Journal of Power Sources 264 (2014) 184e194

Contents lists available at ScienceDirect

Journal of Power Sources

journal homepage: www.elsevier.com/locate/jpowsour

Mathematical model of lithium-ion batteries with blended-electrode

system
Seunghun Jung*
Battery R&D, LG Chem Research Park, 104-1, Moonji-dong, Yuseong-gu, Daejeon 305-380, Republic of Korea

h i g h l i g h t s

Two mathematical models of lithium-ion cell with blended-electrodes are developed.

Equilibrium model can easily calculate equilibrium characteristics of a blended-electrode.

Dynamic model can predict dynamic performance of a lithium-ion cell with blended-electrodes.

Developed model is validated with experiment data and shows good agreements.

a r t i c l e i n f o a b s t r a c t

Article history: Many lithium-ion batteries have electrodes made of multiple types of active materials in order to take
Received 12 March 2014 advantage of each active material. According to the types and blending ratios of constituting active
Received in revised form materials, there can be numerous combinations of blended-electrodes, resulting in different performance
14 April 2014
characteristics. This paper introduces how to numerically model lithium-ion batteries with blended-
Accepted 15 April 2014
Available online 26 April 2014
electrodes, and predict their performance. Further, this paper demonstrates how to apply the devel-
oped model to designing new batteries. First, we propose an equilibrium model which can predict
equilibrium characteristics of a blended-electrode. Second, a physics-based lithium-ion cell model with
Keywords:
Lithium-ion
blended-electrodes is proposed. Developed model is validated with experimental data and shows good
Battery agreements.
Blended electrode Ó 2014 Elsevier B.V. All rights reserved.
Mixed electrode
Modeling

1. Introduction lithium-ion batteries must sustain severe operating conditions and

requirements. Therefore, it requires much more time and efforts to
Although lithium-ion batteries commercialized by SONY have develop large-formatted automotive batteries than small batteries
several merits such as larger capacity, lighter weight, and good for mobile applications.
coulombic efficiency compared to previously used batteries of Typical lithium-ion batteries are composed of negative active
other chemistry such as nickel-metal hydride batteries, there has materials, positive active materials, electrolyte, separator, additives,
been doubt that lithium-ion battery is not safe enough to be used and current collectors. Active materials are the most important
for automotive power source. However, recently, lithium-ion bat- among them since they determine power and energy of a battery.
tery is getting much more public, industrial attention. Several au- Many negative and positive active materials have been searched up
tomakers have launched hybrid electric vehicles (HEV), plug-in to now and each of active material has its own characteristics such
hybrid electric vehicles (PHEV), and battery electric vehicles (BEV) as operating potential profile, capacity and power. For example,
powered by lithium-ion batteries after GM successfully launched LiMn2O4 has good rate capability and high power with relatively
the first commercialized plug-in hybrid vehicle, VOLTÔ which low cost and less safety problem. But, it shows low capacity and
loads LG Chem’s high performance lithium-ion batteries. In Manganese (Mn) dissolution problem also. In contrast, Li(Ni, Mn,
contrast to small batteries for portable electronics, automotive Co)O2 has more capacity compared to LiMn2O4, but shows lower
power due to the layered structure and lower operating potential
window. Olivine material such as LiFePO4 is considered as one of
* Tel.: þ82 42 866 2094; fax: þ82 42 862 1981. promising positive materials due to low cost, improved safety with
E-mail address: [email protected]. large capacity. Many researches to overcome low electronic

http://dx.doi.org/10.1016/j.jpowsour.2014.04.072
0378-7753/Ó 2014 Elsevier B.V. All rights reserved.
 S. Jung / Journal of Power Sources 264 (2014) 184e194 185

conductivity and very low lithium diffusion rate of LiFePO4 (briefly This paper introduces how to predict performance of lithium-
‘LFP’) particle have been conducted [1e4]. As briefly listed above, ion batteries with blended-electrodes by mathematical method.
most of active material have both merits and demerits together. First, we introduce the blended-electrode model in equilibrium
Therefore, it is important to select proper active materials for state, which is useful to predict open-circuit voltage of a blended-
specified purposes. Since large battery capacity is the most electrode with arbitrary combinations of multiple active mate-
important factor for full electric vehicles (EV), Li(Ni1/3Mn1/3Co1/3)O2 rials. Second, the physics-based dynamic model is explained, which
(briefly ‘NMC’) can be selected as positive active material and can simulate battery performance in dynamic operating conditions
graphite can be used for negative electrode. In contrast, high power and explain physico-chemical phenomena inside blended-
is the most required quality for hybrid electric cars (HEV). In this electrodes.
case, spinel materials such as LiMn2O4 (briefly ‘LMO’) and amor-
phous materials such as carbon can be adopted for positive elec- 2. Equilibrium model
trode and negative electrode, respectively. When both capacity and
power are important, multiple active materials can be blended A pure active material has a distinct equilibrium voltage profile
together, so called ‘blended-electrode’. according to lithium intercalation level or lithium stoichiometry,
Several papers about experimental work on blended-electrodes which can be used for us to predict charging/discharging charac-
have been published. Mitigating Manganese dissolution and teristics of a cell made of the active material. By combining such
improving capacity retention by blending different active mate- equilibrium voltage data of several active materials, it is possible to
rials together were considered important by several researchers predict basic characteristics such as discharging profile of a
[5e7]. Fergus [8] extensively reviewed cathode materials including blended-electrode as follows.
composite electrodes. He categorized electrode with different Fig. 1 represents equilibrium voltage profiles of basic active
mixed particles and electrodes with coated particles layered as materials for this study. The basic concept of the present equilib-
composite electrode. Whitacre et al. [9] constructed fully mixed rium model of a blended-electrode is that the blended-electrode is
(blended) electrodes, fully segregated electrodes and layered in thermodynamically equilibrium state. In other words, the surface
electrodes by using LiFePO4 and Li(Li0.17Mn0.58Ni0.25)O2. They tried energy level (or the potential) of the constituting active particles
to take advantages of composite electrode from each active ma- should remain same level by minimizing potential difference be-
terial (high rate capability of LiFePO4 and large capacity of Li(L- tween active particles. This causes each different active particle
i0.17Mn0.58Ni0.25)O2 and they found that segregated active material may have different state of charge (lithium stoichiometry) under
configuration is the most promising solution with those two active equilibrium condition of the blended-electrode. Therefore, we can
materials. find out state of charge (lithium stoichiometry) of each constituting
As briefly reviewed above, it is worthwhile to consider blending active material. Theoretical capacity of LiMn2O4 is about
several active materials together to improve cell performance. 150 mAh g1 while usable capacity is about 107 mAh g1 (150e
However, blending different active materials may cause some 107 ¼ 43 mAh g1 is reserved to protect the crystal structure).
problems in a complex slurry mixing process or electrode coating Therefore, minimum lithium stoichiometry (soc 100%) of LiMn2O4
process, which may result in increased manufacturing cost. In is stoich100 ¼ 1.0 e 107/150 ¼ 0.287 and maximum lithium stoi-
addition, much time and cost would be required to find optimum chiometry (soc 0%) is stoich0 ¼ 1.0 when irreversible loss is ignored.
blending ratio for target performance of a cell. Actually, it takes So, we can setup following functions using Fig. 1 with above
much time and cost to finalize cell chemistry because cell makers relation.
go through several screening tests for candidate materials. Further,
numerous new active materials are being introduced by material 8 capai
8 >
makers. Although, it is good for cell makers to have many options to < ¼ f1 ðcapai Þ > stoichi ¼ capa
>
< max;i
choose, required time and cost for a new cell development may Ui ¼ f2 ðstoichi Þ where
increase as the combination number of materials grows larger. If : >
> stoichi  stoich0;i
¼ f3 ðsoci Þ >
: soci ¼
the performance of a new cell with blended-electrodes can be stoich100;i  stoich0;i
mathematically predicted before the actual cell is constructed, cell (1)
development cost will be significantly saved.
Few publications about modeling work on blended electrodes Inversely, lithium stoichiometry or SOC can be found for the
can be found up to now. Darling and Newman [10] started a given equilibrium potential also by transforming the above func-
modeling work on the electrode with two distinct particle sizes. tions as follows:
Gomadam et al. [11] developed a mathematical model simulating a
composite electrode made of carbon monofluoride (CFx) and silver 8
>
> f11 ðUi Þ ¼ capai
vanadium oxide (SVO) for medical applications. Although their >
>
>
>
>
> 1
model is relatively simplified with many assumptions, their < f 1 ðU Þ ¼ stoich ¼ capai ¼ f1 ðUi Þ
modeling result showed good agreement with the experimental 2 i i
capamax;i capamax;i
result from a small half cell test within moderate C-rate discharging >
>
>
>
condition. Albertus et al. [15e17] conducted more extensive >
> stoichi  stoich0;i f 1 ðUi Þ  stoich0;i
>
> 1 ¼ 2
: f3 ðUi Þ ¼ soci ¼ stoich
modeling work on blended electrodes. They modified their math- 100;i  stoich0;i stoich100;i  stoich0;i
ematical cell model to treat electrodes made of multiple types of
(2)
active particles, and simulated several composition of two types of
active materials (LiNi0.80Co0.15Al0.05O2 and LiMn2O4). They pre- Note that the maximum capacity(capamax), the minimum
sented that their blended electrode showed combined character- lithium-stoichiometry (stoich0), and the maximum lithium-stoi-
istics (such as power and energy) of each pure active material. Jung chiometry(stoich100) of the active material i are known values. In
and Kang [18] introduced a multi-dimensional model of lithium- order to get a full voltage profile of a blended-electrode with n-
ion cells with blended-electrodes. But, their model was an empir- constituting active materials, we calculate accumulated capacity,
ical model which could not be used to predict characteristics of a lithium stoichiometry, and SOC of each material while sweeping
lithium-ion cell without actual experiment data. equilibrium potential from Umax to Umin as follows:
186 S. Jung / Journal of Power Sources 264 (2014) 184e194

Fig. 1. Equilibrium voltage profile of pure active materials used for the present study.

P
n information of each material member from the model. Fig. 3(a)
capablend ¼ capai
i¼1
shows the contribution of each material member to the whole ca-
pacity production by the blended-electrode made of NMC (70%)
capai
capai ¼ xi capai ðUÞ where xi ¼ and LMO (30%). Total capacity of the blended-electrode is about
capablend (3) 143 mAh g1. NMC and LMO produces capacity together until the
stoichi ¼ stoichi ðUÞ net capacity reaches 65 mAh g1. However, after the net capacity
stoichi  stoich0;i reaches 65 mAh g1, LMO does not make any capacity longer while
soci ¼ NMC continues to produce more capacity until the net capacity
stoich100;i  stoich0;i
reaches 143 mAh g1. Fig. 3(b) shows different view of capacity
By applying the scheme explained above, equilibrium potentials production of this blended-electrode. While the blended-electrode
of blended-electrodes can be generated as shown in Fig. 2. Positive discharges from SOC 100%e0%, SOC of LMO component already
blended-electrode made of LiMn2O4 (briefly ‘LMO’ from now) and reaches 0% when SOC of the blended-electrode is around 45%.
Li(Ni1/3Mn1/3Co1/3)O2 (briefly ‘NMC’ from now) gets more capacity Instead, NMC supports capacity production to the last of the dis-
with lowered overall potential when the fraction of NMC increases. charging process. Note the SOC slope of NMC becomes stiffer after
Negative blended-electrodes made of graphite and soft carbon is SOC of LMO reaches 0% in Fig. 3(b) since NMC solely covers the
simulated as shown in Fig. 2(b). Soft carbon has smoothly whole capacity production from this point on. Lithium stoichiom-
decreasing profile whereas graphite has step-decreasing profile etry status of each active material member can be analyzed as
due to lithium staging inside crystal layers (see Fig. 1). Therefore, shown in Fig. 3(c). Starting from 0.3 at SOC 100%, lithium stoichi-
when they are blended together, mixed potential profile is gener- ometry of LMO becomes almost 1.0 when SOC of the blended-
ated. As the present blended-electrode model has information of electrode reaches 45%. In the case of NMC, lithium stoichiometry
each constituting material, it is possible to extract several useful changes from 0.45 to 1.0 when SOC of the blended-electrode moves
 S. Jung / Journal of Power Sources 264 (2014) 184e194 187

solved for each individual active particle. In addition, it is consid-

ered that those particles interact with each other to simulate
physical behaviors of a blended-electrode as shown in Fig. 4(b).

3.1. Formulation

As the present model follows the framework of Newman group’s

[12,13], this paper briefly reviews the governing equations. Fig. 4(a)
shows that there are three mathematical domains in the model. The
following electrolyte-phase species equation is solved in the whole
domain (anode, cathode, and separator).

  o
vð3 ce Þ v vce 1  tþ s
¼ Deff
e þ jLi where Deff
e ¼ 3 De (4)
vt vx vx F
The left hand side term is the transient term, and the right hand
side term consists of the diffusion term and the source term. Here,
Deff
e represents the effective lithium ion diffusivity of the
electrolyte.
The electrolyte-phase potential equation is solved in the entire
domain. Eq. (5) consists of the ohmic term and the ion-diffusion
term with a source term representing the reaction current den-
sity jLi. As the time scale of the potential is much smaller than that
of Li-ion species equation, the transient term is neglected in Eq. (5).

   
v vf v v
keff e þ keff ln c e ¼ jLi (5)
vx vx vx D vx
It is known that the ionic conductivity keff is a strong function of
the salt concentration, temperature, and material composition.
Following relationship is used for finding the ionic conductivity of
electrolyte composed of EC and EMC [19].

k ¼ 3:37115 þ 12:5608m  7:89593m2 þ 3:51922m3

 1:15471m4 þ 18:1863x  6:22756mx  13:6916m2 x
þ 8:43904m3 x  7:83732x2 þ 19:607mx2  18:4529m2 x2
 30:6369x3 þ 29:2mx3  0:0429918T þ 0:180877mT
 0:0836202m2 T þ 0:0230098m3 T þ 0:195946Tx
Fig. 2. Simulated equilibrium potentials of blended-electrodes according to blending
ratio: (a) positive electrode composed of NMC and LMO, (b) negative electrodes made þ 0:0676686mTx  0:14134m2 Tx þ 0:147429Tx2
of graphite and soft carbon.
þ 0:173059mTx2  0:51634Tx3  0:000223097T 2

from 100% to 0%. dQ/dV profile is widely used for analyzing capacity þ 0:000111233mT 2 þ 0:0000495286m2 T 2
production of an electrode according to voltage window. With the þ 0:000952777T 2 x þ 0:00117334mT 2 x
present blended-electrode model, one can separate dQ/dV of each
material member from the total dQ/dV profile as shown in Fig. 3(d).  0:000619157T 2 x2  3:46897  1027T 3  2:75041
It is easily identified that LMO produces capacity in the voltage  1026mT 3  5:57653  1026T 3 x
range of 3.8e4.2 V whereas NMC covers the whole voltage window.
(6)

3. Dynamic model Note that m is electrolyte concentration and x is fraction of EC in

the above equation.
Although the equilibrium model introduced above is useful for The effective ionic conductivity is calculated by considering the
quickly predicting voltage profile and analyzing capacity produc- porosity of the electrode as follows:
tion of a blended-electrode, it cannot simulate physical behavior of
keff ¼ sk (7)
a battery under dynamics operating conditions. By combining the 3

porous electrode theory [12e14] with the present blended-

keff
D is the diffusive conductivity defined as follows [20]:
electrode model, the present dynamic blended-electrode model is
developed. Fig. 4(a) shows the computation domain of the present  
2RT keff  o  d ln fþ
model. Finite volume mesh is constructed in x-direction including keff
D ¼ tþ  1 1 þ (8)
F d ln ce
negative electrode, separator, and positive electrode. Each finite
volume has multiple active particles constituting a blended- The solid-phase potential equation (or electron transport
electrode system. Spherical lithium intercalation equation is equation) which follows Ohm’s law is solved in both electrodes.
188 S. Jung / Journal of Power Sources 264 (2014) 184e194

Fig. 3. Simulated result of a blended-electrode made of NMC (70%) and LMO (30%).

  Therefore, solid-phase lithium concentration in the active parti-

v vf
seff s ¼ jLi (9) cle is acquired by solving a spherical diffusion equation as
vx vx follows:

where seff is the effective electronic conductivity defined as

seff ¼ (1  3 )ss. Again, the transient term is ignored in the above vcs;i v2 cs;i 2Ds;i vcs;i
equation where the electron transport is very fast compared to the ¼ Ds;i 2 þ (11)
vt vr r vr
Li-ion species transport.
As explained in the equilibrium model, open circuit potential The above spherical equation is solved for each particle member
(Ui) of each active material is defined as a function of lithium constituting the blended-electrode. Solid-phase lithium diffusivity
stoichiometry as follows: in active particles Ds is a function of temperature in this study. Eq.
(12) shows lithium intercalation speed will increase at high tem-
   perature, which may result in improved C-rate performance of a
stoichi ¼ cs cs;max i (10) cell.
Maximum lithium concentration cs,max is theoretically deter- ( !)
mined and lithium concentration on an active particle cs is Eact 1 1
Ds;i ¼ Dref
s;i $exp  (12)
calculated by solving a solid-phase lithium diffusion equation. In R Tref T
the present study, all active particles are assumed to be spherical.
 S. Jung / Journal of Power Sources 264 (2014) 184e194 189

Fig. 4. Schematics of the present dynamic blended-electrode model: (a) Mathematical domain of the quasi-2D blended-electrode model, (b) Inter-particular relationship among
three distinctly different active particles located in a single mathematical mesh domain.

    
Loading fraction ji of each active material in a blended- aa;i F ac;i F
electrode is expressed as
jLi
rxn;i ¼ as;i io;i exp h i  exp  hi
RT RT (17)
ua;i X where hi ¼ fs  fe  Ui
ji ¼ where ua ¼ ua;i (13)
ua i¼1 Exchange current density io,i which is a function of lithium
concentration in electrolyte phase and solid phase is expressed as
Average density of active materials is
follows [14]:
!1 h  a i
X
n
j io;i ¼ kðce Þaa cs;max  cs a ðcs Þac (18)
ra ¼ i
(14) i
r
i¼1 i
When a blended-electrode consists of two different active par-
Therefore, porosity of an electrode which has active material, ticles (binary system), two different active particles will face each
binder, and conductive agent can be calculated as follows: other and interactive reaction will occur to build equilibrium be-
      tween them. The driving force of this interactive reaction is voltage
ua ub uc difference between two active particles. If active particle 1 (i ¼ 1)
3 ¼ 1   (15)
ra L rb L rc L and active particle 2 (i ¼ 2) constitute a blended-electrode, the
interactive reaction current density can be defined as follows:
Specific interfacial area (as) which reflects the effective reaction
area of each active material is defined as follows: ( !)
  E 1 1
jLi
12 ¼ k12 as;12 io;12 DU12 $exp act 
8 A ¼ 4pr 2 R Tref T
>
>
i i
>
>
Ai < 4 3 where DU12 ¼ U1  U2 (19)
as;i ¼ Ni where Vi ¼ pri (16)
L > 3
>
> In Eq. (19), as,12 is specific surface area between active particles 1
: N ¼ ui
>
and 2, which is assumed to be the smaller value between two
i
r i Vi
particles. In the same way, exchange current density between two
Reaction kinetics at the interface between each active particle particles (io,12) is determined. k12 is reaction coefficient between
surface and electrolyte is assumed to follow the ButlereVolmer two particles, which is an assumed value in the present study since
equation as follows: it is difficult to measure this value. For each active particle,
190 S. Jung / Journal of Power Sources 264 (2014) 184e194

jLi Li Li In the same way, the present model can be extended to any
1 ¼ jrxn;1 þ j12
(20) blended electrode system with N-different active particles.
jLi Li Li
2 ¼ jrxn;2 þ j21 As the physical behavior of a cell is strongly affected by tem-
In each electrode, total reaction current is the sum of reaction perature, it is necessary to find the cell temperature, especially
current from each active material member. Therefore, the following under high C-rate case where it is difficult to maintain constant
description can be made. temperature. It is assumed that temperature in the cell is same
everywhere due to the small scale of the present model. Therefore,
X
jLi ¼ jLi Li Li the following lumped heat equation is coupled with the present
i ¼ j1 þ j2 (21)
i¼1 model.

This total transfer current density is plugged into the governing

equation Eqs. (4), (5), (9) and (11).
dT
When the blended-electrode is a ternary system which has mcp ¼ hAsurf ðT  Tamb Þ þ q_ (24)
three distinctly different active particles, interactive reaction cur- dt
rent densities among particles are Heat source term q_ consists of the followings [21].

!)
  nE 1 1
act
12 ¼ k12 as;12 io;12 DU12 $exp
jLi
R

Tref T
where DU12 ¼ U1  U2
!)
  nE 1 1
act
jLi
23 ¼ k23 as;23 io;23 DU23 $exp  where DU23 ¼ U2  U3 (22)
R Tref T
!)
  nE 1 1
act
jLi
31 ¼ k31 as;31 io;31 DU31 $exp  where DU31 ¼ U3  U1
R Tref T

P P

q_ þ ¼ ðas io Þi ðfs  fe  Ui Þ þ ðas io Þi T dU

dT
þ seff V2 fs þ keff V2 fe þ keff
D Vln ce Vfe
i i i

q_ s ¼ keff V2 fe þ keff D Vln ce Vfe

(25)
P P

q_  ¼ ðas io Þi ðfs  fe  Ui Þ þ ðas io Þi T dU

dT
þ seff V2 fs þ keff V2 fe þ keff
D Vln ce Vfe
i i i

Current density of the blended-electrode becomes Note each electrode has reaction heat, electronic ohmic heat,
8 ionic ohmic heat whereas separator has only ionic ohmic heat in
> Li Li Li Li
X < j1 ¼ jrxn;1 þ j12 þ j13
> the above expression.
jLi ¼ jLi Li Li Li
i ¼ j1 þ j2 þ j3 where j2 ¼ jrxn;2 þ j21 þ jLi
Li Li Li
Governing equations with boundary conditions for the present
>
>
23
i¼1 : jLi ¼ jLi Li Li model are summarized in Table 1. The model is incorporated into
3 rxn;3 þ j31 þ j32
FORTRAN and solved by finite volume method.
(23)

Table 1
Governing equations and boundary conditions of the Li-ion cell model.

Description Conservation equation Boundary condition

 
Charge (electrolyte phase) eff vfe þ v keff v ln c vfe  v fe 
vx k
v ¼ jLi
vx vx D vx e vx x¼0 ¼ vx x¼L ¼ 0

 
Charge (solid phase in electrode matrix) v seff vfs ¼ j Li v fs  v fs  I
vx vx seff  vx  ¼ seff
þ vx  ¼
x¼0 x¼L A
 
v fs  v fs 
seff  vx  ¼ seff
þ vx  ¼ 0
x¼Ln x¼Lnþs


vce 

vce 
Species (electrolyte phase) vð3 e ce Þ 1tþo Li
vx x¼0 ¼ vx x¼L ¼ 0
v vce
vt ¼ vx e vx þ F j
Deff
Species (solid phase in particles) vcs 2
¼ Ds vvrc2s þ 2D s vcs cs ¼ cs;0 at t ¼ 0
vt r vr
vcs
Ds ¼ 0 at r ¼ 0
vr
vcs jLi
Ds ¼ on particle surface
vr F
Energy mcp dT
dt
¼ hAsurf ðT  Tamb Þ þ q_
 S. Jung / Journal of Power Sources 264 (2014) 184e194 191

Table 2 3.2. Result and discussion

Simulation result of NMC/LMO blended-electrode according to the material
composition.
LMO and NMC are widely used active materials for Li-ion cells.
Total loading NMC loading LMO loading NMC Electrode LMO has advantage of fast lithium intercalation due to spinel
(mg cm2) (mg cm2) (mg cm2) fraction (%) capacity structure along with relatively high operational voltage, which
(mAh cm2)
enables a Li-ion cell to produce high power. For this reason, LMO
16.0 0.0 16.0 0 1.630 will be a good choice as positive material for HEV application. In
16.0 4.8 11.2 30 1.863
contrast, NMC has advantage of larger theoretical capacity
16.0 8.0 8.0 50 2.096
16.0 11.2 4.8 70 2.212 (277 mAh g1) compared to LMO (150 mAh g1) although lithium
16.0 16.0 0.0 100 2.445 intercalation rate is a bit lower compared to LMO. Therefore, it is
reasonable to think that a blended-electrode composed of both
NMC and LMO has intermediate characteristics between them. In
other words, it is possible to acquire both advantages from NMC
and LMO with a blended-electrode.
Discharging performance of NMC/LMO blended-electrodes is
simulated using the dynamic model according to five blending
cases as listed in Table 2. Calculation result shows net electrode
capacity linearly increases according to increment of NMC fraction.
This result is almost identical with the result of equilibrium model
(see Fig. 2(a)). Effect of blending ratio on C-rate capability is shown
in Fig. 5. It is clear that blending more LMO leads to improved C-rate
capability due to fast lithium diffusion in LMO particles, especially
under high C-rate condition over 7C. In addition, LMO helps a cell
produce more power or large current at low temperature.
In order to validate the present dynamic model, three different
full cells with different electrode compositions are built as listed in
Table 3. Baseline model parameters used for the simulation are
listed in Table 4. Three cells are simulated under the given test
conditions, and compared with experimental result as shown in
Fig. 6. By blending different active materials, each cell shows
distinctly different voltage profiles. Cell #1 and cell #2 have most of
their capacity production in the voltage window of 4.1e3.4 V
whereas cell #3 shows slowly decreasing voltage profile between
4.1 and 2.8 V. This result came from anode material selection
(graphite vs. amorphous carbon). As cell #1 and cell #2 share the
same anode, difference between them comes from the cathode

Table 4
Baseline model parameters.

Parameter Unit Value Note

Radius of NMC mm 4.5 Measured

Radius of LMO mm 7.5 Measured
Radius of graphite mm 7.5 Measured
Radius of soft carbon mm 3.8 Measured
Li-ion diffusivity of NMC cm2 s1 5.5  1010 Fitted
Li-ion diffusivity of LMO cm2 s1 2.5  109 Fitted
Li-ion diffusivity of graphite cm2 s1 4.5  1010 Fitted
Li-ion diffusivity of soft carbon cm2 s1 1.7  109 Fitted
Thickness of negative current mm 10.0 Measured
collector
Thickness of positive current mm 20.0 Measured
collector
Thickness of negative electrode mm 50.0 Measured
Thickness of positive electrode mm 55.0 Measured
Thickness of separator mm 25.0 Measured
Fig. 5. Simulated C-rate performance of NMC/LMO blended-electrode: (a) Tcell ¼ 298 K,
Porosity of negative electrode 0.28 Measured
(b) Tcell ¼ 273 K.
Porosity of positive electrode 0.25 Measured
Porosity of separator electrode 0.37 Measured
Ionic conductivity in electrolyte S cm1 Eq. (6) From ref. [6]
Table 3 Transference number 0.363
Active material composition of full-cells constructed for the present study. Faraday’s constant (F) C mol1 96,487
Li-ion diffusivity in electrolyte cm2 s1 2.5  106 From ref. [6]
Full-cell no. Cathode Anode Average concentration of Li-ion M 1.0
in electrolyte
NMC (%) LMO (%) Graphite (%) Carbon (%)
Matrix electronic conductivity S cm1 0.062 Measured
1 70.0 30.0 90.0 10.0 (positive)
2 50.0 50.0 90.0 10.0 Matrix electronic conductivity S cm1 0.1 Measured
3 30.0 70.0 5.0 95.0 (negative)
192 S. Jung / Journal of Power Sources 264 (2014) 184e194

side. Discharging simulation of 10C was successful with cell #2

which has large LMO fraction whereas cell #1 showed maximum
7C discharging performance with the present model. This result
confirms that using much amount of fast diffusion material in the
electrode is beneficial for getting high power from a lithium-ion
cell although there are other factors to control C-rate perfor-
mance of a cell such as electrode thickness. Cell #3 achieved very
high C-rate performance (20C) compared to other cells. This high
power performance was possible with the following three reasons.
First, the cathode consists of large amount of manganese spinel and
the most of active material fillings the anode is amorphous carbon
which has fast solid-phase lithium diffusion speed. Second, elec-
trode thickness of cell #3 is less thick (about 60%) compared to cell
#1 and #2. Third, both electrode of cell #3 is very porous compared
to the other cells, which means amount of active material loading is
low. Consequently, energy density of cell #3 is lower than the other
cells. In contrast, opposite approaches of the above three methods
should be considered when a cell with high energy density is
designed. High energy density cells can be applied to energy stor-
age systems (ESS) which does not require high power capability as
xEVs do. For automotive applications, cell #1 and #2 will be

Fig. 7. Li-stoichiometry profiles of each active material in full-cell #1 during 3C

Fig. 6. Experimental validation of Li-ion cells with blended-electrodes (full-cell #1, #2, #3). discharging.
 S. Jung / Journal of Power Sources 264 (2014) 184e194 193

appropriate for PHEV or EV, whereas cell #3 will be a good choice up lately (see Fig. 8(a)). On the other hand, the positive voltage
for HEV which require high power than large amount of energy. plateau of LPF shows up early and the discharging ends up with
It is possible to analyze physical behavior of each active material positive-limiting condition when the cell operates at low temper-
member in a cell with the present model. Fig. 7 show lithium ature (see Fig. 8(b)). In this simulation case, it was assumed that the
stoichiometry of each active material in cell #1 under discharging temperature dependency of lithium-ion diffusivity in NMC particle
condition. It is observed that lithium stoichiometry window of is stronger than that of graphite particle. Therefore, solid-phase
graphite is wider than that of carbon in the anode while lithium mass-transport in NMC particle is getting slower than that of
stoichiometry window of LMO is wider than that of NMC in the graphite particle at low temperature. As the lithium concentration
cathode. on the surface of NMC particle is quickly saturated due to slow
When a blended-electrode has several constituting active ma- lithium intercalation, operation of LPF starts up earlier under the
terials with quite different physical properties such as temperature low temperature condition than the mild operating condition. In
dependency of lithium-ion diffusivity, the voltage profile may other words, the length of span A and B strongly depends on the
significantly change according to the operating condition. Fig. 8 material properties of the constituting active particles, which can
presents simulated discharging profiles of a lithium-ion cell with be easily predicted with the present dynamic model.
a blended-positive electrode (NMC 85% and LPF 15%) and a negative Another simulation examples of lithium-ion cells with blended-
electrode (graphite 100%). Under the mild operating condition electrodes are shown in Fig. 9. According to the blending ratio and
(room temperature, 298 K), discharging ends up with negative- negative/positive ratio, numerous combinations of cells can be
limiting condition and the positive voltage plateau of LPF shows constructed. When material properties of each active material are
prepared, characteristics and physical behavior of those cells can be

Fig. 8. Simulated 1C discharging profiles of a Li-ion cells with positive electrode

composed of NMC 85%//LFP 15% and negative electrode composed of graphite 100%, (a)
T ¼ 298 K, (b) T ¼ 273 K. Fig. 9. Simulated result of various Li-ion cells with blended-electrodes.
194 S. Jung / Journal of Power Sources 264 (2014) 184e194

predicted with the present model. This will eventually expedite cell f potential (V)
development process which usually takes much time and effort. h surface overpotential of an electrode reaction (V)
k ionic conductivity (S cm1)
4. Conclusion kD diffusive conductivity of a species (A cm1)
r density
Two mathematical models of lithium-ion cells with blended- s electronic conductivity (S cm1)
electrodes are developed. First, the equilibrium model can simu- s Bruggeman tortuosity exponent
late any combination of active materials in a blended-electrode by a j loading fraction
quick and easy method. The equilibrium model enables us to u loading amount (g cm2)
decompose a blended-electrode into each active material, x capacity fraction
analyzing steady-state characteristics of each constituting member
such as voltage window, and lithium stoichiometry. Second, the
physics-based dynamic model can conduct simulation under Superscripts and subscripts
various operating conditions such as C-rate and temperature by a active material
solving physico-chemical governing equations. Simulation results b binder
show good agreement with the experiment data, confirming that amb ambient
the present model is useful for checking candidate active material c conductive agent
compositions at the early stage of battery development. diff diffusion
e electrolyte phase
Nomenclature eff effective
eq equilibrium
i ith member
Acronyms and abbreviations Li lithium
as active surface area per electrode unit volume (cm1) ref reference
A area (cm2) rxn reaction
c concentration of lithium (mol cm3) s solid phase
cp specific heat (J K1) surf surface
capa capacity
D diffusivity of lithium (cm2 s1)
Eact activation energy References
h heat transfer coefficient (W cm2 K1) [1] S.Y. Chung, J.T. Bloking, Y.M. Chiang, Nat. Mater. 1 (2002) 123.
i0 exchange current density (A cm2) [2] Y. Xia, M. Yoshio, H. Noguchi, Electrochem. Acta 52 (2006) 240.
I applied current (A) [3] C.R. Sides, F. Croce, V.Y. Young, C.R. Martin, B. Scrosatt, Electrochem. Solid
j volumetric current density (A cm3) State Lett. 8 (2005) A484.
[4] Z. Bakenov, I. Taniguchi, in: 216th Meeting of the Electrochemical Society,
k reaction coefficient 2009.
L thickness (cm) [5] T. Numata, C. Amemiya, T. Kumeuchi, M. Shirakata, M. Yonezawa, J. Power
m mass (g) Sources 97-98 (2001) 358.
[6] A. Manthiram, W. Choi, Electrochem. Solid State Lett. 10 (2007) A228.
N number of active particles [7] S.T. Myoung, M.H. Cho, H.T. Hong, T.H. Kang, C.S. Kim, J. Power Sources 146
r radial coordinate, radius (cm) (2005) 222.
R universal gas constant (8.314 J mol1 K1) [8] J.W. Fergus, J. Power Sources 195 (2010) 939.
[9] J.F. Whitacre, K. Zaghib, W.C. West, B.V. Ratnakumar, J. Power Sources 177
soc state of charge (2008) 528.
stoich lithium stoichiometry [10] R. Darling, J. Newman, J. Electrochem. Soc. 144 (1997) 4201.
t time (sec) [11] P.M. Gomadam, D.R. Merritt, E.R. Scott, C.L. Schmidt, P.M. Skarstad,
o J.W. Weidner, J. Power Sources 174 (2007) 628.
tþ transference number of lithium ion with respect to the [12] M. Doyle, T.F. Fuller, J. Newman, J. Electrochem. Soc. 140 (1993) 1526.
velocity of solvent [13] T.F. Fuller, M. Doyle, J. Newman, J. Electrochem. Soc. 141 (1994) 1.
T temperature (K) [14] W.B. Gu, C.Y. Wang, in: Lithium Batteries, PV99e25, 748, The Electrochemical
Society Proceeding Series, Pennington, NJ, 2000.
V volume (cm3)
[15] P. Albertus, J. Newman, in: 214th Meeting of the Electrochemical Society,
x thickness coordinate (cm) 2008.
U open circuit potential (V) [16] P. Albertus, J. Newman, in: 215th Meeting of the Electrochemical Society,
2009.
q_ heat source (W)
[17] P. Albertus, J. Christensen, J. Newman, J. Electrochem. Soc. 156 (2009) A606.
[18] S. Jung, D. Kang, J. Power Sources 248 (2014) 498.
Greek letters [19] M.S. Ding, K. Xu, S.S. Zhang, K. Amine, G.L. Henrikent, T.R. Jow, J. Electrochem.
aa, ac anodic and cathodic transfer coefficient for an electrode Soc. 148 (2001) A1196.
[20] J. Newman, K.E. Thomas-Alyea, Electrochemical Systems, third ed., Wiley
reaction Inter-Science Inc., Hoboken, NJ, 2004.
3 porosity [21] L. Rao, J. Newman, J. Electrochem. Soc. 144 (1997) 2697.