See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/244689181

Multimodal Physics-Based Aging Model for Life Prediction of Li-Ion Batteries

Article  in  Journal of The Electrochemical Society · January 2009

DOI: 10.1149/1.3043429

CITATIONS READS
367 2,928

4 authors, including:

Mohammadhosein Safari Mathieu Morcrette

Hasselt University Université de Picardie Jules Verne
54 PUBLICATIONS   2,215 CITATIONS    146 PUBLICATIONS   13,121 CITATIONS   

SEE PROFILE SEE PROFILE

Charles Delacourt
French National Centre for Scientific Research
100 PUBLICATIONS   8,722 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Na-ion batteries View project

Solid electrolyte interphase View project

All content following this page was uploaded by Mohammadhosein Safari on 02 December 2014.

The user has requested enhancement of the downloaded file.

 Journal of The Electrochemical Society, 156 共3兲 A145-A153 共2009兲 A145
0013-4651/2008/156共3兲/A145/9/$23.00 © The Electrochemical Society

Multimodal Physics-Based Aging Model for Life Prediction

of Li-Ion Batteries
M. Safari,a M. Morcrette,a A. Teyssot,b and C. Delacourta,*,z
a
Laboratoire de Réactivité et Chimie des Solides, Université de Picardie Jules Verne,
80039 Amiens Cedex, France
b
Renault Research Department, 78288 Guyancourt, France

An isothermal, multimodal, physics-based aging model for life prediction of Li-ion batteries is developed, for which a solvent-
decomposition reaction leading to the growth of a solid electrolyte interphase 共SEI兲 at the carbonaceous anode material is
considered as the source of capacity fade. The rate of SEI film growth depends on both solvent diffusion through the SEI film and
solvent-reduction kinetics at the carbon surface. The model is able to simulate a wide variety of battery aging profiles, e.g.,
open-circuit and constant-voltage storage, charge/discharge cycling, etc. An analysis of capacity-fade data from the literature
reveals that the same set of aging parameters may be used for predicting cycling and constant-voltage storage. The use of this set
of parameters for predicting storage under open-circuit voltage points out that part of the self-discharge is reversible.
© 2008 The Electrochemical Society. 关DOI: 10.1149/1.3043429兴 All rights reserved.

Manuscript submitted October 3, 2008; revised manuscript received November 14, 2008. Published December 29, 2008.

Aging of Li-ion cells is becoming more and more a concern ode surface. Ploehn et al. presented a “solvent-diffusion” model for
because of their potential use in long-life applications such as elec- capacity-fade predictions during storage under constant voltage at
tric vehicles 共EVs兲 and hybrid EVs, for which lifetimes of 10 and various temperatures.5 They came out with a correlation relating the
15 years are expected, respectively.1 In terms of cycle life, a lifetime fractional capacity loss to the square root of storage time. In this
up to 3000 deep cycles can be requested for high-energy applica- study, the kinetics for the side reaction was not considered. Ra-
tions 共typical for EVs兲. As a result of these drastic conditions, there masamy et al. developed a mathematical model that predicts the
is a need for lifetime-prediction methods that cannot be substituted capacity loss of the carbon electrode during open-circuit voltage
for test programs. Among those methods, some are based on a 共OCV兲 storage due to a continuous reduction of organic solvent and
purely statistical approach, like the standard Wohler method,2 while concomitant de-intercalation of lithium from the carbon.6 Their
others are based on a physical description of the aging phenomena model was able to predict the anode potential rise during OCV stor-
and are referred to as physics-based models. age. Just like in the paper by Ramadass,4 no transport limitation
Physics-based models fall into two groups. Some of them are across the SEI was taken into account.
referred to as “phenomenological,” because aging sources 关e.g., a In almost all the above-mentioned studies but the work by Dar-
side reaction leading to the growth of a solid electrolyte interphase ling and Newman,3 the solvent decomposition at the anode surface
共SEI兲兴, explicitly show up in the set of governing equations and are was considered as the side reaction that dominates the capacity fade
in direct interaction with the other model features. The second group of the cell. Furthermore, all of the studies reported so far were
limited to the description of the capacity fade of the battery in a
gathers mathematical models that are essentially similar to those for
single mode of operation, and the choice of neglecting either the
a pristine cell, and some of the model parameters 共e.g., SEI film
solvent transport across the SEI or the side reaction kinetics was not
resistance or thickness, volume fraction of active material, etc.兲 are
obvious.
updated using some isolated empirical relations or curve-fitting pro-
Having a physics-based model with the ability to simulate the
cedures.
cell performance and aging under different modes of operation 共i.e.,
Phenomenological physics-based models constitute the most so- multimodal兲 is much desired for the purpose of battery-life predic-
phisticated but also the most complex approach, because an exten- tion. Furthermore, experimental data of different types 共charge/
sive understanding of the aging mechanisms is required. Once the discharge cycling, OCV storage, constant voltage storage, etc.兲 can
model is validated, it allows for an analysis of a wide variety of be used to refine the set of input parameters for the model, thus
operating conditions and control strategies and can even be used by greatly improving its reliability, predictability, and applicability.
battery manufacturers for improving the battery design. So far, most In this study, a physics-based phenomenological multimodal
of the physics-based models have embedded only a single aging model is developed for battery-life prediction, with the aging con-
source due to both the extreme complexity of real systems and the sisting in a solvent-decomposition side reaction leading to the
lack of theoretical background in the modeling of some aging growth of an SEI film at the carbonaceous anode material. The
sources like the structural degradation of active materials. model is used to analyze capacity-fade data of LiCoO2/graphite Li-
Let us give a brief overview of the phenomenological models ion cells obtained under several aging procedures 共charge/discharge
reported so far in the literature for battery aging. Darling and New- cycling, constant-voltage, and OCV storage兲. Finally, the model is
man made a first attempt to model side reactions in lithium-ion used to assess the influence of the SEI porosity on capacity fade, and
batteries by incorporating a solvent oxidation reaction into a the SEI growth law is studied under two storage modes, namely,
lithium-ion battery model.3 Even though a simplified treatment of constant-potential and OCV.
the oxidation reaction was used, their model was able to draw sev-
eral interesting conclusions about self-discharge processes in the Model Development
cells under consideration and the impact on positive electrode state- Physical picture of SEI growth.— In the case of graphite anodes
of-charge. Ramadass et al. developed a capacity-fade model by as- in Li-ion batteries, there are continuous small-scale surface reactions
cribing the loss of active lithium during charge/discharge cycling to between lithiated graphite and solution species that lead to a gradual
a continuous SEI film formation over the surface of the negative decrease in cyclable lithium and to an increase in the electrode im-
electrode.4 No transport limitation for the solvent across the SEI was pedance upon cycling.7-9 These small-scale reactions take place
considered, i.e., the solvent was assumed to be abundant at the an- when the SEI film has an “open structure” that can be due to a
nonoptimized SEI formation step during the first few cycles, un-
avoidable volume changes in the graphite particles during lithium
* Electrochemical Society Active Member. insertion–deinsertion, SEI-film partial dissolution, side reactions
z
E-mail: [email protected] leading to gas evolution, etc.

Downloaded on 2014-03-24 to IP 129.97.80.157 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 A146 Journal of The Electrochemical Society, 156 共3兲 A145-A153 共2009兲

O O
_
O + e O
O (graph) O

Scheme 1.

Regarding the solvent-reduction mechanisms 关for example, for

ethylene carbonate 共EC兲 or propylene carbonate兴 Dey initially pro-
posed a two-electron reduction process, bringing about a precipitate
of Li2CO3 in the organic electrolyte and ethylene or propylene
gas.10 After extensive studies of solvent- and salt-reduction pro-
cesses in the 1990s, Aurbach et al. concluded that lithium ethylene
dicarbonate 共CH2OCO2Li兲2 and ethylene gas resulting from a one-
electron reduction process of EC are the dominant products; how-
ever, the dicarbonate is highly sensitive to water traces and rapidly
reacts with them to form Li2CO3.11 Yang et al. also investigated the
passive SEI-film composition on the carbon anode surface in contact Figure 1. 共Color online兲 Physical picture for SEI growth at the anode par-
with an EC-based electrolyte using Fourier transform infrared and ticle surface.
mass spectroscopy.12 They observed that the passive SEI film con-
tains chiefly 共CH2OCO2Li兲2 not only for a single EC solvent but
also for binary EC/diethylcarbonate 共DEC兲 共1:1兲 regardless of the
the model simple enough for fast convergence. It is composed of
EC concentration. Only EC is decomposed in a binary solvent con-
共CH2OCO2Li兲2 as the major component, plus minor species: EC,
sisting of EC/DEC or EC/dimethylcarbonate 共DMC兲. DEC and
DMC do not seem to be decomposed in the presence of EC. Using ethylene, and defects that act as charge carriers 共lithium vacancies
high-level density functional calculations, Wang et al. concluded V−Li and interstitial Li+i 兲, etc. Dilute solution theory is used to de-
that reductive decomposition of EC initially proceeds through an scribe the transport properties of the minor species. For the sake of
ion-pair intermediate.13 It then homolytically cleaves via an simplicity, the entire SEI is assumed to be electronically insulating.
11.5 kcal/mol barrier to generate a radical anion, which undergoes a A schematic of the physical picture for the model is shown in Fig. 1.
secondary reaction by barrier-free self-dimerizing. Governing equations.— For aging simulation purpose, the use of
In this study, the SEI is considered to be made of lithium ethyl- simplified models saves a lot of time in solving procedures, because
ene dicarbonate 共CH2OCO2Li兲2. The rate-determining step 共RDS兲 is time-consuming simulations for thousands of cycles or several years
considered to be the radical anion formation as seen in Scheme 1. of storage need to be run. Therefore, and also for the sake of sim-
Then, as in Scheme 2 and 3, several possible follow-up reactions plicity, a single particle 共SP兲 model was used in this study. The SP
may occur that lead to the formation of the ethylene dicarbonate approach was originally developed by Atlung et al.,18 used by Haran
anion. Finally, the precipitation of solid lithium ethylene dicarbonate et al.19 for the metal-hydride battery, and extended to the lithium-ion
occurs as seen in Scheme 4. battery later on. A detailed description of this model can be found in
The SEI film on the negative electrode is generally recognized as the work by Santhanagopalan et al.20 SP models do not consider the
having a complex structure consisting of a thin dense part in the transport limitations in the liquid phase and therefore are not accu-
vicinity of the particle surface of a few nanometers and a thicker rate for high charge/discharge rates.
secondary porous part exposed to the solution. It is considered that Li insertion/deinsertion in the active materials occurs through a
the outer part of the SEI, for which the pores are filled with electro- surface reaction that takes place at the interface between the graphite
lyte, controls the progress of SEI growth and, consequently, the and the SEI and that follows Butler–Volmer kinetics

再 冉 冊
extent of capacity fade and impedance rise. The porous structure of
the SEI film allows for the permeation of solvent molecules toward 0,app ␤ max 共1 − ␤兲F
the active particle surface and therefore controls the rate of solvent iint = Fkint cLi共cLi − cLi兲1−␤ exp 共⌽1 − U − RSEIit兲
RT

冉 冊冎
decomposition reaction. Some previous works consider the transport
of electrons through the SEI as the rate-determining process for SEI ␤F
− exp − 共⌽1 − U − RSEIit兲 关1兴
growth.14,15 Using this assumption, Peled derived a parabolic- RT
growth law for the SEI on lithium-metal electrodes.14 The partial
The variables are as defined in the list of symbols. The potential of
electronic conductivity of the SEI was ascribed to the deposition of
the ion-conductive phase ⌽2 at the interface between the graphite
lithium metal within the bulk of the SEI due to local excessive
and the SEI 共where the reaction takes place兲 is related to the poten-
heating and the high electric field within the film.14 In the case of
tial of the electrolyte phase 共taken to zero because of the SP assump-
SEIs on graphite, lithium-metal deposition within the SEI is less
tion兲 using Ohm’s law and is thus equal to RSEIit, with RSEI共t兲
prone to occur, and therefore, as a first approximation, no electron
transport through the SEI is considered, except perhaps in the thin = ␦共t兲/␬SEI the ionic resistance of the SEI. Note that because we
inner part of the SEI in contact with the graphite for which electron assume a constant ionic conductivity ␬SEI throughout the SEI, the
tunneling is possible. resistance increase is solely dependent on the thickness increase.
From the above considerations, let us describe the physical pic- The expressions for the equilibrium potentials of cathode and anode
ture of the SEI we defined. We decided to model the SEI as a single are given in the Appendix.
phase covering uniformly the surface of the carbon particles. Such Solid-state lithium diffusion in the spherical active particles fol-
an assumption has been extensively used for ion-exchange mem- lows Fick’s second law
brane modeling, for instance, even though a more accurate treatment
would require the consideration of a two-phase system.16,17 The
main advantage is that it considerably reduces the number of gov-
⳵cLi 1 ⳵
⳵t
= 2
r ⳵r
冉
DLir2
⳵cLi
⳵r
冊 关2兴

erning equations and input parameters, which is of interest because Two boundary conditions are required for this equation. One is de-
of the lack of input parameter values and because we want to keep duced from symmetry at the center of the spherical active particles

Downloaded on 2014-03-24 to IP 129.97.80.157 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 Journal of The Electrochemical Society, 156 共3兲 A145-A153 共2009兲 A147

_ _
O O O O
O
O + O
- + H2C CH2
O O -
O O O

Scheme 2.

冏 冏⳵cLi
⳵r r=0
=0 关3兴
as its dimension is significantly smaller than that of active-material
particles. The material balance of solvent 共EC兲 inside the SEI film
takes the following form
and the other results from relating the lithium flux density at the
particle surface to the intercalation current density ⳵cEC ⳵NEC
=− 关7兴

冏 冏
⳵t ⳵r
⳵cLi iint
=− 关4兴 with the solvent flux density defined as
⳵r r=R FDLi
⳵cEC
The RDS for the formation of the SEI product was assumed to be NEC = − DEC + ␯cEC 关8兴
the formation of EC radical anions through Scheme 1. This reaction ⳵r
takes place at the same interface as Li insertion/deinsertion reaction, The second term on the right side of Eq. 8 represents the convection
and the kinetic expression for this side reaction reads term, with ␯ the SEI film velocity defined as

is = −Fkf,scEC exp − 冋 ␤ sF
RT
共⌽1 − RSEIit兲 册 关5兴 ␯=
d␦
dt
关9兴

As the radical anion is consumed quickly to form irreversibly the with ␦ the SEI film thickness. Substitution of Eq. 9 into Eq. 8 and of
final SEI product through the possible reaction schemes described Eq. 8 in Eq. 7 leads to the final form of the material balance on
above, it seems reasonable to use a Tafel-like expression. Note, solvent
however, that Eq. 5 does not have a regular form, as the exponential
term does not show the presence of the equilibrium potential. This ⳵cEC ⳵2cEC ⳵cEC
= DEC 2 − ␯ 关10兴
alternative form was chosen to have only a single adjustable param- ⳵t ⳵r ⳵r
eter kf,s instead of two 共an exchange current density and an equilib- The boundary condition on the solvent flux at the particle–SEI in-
rium potential兲. Equilibrium potentials are not well-known for those
terface 共r = R兲 in the case where the follow-up reaction is the cou-
side reactions, even though some potential ranges were reported in
pling between two radical anions 共Scheme 2兲 reads
the literature for the SEI formation reaction. Mori et al. reported the
main SEI formation step to be occurring between 0.55 and 0.2 V vs is
Li electrode.21 Aurbach et al. stated that the major reduction pro- 兩NEC兩r=R = 关11兴
F
cesses occur at around 1 V vs Li electrode and continue gradually as
the potential is lowered.9 Naji et al. reported that the formation of an The rate of growth of the SEI layer due to the accumulation of
EC radical anion occurs below 0.8 V vs Li electrode.22 In almost all lithium ethylene dicarbonate is related to side-reaction current den-
the previous modeling works, a value of 0.4 V vs Li electrode was sity through
chosen as the equilibrium potential of the SEI-formation side reac- d␦ is M SEI
tion at the anode.4,23-26 =− 关12兴
The current balance for the negative electrode also should be dt 2F ␳SEI
updated to include the side-reaction current density 共is in Eq. 5兲 Another boundary condition needs to be set at the SEI film–
electrolyte interface 共r = R + ␦兲. Assuming the abundance of elec-
it = iint + is 关6兴
trolyte at this interface and no change in electrolyte composition
it is the total current density in the negative electrode 共which refers during the battery life, it is reasonable to have a constant solvent
to the total interfacial surface area of the anode兲 and iint stands for concentration at this boundary. This concentration is not known and
the part of current density that takes part in the intercalation process. depends on the physical properties of the SEI film. Because we have
Next, the transport of solvent molecules 共EC兲 through the SEI chosen to describe the SEI as a single phase, the solvent concentra-
film is addressed. The SEI is considered to have a planar geometry, tion 共EC兲 in the SEI cEC at the SEI/electrolyte interface is related to

_
O O O O
_ O
O + O + e
- + H2C CH2
(graph) O O -
O O O

Scheme 3.

O O O O
O + O
- O -
+ 2 Li O
O O SEI Li O O Li

Scheme 4.

Downloaded on 2014-03-24 to IP 129.97.80.157 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 A148 Journal of The Electrochemical Society, 156 共3兲 A145-A153 共2009兲

that in the electrolyte cEC,sln through a partition coefficient which is Table I. Summary of the equations used in the aging model.
a thermodynamic quantity

兩cEC兩r=R+␦ Governing equations

KEC = 关13兴
cEC,sln
However, the value of such a partition coefficient is not known.
0,app ␤
iint = Fkint cLi共cmax
Li − cLi兲
1−␤
兵 关
exp
共1 − ␤兲F
RT 共 ⌽1 − U −
␦
i
␬SEI t 兲兴
兵关 共 兲兴其
Although the SEI film is taken as a single phase, in order to get a ␤F ␦
feel for the range of cEC, we can use the following trick. Let us −exp − ⌽ −U− i
RT 1 ␬SEI t
consider the SEI as a two-phase medium, where the liquid electro-
lyte fills the porosity ␧SEI of the SEI. In this case, the volume-
averaged EC concentration in the SEI would be ␧SEIcEC,sln. Compar-
ing the two physical pictures, the partition coefficient in the SP
model can be regarded as the SEI porosity in the two-phase model
⳵cLi 1 ⳵
⳵t
= 2 D r2
r ⳵r Li 共
⳵cLiLi
⳵r 兲
KEC = ␧SEI 关14兴
As initial conditions, a flat lithium concentration profile correspond-
ing to the initial material utilization was set in the active particles
is = −Fkf,scECexp − 关 共
␤ sF
⌽ −
␦
RT 1 ␬SEI t
i 兲兴
and a flat solvent concentration equal to KECcEC,sln was set in the
SEI. it = iint + is
Differential algebraic equations were converted into dimension-
less forms and cast in finite-difference forms. The Crank–Nicholson
implicit method was used to evaluate the time derivatives. All the
equations were solved simultaneously using the FSOLVE subroutine ⳵cEC ⳵2cEC d␦ ⳵cEC
= DEC 2 −
embedded in the MATLAB software. A summary of the governing ⳵t ⳵r dt ⳵r
equations and boundary conditions is presented in Table I.

Results and Discussion d␦ is M SEI

=−
dt 2F ␳SEI
In this section, the model is used to study the capacity fade in
Boundary conditions
different modes of operation, namely, charge/discharge cycling,
constant-voltage storage, and OCV storage. The LiCoO2/graphite ⳵cLi
chemistry was picked up as the case study. Lithium-insertion rate 兩 兩 =0
⳵r r=0
constant and lithium-diffusion coefficient for both electrodes were
refined from charge/discharge cycling data 共0.5 C兲 from a pristine
18650 Sony cell27 and are given in Table II. These values were ⳵cLi iint
unchanged in the following aging data analysis. Note that the rate- 兩 兩 =−
⳵r r=R FDLi
constant values are generally slightly underestimated compared to a
full model27 because of the SP assumption of no liquid-phase trans-
port limitations. The equilibrium potential functions for LiCoO2 and
⳵cEC is
graphite were taken from Ref. 27 and are reproduced in the Appen- −DEC兩 兩 + ␯兩cEC兩r=R =
dix. In the following, only the parameters dealing with the SEI ⳵r r=R F
growth on anode, i.e., the side-reaction rate constant kf,s and the
solvent-diffusion coefficient DEC, were refined. The charge-transfer
兩cEC兩r=R+␦ = ␧SEIcEC,sln
coefficient, ␤S, was set arbitrarily to 0.5, as for the lithium insertion
reaction. The physical properties of the SEI product,
共CH2OCO2Li兲2,
concentration in the bulk of electrolyte for all the simulations. The
are taken from the work by Borodin et al.28 They used a quantum-
model parameters for the SEI discussed above are gathered in Table
chemistry-based many-body polarizable force field to predict the
II.
physical properties of lithium ethylene dicarbonate and lithium me-
thyl carbonate. Physical properties for 共CH2OCO2Li兲2 used in the Capacity fade during charge/discharge cycling.— There are only
simulations are presented in Table II. The initial SEI resistance was a little long-term aging data for Li-ion batteries in the literature. The
taken to be 0.001 ⍀ m2 based on the work by Zhang et al.29 This Sony 18650 cell with the LiCoO2/graphite chemistry 共rated capacity
value of resistance corresponds to a thickness of 5 nm based on the of 1.8 Ah兲 has been the most studied, especially for modeling aging.
ionic conductivity of 共CH2OCO2Li兲2, which seems reasonable for Most of the aging data for this chemistry were obtained under
an electrode after a few formation cycles. Notice that the presence of charge/discharge cycling 共0.5C with a potential range 2–4.2 V and a
a thin inner SEI layer introduced above, which is usually 1.5–2 nm conventional constant current–constant voltage charge兲 rather than
thick according to Christensen and Newman,15 is discarded in this under storage or other aging profiles. Before getting into simula-
work. The SEI porosity is another parameter that needs to be set. tions, we try to estimate the anode contribution to the total capacity
Based on preliminary adjustments, a value of 5% was chosen. This fade of Sony 18650 cells. Ramadass et al. quantified the various
parameter reflects the quality of the SEI formed and also is subject contributions to capacity fade of Sony 18650 batteries for various
to variations upon aging of the battery. Those variations are not cycle numbers and temperatures but not for each electrode.30 The
considered in the simulations and could be the subject of future total capacity loss of 539 mAh after 800 cycles at room temperature
studies. The bulk concentration of EC molecules in the electrolyte is split into 71 mAh due to rate-capability loss, 392 mAh due to
also needs to be set in the model. There were no data for EC con- secondary capacity loss 共structural degradation of active material兲,
centration for the Sony 18650 cell; however, Ploehn et al. estimated and 76 mAh due to primary capacity loss 共active lithium consumed
the EC concentration in an EC–DEC–DMC 共1:1:1兲 electrolyte with in side reactions兲. See Ref. 30 for the detailed calculation method-
1 M LiPF6 to be 4.541 mol/L.5 This value was used for the EC ology. It is reasonable to assume that the primary capacity loss

Downloaded on 2014-03-24 to IP 129.97.80.157 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 Journal of The Electrochemical Society, 156 共3兲 A145-A153 共2009兲 A149

Table II. Model parameters for the two commercial cells and the three aging modes used in this study (default values unless otherwise stated).a

Sony 18650 cell SAFT MP prototype

Anode Cathode Anode

R 共␮m兲 2 共Ref. 27兲 2 共Ref. 27兲 1es

Li 共mol/m 兲
cmax 3 30555 共Ref. 27兲 51555 共Ref. 27兲 30555 共Ref. 27兲
A 共m2兲 4.38 共Ref. 27兲 4.76 共Ref. 27兲 21.5 共Ref. 5兲
DLi 共m2 /s兲 2 ⫻ 10−14 ad 1 ⫻ 10−14 ad 2 ⫻ 10−14 ad
0,app
kint 共m/s兲 2.07 ⫻ 10−11 ad 1.04 ⫻ 10−11 ad 1.04 ⫻ 10−11 ad
␤ 0.5a 0.5a 0.5a
x0 and y 0 0.74 共cycling兲 共Ref. 27兲 0.5 共Ref. 27兲 共cycling兲 0.9es 共constant-potential storage兲
0.99 共OCV storage兲 共Ref. 6兲
Anode SEI parameters

cEC,sln 共mol/L兲 4.541 共Ref. 5兲

0
RSEI 共⍀ m2兲 0.001 共Ref. 29兲
M SEI 共kg/mol兲 0.162 共Ref. 28兲
␳SEI 共kg/m3兲 1690 共Ref. 28兲
␦0 共nm兲 5es
␬SEI 共S/m兲 5 ⫻ 10 共Ref. 28兲
−6

␧SEI 0.05a
␤s 0.5a
DEC 共m2 /s兲 6.8 ⫻ 10 −21 ad
共diffusion-limited case兲
3.7 ⫻ 10−19 ad 共adjusted from anode potential rise during OCV storage兲
kf,s 共m/s兲 1.36 ⫻ 10−12 ad 共kinetic-limited case兲
1.36 ⫻ 10−7 ad 共adjusted from anode potential rise during OCV storage兲
a
ad: adjusted; a: assumed; es: estimated.

共76 mAh兲 in the cell comes mostly from the side reactions at the Simulations were carried out for the two extreme cases of diffusion-
anode as shown in various studies,7,8,30-33 which implies that no and kinetic-limited SEI growth using the parameter values given in
lithium-based passivating SEI films are formed at the cathode. Be- Table II. The discharge profiles for various cycle numbers are plot-
cause structural changes can often be neglected for the graphite, we ted in Fig. 2, and the cell capacity and SEI resistance as a function
assume that the secondary capacity loss is attributed completely to of the cycle number is shown in Fig. 3. The SEI resistance increase
the cathode. Assuming the fraction of capacity losses due to the with time is linear for the kinetic-limited regime and quadratic for
decrease in rate capability for each electrode to be proportional to the diffusion-limited case. However, both regimes satisfy the opti-
the interfacial resistance increase for the same 共i.e., 11- and 16-fold mization objectives, and therefore it is not possible to distinguish
increase for anode and cathode as measured by Zhang et al. by whether the SEI growth is more under diffusion or kinetic control at
three-electrode impedance spectroscopy in the same aging condi- this point of analysis. This may be possible if more experimental
tions, respectively29兲, 41% of the total capacity loss due to rate- data for the anode contribution to the capacity fade are available.
capability decrease 共71 mAh兲 is attributed to the anode part
共29 mAh兲. In the anode, it is reasonable to think that this source of Capacity fade upon storage.— Physics-based models of the ag-
capability loss is connected to an SEI growth. Because we did not ing of lithium-ion batteries during storage were reported by
consider any lithium consumption at the cathode, the fraction of Ramasamy et al.6 and Ploehn et al.5 The first authors used a kinetic
expression for the solvent reduction at the carbon particle/SEI inter-
capacity loss due to a decrease in rate capability 共42 mAh兲 could be
face to describe the self-discharge of a graphite anode from a Sony
due to a surface decomposition of the layered material, leading to
18650 lithium-ion cell. Solvent molecules and lithium were assumed
thin insulating layers at the surface of the particles 共secondary ca-
to be in excess for the solvent-decomposition reaction. Their model
pacity loss兲 or to particle disconnection from the electron conductive
was able to account for the anode potential rise during OCV storage
matrix of the electrode. To summarize, given our assumptions, the
of the cell, in agreement with the experimental data. Ploehn et al.5
anode contribution to capacity losses would be 共76 + 29兲/539 developed a “solvent-diffusion” model to describe the SEI growth in
= 19.5% of the total capacity loss after 800 cycles, which itself cor- SAFT batteries during constant-voltage storage. In contrast with the
responds to 31% of the initial capacity. The fraction of capacity loss work by Ramasamy et al.,6 they considered the SEI growth to be
due to the anode is thus estimated to 0.195 ⫻ 0.31 = 6%. A sum-
mary of this capacity fade analysis is presented in Table III.
In the published capacity-fade models for charge/discharge cy-
cling so far, the SEI growth in the anode was considered to occur Table III. Contributions to capacity fade of a Sony 18650 cell
through a side reaction following a Tafel kinetic expression having after 800 cycles at room temperature. Data for complete cell are
an equilibrium potential equal to 0.4 V. Transport of solvent mol- taken from Ref. 30.
ecules across the SEI was simply neglected. In other words, the SEI
growth was assumed to be only kinetic-limited.4,23-26 In the model Secondary Primary
we developed, however, both kinetic and diffusion limitations are capacity loss Capacity loss due capacity loss
共loss of active to rate-capability 共active lithium loss兲
considered. material兲 共mAh兲 decrease 共mAh兲 共mAh兲
The objective of the parameter refinement is to account for ca.
6% capacity fade together with an increase of ca. 11-fold in the SEI Complete 392 71 76
resistance after 800 cycles at room temperature. This refinement cell
process is not straightforward and is time-consuming; therefore, the Anode 0 29 76
parameters reported hereafter are not necessarily the best ones. Cathode 392 42 0

Downloaded on 2014-03-24 to IP 129.97.80.157 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 A150 Journal of The Electrochemical Society, 156 共3兲 A145-A153 共2009兲

Figure 2. Simulations of the discharge profiles during cycling of a Sony

18650 cell at room temperature and 0.5C rate for kinetic- and diffusion-
limited regimes.

limited by the solvent diffusion, and no kinetic expression for the

solvent decomposition was considered. Although the two above-
mentioned models successfully fitted the experimental data, in both
cases some of the assumptions made were not proved to be realistic,
such as the choice of a pure kinetic limitation or a pure diffusion
limitation of the SEI growth. In order to get more insight in the
capacity fade during storage conditions, some of the data reported in
those two references were reanalyzed using our model.
Constant-voltage storage.— The parameters refined for aging
during cycling and discussed in the previous section are used to Figure 4. Experimental and simulated capacity fade 共a兲 and simulated SEI
simulate aging under storage at constant potential, with the aim of resistance 共b兲 of an MP prototype cell under constant-voltage storage 共3.9 V兲
sensing how different the capacity fade between the two different at 30°C for kinetic- and diffusion-limited regimes 共data from Ref. 5兲.
aging modes can be and whether a universal set of side-reaction
parameters can describe the capacity fade irrespectively of the op-
eration mode.
The capacity loss of a 5 Ah prismatic “MP” 共LiCoO2/graphite兲 aging when the cell voltage is set, it is assumed here that this does
prototype cell was reported by Broussely et al.32 for constant- not affect the cathode potential significantly, and therefore simula-
tions were run for the graphite half-cell only 共i.e., the graphite elec-
voltage storage conditions 共3.9 V兲 at 30°C. The cell specifications
trode potential was set兲. Using a trial-and-error method as well as
for this battery were not reported by Broussely et al.; instead, Ploehn
the equilibrium potential functions introduced in Appendix A, the
et al.5 estimated some parameters like the carbon surface area, the
constant cell voltage of 3.9 V was estimated to correspond to an
anode current collector surface area, etc. Similar to the paper of
initial SOC of 90%, which leads to an open-circuit potential of
Ploehn et al.,5 it is assumed that the capacity fade is entirely due to
0.0664 V vs Li electrode for the graphite electrode. The percentage
an SEI growth on the anode. For the sake of simplicity, although the
of capacity loss was calculated by considering the initial anode ca-
state of charge 共SOC兲 of the cathode is expected to vary because of
pacity to be equal to the theoretical one. The estimated and assumed
cell parameters used in the simulation of the constant-potential stor-
age of the graphite are summarized in Table II. The values for side-
reaction parameters are the same as those refined from the cycling
mode. The simulation results for 450 days of storage are overlaid to
the experimental data in Fig. 4a. The diffusion-limited case is in
much better agreement with the experimental data as compared with
the kinetic-limited case. This might be proof that the SEI growth is
more under diffusion control. Furthermore, the suitability of the
model parameters 共diffusion-limited case兲 from cycling analysis to
the constant-potential storage situation tends to prove that aging
proceeds through the same mechanism regardless of the mode.
The SEI resistance increase during storage is displayed in Fig.
4b. A 16-fold increase in SEI resistance during the 450 days of
storage is observed for the diffusion-limited case. A deeper under-
standing of the capacity-fade process can be obtained by looking at
the current densities for lithium insertion/deinsertion and SEI
growth 共Fig. 5兲. The magnitudes are small and, for more clarity, the
log of absolute values of current densities are plotted. Note that the
current density for the side reaction is negative, while that for the
Figure 3. Simulations of capacity fade and SEI resistance increase during main reaction is positive 共deinsertion兲. The main reaction current,
cycling of a Sony 18650 cell at room temperature and C/2 for kinetic- and although several orders of magnitudes lower than that for the side
diffusion-limited regimes. reaction 共in absolute value兲, is not zero because of the ohmic drop

Downloaded on 2014-03-24 to IP 129.97.80.157 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 Journal of The Electrochemical Society, 156 共3兲 A145-A153 共2009兲 A151

Figure 5. Simulation of side-reaction and intercalation current densities at

the anode of an MP prototype cell under constant-voltage storage 共3.9 V兲 at Figure 6. Anode potential rise and capacity loss simulations of a Sony
30°C 共diffusion-limited case兲. 18650 cell during OCV storage at room temperature with an initial SOC
= 99% 共data taken from Ref. 6兲.

within the SEI, which is mainly caused by the side-reaction current

共RSEI兩it兩 ⬇ RSEI兩is兩兲 and which yields a positive overpotential for the SEI quality and capacity fade.— The SEI porosity is believed to
main reaction. have quite a strong influence on Li-ion batteries capacity fade for
both aging cycling and storage. It is a key factor in the quality of the
OCV storage.— OCV storage of Li-ion batteries leads to a cell- SEI and consequently controls the rate of capacity fade of a battery.
voltage decrease together with a capacity loss generally composed It is well-known that the SEI film has a limited elasticity and is
of a reversible and an irreversible part. The causes for the reversible prone to crack. According to Aurbach et al.,7 Li-ion insertion into
self-discharge are not generally clear. For instance, Yazami and Rey- graphite particles during charge causes an increase in the lattice
nier suggested the formation of a metastable electron-ion-solvent volume that may partially break the SEI film and therefore alters its
complex that forms and adsorbs on the graphite during OCV passivity by exposing more of the underlying carbon to the electro-
storage.34 Upon cell discharge, the complex is stable enough and lyte.
accounts for the capacity loss. Upon cell recharge 共graphite lithia- A fracture model for the SEI evolution that could be used in a
tion兲, the complex dissociates, and the capacity is thus recovered. physics-based capacity-fade model has not been proposed so far.
According to Spotnitz,35 acidic impurities could dissolve some of Nevertheless, the porosity of the SEI was involved in the model
the anode SEI species that could be further reoxidized at the positive developed in this study, and even though porosity changes during
electrode, thus accounting for a reversible capacity loss as well. The aging were not accounted for, it is worth looking at the effects of
reversible part of the capacity fade during OCV storage is not con- SEI-film porosity on the SEI growth rate. Using the parameters re-
sidered in the present model; however, using the experimental data fined for the capacity loss during cycling and constant-potential stor-
provided by Ramasamy et al.,6 a parameter-sensitivity analysis of age 共diffusion-limited case, Table II兲, a sensitivity analysis of the
capacity loss during storage is carried out. Ramasamy et al. provided porosity effect on the capacity fade was performed by simulating
the experimental data for the potential rise of a graphite electrode of case studies for which the SEI is either of high 共1% porosity兲 or low
a Sony 18650 cell upon OCV storage during a year at 25°C.6 As for 共10% porosity兲 quality. Quality-signature curves that relate the stor-
the other modes investigated, the capacity loss was assumed to be age time, the SOC, and the graphite potential to each other are
solely caused by the solvent decomposition leading to the growth of presented in Fig. 7. The anode with a high-quality SEI 共1% porosity兲
a SEI and therefore was totally irreversible. Because we have shown loses only 9% of initial capacity 共SOC ⬇ 90%兲 after 10 years of
that the model parameters for a diffusion-limited regime of the side storage, while the low-quality SEI suffers from much more lithium
reaction are valid for aging upon cycling and upon storage at con- exhaustion after the same amount of time 共SOC ⬇ 68%兲. After
stant potential, the same set was also used for the simulation of
OCV storage data. Besides, a new set of model parameters, given in
Table II, was obtained from a refinement of the OCV storage data.
The results of the two types of simulation are displayed in Fig. 6. No
experimental data are reported in Ref. 6 for the capacity loss, which
is why only the simulation is presented. Although the new set of
refined parameters 共Table II兲 is able to predict the potential-rise data,
the corresponding simulated capacity fade amounts to 60%, which
seems high. For comparison, Spotnitz reported an irreversible capac-
ity loss of 17% for the same type of cell after 1 year of storage at
20°C.35 The diffusion coefficient 共3.7 ⫻ 10−19 m2 /s兲 is nearly 2 or-
ders of magnitude larger than that obtained for the cycling and
constant-potential storage situations. When the parameters obtained
from cycling and constant-potential aging are used for the OCV
storage, the potential rise is largely underestimated, while the capac-
ity loss is in a more reasonable range 共8%兲. From these trends, it is
proposed that the cycling and constant-potential parameters are rep-
resentative of the irreversible capacity loss during OCV storage.
Indeed, if a mechanism such as the one proposed by Spotnitz is
assumed,35 reversible capacity fade cannot be observed during cy- Figure 7. Quality-signature plots for an SEI with 1 or 10% porosity, simu-
cling and constant-potential storage, and therefore only an irrevers- lated from graphite OCV storage at room temperature with an initial SOC of
ible capacity fade is expected. 99% using the aging parameters from the diffusion-limited case.

Downloaded on 2014-03-24 to IP 129.97.80.157 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 A152 Journal of The Electrochemical Society, 156 共3兲 A145-A153 共2009兲

respectively, while Fig. 8b and 9b exhibit the order of t in the SEI-

growth equation 关␦共t兲 − ␦共0兲 = bt␣兴 as a function of time. As ex-
pected, no difference is observed between the two modes of opera-
tion for the diffusion-limited case; the SEI grows following a
parabolic growth law, with the order ␣ that decreases quickly at
short times, probably because a solvent concentration needs to build
up before solvent-diffusion limitations occur, and which eventually
levels off toward a value close to 0.5. However, the kinetic-limited
case is greatly affected by the operation mode. SEI growth is linear
with time during potentiostatic aging, whereas it bends over with
time during OCV mode, which stems from the anode potential
change with time that leads to a decrease of the overpotential for the
side reaction. The mixed regime lies between the two extreme cases,
although it looks closer to the diffusion-limited one. Note that a
similar parabolic growth behavior was reported by Peled et al.,
which was due to electron transport limitation in SEIs forming on
lithium metal electrodes.14 Later on, a more detailed study of the
Figure 8. Dimensionless SEI thickness increase 共a兲 and evolution of the growth law was reported by Christensen and Newman for SEIs hav-
growth law exponent 共b兲 during the constant-potential storage of graphite ing a nonzero electron mobility as well.15
electrode 共0.064 V vs Li electrode兲 at room temperature for three different
regimes of growth 关diffusion-limited 共DEC = 6.8 ⫻ 10−21 m2 /s, kf,s Conclusion
= 10−7 m/s兲, kinetic-limited 共DEC = 1.3 ⫻ 10−15 m2 /s, kf,s = 1.36
⫻ 10−12 m/s兲, and mixed regime 共DEC = 13.6 ⫻ 10−21 m2 /s, kf,s = 3.4 A multimodal physics-based aging model that can be used both
⫻ 10 m/s兲兴.
−12 in cycling and storage condition was developed. The model is based
on the continuous, small-scale growth of an SEI layer on the surface
of the anode active particles. The solvent decomposition kinetics
10 years, the potential reaches 0.07 V vs Li electrode in the case of and solvent diffusion through the SEI layer were both considered in
the high-quality SEI and 0.1 V vs Li electrode for the low-quality the model. The LiCoO2/graphite cell was chosen as a case study, and
one. using the various experimental data from the literature, the aging of
the cell upon both cycling and storage was simulated.
SEI growth and evolution of the growth law exponent.— In or- In cycling mode, the contribution of the anode SEI growth to the
der to look at the SEI growth law in more detail, a dimensionless total capacity fade in the cell was first estimated together with the
parameter ␶ = ␦kf,s /DEC is defined. Large values of ␶ are represen- interfacial resistance increase. From simulation, it was shown that
tative of diffusion-limited cases, while low and intermediate values both diffusion- and kinetic-limited regimes were able to meet the
of ␶ are for kinetic-limited and mixed regimes, respectively. Note estimated values. More experimental data are required to assess
that although changes in the growth regime may occur at long aging whether this is more kinetic- or diffusion-limited or to refine simul-
times due an increase of ␦, here we focus on a comparative study of taneously the two side-reaction parameters 共i.e., mixed regime兲. The
the relative values of kf,s and DEC, and therefore ␶0 = ␦0kf,s /DEC is same set of side-reaction parameters was used to simulate the ca-
introduced instead of ␶. Aging simulations for 1500 days under ei- pacity fade during constant-potential storage, and contrary to the
ther constant anode potential 共0.064 V vs Li electrode兲 or OCV cycling mode, the kinetic-limited case failed to describe the capacity
共SOC0 = 0.99兲 were run for the three SEI growth regimes 共␶0 fade. This proves that SEI growth is more likely to be a diffusion-
= 7.3 ⫻ 104, ␶0 = 5 ⫻ 10−6, and ␶0 = 1.3 ⫻ 100兲. Dimensionless limited process rather than a kinetic-limited one. For OCV storage,
SEI thicknesses 关␦共t兲 − ␦共0兲/␦共1500兲 − ␦共0兲兴 vs storage time are it was shown that the set of model parameters refined for the other
presented in Fig. 8a and 9a for constant-potential and OCV modes, modes could not satisfy the experimental data, pointing out that part
of the capacity loss is reversible.
The SEI porosity was shown to be a critical parameter that re-
flects the quality of the SEI in terms of passivation properties. Using
the simulation results, it was shown that an SEI layer with 10%
porosity leads to a capacity fade more than 3 times higher than for
an SEI with 1% porosity.
Finally, the growth law of SEI was studied for the two storage
modes in a more systematic way by looking at different extreme
cases. While the SEI grows according to a parabolic law regardless
of the storage mode when the growth is under solvent-diffusion
limitations, the kinetic-limited growth of the SEI leads to different
behaviors depending on the operation mode.
Besides the work on anodes, the necessity of modeling structural
degradations in the cathode is an important step in the development
of battery-life prediction models. Addition of thermal concerns to
the present model and considering composition and porosity
changes of the SEI as well as SEI dissolution could be the subject of
a future work.

Figure 9. Dimensionless SEI thickness increase 共a兲 and evolution of the Acknowledgments
growth law exponent 共b兲 during the OCV storage of graphite electrode M.S. acknowledges the financial support of the E.U. through the
共SOC0 = 0.99兲 at room temperature for three different regimes of growth MESC Master Course granted as an Erasmus Mundus programme.
关diffusion-limited 共DEC = 6.8 ⫻ 10−21 m2 /s, kf,s = 10−7 m/s兲, kinetic-limited
共DEC = 1.3 ⫻ 10−15 m2 /s, kf,s = 1.36 ⫻ 10−12 m/s兲, and mixed regime Université de Picardie Jules Verne assisted in meeting the publication
共DEC = 13.6 ⫻ 10−21 m2 /s, kf,s = 3.4 ⫻ 10−12 m/s兲兴. costs of this article.

Downloaded on 2014-03-24 to IP 129.97.80.157 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
 Journal of The Electrochemical Society, 156 共3兲 A145-A153 共2009兲 A153

Appendix Subscript

For the graphite electrode27 s side reaction

⫹ variables referred to cathode
0.0172 0.0019
U− = 0.7222 + 0.1387x + 0.029x 0.5
− + + 0.2808 exp共0.9 − 15x兲 ⫺ variables referred to anode
x x1.5
Superscript
− 0.7984 exp共0.4465x − 0.4108兲
For the LiCoO2 electrode27 0 initial value
2 4 6 8 10
U+ =
− 4.656 + 88.669y − 401.119y + 342.909y − 462.471y + 433.434y References
− 1 + 18.933y 2 − 79.532y 4 + 37.311y 6 − 73.083y 8 + 95.96y 10 1. http://www.uscar.org/guest/publications.php?title_filter⫽EV, last accessed Nov 17,
2008.
2. D. U. Sauer and H. Wenzl, J. Power Sources, 176, 534 共2008兲.
List of Symbols 3. R. Darling and J. Newman, J. Electrochem. Soc., 145, 990 共1998兲.
4. P. Ramadass, B. S. Haran, P. M. Gomadam, R. White, and B. N. Popov, J. Elec-
A total interfacial surface area of the electrode, m2 trochem. Soc., 151, A196 共2004兲.
b constant in the SEI growth law, ␦共t兲 − ␦共0兲 = bt␣ 5. H. J. Ploehn, P. Ramadass, and R. White, J. Electrochem. Soc., 151, A456 共2004兲.
cEC solvent concentration in the SEI film, mol/m3 6. R. P. Ramasamy, J. W. Lee, and B. N. Popov, J. Power Sources, 166, 266 共2007兲.
7. D. Aurbach, J. Power Sources, 89, 206 共2000兲.
cEC,sln solvent concentration in the bulk of the electrolyte, mol/m3
8. D. Aurbach, B. Markovsky, A. Rodkin, M. Cojocaru, E. Levi, and H. J. Kim,
cLi concentration of Li inside the active material, mol/m3 Electrochim. Acta, 47, 1899 共2002兲.
max
cLi maximum concentration of Li inside the active material, mol/m3 9. D. Aurbach, E. Zinigrad, Y. Cohen, and H. Teller, Solid State Ionics, 148, 405
DEC solvent diffusion coefficient inside the SEI layer, m2 /s 共2002兲.
DLi lithium diffusion coefficient inside active particles, m2 /s 10. A. N. Dey and B. S. Sullivan, J. Electrochem. Soc., 117, 222 共1970兲.
F Faraday’s constant, 96,487 C/mol 11. D. Aurbach, Y. Ein-Eli, O. Chusid, Y. Carmeli, M. Babai, and H. Yamin, J. Elec-
iint intercalation current density referred to interfacial surface area of trochem. Soc., 141, 603 共1994兲.
anode, A/m2 12. C. R. Yang, Y. Y. Wang, and C. C. Wan, J. Power Sources, 72, 66 共1998兲.
is side-reaction current density referred to interfacial surface area of 13. Y. Wang, S. Nakamura, M. Ue, and P. B. Balbuena, J. Am. Chem. Soc., 123, 11708
anode, A/m2 共2001兲.
it total current density referred to interfacial surface area of anode, 14. E. Peled, J. Electrochem. Soc., 126, 2047 共1979兲.
A/m2 15. J. Christensen and J. Newman, J. Electrochem. Soc., 151, A1977 共2004兲.
16. A. Z. Weber and J. Newman, J. Electrochem. Soc., 150, A1008 共2003兲.
KEC partition coefficient of solvent between the electrolyte and the
17. A. Z. Weber and J. Newman, J. Electrochem. Soc., 151, A311 共2004兲.
SEI
18. S. Atlung, K. West, and T. Jacobsen, J. Electrochem. Soc., 126, 1311 共1979兲.
kf,s rate constant of side reaction, m/s
0,app 19. B. S. Haran, B. N. Popov, and R. E. White, J. Power Sources, 75, 56 共1998兲.
kint apparent rate constant for lithium intercalation reaction, m/s 20. S. Santhanagopalan, Q. Guo, P. Ramadass, and R. White, J. Power Sources, 156,
M SEI molecular weight of SEI, kg/mol 620 共2006兲.
NEC solvent flux density, mol/共m2 s兲 21. S. Mori, H. Asahina, H. Suzuki, A. Yonei, and K. Yokoto, J. Power Sources, 68,
r radial distance in active particle and across the SEI, m 59 共1997兲.
R active particles radius 共m兲 and ideal gas constant, 22. A. Naji, J. Ghanbaja, B. Humbert, P. Willmam, and D. Billaud, J. Power Sources,
8.314 J/共mol K兲 63, 33 共1996兲.
RSEI SEI film resistance referred to interfacial surface area of anode, 23. P. Arora, M. Doyle, and R. White, J. Electrochem. Soc., 146, 3543 共1999兲.
⍀ m2 24. G. Sikha, B. N. Popov, and R. White, J. Electrochem. Soc., 151, A1104 共2004兲.
t time, s 25. S. Santhanagopalan, Q. Guo, P. Ramadass, and R. White, J. Power Sources, 156,
T temperature, K 620 共2006兲.
U equilibrium potential vs Li electrode, V 26. G. Ning and B. N. Popov, J. Electrochem. Soc., 151, A1584 共2004兲.
␯ SEI film velocity, m/s 27. P. Ramadass, B. Haran, R. White, and B. N. Popov, J. Power Sources, 123, 230
x lithium stoichiometry in anode 共2003兲.
y lithium stoichiometry in cathode 28. O. Borodin, G. D. Smith, and P. Fan, J. Phys. Chem. B, 110, 22773 共2006兲.
29. D. Zhang, B. S. Haran, A. Durairajan, R. White, Y. Podrazhansky, and B. N.
␣ power law exponent for SEI growth
Popov, J. Power Sources, 91, 122 共2000兲.
␤ cathodic charge-transfer coefficient for the intercalation reaction
30. P. Ramadass, B. Haran, R. White, and B. N. Popov, J. Power Sources, 112, 614
␤s cathodic charge-transfer coefficient for the side reaction
共2002兲.
␦ SEI film thickness, m 31. M. Broussely, P. Biensan, F. Bonhomme, P. Blanchard, S. Herreyre, K. Nechev,
␧SEI SEI film porosity and R. J. Staniewicz, J. Power Sources, 146, 90 共2005兲.
⌽1 potential of the electron-conductive phase, V 32. M. Broussely, S. Herreyre, P. Biensan, P. Kasztejna, K. Nechev, and R. J. St-
⌽2 potential of the ion-conductive phase, V aniewicz, J. Power Sources, 97–98, 13 共2001兲.
␬SEI SEI ionic conductivity, S/m 33. G. Sarre, P. Blanchard, and M. Broussely, J. Power Sources, 127, 65 共2004兲.
␳SEI SEI density, kg/m3 34. R. Yazami and Y. F. Reynier, Electrochim. Acta, 47, 1217 共2002兲.
␶ dimensionless parameter determining SEI growth regime 35. R. Spotnitz, J. Power Sources, 113, 72 共2003兲.

Downloaded on 2014-03-24 to IP 129.97.80.157 address. Redistribution subject to ECS terms of use (see ecsdl.org/site/terms_use) unless CC License in place (see abstract).
View publication stats