A1.5]
mbat ma  mc  mo
Eliminating and on the right-hand side of Eq. A1.5 by using Eqs. A1.1- A1.3, the
specific capacity of battery can be expressed as
Qbat qa qc
 . [1.18]
mbat  mo 
qa  1   qc
 ma 
227

228

Appendix II. Derivation of the tunneling probability

Appendix II
Derivation of electron tunneling probability
For a one-dimensional rectangular barrier model (see Fig. 7.2), the space can be divided
into 3 parts: region I corresponds to graphite electrode, where we suppose 0; region II
corresponds to the inner SEI, where 0 ; region III corresponds to the solvent side,
where . The wave function of electrons in various regions is calculated according to the
stationary Schrödinger equation.
In region I, the Schrödinger equation takes the following form
d 2 1 2m
 2  E f  U I  1  0 , [A2.1]
dx 2 
where is the electron wave function in region I. The solution of Eq. A2.1 yields
 1  eik x  re  ik x ,
1 1
[A2.2]
and therefore
d 1
 ik1  e ik1 x  re  ik1 x  , [A2.3]
dx
2m  E f  U I 
where k1  and are constants to be determined from the boundary conditions.
2
Similarly, the wave function in region II can be calculated, according to
d 2 2m
2
 2  E f  U II   0 , [A2.4]
dx 
where is the electron wave function in region II. Solving Eq. A2.4 leads to
  aei x  be  i x , [A2.5]
and
d
 i  ae i x  be  i x  , [A2.6]
dx
229


2 m U II  E f ,
where   2
and are constants.

In region III, the Schrödinger equation can be written as
d 2 2 2m
 2  E f  U III  2  0 , [A2.7]
dx 2 
where is the electron wave function, which leads to
 2  ce ik x ,2
[A2.8]
and
d 2
 ik 2 ceik2 x , [A2.9]
dx
2 m  E f  U III 
where k 2  and are constants.
2
Considering that both the wave function and its derivative are continuous, the continuity
conditions must be applied at the boundary. As a result, the following system of equations holds
  1  x  0    x  0

   x  l  2  x  l 
in in
 d
 d [A2.10]
 dx
1
 x  0   x  0
dx
 d d 2

 dx
 x  l in  
dx
 x  l in 
Substituting the corresponding boundary values, described by Eqs. A2.2, A2.3, A2.5, A2.6,
A2.8 and A2.9, into Eq. A2.10 leads to the following system of equations
r  a  b  1
k1r  a  b  k1
in in in . [A2.11]
aei l  be i l  ceik2l  0
in in in
a ei l  b e i l  ck2eik2l  0
Eqs. A2.11 can be written in the form of
, [A2.12]
230


where
 1 1 1 0  r 1
k   0  a  
 1  , V    , Z   k1  . [A2.13]
 0 e i l
in
e  i l
in
 e ik2l 
in
b 0
     
 0  e i l in  e  i l
in
 k 2 e ik2l 
in
c 0

A straightforward calculation leads to
4 k1e 
  k 2i l in
c , [A2.14]
 
  k1  k 2  e 2 l  1  i  k1k 2   2  e 2 l  1
in
 in

which is known as ‘transmission coefficient’. Note that the magnitude of the complex number
, given in Eq. A2.14, determines the amplitude of the complex function according to Eq.
A2.8. However, the function of interest is not the transmission coefficient itself but rather a
transmission probability, given by
k2
P cc , [A2.15]
k1
where ̅ is a complex conjugate of , i.e.
4 k1e 
  k 2 i l in
c  . [A2.16]
 
  k1  k 2  e 2 l  1  i  k1k 2   2  e 2 l  1
in
 in

Straightforward multiplication leads to the tunneling probability of electrons on their Fermi
level
in
k2 16 2k1k2e2l
P  cc  2 , [A2.17]

  k1  k2  e  1   k1k2    e  1   
2
k1 2 2 2l in 2 2 2l in
where
1
k1  2m  E f  x   U I  , [A2.18]

1
k2  2 m  E f  x   U III  , [A2.19]

231


1
 2 m U II  E f  x   , [A2.20]

and m is mass of an electron, is corresponding to the graphite Fermi level, 4.4 ,

is the solvent LUMO, which is determined by the state of the solvent, e.g. when 3 EC
molecules combine with one Li ion, then the LUMO of each solvent is -2.99 eV.
corresponds to the energy level of free electrons, here assumed 0, is the electron
fermi level of the LixC6 electrode.
Note that
in
16 2k1k2 e 2 l
P , [A2.21]
 
 2  k1  k2  e 2 l  1   k1k2   2  e 2 l  1  
2 in 2 2 in 2
therefore
16 2k1k2 in
P e2 l , [7.11]
 2  k1  k2    k1k2   2 2

2
thus
in
P  P0 e 2 l , [7.12]
where
16 2k1k2
P0  . [7.13]
 2  k1  k2    k1k2   2 
2 2
Eq. 7.12 has much simpler form, widely accepted in the literature, but usually assumes 1,
which clearly violates Eq. 7.13. It is interesting to find out the dependence of as a function
of the underlying parameters , and . Eq. 7.13 can be rewritten in the following form
16
P0  . [A2.22]
k1 k 2 k1k2  2
  
k 2 k1  2 k1k2
From Eq. A22 it can be seen that depends only on two combinations of parameters. Defining
k1 kk
x and y  1 22 will lead to
k2 
232


16
P0  . [A2.23]
1 1
x  y
x y
The behavior of as function of and is illustrated in Fig. A2.1.

It can be seen that is symmetric with respect to and . It varies between 0 and 4,
reaching the maximal value when 1 , which corresponds to the symmetric case
. This implies that using the simplified Eq. 7.12 with 1 can lead to considerable
errors. This error can overestimate or underestimate the real tunneling probability.
2
P0
0
0
2 10
4 8
6
6 4
8 2
x 10 0 y

Fig. A2.1. calculated according to Eq. A2.23 as function of and .
233

234

Appendix III. Derivation of Eq. 7.34 from Eq. 7.33

Appendix III
Derivation of Eq. 7.34 from Eq. 7.33

st
 t   2 mE 
 2  l  
dQSEI  t   6  x  F  C6 ve AC6
st
  0
 in
A w in
F 
P0 exp    
C Li
 6
. [7.33]
dt 4 M C6 
 
 
 
Moving the capacity term to the left-hand side and to the right-hand side results in
 2 M Li QSEI
st
 t  2 mE  st  6  x  F C6 ve AC6  2l in 2mE 
exp  
  dQSEI  t   P0 exp   0  dt . [A3.1]
  AC6 wLiin F 
in
 4 M C6   
When 0, the capacity loss 0 0, and integrating both sides of Eq. A3.1 this yields
 2 M Li QSEI  t  2mE  st t  6  x  F  C ve AC  2l0in 2mE 

st
st
QSEI t 
 exp  
  dQSEI  t   0 6 6
P0 exp    dt ,
0
in
 4 M C6   
[A3.2]
therefore
 in AC wLiin F    2 M Li Q SEI

st
 t  2 mE    6  x  F  C6 ve AC6 t  2l0in 2 m  E 
 exp     
 
6
1 P exp 
2 M Li 2 m  E    AC6 wLiin F 
in
  4 M C 6
0
  
[A3.3]
Shifting the multiplier from the left-hand side to the right-hand side leads to
 2 M Li QSEI
st
 t  2 mE   6  x  C6 ve M Li t 2mE  2l0in 2mE 
exp 
   1  P exp   . [A3.4]
in
 2 M C6
 in in
wLi 
0
  
235

Appendix III. Derivation of Eq. 7.34 from Eq. 7.33

Moving the second term from the left-hand side to the right-hand side, this leads to
 2 M Li QSEI
st
 t  2mE   6  x  C6 ve M Li t 2mE  2l0in 2mE 
exp    

1 P exp  . [A3.5]
  AC6 wLiin F 
in
2 M  in in
w 
0

  C6 Li  
Taking the logarithm of both sides results in
2 M Li QSEI
st
 t  2mE  ln  1   6  x  C6 ve M Li t 2mE P exp   2l0in 2mE 
    . [A3.6]
 in AC6 wLiin F   2 M  in in
w 
0
  
 C6 Li
Finally, the capacity loss is obtained as a function of time, according to
 in wLiin AC F  6  x   C ve M Li P0 2mE  2l0in 2mE 

st
QSEI t   6
ln  1  6
t exp    . [7.34]
2 M Li 2mE  2 M C6  in wLiin    
236

List of publications
Journal publications
1. D. Li, D. Danilov, Z. Zhang, H. Chen, Y. Yang and P.H.L. Notten, Electron tunneling based
SEI formation model, ECS Transactions, 62 (2014) 1-8.
2. D. Li, D. Danilov, Z.R. Zhang, H.X. Chen, Y. Yang and P.H.L. Notten, Modeling the SEI-
Formation on Graphite Electrodes in LiFePO4 Batteries, Journal of the Electrochemical
Society, 162 (2015) A858-A869.
3. D. Li, D. Danilov, J. Xie, L. Raijmakers, L. Gao, Y. Yang, P.H.L. Notten, Degradation

Mechanisms of C6/LiFePO4 Batteries: Experimental Analyses of Calendar Aging,
4. D. Li, D. Danilov, L. Gao, Y. Yang, P.H.L. Notten, Degradation Mechanisms of Li-ion

Batteries: Experimental Analyses of cycling-induced Aging on C6/LiFePO4 Batteries,
5. D. Li, D. Danilov, L. Gao, Y. Yang, P.H.L. Notten, Degradation Mechanisms of the graphite
electrode in C6/LiFePO4 Batteries Unraveled by A Nondestructive Approach, Journal of the
6. D. Li, D. Danilov, Y. Yang, P.H.L. Notten, Degradation Mechanisms of Li-ion Batteries:

Modeling of Calendar and Cycling Aging on C6/LiFePO4 Batteries, Electrochimica Acta,
to be submitted (2016).
7. J. Xie, J.F.M. Oudenhoven, P.-P.R.M.L. Harks, D. Li, and P.H.L. Notten, Chemical Vapor
Deposition of Lithium Phosphate Thin-Films for 3D All-Solid-State Li-Ion Batteries,
237

8. J. Xie, P.R.M.L. Harks, D. Li, L.H.J. Raijmakers, P.H.L. Notten, Planar and 3D deposition
of Li4Ti5O12 thin film electrodes by MOCVD, Solid State Ionics 287 (2016) 83–88
9. J. Xie, J.F.M. Oudenhoven, D. Li, C. Chen, R-A. Eichel, and P.H.L. Notten, High Power
and High Capacity 3D-structured TiO2 Electrodes for Lithium-ion Microbatteries, Journal
of the Electrochemical Society, 163 (10) A2385-A2389 (2016).
10. C. Chen, D. Li, L. Gao, P.P.R.M.L. Harks, R-A Eichel, P.H.L. Notten, Carbon-coated core-
shell Li2S@C nanocomposites as high performance cathode material for Lithium-Sulfur
batteries, Journal Materials Chemistry A, accepted, (2016).
Conference contributions
11. D. Li, D. Danilov, Y. Yang, P.H.L. Notten, Electron tunneling based SEI formation model,
17th IMLB (2014), Como, Italy, Oral presentation.
12. D. Li, D. Danilov, Y. Yang, P.H.L. Notten, Modeling the SEI formation on graphite
electrodes in LiFePO4 batteries, Conference on Electrochemical Energy Science and
Technology (EEST2014), Shanghai, China, Oral presentation.
13. D. Li, D. Danilov, Y. Yang, P.H.L. Notten, A new approach to distinguish between calendar
ageing and cycling-induced ageing of Li-ion batteries, 229th ECS meeting (2016), San Diego,
United States, Abstract.
14. D. Li, D. Danilov, Y. Yang, P.H.L. Notten, Aging mechanisms of Li-ion batteries,
International Symposium on Advanced Battery Power 2016, Muenster, Germany, Oral
presentation.
15. D. Li, D. Danilov, Y. Yang, P.H.L. Notten, A New Approach to Distinguish Two Different
Degradation Mechanisms in Cycling Aging of Li-Ion Batteries, 18th IMLB (2016), Chicago,
America, Poster presentation.
238

Acknowledgement
Prof.dr. Peter H.L. Notten (Eindhoven University of Technology) and Prof.dr. Yong Yang
(Xiamen University) gave me the opportunity to study Li-ion battery modeling at the Eindhoven
University of Technology (TU/e). Since my first day I arrived in Eindhoven I was deeply
fascinated by the kindness and friendliness of the people there. I really felt myself at home at
the TU/e during the past 4 years. The university provided unique facilities and opportunities in
supporting my research work. Throughout these years I have not only learned how to make
battery models, but also improved my language, enriched my knowledge and reshaped my way
of thinking. This thesis represents not only my work at the keyboard. It can better be seen as a
milestone in my research life. Here, I wish to thank all people who have supported and assisted
me during my PhD study.
First and foremost, I would like to express my whole-hearted gratefulness to my first
promotor, Prof.dr. Peter Notten, leader of the Energy Material and Device (EMD) group at the
TU/e. Thanks to him I had this opportunity to study battery modeling in his group. As a scientist,
he is intelligent and energetic. His inspiration, enthusiasm, strictness and devotion to scientific
research activities in the field of energy conversion and storage have extremely inspired me to
improve myself to be a qualified scientist. As a supervisor, he is supportive and responsible.
Peter has been supportive since the first day I started working there. He has supported me not
only academically but also emotionally through the rough road to complete this thesis. The
countless helpful discussions and valuable encouragements have been a great drive for my
progress. Furthermore, he gave me a free environment to be able to think independent and to
develop my initiative.
Secondly, I would like to express my extreme appreciations to my second promotor, Prof.dr.
Yong Yang. As one of the most famous battery scientists in China, Prof.dr. Yang is not only
wise and knowledgeable but also thoughtful and patient. I still remember 5 years ago when I
joined his group. Back then I had no background related to Li-ion battery. He patiently
explained me the battery fundamentals and encouraged me to follow the useful courses. His
encouragement and trust greatly rebuilt my confidence. Furthermore, his openness and easy
characteristics facilitated many collaborations which provided me excellent opportunities to
239

learn battery industries and to open new research possibilities. Thanks to his support, I obtained
a funding from CSC and conducted the studies at the TU/e.
Thirdly, I wish to express my sincere gratitude to my daily supervisor, Dr. Dmitri Danilov.
He spent so many efforts in teaching me the program coding, revising papers and writing the
thesis. His guidance helped me a lot during the course of my project and by writing this thesis.
Besides my advisors, I would like to thank the other committee members of my defense
committee: Prof. dr. ir. E.J.M. Hensen, Prof. dr. D.U. Sauer, Prof. dr. R.-A. Eichel, Prof. dr. ir.
H.J. Bergveld and Prof. dr. F. Roozeboom. I appreciate the time they spent reading my thesis.
Their encouragements, insightful comments, and valuable questions are useful to improve the
thesis quality and widen my thoughts from various perspectives. Special acknowledgement to
Prof. dr. ir. Hensen for chairing my defense ceremony.
In EMD group, I would like to acknowledge Dr. Lu Gao, who assisted me measuring Raman
and XPS. Lu also taught me how to analyze these experimental data. Many thanks to my
officemate, Luc Raijmakers, who helped me a lot in Matlab coding. We also had extensive
valuable discussions related to principles but also to experimental details. Jie Xie, my Chinese
officemate, is acknowledged for his assistance in SEM and XPS measurements. We had many
interesting discussions, but not limited to only Li-ion batteries. I also would like to express my
gratitude to Chunguang Chen and Lei Zhou who organized my thesis printing. Many thanks to
Peter-Paul Harks, our funny Dutch colleague, for his warmhearted assistance in daily activities.
Xuefei Chen, the former master student, is also acknowledged for the pleasant time she brought
us. Our former technicians, Anneke Delsing, Jan Verwijlen and Marcel Vliex are acknowledged
for their technical supports in our lab. Sincere gratitude to our former secretaries Jose Jong and
Agnes Hemert for their kind assistances in administrative matters. Our other colleagues Thiru,
Kamil, Joanna in the EMD group are also acknowledged.
Furthermore, I would like to appreciate our colleagues beyond the TU/e. My sincere
gratitude to Prof.dr. Fichtner Maximilian and Barbara Zwikirsch, our German colleagues from
the Helmholtz-Institute Ulm (HIU), for their efforts in ICP measurements. Dr. Zhongru Zhang,
Prof.dr. Zhengliang Gong, Dr. Yixiao Li, Dr. Jun Gao and Huixin Chen from Xiamen
University (XMU) are also acknowledged for their guidance in battery research. Prof.dr. Jun Li
is acknowledged for his kind teaching on battery modeling. Special appreciations to Dr. Jiang
Zhou for providing me so many excellent NMC batteries. The discussions about battery
manufacturing with him was also helpful in my modeling activities. My friends Jin Kang, Gang
Chang, Lisong Deng, Minglian Lin, Yuwei Zhou, Juan Qiao, Yangjuan Li, Zhiying Deng, Liqin
Mo, Yajie Gu, Qian Cai, Chuangneng Cai and Yanpei Wang are also acknowledged for their
240

inspiring encouragements. Special gratitude to my supervisor Prof.dr. Quanming Wang for his
support and encouragement before I started my PhD study.
Last but not the least, I want to thank my family. It was a tough period for my parents during
the past four years I studied abroad. I highly appreciate their understanding and immense
encouragements! Special appreciation also goes to my wife Min Lin. Her spiritual support and
encouragement are the source of motivation. My sisters and brothers are also appreciated for
their kind understanding and encouragements.
241

242

Curriculum Vitae
Dongjiang Li was born on 13th October, 1984 in Shandong province, China. He was educated
in Applied Chemistry at the Qingdao University from 2003 onwards and received his BSc
degree in 2007. In 2008 he started his master’s study in the State Key Laboratory of Physical
Chemistry of Solid Surfaces, in Xiamen University. He graduated on the topic of Heterometallic
Coordination Polymers in 2011. After obtaining his MSc degree, he joined the research group
of Prof.dr. Yong Yang and started the research of Li-ion batteries. In 2012, the collaboration
between Xiamen University (XMU) and Eindhoven University of Technology (TU/e) on the
topic of modeling of battery aging has facilitated his study in the Energy Materials and Devices
group of TU/e headed by Prof.dr. Peter H.L. Notten. His research output is presented in this
dissertation.

243

Aging Mechanisms of Li-Ion Batteries: Seen From An Experimental and Simulation Point of View

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

Aging Mechanisms of Li-Ion Batteries: Seen From An Experimental and Simulation Point of View

Uploaded by

Copyright:

Available Formats

Aging mechanisms of Li-ion batteries : seen from an

experimental and simulation point of view

Document status and date:

Please check the document version of this publication:

Take down policy

Download date: 01. Jun. 2019

Aging Mechanisms of Li-ion Batteries

ter verkrijging van de graad van doctor aan de Technische Universiteit

geboren te Shandong, China

voorzitter: prof.dr.ir. E.J.M. Hensen

Het onderzoek of ontwerp dat in dit proefschrift wordt beschreven is

Aging Mechanisms of Li-ion Batteries

Cover design: Dongjiang Li

Copy right © 2017 by Dongjiang Li

A catalogue record is available from the Eindhoven University of Technology Library

6.1 Introduction ....................................................................................................... 122

Charge efficiency at the anode -

H+ concentration in the electrolyte at 0 mol·L-1

H+ concentration in the electrolyte at time mol·L-1

/ Number of electrons reaching the graphite surface in unit time -

Gravimetric energy density Wh·kg-1

Volumetric energy density Wh·L-1

Fermi level of LiC6 electrode eV

Fermi level of the anode eV

Fermi level of the cathode eV

Faraday constant 96500 C·mol-1

Total dissolution flux of Fe2+ from the cathode mol·s-1

Current density of SEI formation A·m-2

Dissociation reaction constant of LiH2PO4 depends on parameter

Exchange reaction constant depends on parameter

Thickness of the inner SEI layer on C6 m

Initial thickness of the inner SEI layer on C6 m

Total thickness of the inner SEI layer on C6 m

Mole mass of Fe 55.85 g·mol-1

Total mass of the battery g

Δ Li mass in the increased inner SEI layer g

Δ Increased mass of the inner SEI layer g

Δ Moles of Li consumed in the increased inner SEI layer mol

Charge capacity in cycle Ah

, Width of the first plateau of the graphite electrode Ah

, Width of the second plateau of the graphite electrode Ah

, Width of the third plateau of the graphite electrode Ah

Discharge capacity in cycle Ah

, Capacity loss caused by Fe dissolution and deposition Ah

Total capacity of the LiFePO4 electrode Ah

Maximum LiFePO4 electrode capacity (pristine state) Ah

, Total amount of Li on the first plateau of the graphite electrode Ah

, Total amount of Li on the second plateau of the graphite electrode Ah

, Total amount of Li on the third plateau of the graphite electrode Ah

Maximum battery discharge capacity Ah

Capacity loss due to the SEI formation on in each cycle Ah

Capacity consumed by inner SEI growth Ah

, Irreversible capacity losses caused by SEI formation on Fe Ah

Specific capacity of the anode mAh·g-1

Δ Cathode capacity loss Ah

Δ , Decline of the first plateau of the graphite electrode Ah

Δ , Decline of the second plateau of the graphite electrode Ah

Δ , Decline of the third plateau of the graphite electrode Ah

Δ , Li loss on the first plateau of the graphite electrode Ah

Δ , Li loss on the second plateau of the graphite electrode Ah

Δ , Li loss on the third plateau of the graphite electrode Ah

Δ Irreversible capacity loss Ah

Δ Total irreversible capacity loss induced by cycling aging Ah