Journal of The Electrochemical

Society

OPEN ACCESS You may also like

- Soil physico-chemical properties are
Parameterization of a Physico-Chemical Model of critical for predicting carbon storage and
nutrient availability across Australia
a Lithium-Ion Battery: I. Determination of Jinquan Li, Ming Nie, Jeff R Powell et al.

- Influence of Model Detail Level on the

Parameters Prognosis Quality of the Perceived Sound
of an Electrified Passenger Car Powertrain
P Drichel, M Jaeger, M Möller-Giebeler et
To cite this article: Madeleine Ecker et al 2015 J. Electrochem. Soc. 162 A1836 al.

- Parameterization of a Physico-Chemical
Model for Lithium-Ion Batteries
Madeleine Ecker, Stefan Käbitz, Dung
Tran et al.
View the article online for updates and enhancements.

This content was downloaded from IP address 201.71.169.193 on 26/10/2023 at 20:39

A1836 Journal of The Electrochemical Society, 162 (9) A1836-A1848 (2015)

Parameterization of a Physico-Chemical Model of a Lithium-Ion

Battery
I. Determination of Parameters
Madeleine Ecker,a,b,z Thi Kim Dung Tran,a,b Philipp Dechent,a,b Stefan Käbitz,a,b
Alexander Warnecke,a,b and Dirk Uwe Sauera,b,c
a Chair for Electrochemical Energy Conversion and Storage Systems, Institute for Power Electronics and Electrical
Drives (ISEA), RWTH Aachen University, 52066 Aachen, Germany
b Juelich Aachen Research Alliance, JARA-Energy, Germany
c Institute for Power Generation and Storage Systems (PGS) @E.ON ERC, RWTH Aachen University,
52056 Aachen, Germany

The parameterization of a physico-chemical model constitutes a critical part in model development. Conclusions about the internal
state of a battery can only be drawn if a correct set of material parameters is provided for the material combination under consideration.
In this work, parameters to fully parameterize a physico-chemical model for a 7.5 Ah cell produced by Kokam are determined and
are compared with existing literature values. The paper presents parameter values and procedures to determine the parameters. Cells
were opened under argon atmosphere and the geometrical data were measured. Hg-porosimetry was conducted to determine porosity,
particle radius as well as tortuosity of the electrodes and the separator. Conductivity and diffusion constants of the electrolyte as
well as the electronic conductivity of the active material were measured detecting the voltage response to a dc current. Finally,
electrochemical measurements were conducted on laboratory-made coin cells in order to determine open circuit voltage curves,
diffusion coefficients and charge transfer kinetics of the electrodes as well as their balancing. In part II of this paper, a physico-
chemical model is introduced and a validation of the measured parameter set is given by comparing model results with experiments.
© The Author(s) 2015. Published by ECS. This is an open access article distributed under the terms of the Creative Commons
Attribution Non-Commercial No Derivatives 4.0 License (CC BY-NC-ND, http://creativecommons.org/licenses/by-nc-nd/4.0/),
which permits non-commercial reuse, distribution, and reproduction in any medium, provided the original work is not changed in any
way and is properly cited. For permission for commercial reuse, please email: oa@electrochem.org. [DOI: 10.1149/2.0551509jes]
All rights reserved.

Manuscript submitted January 13, 2015; revised manuscript received May 14, 2015. Published June 29, 2015.

Physico-chemical models are based on fundamental equations de-
scribing migration and diffusion processes as well as intercalation
kinetics. They can be used to gain understanding of internal processes
of batteries and to optimize material development. Several papers have
been published developing physico-chemical simulation models that
are based on the work of Newman and Tiedemann 1975,1 amongst
others.2–6
However, one crucial part of physico-chemical models is the model
parameterization. Especially if conclusions about the internal state of
the battery are drawn, it is of utmost importance to choose the right
parameters for the materials under consideration. To the knowledge of
the authors no work exists where a simulation model was completely
parameterized using the geometric data and the parameters of the
materials of one commercial cell in total. In most works dealing
with physico-chemical models, values from supplementary literature
sources were used; parameters were fitted or even guessed, e.g.2–6
There are publications focusing on the determination of certain
parameters for certain material combinations. Park et al.7 for example
gathered diffusion constants as well as conductivities investigated for
different materials used in lithium-ion batteries in literature. Several
authors also determined exchange currents for graphite materials.8–11
The problem here is that these parameters are usually not measured
with the purpose to parameterize a battery model. This leads to mea-
surement setups and finally to parameters that are not applicable in
battery models. The exchange current, for example, is usually not
scaled with the active surface area, which makes a transfer to a ma-
terial with a different surface structure impossible. Another example
is the determination of the electronic conductivity of active materials.
The conductivity in literature is usually not measured for a whole
electrode setup, including the filler, binder and porous structure. This
makes it difficult to apply these values in models.
Furthermore, there are parameters that have (to the knowledge
of the authors) not yet been investigated in literature for materials
used in lithium-ion batteries. No data was found for example for the
charge transfer coefficients of the materials, the exchange current, the
z
E-mail: batteries@isea.rwth-aachen.de
activation energy of diffusion or the electronic conductivity of the
cathode material used in this work.
Another problem with material parameters from literature is that
the materials used in different works are not identical. The diffu-
sion coefficient of graphite e.g. depends on the structure of the used
graphite or in the case of the cathode on the exact cathode compo-
sition. Additionally, the materials cannot be considered separately in
battery systems as they act together as a system. The exchange cur-
rent, for example, does not only depend on the active material used
in the cell but also on the electrolyte system8 which determines the
formation of the solid electrolyte interphase (SEI). Therefore, data
measured for one cell system cannot be transferred to another, even if
it was a similar one.
In this work, a commercially available cell is analyzed for which
the authors have no construction plans or any detailed information
on the used materials and components. The authors aim to parame-
terize a physico-chemical model from the roots. The parameters are
determined by results of experiments conducted for the material of
the cell under investigation. For this purpose, the cell was opened un-
der argon atmosphere; material samples were collected and analyzed
using different methods. This part of the paper deals with the realisa-
tion of parameter determination. Methods are discussed for effective
model parameterization. In the second part of the paper,12 the simula-
tion model is introduced and a validation of the model including the
measured parameters is given. It is discussed to which extent kinetic
parameters as well as the balancing obtained in laboratory cells can
be transferred to the original cell from which the electrodes for the
laboratory cells were extracted.
Experimental
Considered cell system.— For model parameterization, a commer-
cial high energy pouch lithium-ion polymer battery manufactured by
Kokam, labelled SLPB 75106100 has been used. The cell has a nom-
inal capacity of 7.5 Ah. A voltage between 2.7–4.2 V is allowed by
the manufacturer with a nominal voltage of 3.7 V. Charging with
1 C in a temperature range of 0–40◦ C, as well as discharging with
 Journal of The Electrochemical Society, 162 (9) A1836-A1848 (2015)
2 C between −20 – 60◦ C is approved for the cell by the manufacturer.
According to the manufacturer, the materials comprise carbon on the
negative electrode, Li(NiMnCo)O2 on the positive electrode and an
EC/EMC mixture with LiPF6 as electrolyte. However, Induced Cou-
pled Plasma (ICP) spectroscopy (using a Varian 725- ICP-OES) on the
electrode material of the 7.5 Ah cell reveals that the Li(NiMnCo)O2
material does not comprise manganese. According to the measure-
ment results, the positive electrode consists of 60% cobalt and 40%
nickel. Therefore, in the following the cathode material is labelled as
Li(Ni0.4 Co0.6 )O2 .
Sample extraction and preparation.— To gather samples for pa-
rameter identification, Kokam cells were discharged to 0% state of
charge (SOC) with a 1 C discharge to 2.7 V and disassembled under
argon atmosphere to avoid undesired film formation on the electrodes
due to air contact. The double-sided electrodes as well as the separator
were separated from each other. No further preparation was conducted
for Hg-porosimetry and conductivity measurements of the electrodes.
For conductivity measurement of the electrolyte as well as for
the electrochemical measurements an electrolyte produced by BASF
(LP50) containing 1M LiPF6 in ethylene carbonate/ethyl methyl car-
bonate EC:EMC (1:1) (wt) was used. LP50 is assumed to resemble the
applied electrolyte in the Kokam cell most, as the manufacturer makes
declarations about the content of EC/EMC with LiPF6 as conducting
salt. However, no declarations were made about the quantity of the
ingredients or additives used by the manufacturer. The electrolyte con-
ductivity is not expected to change significantly with proportion of the
ingredients. For the electrochemical measurements it is expected that
only the determination of the exchange current density is crucially
influenced by the choice of electrolyte through film formation. If sur-
face films get destroyed during the assembling process, they have to
be restored during cycling within the laboratory cell and a change
in morphology gets possible. This effect is discussed in the second
part.12
For the characterization of the single components via electrochem-
ical measurements, cell materials extracted from the Kokam cell were
assembled in coin cells under argon atmosphere. For this purpose,
one side of the coating was removed from the electrodes using N-
Methyl-2-pyrrolidone (NMP). Different types of laboratory cells were
produced:
r Coin full cells: consisting of anode (16 mm), cathode (16 mm)
and separator (18 mm) coming from the Kokam cell together with
100 μl of LP50 from BASF.
r Coin half cells: consisting either of anode or cathode (16 mm)
and separator (18 mm) coming from the Kokam cell together with a
metallic lithium foil (16 mm) as counter electrode and 100 μl of LP50
from BASF. No reference electrode has been used in this setup.
r Li/Li coin cells: consisting of two metallic lithium foils
(16 mm), separator coming from the Kokam cell together with 100 μl
of LP50 from BASF.
After assembling, all cells were subject to additional initialization
cycles in order to restore the SEI. The coin cells showed a good
reproducibility and were comparable to the 7.5 Ah cell when scaling
the capacity and resistance according to the geometric area of the
cells. All electrochemical measurements on coin cells in this work
were performed with a BaSyTec battery test device at 23◦ C. For
deviating temperatures a climate chamber by Binder (MK53, −40◦ C
to +180◦ C) was used for temperature control.
Measurement techniques.– Hg-porosimetry.— For measurements
of porosity a porosimeter of the series Pascal 140 combined with a
Pascal 440 by Thermo Scientific was used.13 The active materials
were measured as a whole sample, including current collector and the
two-sided active material films to avoid possible structural destruction
during removal of the coating. As only the parameters of the active
material are of interest, the current collector adds an error to the
measurement. Therefore, porosity, particle radius as well as tortuosity
A1837
factor were corrected according to the real sample mass and volume.
For measurement purposes, the samples were exposed to air for a short
time period.
Conductivity of electrolyte.— For the conductivity measurement
of the electrolyte, a conductivity measurement device ‘SevenMulti’
by Mettler Toledo was used, which determines the conductivity by
applying an alternating current. The measurement was conducted un-
der nitrogen atmosphere. The ambient temperature was adjusted in a
temperature chamber from Binder (MK53, −40◦ C to +180◦ C).
Electronic conductivity of electrodes.— To measure the electronic
conductivity of the electrode materials, a Keithley Model 2601A Sys-
tem SourceMeter was used. The instrument applies a DC current to
the probe and measures the voltage response. The instrument is able
to apply currents up to 3 A and measure potential differences down to
1 μV.14
A four-point method was used for contacting the samples in order
to avoid that the result is influenced by the resistance of the setup. The
measurement design assures that the contact resistance between the
setup and the sample is not considered in the measurement. Figure 1
shows the design of the measurement setup. A sample (in this case the
two-sided coated electrode) with a diameter of 20 mm is fixed between
two copper cylinders of the same diameter. Different currents from
−0.1 A to 0.1 A were applied to the sample. Different currents lead
to similar results, indicating the reliability of the measurement. The
measurement was conducted under argon atmosphere.
OCV measurement.— To measure the OCV, the coin cells were
discharged to a minimum voltage using a CC-CV discharge, followed
by a stepwise charge process. In each step, the OCV was identified
after a break of 5 h. The measurement was performed at 23◦ C.
Electrochemical Impedance Spectroscopy (EIS).— To identify the
exchange current density and the solid-state diffusion coefficients of
the materials, EIS measurements were conducted on coin half cells
in potentiostatic mode with maximum voltage amplitude of 5 mV.
A frequency range between 10 kHz-500 μHz was chosen. All mea-
surements were conducted without a superposition of DC current.
Impedance spectra were measured at different temperatures and SOC
for graphite and Li(Ni0.4 Co0.6 )O2 coin half cells.
In order to identify the charge transfer coefficient of the charge
transfer reaction, measurements were performed with DC superposi-
tion in galvanostatic mode with maximum AC amplitude of 100 μA
using different DC currents. These spectra were measured in a fre-
quency range between 10 kHz-100 mHz at 23◦ C.
The Zennium impedance spectroscope by Zahner was used for all
EIS measurements.
Figure 1. Setup of the conductivity measurement. Two copper cylinders are
fixed in a holder with a spring and are used to apply a current to the sample.
In the middle of the cylinder is a hole where a thin cylindrical stick is placed
electrically insulated from the main cylinder. Furthermore, the stick touches
the surface of the sample and is used for voltage measurement.
A1838 Journal of The Electrochemical Society, 162 (9) A1836-A1848 (2015)

Table I. Model parameter.

Cathode Anode Separator Electrolyte

Geometrical Data
Material Properties
stoichiometric Li+ concentration
Transport Properties
Electrode Kinetics
length & width length & width length & width
layer thickness layer thickness layer thickness
porosity porosity porosity
inactive part inactive part
particle radius particle radius
electrical conductivity electrical conductivity electrical conductivity
stoichiometric Li+ concentration equilibrium concentration
Utilization portion of SEI
OCV curve OCV curve
diffusion coefficient diffusion coefficient diffusion coefficient
transport number Li+
tortuosity tortuosity tortuosity
exchange current density exchange current density
charge transfer coefficient charge transfer coefficient

Galvanostatic Intermittent Titration Technique (GITT).— For de-
termination of the diffusion coefficient, GITT was used.15,16 For the
graphite material a current pulse of 0.05 C for 150 s was applied to coin
half cells at different temperatures and SOC. For the Li(Ni0.4 Co0.6 )O2
material, a current pulse of 0.1 C for 150 s was used. Breaks of 4 h
after adjustment of a certain SOC as well as 1 h after the pulse were
performed.
Model-Parameterization
The physico-chemical model used in this work is introduced in the
second part of this paper.12 The model relies on the governing physical
and chemical processes in a lithium-ion battery comprising diffusion
and migration processes in the electrolyte and the solid material as
well as the charge transfer process and has a 1D spatial resolution
within the electrolyte as well as a 1D resolution within the spherical
active material particles. It is able to simulate the externally accessible
voltage (current) response of a battery to a given current (voltage) as
well as the time evolution of internal parameters of the battery like
local potentials or concentration distribution of lithium ions in the
electrolyte or the active material. It is based on the porous electrode
theory originally introduced by Newman and Tiedemann 1975.1,17
The parameters needed for model parameterization are listed in
Table I. In total more than 35 material properties or geometric data
must be determined, most of them with dependencies on temperature
or state of charge. In the following the different experimental methods
to determine the material parameters are discussed.
Geometrical data.— The cell consists of 24 layers of double-sided
coated anode sheets, 23 layers of double-sided coated Li(Ni0.4 Co0.6 )O2
sheets as well as 2 single-sided coated Li(Ni0.4 Co0.6 )O2 sheets. The
layer thicknesses of the electrode foils and the separator were mea-
sured with a micrometer screw. Also the dimensions of the electrode
foils were measured and the values are listed in Table II. The thick-
ness of the current collector was determined via ICP spectroscopy
measurements, using the densities of copper and aluminum.
Hg-porosimetry.— The porosity of the different parts of a cell
strongly affects the behavior of the system. The surface area deter-
Table II. Geometrical data of the 7.5 Ah Kokam cell.
mines especially the chemical reaction. Mercury porosimetry was
used to determine the porosity, the particle radius as well as the tortu-
osity factor of the system. The porosity of a sample ε is defined as the
ratio of the total pore volume and the volume of the sample. The dis-
tribution of particle radius is derived using an approach of Mayer and
Stowe.18 The tortuosity factor is calculated according to Carniglia.19
Porosity, mean particle radius as well as tortuosity factor were mea-
sured for the anode and cathode material as well as for the separator.
The resulting values corrected for the current collector are displayed
in Table III. Figure 2 shows the distribution of the particle radius for
the two materials before correcting for the current collector. For both
electrode materials, two particle sizes can be distinguished. It is not
clear to the authors whether two particle sizes exist in the sample or
whether the larger particle is an agglomerate of smaller particles.
Conductivity and diffusion of electrolyte.— An important parame-
ter determining the inhomogeneity inside the active material of a cell
is the ionic conductivity of the electrolyte. The ionic conductivity in
dependency of the salt concentration as well as the temperature de-
pendency was measured for the electrolyte LP50 by BASF. Figure 3
shows the electrolyte conductivity in dependency on salt concentra-
tion (a). This dependency is of importance in the model as during
operation, concentration gradients are formed inside the electrolyte.
Similar to Ding et al.20 a polynomial function is used to fit this de-
pendency. The fit is performed for the data at 23◦ C with the following
result:
σe = 2.667 · x 3 − 12.983 · x 2 + 17.919 · x + 1.726 [1]
where σe is the electrolyte conductivity in [mS/cm] and x the salt
concentration in [mol/l]. The fitting function is valid for 0.5 ≤ x ≤1.5.
To include the temperature dependency in the model as well, the
activation energy of the conductivity has to be determined. In Fig-
ure 3b, the logarithm of conductivity times temperature over inverse
of temperature is displayed. Comparison with data of the data sheet21
of the electrolyte shows good agreement. The Arrhenius behavior of
the conductivity is clearly visible in the graph:
1 −E a,σ
σe =· σ0 · e R·T [2]
T
R is the gas constant. The slope of the graph can be used to determine
the activation energy of the process Ea,σe = 17.12 ± 0.13 kJ/mol.

total Table III. Porosity, particle radius and tortuosity of the 7.5 Ah
thickness thickness surface Kokam cell.
of one current length width area
sheet [μm] collector [μm] [mm] [mm] [mm2 ] Porosity [%] Median particle radius [μm] Tortuosity factor
Anode 162 ± 2 14.7 ± 0.06 103 87 8961 Anode 32.9 ± 0.5 8.7 ± 0.9 2.03 ± 0.006
Cathode 124 ± 1 15.11 ± 0.01 101 85 8585 Cathode 29.6 ± 0.7 6.5 ± 0.1 1.94 ± 0.006
Separator 19 Separator 50.8 ± 2 - 1.67 ± 0.02
 Journal of The Electrochemical Society, 162 (9) A1836-A1848 (2015) A1839

Figure 2. Particle size distribution obtained by Hg-porosimetry for graphite material (a) and Li(Ni0.4 Co0.6 )O2 material (b) before correction for the current
collector.

The Einstein relation can finally be used to derive the diffusion
constant of the electrolyte from the conductivity measurement:22
σe · k B · T
De =
e2 · N A · c
kB is the Boltzmann constant, e the elementary charge, NA the Avo-
gadro constant and c the lithium concentration within the electrolyte.
Following Equation 3 for a 1 molar electrolyte at 23◦ C, a diffusion
constant of De = (2.44 ± 0.05) 10−6 cm2 /s can be derived. This is sim-
ilar to the values found by Capiglia et al. 199923 of about 2.25 · 10−6
cm2 /s using NMR and Nyman et al. 200824 who derived an apparent
diffusion constant of about 3.7 · 10−6 cm2 /s at 25◦ C for a 1M LiPF6 -
EC-EMC electrolyte. Nyman et al. 200824 also derived a transport
number for Li+ in this electrolyte with respect to the solvate velocity
of t0 + = 0.26. This value will be used in the following for model
parameterization. The activation energy of the diffusion in the elec-
trolyte follows the activation energy of the ionic conductivity and
yields Ea,De = Ea,σe = 17.12 ± 0.13 kJ/mol. It has to be mentioned
here that the Einstein relation only holds for dilute solutions. For bat-
tery electrolytes, concentrated solution theory holds, but the Einstein
relation is considered as an approximation here. As comparison with
values from literature, derived using concentrated solution theory, as
well as the simulation validation in12 show, this approximation seems
to be justified for the material and the currents up to 5 C consid-
ered in this work. Further technics to characterize electrolytes using
[3] concentrated solution theory are described in.25–27
Electronic conductivity of electrodes.— In literature the values for
the electronic conductivity of electrode materials vary over many or-
ders of magnitudes. Variations of 4 orders of magnitude are quoted
for example for graphite materials.7 Furthermore, most of the mea-
surements do not take the total electrode setup into account nor the
porous structure of an electrode. Therefore, it is necessary to mea-
sure the conductivity of the material and the electrode setup under
consideration.
Electronic conductivities of the electrode materials determined in
a 4-point setup by applying DC currents of −0.1 A and 0.1 A are
listed in Table IV. A huge scatter of more than 60% is observed
between different samples. Compared with results in literature,28,29
this scatter seems to be a common phenomenon. Electronic conduc-
tivity measured for the Li(Ni0.4 Co0.6 )O2 material is in good accordance
with results obtained by Chen et al. 2010,28 who measured the elec-
trical conductivity of a NMC electrode. The measured value for the
graphite material seems to be a bit too small. In literature, values for

Figure 3. Electrolyte conductivity vs. salt concentration (a) for different temperatures. The results of the fitting (Equation 1) are shown in black. The fitting results
for 20◦ C and 25◦ C are calculated using Arrhenius Equation 2. Arrhenius plot (b). The logarithm of conductivity times temperature vs. inverse of temperature is
displayed.
A1840 Journal of The Electrochemical Society, 162 (9) A1836-A1848 (2015)
Table IV. Electronic conductivities of the electrode materials using
DC measurement method.
Electronic Conductivity [mS/cm]
Anode 1.11 ± 0.7
Green anode, prior to formation 139.91 ± 34.2
process (different manufacturer)
Cathode 680.67 ± 442.7
graphite crystals are 3 to 7 orders of magnitude higher compared to
the measured value.7
One problem of the determination of the anode conductivity with
the measurement setup described here could be that the graphite par-
ticles are covered by SEI layer. Therefore, the conductivity of the
SEI layer is measured as well, instead of only the conductivity of the
pure graphite material. To support this theory, a green graphite elec-
trode (complete and already calendered electrode prior to formation
process) without SEI was measured as well. Unfortunately, no green
electrode of the cell considered in this work has been available. There-
fore, a graphite electrode of a different cell supplier was used. The
measurement revealed a conductivity of 139.91 ± 34.2 mS/cm, which
is 2 orders of magnitudes higher, compared to the already formed
anode considered in this work. As for model parameterization, only
the pure electronic conductivity of the electrode material is of interest,
this value is used for further considerations.
To discuss the limitations of this measurement setup, it has to be
mentioned that it is also not completely clear how the voltage measur-
ing tip on the probe influences the current flow through the probe. It
could be possible that the potential lines close to this tip are not exactly
parallel, which could manipulate the measurement result. Addition-
ally, the method does not determine the pure electronic conductivity
of the electrode material but includes the contact resistance between
the active material and the current collector as well. This adds a small
error to the model parameter. However, the method seems to be an
adequate way to derive the order of magnitude of the parameter.
Electrochemical measurements.— To determine the dynamic
properties of the materials, electrochemical measurements were per-
formed and are discussed in the following.
OCV measurement and determination of cell balancing.—During op-
eration of a coin full cell, only parts of the capacities of the two
electrode materials are used. In a coin full cell the graphite material
Lix C6 is used from x = 0 in the fully discharged state to x < 1
in a fully charged state of the coin full cell. Graphite is usually
not used in a fully intercalated state to prevent lithium plating. The
Lix (Ni0.4 Co0.6 )O2 material also does not reach neither reaches max-
imum loading with lithium intercalation at fully discharged state of
4.2
4
B=2.925
3.8
A=1.425
OCV [V]
3.6
3.4
3.2
3
2.8
0 1 2 3 4 0
(a) Charge Capacity Q [mAh]
Figure 4. Results of OCV measurement for a coin full cell at 23◦ C. OCV vs. charge capacity (a). Differential voltage vs. charge capacity (b). The two most
striking inflection points A and B are marked.
the coin full cell due to SEI formation. The Li(Ni0.4 Co0.6 )O2 material
is used from x < 1 in the fully discharged state to x > 0 in a fully
charged state of the full cell. Complete de-intercalation of lithium is
usually avoided due to thermodynamic instabilities at low intercala-
tion states. This so-called balancing of the cell is an important input to
physico-chemical simulation models. In this section, the balancing of
a coin full cell is determined. To which extent it is possible to transfer
this balancing to the Kokam cell is discussed in.12
To determine the balancing of the coin cell, OCV measurements
were performed on graphite, respectively Li(Ni0.4 Co0.6 )O2 -based coin
half cells, as well as on coin full cells with both electrodes originated
from the Kokam cell. Figure 4, Figure 5 and Figure 6 display the OCV
vs. SOC of the coin half as well as full cells. Characteristic points can
be detected in all three curves.
These inflection points observed in the OCV curve of the coin full
cell result from staging phenomena in the anode as well as from the
cathode material. To determine the cell balancing, the inflection points
in the OCV curve of the coin full cell are compared with those in the
OCV curve of the two coin half cells. Figure 4 displays the points of
the most striking inflection points A and B in the OCV curve of the
coin full cell. Careful analysis of Figure 4, Figure 5 and Figure 6 leads
to the conclusion that the characteristic point A in the OCV curve
of the coin full cell corresponds to the characteristic point a in the
OCV curve of the Li(Ni0.4 Co0.6 )O2 material. Similarly, point B of the
coin full cell corresponds to the characteristic point b in the graphite
material. Following Kumaresan et al. 2006,30 the lithium content of
the electrode materials at end of discharge xEoD as well as the lithium
content at end of charge xEoC can be calculated.
Table V displays the results for the calculated lithium content at end
of discharge and end of charge of a coin full cell in the Li(Ni0.4 Co0.6 )O2
and the graphite material, as well as the corresponding capacities of
the coin cells during OCV measurement Canode , Cfull and Ccathode used
for the calculation. The anode is cycled in the coin full cell in the range
0.04 < x < 0.75, the cathode between 0.26 < x < 0.86. The same
result for the lithium content of the Li(Ni0.4 Co0.6 )O2 material at end of
discharge is obtained using ICP spectroscopy, confirming the method
used above. This leads to the conclusion that 14% of the lithium was
bound in the SEI during formation. The cathode was designed for an
utilization of around 74%. These are important input parameters for
the simulation model.
For model parameterization, the inactive part of the electrode ma-
terial like filler and binder has to be determined as well. The inactive
part can be estimated using the measured Cmeas and the theoretical
capacity Ctheo according to
 
Cmeas
inactive par t [%] = 1 − · 100 [4]
Ctheo
0.5
0.4
B
0.3
dV/dQ
0.2
0.1 A
0
1 2 3 4
(b) Charge Capacity Q [mAh]
 Journal of The Electrochemical Society, 162 (9) A1836-A1848 (2015) A1841

1.5 0

-0.5

1
-1
OCV [V]

dV/dx
b

-1.5
0.5

-2
b=0.62

0 -2.5
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
(a) x in LixC6 (b) x in LixC6

Figure 5. Results of OCV measurement for a graphite coin half cell at 23◦ C. OCV vs. charge capacity (a). Differential voltage vs. charge capacity (b). The most
striking reflection point b is marked.

5 0
a
-0.5
4.5

-1
OCV [V]

dV/dx

4
-1.5
a=0.6192
3.5
-2

3 -2.5
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
(a) x in Lix(Ni0.4Co0.6)O2 (b) x in Li (Ni Co )O
x 0.4 0.6 2

Figure 6. Results of OCV measurement for a Li(Ni0.4 Co0.6 )O2 coin half cell at 23◦ C. OCV vs. charge capacity (a). Differential voltage vs. charge capacity (b).
The most striking reflection point a is marked.

The theoretical capacity of the material can be estimated by Therefore, this parameter cannot be considered to be totally reliable
for model parameterization and has to be used with care.
ρ · V · (1 − ε) · N A · e
Ctheo = [5] Solid-state diffusion in the electrode material.— For lithium-ion bat-
M tery performance, the solid-state diffusion of lithium inside the active
with the density of a crystal of the material ρ, the volume of the used material is of great relevance. For battery applications, the chemical
electrode, the porosity of the electrode material ε, the Avogadro con- diffusion coefficient is of importance; it can be determined by dif-
stant NA , the elementary charge e and the molar mass of the material ferent methods like electrochemical impedance spectroscopy (EIS),
M. As density for the Li(Ni0.4 Co0.6 )O2 the value for Li(NiMnCo)O2 galvanostatic intermittent titration technique (GITT) or potentiostatic
material is used according to Yoshizawa and Ohzuku 2007,31 due to intermittent titration technique (PITT). In this work, the focus is on
the lack of knowledge of the density of the actual material. All results the first two techniques.
are listed in Table VI. The resulting values for the inactive parts seem GITT.— GITT is a technique where the battery is excited with a
to be very high. However, for lithium-ion polymer batteries up to 30% short constant current pulse in equilibrium state at a certain SOC.
(w) of inactive parts in MCMB as well as in Li(Ni0.82 Co0.18 )O2 was According to Weppner and Huggins,15,16 the IR drop only acts as a
reported.32 Additionally, the estimation of the theoretical capacity of constant that shifts the measured voltage, but does not change the
the material adds an uncertainty to the calculation of the inactive part. shape of the curve. After subtracting the IR drop, the change of volt-
age is assumed to be due to formation of a concentration gradient
in the solid state material. The influence of the reaction kinetic is
Table V. Lithium content at end of discharge and end of charge of
a coin full cell in the Li(Ni0.4 Co0.6 )O2 and the graphite material
as well as the capacity of the different coin cells measured during Table VI. Theoretical capacities of the cells, the resulting inactive
OCV. part of the electrode material as well as the density used for
calculation.
xEoC,cathode xEoD,cathode xEoC,anode xEoD,anode
Theoretical
0.26 0.86 0.75 0.04
Capacity Inactive Part Assumed Material
Capacity measured during OCV
[mAh] (v) [%] Density [g/cm3 ]
Li(Ni0.4 Co0.6 )O2 Ccathode = 6.11 mAh
LiC6 Canode = 5.07 mAh Li(Ni0.4 Co0.6 )O2 10.04 39.17 4.75
Coin full cell Cfull = 3.60 mAh LiC6 8.49 40.37 2.3
A1842 Journal of The Electrochemical Society, 162 (9) A1836-A1848 (2015)

x=1 / V=0.020 V x=1 / V=3.45 V

x=0.73 / V=0.089 V x=0.85 / V=3.704 V
x=0.35 / V=0.131 V -35 x=0.7 / V=3.887 V
Figure 7. Impedance spectra for a graphite coin
-80 x=0.35 / V=0.131 V repetition x=0.7 / V=3.887 V repetition
-30 half cell (a) and a Li(Ni0.4 Co0.6 )O2 coin half cell
x=0.15 / V=0.200 V x=0.25 / V=4.279 V (b), measured at 23◦ C at different SOC (x denotes
-60 -25 the amount of intercalated lithium, where SOC =
x∗100 and V the voltage of the half cells). The
Im(Z ) / Ω

Im(Z ) / Ω
spectrum marked with ‘repetition’ is a spectrum
-20
-40 that was repeated after all other spectra had been
performed in order to identify cell ageing during
-15 the measurement process. For the graphite the fre-
-20 quencies of the spectrum at 35% SOC are shown
-10 for the centre of the semi-circle and for the knee
where diffusion starts. For the Li(Ni0.4 Co0.6 )O2
0 -5 these frequencies are given for the spectrum at
70% SOC. Additionally the frequency of the knee
0 where charge transfer starts is shown.
20 40 60 80 100 10 20 30
Re(Z ) / Ω Re(Z ) / Ω
180 Hz
3 Hz
(a) (b) 1 Hz
240 Hz 15 Hz

neglected. Assuming 1-dimensional diffusion and relying on Fick’s
law, the diffusion coefficient can√be derived. For small currents and
a potential E that is linear over t, the diffusion coefficient can be
calculated by:
     2
4 · r2 E S 2 r
DG I T T = · if t  [6]
π·t E t D
r is the particle radius, t the pulse duration, Es the change in equi-
librium voltage during the current pulse, and Et the total voltage
change during the current pulse after subtracting the IR drop.
Dees et al.33 evaluated the limitation of the GITT technique by
comparing the results obtained by GITT with the ones obtained by
fitting of a simulation model. They found the method not to be the
most appropriate for model parameterization. The approach assumes
a uniform current distribution in the composite electrode. This is
certainly an invalid assumption, not given in a porous system where
the electronic conductivity exceeds by far the ionic conductivity and
leads to deviations in measurement. In the following, results obtained
by EIS are compared to the GITT results in order to investigate the
difference of the two techniques.
EIS.— EIS is a measurement technique based on alternating volt-
age and current, where the resulting impedance at different frequencies
is analyzed. It is a suitable in-situ method to identify processes oc-
curring in batteries. Electric circuit diagrams can be used to describe
such a spectrum, which was described by Aurbach et al. 1999.34
The mid-frequency semi-circle accounting for effects of the surface
films as well as for the charge transfer of the two electrodes and the
coupled double layer capacities is described by ZARC elements ZZarc
in this work. A ZARC element consists of a resistance R and a constant
phase element (ZCPE = 1/C(jω)ξ ) in parallel:35
R with Figure 7 leads to the conclusion that the charge transfer circle
Z Z ar c =
1 + R · C · ( j · ω)ξ
The parameter ξ describes the magnitude of depression.
The diffusion branch accounting for solid state diffusion is de-
scribed by a ‘Warburg’ element for planar particles with finite-length
diffusion in the following, similar to the work of Levi et al. 1999.36
The ‘Warburg’ element can then be written as:
  
R·T tanh l · j · ωD
Z W ar = · √ [8]
c · z2 · F 2 · S j ·ω· D
R is the gas constant, T the temperature, c the lithium concentration
in the bulk, z the charge number (for lithium-ion battery z = 1),
F the Faraday constant, l the diffusion distance (in this case the particle
radius) and D the diffusion coefficient. The contact area S between
electrode and electrolyte is calculated by the total electrode volume
Velectrode , the porosity ε, the inactive part of the material inactivepart , the
volume of a single particle V1particle and the particle radius r, following:
 
Velectr ode · (1 − ε) · 1 − inactive par t
S= · 4 · π · r2 [9]
V1 par ticle
Impedance spectra were measured for cathode and anode at dif-
ferent SOC and temperatures. Figure 7 shows impedance spectra for
a graphite coin half cell (a) and a Li(Ni0.4 Co0.6 )O2 coin half cell (b),
measured at 23◦ C at different SOC. The spectrum at 35% SOC for the
anode and 70% SOC for the cathode was repeated after finishing with
all spectra in order to identify cell aging during the measurement pro-
cess. In both cases no severe changes in the spectra occurred during
measurement.
The anode spectra show one semi-circle (in some cases there is
a hint of a second semi-circle below the dominant one) in the mid-
frequency range, followed by a diffusion branch and the vertical part
of the intercalation capacitance. As graphite exhibits voltage plateaus
in the medium SOC ranges, the intercalation capacitance is only rarely
visible in these spectra. Furthermore, the radius of the mid-frequency
semi-circle changes with SOC. In the region of the graphite plateaus,
the change in radius is small, while for high intercalation states the
circle becomes larger. As neither the charge transfer semi-circle of the
lithium is expected to change size with changing SOC (or is at least
very small, see Figure A1 in the appendix) nor the one of the surface
layer, it can be assumed that the dominant part of the semi-circle ori-
gins from the charge transfer of the graphite material. This conclusion
is supported by the impedance spectra obtained for symmetrical Li/Li
cells in the appendix. Figure A1 shows that the charge transfer of
the lithium occurs around 300 Hz with a circle diameter of around
12 . This leads to a diameter of 6  in a coin half cell. Comparison
[7] of the lithium is hidden below the dominant charge transfer circle of
the graphite and has only minor contributions to the total diameter of
the main circle. A fit of a ‘Warburg’ impedance (Equation 8) to the
diffusion branch reveals an additional semi-circle located below the
diffusion branch. It is not clear to which physical process this semi-
circle belongs. It can be suspected to originate from the electrolyte
diffusion. However, it is clear that it cannot be attributed to the charge
transfer due to the low frequency range. According to the shape of the
impedance spectra at different SOC, an electric circuit model shown
in Figure 8 is used to fit the graphite spectra. It consists of a serial
connection of an ohmic resistor, two ZARC elements, a ‘Warburg’
element following Equation 8 and an intercalation capacitance.
A similar picture arises from the spectra of the cathode material.
Here, the mid-frequency semi-circle is also dependent on the SOC of
the battery and can therefore be attributed to the charge transfer process
 Journal of The Electrochemical Society, 162 (9) A1836-A1848 (2015)
R,ohm Zarc,ct Zarc,2 Z,War
Figure 8. Electric circuit diagram used to fit the anode and cathode spectra.
of the cathode material. The semi-circle is followed by a diffusion
branch and an intercalation capacitance. For the cathode material, an
additional semi-circle is also observed which is overlapped by the
diffusion branch. In this case as well, the attributed process is not
clear, but the frequency is too low in order to refer this circle to the
charge transfer. Additionally, on the cathode side a third semi-circle
is indicated for very high frequencies. The circle seems to be located
at frequencies above 180 Hz and no SOC dependency is observed.
Comparing the results with the spectra of a symmetrical Li/Li cell in
the appendix in Figure A1, this semi-circle can be attributed to the
lithium counter-electrode occurring at around 300 Hz. Neglecting this
high frequency circle, the same electric circuit diagram can be used
as for the anode material (see Figure 8).
In order to identify the diffusion constant and the exchange current
density of the material under consideration, the electric circuit dia-
gram in Figure 8 together with the ‘Warburg’ impedance of Equation 8
is used to fit the impedance spectra of the anode and cathode material.
The fitting is performed with the Matlab function easyfit.m which uses
the ‘Nelder-Mead Simplex’ method for minimization. In the follow-
ing, the resulting fitting parameters of the ‘Warburg’ element are used
to determine the diffusion coefficient of the active material. It has to
be mentioned in this context that depending on SOC and temperature
the fitting results were not always unambiguous. Depending on the
starting parameters, different results can be obtained by the fitting
procedure.
Results.— Figure 9 displays the results of the GITT and EIS mea-
surements for the solid-state diffusion at 23◦ C. The diffusion co-
efficient is shown at different states of charge for graphite (a) and
Li(Ni0.4 Co0.6 )O2 (b). The results show a good accordance between
the diffusion coefficients obtained by the different techniques GITT
and EIS. The diffusion coefficients of both materials are strongly
SOC dependent. The diffusion coefficient of graphite decreases
with increasing amount of intercalated lithium, the diffusion coef-
ficient of the Li(Ni0.4 Co0.6 )O2 material shows a minimum at around
60% SOC.
The diffusion constant of the graphite correlates with the staging of
the graphite. In the transition between two plateaus at 60% SOC a peak
in diffusion constant is observed, whereas in the regions of coexisting
phases the dependency of diffusion constant on SOC is flat, similar to
the OCV. The results for the graphite material are in good accordance
with results by Piao et al. 1999.37 By using impedance spectroscopy
and chromamperometric method on commercially available graphite
-8.5
log (Diffusion Coefficient) [cm /s]
2
-9
-9.5
-10
-10.5
-11
-11.5
0 0.2 0.4 0.6
(a) x in LixC6
Figure 9. Base 10 logarithm of diffusion coefficient vs. intercalated lithium measured with EIS and GITT at 23◦ C for graphite (a) and Li(Ni0.4 Co0.6 )O2 (b) at
23◦ C.
A1843
electrodes, they found a decreasing diffusion coefficient with increas-
ing lithium content in the range of 3 × 10−9 –3 × 10−12 cm2 /s, similar
C,inter to the results shown here. They also observed two domains over SOC.
For x < 0.4 the slope of the curve is very steep, whereas the diffusion
constant becomes more or less constant for x > 0.4. However, they
did not find a correlation with phase transitions. Levi et al.38 also
observed a correlation between the staging of the graphite and the
diffusion coefficient similar to the results shown in Figure 9. Funabiki
et al. 199811 found this correlation on the staging as well, but obtained
values for the diffusion coefficients that were orders of magnitudes
higher.
It has to be mentioned that the results of the diffusion coefficient
for graphite show a large spread in literature, depending on the mea-
surement technique but also on the structure of the material and the
SOC under consideration. Park et al. 20107 give an overview of the
diffusion coefficients measured for different materials. Concerning
model parameterization, these results clearly show the importance of
measuring the diffusion of the special material under consideration.
The high variation of diffusion coefficients over SOC illustrates that
it is not suitable to parameterize the model with a single value, inde-
pendent of SOC, as done in most works dealing with physic-chemical
models (e.g. Ref. 39). In order to derive reliable results of the internal
battery parameters the SOC dependency of diffusion coefficients must
not be neglected.
The Li(Ni0.4 Co0.6 )O2 material shows a SOC dependency similar
to the one measured by Montoto and Rosolen 200440 with a min-
imum diffusion coefficient at x = 0.6. The measured magnitudes
are slightly different, probably due to the difference in composi-
tion (Li(Ni0.5 Co0.5 )O2 ). The minimum in diffusion coefficient coinci-
dences with the inflection point observed in 0. However, this minimum
does not coincidence with the order/disorder transition observed in
Ref. 40. In literature values for Li(NiCo)O2 vary from 10−8 – 10−12
S/cm2 .
As the results obtained by GITT as well as EIS seem to be reliable,
due to the greater amount of measurement points, the GITT results
are used as lookup for model parameterization in Ref. 12.
Additionally to the measurement at 23◦ C, GITT as well as EIS
was performed at different temperatures at 15% SOC for a graphite
coin half cell and at 70% SOC for a Li(Ni0.4 Co0.6 )O2 coin half cell.
15% SOC has been chosen for the graphite material in order to pre-
vent a measurement in a graphite plateau where only small changes
in voltage arise. In the impedance spectra, the intercept with the x-
axis as well as the radius of the charge transfer circle of the ma-
terials change with temperature. For the Li(Ni0.4 Co0.6 )O2 material,
additionally the charge transfer circle of the lithium changes with
temperature.
Figure 10 displays the resulting Arrhenius plot of the diffusion
coefficients for graphite (a) and Li(Ni0.4 Co0.6 )O2 (b). For both mate-
rials the GITT results confirm Arrhenius behavior, whereas the EIS
-8
log (Diffusion Coefficient) [cm /s]
GITT
EIS
-8.5
-9
-9.5
-10 GITT
EIS
-10.5
0.8 1 0 0.2 0.4 0.6 0.8 1
(b) x in Lix(Ni0.4Co0.6)O2
A1844 Journal of The Electrochemical Society, 162 (9) A1836-A1848 (2015)

-19 -18
log (Diffusion Coefficient) [cm /s]

log (Diffusion Coefficient) [cm /s]

GITT GITT
2

2
-20 EIS -9691*x+10.85 EIS
-20
-5886*x-1.78
-21

-22 -22
-3816*x-9.21
-4912*x-5.67
-23 -24
-24
-26
-25

-26 -28
3 3.2 3.4 3.6 3.8 4 3 3.2 3.4 3.6 3.8 4
(a) 1/T [1/K] x 10
-3 (b) 1/T [1/K] -3
x 10

Figure 10. Arrhenius plot of solid-state diffusion coefficient obtain by EIS and GITT technique. Logarithm of diffusion coefficient vs. inverse of temperature is
shown for graphite at 15% SOC (a) and Li(Ni0.4 Co0.6 )O2 at 70% SOC (b). The dashed lines indicate the Arrhenius fit with the resulting fitting parameter.

results slightly deviate from Arrhenius behavior. The deviation of the
EIS results from Arrhenius can be due to the fact that the fitting re-
sults for some temperatures have been ambiguous depending on the
starting parameters. Furthermore, the results obtained by EIS differ
strongly from the results obtained by GITT technique. The GITT
results exhibit a stronger temperature dependency of the graphite
compared to the Li(Ni0.4 Co0.6 )O2 material. The resulting slopes of
the Arrhenius plot are used to calculate the activation energies, yield-
ing Ea,Dn = 48.9 kJ/mol for graphite and EaDp = 31.7 kJ/mol for the
Li(Ni0.4 Co0.6 )O2 material for the GITT measurement. Contrariwise,
according to the EIS results the Li(Ni0.4 Co0.6 )O2 material exhibits a
stronger temperature dependency compared to the graphite. Activa-
tion energies of Ea,Dn = 40.8 kJ/mol for graphite and EaDp = 80.6
kJ/mol for the Li(Ni0.4 Co0.6 )O2 material are obtained by EIS. For the
Li(Ni0.4 Co0.6 )O2 material this means a difference of 40% between
the two measurement techniques. Table VII summarizes the activa-
tion energies obtained by different measurement techniques. To the
knowledge of the authors, in literature only Kulova et al. 200641 is
known dealing with the determination of activation energy of solid-
state diffusion for the materials considered in this work. They obtained
35 kJ/mol for graphite material, which is in the range of the activation
energy measured with EIS in this work.
As discussed before, both methods have their weaknesses. For EIS,
the ambiguity of the fitting results depending on the start parameters
poses a problem, while the GITT theory assumes a uniform current
distribution in the composite electrode, which is not given in battery
electrodes. Furthermore, the influence of the charge transfer can be
separated in EIS, but not in GITT measurements. Both methods can
lead to a change in SOC and temperature of the battery due to the
applied currents. The adjustment of SOC can also lead to deviations.
Due to the uncertainties in the determination of this parameter, the
activation energy of solid-state diffusion has been left as a free fitting
parameter in the model (see12 ). For the measurement of the activation
energy of the diffusion coefficient only one state of charge was consid-
ered. As the diffusion coefficient varies strongly with SOC it is likely
that the corresponding activation energy also exhibits a SOC depen-
dency. Further measurements at different SOC have to be performed
in future works.
Table VII. Activation energies for solid state diffusion coefficients
obtained by different measurement techniques.
GITT
graphite 48.9 kJ/mol
Li(Ni0.4 Co0.6 )O2 31.7 kJ/mol
Charge transfer.—Finally, to parameterize a physico-chemical model,
the reaction kinetics of the electrodes are of importance. To determine
the exchange current density and the charge transfer coefficient of
the charge transfer reaction of the two electrodes, the Butler-Volmer
equation can be used, giving the relation between overpotential ηD
and current iD :
    
α·z· F (1 − α) · z · F
i D = i 0 · ex p · η D − ex p − · ηD
R·T R·T
[10]
i0 is the exchange current and α the charge transfer coefficient. For
small overpotentials (which is the case in EIS measurements), the
Butler-Volmer equation can be approximated using a Taylor series,
leading to the exchange current density:
R·T
j0 = [11]
α · z · F · S · Rct
where S is the contact area between electrode and electrolyte (see
Equation 9) and Rct the charge transfer resistance.
The intensity of charge transfer depends on the concentration of the
reactants. For a lithium electrode with Li ↔ Li+ + e− the exchange
current density depends on the lithium concentration in the electrolyte
ce , leading to:
j0 = F · k0 · ceα [12]
For an insertion electrode with Li ↔ Li+ +  + e− , where  rep-
resents the solid lattice, the exchange current density depends on the
lithium concentration in the electrolyte ce , the lithium concentration
in the solid lattice cs and the concentration of unoccupied sites in the
lattice (cs,max -cs ) according to:
 (1−α) α α
j0 = F · k0 · cs,max − cs · cs · ce [13]
As the charge transfer coefficients α of the graphite, the
Li(Ni0.4 Co0.6 )O2 and the lithium material determined in section
Charge transfer are all close to 0.5, α = 0.5 is assumed in the
following.17
The EIS spectra measured at different SOC and temperature are
evaluated for the anode and the cathode. Fitting results of the electric
circuit model in Figure 8 are used to determine the charge transfer
resistance Rct of the ZARC element Zarcct following Equation 7. The
exchange current density is obtained using Equation 11.
As already discussed in section Solid-state diffusion in the elec-
trode material, the dominant semi-circle in the anode and cathode
spectra seem to origin from the charge transfer of the electrode. How-
ever, in both cases the semi-circle of the surface layer is not clearly
EIS Literature
visible in the spectra. Especially for the anode a semi-circle due to
40.8 kJ/mol 35 kJ/mol41 surface layer is expected. It is likely that the smaller circle of the sur-
80.6 kJ/mol - face layer is placed below the dominant charge transfer one. Therefore
 Journal of The Electrochemical Society, 162 (9) A1836-A1848 (2015) A1845

-4 -4
x 10 x 10
1 2.4

Exchange Current Density [A/cm ]

Exchange Current Density [A/cm ]

2
2

Measurement Measurement
Fit Fit
0.8 2.2

0.6 2

0.4 1.8

0.2 1.6

0 1.4
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
(a) x in LixC6 (b) x in Lix(Ni0.4Co0.6)O2

Figure 11. Exchange current density vs. intercalated lithium measured with EIS at 23◦ C for graphite (a) and Li(Ni0.4 Co0.6 )O2 (b). The dashed lines show fits of
the SOC dependency following Equation 13.

it has to be mentioned here that the resistance, obtained by fitting this
circle to a ZARC element, probably contains parts of the surface layers
additionally to the charge transfer.
Figure 11 shows the resulting exchange current density over inter-
calated lithium for graphite (a) and Li(Ni0.4 Co0.6 )O2 (b). The exchange
current density of the anode material is smaller compared to the cath-
ode. A small variation of exchange current density with SOC can
be observed. Compared to the variation in the diffusion coefficient
the differences are small. Equation 13 is used to fit the dependency
of exchange current density on SOC. The fitting results are shown
in Figure 11 as well. √
The free fitting parameter k0 yields a value
2
of 4.58 · 10−9 [ dms √moldm ] for the graphite and a value of 9.53 · 10−9
2
√ LiMn2 O4 exist yielding values of 8 · 10−5 A/cm2 and 1.3 · 10−6 A/cm2
[ dms √moldm ] for the Li(Ni0.4 Co0.6 )O2 material. Therefore an exchange
current density at 50% SOC of j0n = 7.05 · 10−5 A/cm2 for graphite
and j0p = 2.23 · 10−4 A/cm2 for the Li(Ni0.4 Co0.6 )O2 material is ob-
tained. It is difficult to find values in literature to compare with. There
are several approaches to determine the charge transfer resistance for
graphite.8–10 But only Funabiki et al. 199811 scaled the resulting ex-
change current with the active surface of the material. Without scaling
with the contact area between electrode/electrolyte the measured value
cannot be applied to or compared with the material under considera-
tion. Funabiki et al. 199811 obtained a charge transfer resistance for a
natural graphite powder which is one order of magnitude higher com-
pared to the value measured in this work. They also used impedance
ln (Exchange Current Density) [A/cm ]
spectroscopy for its determination. This fact already suggests that,
depending on the used material, the exchange current density can
exhibit large differences. Also Smart et al. 20118 showed the strong
dependency of the exchange current of graphite on the electrolyte
and the formed surface layer. They obtained a difference in exchange
current of factor 7 only by changing the percentage of the used sol-
vent components. This fact points out once more the importance of
measuring the exchange current density of the special material combi-
nation under consideration for model parameterization. The influence
of the electrolyte on the cathode material (Li(Ni0.8 Co0.2 )O2 ) used by
Smart et al.8 on the other hand is negligible. On the cathode the
material composition itself seems to be crucial. Literature values for
and also reveal large differences,42,43 depending on the considered
material.
Figure 12 displays the Arrhenius plot of the exchange current
density for graphite (a) and Li(Ni0.4 Co0.6 )O2 (b). In both cases Ar-
rhenius behavior can be confirmed. The graphite exhibits a slightly
stronger temperature dependency, meaning that the anode limitation
is getting stronger with lower temperature. The resulting slopes of
the Arrhenius plot are used to calculate the activation energies, yield-
ing Ea,i0n = 53.4 kJ/mol for graphite and Ea,i0p = 43.6 kJ/mol for
the Li(Ni0.4 Co0.6 )O2 material. Similar results were obtained by Smart
et al. 2011.8 Depending on the electrolyte they found activation ener-
gies of the exchange current of graphite in the range of 45–60 kJ/mol
ln (Exchange Current Density) [A/cm ]

2
2

-7 -6
y = - 6418.1*x + 11.848 y = - 5250.3*x + 8.7789
-8 -7

-9 -8

-10 -9

-11 -10

-12 -11

-13 -12
Graphite Lix(Ni0.4Co0.6)O2
-14 -13
3 3.2 3.4 3.6 3.8 4 3 3.2 3.4 3.6 3.8 4
-3 1/T [1/K] -3
(a) 1/T [1/K] x 10 (b) x 10

Figure 12. Arrhenius plot of exchange current density. Logarithm of exchange current density vs. inverse of temperature is shown for graphite (a) and
Li(Ni0.4 Co0.6 )O2 (b). The exchange current densities are obtained by fitting of spectra of a graphite coin half cell at 55% SOC and for a Li(Ni0.4 Co0.6 )O2
coin half cell at 70% SOC. The dashed line indicates the Arrhenius fit with the resulting fitting parameter.
A1846 Journal of The Electrochemical Society, 162 (9) A1836-A1848 (2015)
0 C DC
-10 -0.1 C DC
-0.25 C DC
-0.5 C DC
-5 -1 C DC
Im(Z ) / Ω
Im(Z ) / Ω
0 Li counter
electrode 240 Hz Li counter
(300 Hz) 2 electrode
x=0.55 / V=0.125 V
5 (300 Hz)
8 10 12 14 16 18 20 22 6 8 10 12 14 16 18
(a) Re(Z ) / Ω (b)
Figure 13. Impedance spectra for a graphite coin half cell (a) and a
Li(Ni0.4 Co0.6 )O2 coin half cell (b), measured at 23◦ C using different super-
imposed DC currents. The anode spectra were measured at 55% SOC, the
cathode spectra at 70% SOC. In both graphs the location of the charge transfer
semi-circle of the lithium counter electrode at around 300 Hz is sketched.
and of the Li(Ni0.8 Co0.2 )O2 material of 34–48 kJ/mol. But also higher
(67 kJ/mol)10 and lower (40–50 kJ/mol)9 values were reported in lit-
erature for graphite. In this context, the importance of measuring the
activation energies of the special material combination under consid-
eration for model parameterization becomes obvious again.
In order to determine the exact values of the charge transfer coef-
ficients α, which could deviate from the beforehand assumed value of
0.5, impedance spectra were measured using different superimposed
DC currents. Figure 13 displays the impedance spectra for a graphite
coin half cell (a) and a Li(Ni0.4 Co0.6 )O2 coin half cell (b), measured
at 23◦ C with different superimposed DC currents. For both materi-
als the dominant mid-frequency circle is visible that was identified as
charge transfer process of the corresponding electrode material in sec-
tion Solid-state diffusion in the electrode material. Figure 13 shows
that this circle depends on the applied current rate which confirms
the assignment to the charge transfer of the electrode materials. For
the graphite material one has to keep in mind that the charge trans-
fer semi-circle of the lithium counter electrode is overlapped by the
dominant charge transfer semi-circle of the graphite. As the lithium
semi-circle is much smaller compared to the graphite one and is lo-
cated at higher frequencies it can be assumed that the total radius of
the graphite semi-circle is not influenced by the current dependency
of the lithium semi-circle.
As before, the spectra are fitted using the electric circuit model
in Figure 8 to determine the charge transfer resistance Rct from the
ZARC element Zarcct following Equation 7. Figure 14 shows the
resulting charge transfer resistance vs. current for graphite (a) and
Li(Ni0.4 Co0.6 )O2 (b). The charge transfer resistances clearly show the
12.5
Charge Transfer Resistance [Ohm]
12
11.5
11
10.5
10
-4 -2
(a) Current [mA]
Figure 14. Charge transfer resistance for graphite (a) and Li(Ni0.4 Co0.6 )O2 (b) vs. superimposed DC current obtained by fitting the spectra shown in Figure 13.
The blue triangles indicate the measured results, the red dots the fitting result using the Butler-Volmer Equation 10. The applied currents correspond to 0 C, ±0.1
C, ±0.25 C, ±0.5 C and ±1 C.
0 C DC Butler-Volmer relation. Therefore, Equation 10 is used to fit the result
-8 to obtain the exchange currents of the electrodes as well as the charge
0.1 C DC
-6
0.25 C DC transfer coefficients. For the anode material a value of αn = 0.489
0.5 C DC is obtained, for the cathode αp = 0.517. It has to be mentioned that
1 C DC
-4 the Butler-Volmer equation for the Li(Ni0.4 Co0.6 )O2 does not fit well
-2
(see Figure 14). Therefore, the resulting charge transfer coefficient
has to be considered with care. According to the knowledge of the
0 authors, no work is available in literature determining the charge
15 Hz transfer coefficient of the charge transfer reaction of materials applied
in lithium-ion batteries. In simulative works usually a charge transfer
x=0.7 / V=3.887 V
coefficient of 0.5 is assumed.39,17,5,44,6
Re(Z ) / Ω
Conclusions
In this work all parameters needed to parameterize a physico-
chemical model have been determined experimentally for a 7.5 Ah
pouch cell produced by Kokam. The methods are summarized and
discussed for effective parameter determination. The measured values
can be found in summary in the second part of this publication,12
where they are applied and validated in a physico-chemical model. The
parameters have been compared with existing literature values. The
results illustrate the importance of the determination of the parameters
for the special materials of the considered system.
Geometrical data as well as porosity, particle radius and tortuosity
have been measured. These material parameters are specific for the
materials under consideration and can thus not be taken from literature
for model parameterization.
The ionic conductivity and diffusion coefficient of the
EC/EMC/LiPF6 electrolyte has been found to match well with values
taken from literature. Furthermore, both parameters do not seem to
change strongly due to different compositions. Therefore, values can
be taken from literature for model parameterization.
The measured values of the electronic conductivity of the electrode
materials showed large scatter. Also literature values exhibit scatters
of the same magnitude. However, depending on the sample under in-
vestigation, variations of 4 orders of magnitudes occur. Furthermore,
literature values for the electronic conductivity of the active material
usually do not include results for a whole electrode setup, including
the filler, binder and porous structure. Therefore, for model parame-
terization a simple parameter estimation using a DC current signal is
suggested.
Laboratory made coin cells have been used to determine OCV
curves of the materials. The balancing of the cell has been derived by
OCV measurement. The balancing of the system is unique and cannot
be taken from literature.
The most crucial parameter for model parameterization is the
solid-state diffusion coefficient and its temperature and concentration
8.5
Charge Transfer Resistance [Ohm]
Measurement
Fit
8
7.5
7
6.5
Measurement
Fit
6
0 2 4 -4 -2 0 2 4
(b) Current [mA]
 Journal of The Electrochemical Society, 162 (9) A1836-A1848 (2015) A1847

dependency. It has been determined using GITT and EIS. Both tech-
niques show good accordance in the determination of the concentra-
tion dependency of the diffusion coefficients, but results differ strongly
considering the temperature dependency. The results show that the
diffusion coefficients depend strongly on the lithium concentration
of the material. Therefore, it is not suitable to parameterize a model
with a single value, independently of concentration. In order to derive
reliable results of the internal battery parameters, the concentration
dependency of diffusion coefficients cannot be neglected. The results
of diffusion coefficients for certain materials show a large spread in
literature, depending on the measurement technique, but also on the
structure of the material and the present lithium concentration in the
material which makes it necessary to determine the value for the
special material under consideration.
Finally, the exchange current density has been determined. The
importance to measure this value for model parameterization for the
material combination used in the considered cell has been illustrated.
Altogether this work proposes a parameterization procedure that can
be applied by model developers, to adopt their model to new materials.
Acknowledgment
This work has been performed in the framework of the research ini-
tiative “Modellierung von Lithium-Plating”. The IGF-project LN 15
of the Research Association FKM, Lyon Strasse 18, 60528 Frankfurt
am Main was financed via the AiF within a program to promote indus-
trial research (IGF) by the Federal Ministry of Economic Affairs and
Energy based on a decision by the German Bundestag. Responsibility
for the content of this publication lies with the authors.
Appendix
Investigation of symmetrical lithium metal coin cells.— The laboratory cell design
used in this work for parameter determination is a coin half cell consisting of the electrode
under investigation and a metallic lithium foil as counter electrode. To determine the
influence of the lithium metal in the used electrolyte system on the measurement results,
symmetrical Li/Li coin cells were investigated. The cells were cycled by charging them
with 1 mA/cm2 for 30 min followed by a discharge with −1 mA/cm2 for 30 min. All tests
on the symmetrical Li/Li coin cell were conducted at 25◦ C. The resulting voltage curve
is shown in Figure A1a. The resistance of the cells changes with SOC, but this change
is small and can therefore be neglected. Impedance spectra were recorded before cycling
and after different cycle intervals. The times where spectra were measured are marked
with red crosses in Figure A1a. Figure A1b shows the resulting spectra. After six cycles
equilibrium seems to establish, where all surface film formation appears to reach a stable
state. At frequencies around 300 Hz a dominant semi-circle is visible corresponding to the
charge transfer of the system. There is a smaller circle hardly recognisable in Figure A1
at higher frequencies due to surface films. In the lower frequency ranges the electrolyte
diffusion can be observed.
References
1. J. Newman and W. Tiedemann, AIChE J., 21, 25 (1975).
2. T. F. Fuller, M. Doyle, and J. Newman, J. Electrochem. Soc., 141, 1 (1994).
3. M. Doyle, J. Newman, A. S. C. N. Gozdz, A. S. Gozdz, and J.-M. Tarascon, J. Elec-
trochem. Soc., 143, 1890 (1996).
4. P. Arora, M. Doyle, A. S. Gozdz, R. E. White, and J. Newman, J. Power Sources, 88,
219 (2000).
5. M. Doyle and Y. Fuentes, J. Electrochem. Soc., 150, A706 (2003).
6. Y. Ji, Y. Zhang, and C.-Y. Wang, J. Electrochem. Soc., 160, A636 (2013).
7. M. Park, X. Zhang, M. Chung, G. B. Less, and A. M. Sastry, J. Power Sources, 195,
7904 (2010).
8. M. C. Smart and B. V. Ratnakumar, J. Electrochem. Soc., 158, A379 (2011).
9. O. S. Mendoza-Hernandez, H. Ishikawa, Y. Nishikawa, Y. Maruyama, Y. Sone, and
M. Umeda, Electrochim. Acta, 131, 168 (2014).
10. T. R. Jow, M. B. Marx, and J. L. Allen, J. Electrochem. Soc., 159, A604
(2012).
11. A. Funabiki, J. Electrochem. Soc., 145, 172 (1998).
12. M. Ecker, S. Käbitz, I. Laresgoiti, and D. U. Sauer, J. Electrochem. Soc.,
In press.
13. Porotec Solid 1.3.4 Manual.
14. I. Keithley Instruments, Reference Manual: Series 2600A System SourceMeter,
(2011).
15. W. Weppner, J. Electrochem. Soc., 124, 1569 (1977).
16. R. A. Huggins, Solid State Ionics, 86–88, 41 (1996).
17. W. van Schalkwijk and B. Scrosati, Advances in Lithium-Ion Batteries. Springer
(2002).
18. R. P. Mayer and R. A. Stowe, J. Colloid Science, 20, 893 (1965).
19. S. Carniglia, J. Catalysis, 102, 401 (1986).
20. M. S. Ding, K. Xu, S. S. Zhang, K. Amine, G. L. Henriksen, and T. R. Jow, J.
Electrochem. Soc., 148, A1196 (2001).
21. BASF, “Selectilyte lp,” tech. rep., BASF (2012).
22. C. Czeslik, H. Seemann, and R. Winter, Basiswissen Physikalische Chemie. Studi-
enbücherChemie, Vieweg+TeubnerVerlag (2010).
23. C. Capiglia, Y. Saito, H. Kageyama, P. Mustarelli, T. Iwamoto, T. Tabuchi, and
H. Tukamoto, J. Power Sources, 81–82, 859 (1999).
24. A. Nyman, M. Behm, and G. Lindbergh, Electrochim. Acta, 53, 6356
(2008).
25. L. O. Valoen and J. N. Reimers, J. Electrochem. Soc., 152, A882 (2005).
26. S. G. Stewart and J. Newman, J. Electrochem. Soc., 155, F13 (2008).
27. S. Stewart and J. Newman, J. Electrochem. Soc., 155, A458 (2008).
28. Y.-H. Chen, C.-W. Wang, X. Zhang, and A. Sastry, J. Power Sources, 195, 2851
(2010).
29. C.-W. Wang, A. M. Sastry, K. A. Striebel, and K. Zaghib, J. Electrochem. Soc., 152,
A1001 (2005).
30. K. Kumaresan, Q. Guo, P. Ramadass, and R. E. White, J. Power Sources, 158, 679
(2006).
31. H. Yoshizawa and T. Ohzuku, J. Power Sources, 174, 813 (2007).
32. K. M. Kim, J.-C. Kim, N.-G. Park, K. S. Ryu, and S. H. Chang, J. Power Sources,
123, 69 (2003).
33. D. W. Dees, S. Kawauchi, D. P. Abraham, and J. Prakash, J. Power Sources, 189, 263
(2009).
34. D. Aurbach, B. Markovsky, M. Levi, E. Levi, A. Schechter, M. Moshkovich, and
Y. Cohen, J. Power Sources, 81–82, 95 (1999).
35. E. Barsoukov and J. R. Macdonald, Impedance Spectroscopy: Theory, Experiment,
and Applications. Wiley (2005).
36. M. D. Levi, J. Electrochem. Soc., 146, 1279 (1999).

before cycling
-60
0.04 25°C after 3 cycles
after 6 cycles
Voltage [V]

0.02
0.15 0 -40 after 16 cycles
after 26 cycles
-0.02
Im(Z ) / Ω

0.1 -0.04
-20
28 29 30
Time [h]
Voltage [V]

0.05
0

0
20
-0.05
0 20 40 60 80
-0.1 Re(Z ) / Ω
0 10 20 30
(a) Time [h] (b) ≈300 Hz

Figure A1. (a): Voltage response of a symmetrical Li/Li coin cell while cycling by charging them with 1 mA/cm2 for 30 min followed by a discharge with
−1 mA/cm2 for 30 min. Red crosses mark time point where impedance spectra were recorded. (b): Impedance spectra of the cell at 25◦ C recorded after different
number of cycles.
A1848 Journal of The Electrochemical Society, 162 (9) A1836-A1848 (2015)

37. T. Piao, S.-M. Park, C.-H. Doh, and S.-I. Moon, J. Electrochem. Soc., 146, 2794
(1999).
38. M. D. Levi and D. Aurbach, J. Phys. Chem. B, 101, 4641 (1997).
39. M. Doyle, T. F. Fuller, and J. Newman, J. Electrochem. Soc., 140, 1526
(1993).
40. L. Montoro and J. Rosolen, Electrochim. Acta, 49, 3243 (2004).
41. T. Kulova, A. Skundin, E. Nizhnikovskii, and A. Fesenko, Russian J. Electrochem.,
42, 259 (2006).
42. A.-K. Hjelm and G. Lindbergh, Electrochim. Acta, 47, 1747 (2002).
43. H. Kanoh, J. Electrochem. Soc., 143, 2610 (1996).
44. G. B. Less, J. H. Seo, S. Han, A. M. Sastry, J. Zausch, A. Latz, S. Schmidt, C. Wieser,
D. Kehrwald, and S. Fell, J. Electrochem. Soc., 159, A697 (2012).