Journal of Power Sources 316 (2016) 215e223

Contents lists available at ScienceDirect

Journal of Power Sources

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

AC impedance analysis of ionic and electronic conductivities in

electrode mixture layers for an all-solid-state lithium-ion battery
Zyun Siroma a, *, Tomohiro Sato b, Tomonari Takeuchi a, Ryo Nagai b, Akira Ota b,
Tsutomu Ioroi a
a
Research Institute of Electrochemical Energy, National Institute of Advanced Industrial Science and Technology (AIST), Midorigaoka, Ikeda, Osaka 563-
8577, Japan
b
Lithium Ion Battery Technology and Evaluation Center (LIBTEC), Midorigaoka, Ikeda, Osaka 563-8577, Japan

h i g h l i g h t s g r a p h i c a l a b s t r a c t

Ionic and electronic effective con-

ductivities of electrode layers were
measured.

TLMs with electronic resistance were
used for the analysis of EIS.

An alternative way to use TLMs,
named “electron-electron connec-
tion”, is proposed.

The obtained ionic conductivities
were compatible with those by
traditional way.

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

Article history: The ionic and electronic effective conductivities of an electrode mixture layers for all-solid-state lithium-
Received 22 December 2015 ion batteries containing Li2SeP2S5 as a solid electrolyte were investigated by AC impedance measure-
Received in revised form ments and analysis using a transmission-line model (TLM). Samples containing graphite (graphite
26 February 2016
electrodes) or LiNi0.5Co0.2Mn0.3O2 (NCM electrodes) as the active material were measured under a
Accepted 17 March 2016
“substrate j sample j bulk electrolyte j sample j substrate” configuration (ion-electron connection) and a
“substrate j sample j substrate” configuration (electron-electron connection). Theoretically, if the elec-
tronic resistance is negligibly small, which is the case with our graphite electrodes, measurement with
Keywords:
Electrochemical impedance
the ion-electron connection should be effective for evaluating ionic conductivity. However, if the elec-
All-solid-state lithium-ion battery tronic resistance is comparable to the ionic resistance, which is the case with our NCM electrodes, the
Effective ionic conductivity results with the ion-electron connection may contain some inherent inaccuracy. In this report, we
Effective electronic conductivity theoretically and practically demonstrate the advantage of analyzing the results with the electron-
Transmission-line model electron connection, which gives both the ionic and electronic conductivities. The similarity of the
behavior of ionic conductivity with the graphite and NCM electrodes confirms the reliability of this
analysis.
© 2016 Elsevier B.V. All rights reserved.

1. Introduction

All-solid-state lithium-ion batteries are an important subject in

* Corresponding author.
the field of next-generation secondary batteries [1e5]. Their prac-
E-mail address: [email protected] (Z. Siroma). tical application presents challenges inherent in all-solid-state

http://dx.doi.org/10.1016/j.jpowsour.2016.03.059
0378-7753/© 2016 Elsevier B.V. All rights reserved.
216 Z. Siroma et al. / Journal of Power Sources 316 (2016) 215e223

conditions, such as contact between the active material and elec- of a cell with the configuration shown in Scheme 2. Instead, here
trolyte in the electrode mixture layers [6]. Contact among the we consider another possibility that accurate measurements of
grains of the solid electrolyte might also be a serious problem both the ionic and electronic conductivities are theoretically
compared to the situation in a normal battery which uses a liquid possible from the impedance of a more simple cell of
electrolyte. Insufficient contact causes a low effective ionic con-
ductivity in the electrode and inhomogeneous discharge in the substrate j sample j substrate; (3)
direction of thickness, and ends in a lower capacity than that ex-
pected based on the materials. For batteries with a liquid electro- which is the same configuration as that used for the DC resistivity
lyte, it is important to know the effective ionic conductivity in the measurements to estimate the electronic conductivity in this work.
electrode to understand and optimize the microstructure [7e9], This method is only effective for a sample with a relatively low
and this is even more the case for batteries with a solid electrolyte. electronic conductivity. This report presents our experimental re-
Ogihara et al. [10] demonstrated how electrochemical imped- sults with the “NCM electrodes” using the configuration shown in
ance spectroscopy could be used to obtain the effective ionic con- Scheme 3 and provides an interpretation using a TLM with elec-
ductivity in an electrode for lithium-ion batteries with a liquid tronic resistance.
electrolyte by using a transmission-line model (TLM) as an equiv-
alent circuit for the analysis. In this study, we tried to apply their 2. Theory
approach to electrodes containing a solid electrolyte. The TLM used
in this method, and which is also often used to analyze porous 2.1. TLMs without electronic resistance
electrodes [11e18], is based on the assumption that the electronic
conductivity is sufficiently high. This means that the model consists Fig. 1(a) shows a drawing of a TLM represented as a ladder
of two rows of the component elements: ionic resistance and network. The model is actually continuous with an infinite number
interface impedance. Under an additional assumption that the of infinitesimal steps. (To simplify the drawing, an array of infini-
interface impedance in this TLM without electronic resistance is tesimal elements is represented by a gray area in the following
purely capacitive, the real part of the total impedance of the elec- figures.) Two rows of elements labeled A and B represent the ionic
trode approaches one-third of the ionic resistance when the fre- resistance and interface impedance, with dimensions of [U cm] and
quency approaches zero. This characteristic is generally used to [U cm3], respectively. A general solution for the total impedance of
obtain the effective ionic conductivity [10,13,14]. this TLM without electronic resistance is given [15] as
In principle, the electrochemical impedance of an electrode is pffiffiffiffiffiffiffiffiffiffiffiffi
pffiffiffiffiffiffiffiffiffiffi
measured using a three-electrode configuration with a reference Z¼ zA zB coth L zA =zB ; (4)
electrode:
where L represents the thickness of the electrode. For the condition

counter electrode j bulk electrolyte ðwith reference electrodeÞ j sample j substrate (1)

in which the sample is an ideal polarized electrode and therefore

However, it is not always easy to introduce a reference electrode. the interface impedance is purely capacitive, a specified equation
Alternatively, a symmetric cell: can be obtained by substituting

1
substrate j sample j bulk electrolyte j sample j substrate (2) zA ¼ rion ¼ (5)
seff
ion
can also be used to avoid the necessity of a reference electrode to
obtain an impedance response without interference from the 1
zB ¼ (6)
counter electrode [10,19,20]. In this work, we measured our sam- ju cdl
ples containing graphite powder (hereafter named as “graphite
electrodes”) using this configuration, and adopted the traditional into Eq. (4), where cdl represents the double-layer capacitance per
TLM without electronic resistance for the analysis. unit volume, with a dimension of [F cm3]. As shown in Fig. 2(a), a
On the other hand, preliminary DC resistivity measurements Nyquist plot shows a 45 line in the high-frequency region and a
showed that our samples containing LiNi0.5Co0.2Mn0.3O2 powder perpendicular line in the low-frequency region. Mathematically,
(hereafter named as “NCM electrodes”) have relatively low elec- the position of the perpendicular line is given by
tronic conductivities. This means that the assumptions in the
traditional analysis are not satisfied. To cope with such samples, we 1 1
lim Z 0 ¼ r L ¼ Rion ; (7)
use a TLM with three element rows, i.e., ionic resistance, interface u/0 3 ion 3
impedance, and electronic resistance. Although mathematical so-
lutions for such a TLM with electronic resistance have been pre- where Z0 and Rion are the real part of Z and the total ionic resistance
sented [21e23], there have been few reports about the behaviors of in the direction of the thickness. Therefore, the “width” of this
their Nyquist plots depending on the ionic and electronic resistance feature in the Nyquist plot is used to obtain the effective ionic
[24]. Therefore, we examined the theoretical characteristics of conductivity [13,14]
TLMs with electronic resistance beforehand. Based on our numer- The interface impedance, especially for a battery electrode, is
ical calculations for some equivalent circuit models, we found that often not simply capacitive, which results in deviation from a
it might be possible, albeit with a problem concerning accuracy, to perpendicular line in the low-frequency region. Fig. 2(b) and (c)
obtain the ionic and electronic conductivities from the impedance show the behaviors when the double-layer capacitance is
 Z. Siroma et al. / Journal of Power Sources 316 (2016) 215e223 217

perpendicular line (or arc) region [10].

Here, the dimension of each infinitesimal element (zA and zB) in
TLMs is explained in detail: zA corresponds to the specific resistance
with a dimension of [U cm]. This is the inverse of the specific
conductivity s [S cm1]. The total in the thickness direction is ob-
tained by multiplying with the thickness L [cm], resulting in an
amount with a dimension of [U cm2], which is identical with that of
the whole TLM. (The dimension of an actual measurement value,
[U], is obtained by dividing by the geometric electrode surface area
S [cm2].) Oh the other hand, zB represents interfacial impedance per
unit volume with a dimension of [U cm3]. From a microscopic
viewpoint, this value is obtained by dividing the actual interfacial
impedance per unit surface area zIF [U cm2] by the volumetric
specific surface area s [cm2/cm3]. The total in the thickness direc-
tion is obtained by dividing by the thickness L [cm], resulting in an
amount with a dimension of [U cm2] again.
It should be noted that cdl (or c*dl) value in zB affects the fre-
quency regions, but does not affect the overall shape of the Nyquist
plot in any case shown in Fig. 2. Therefore, although possible values
for the double-layer capacitance might be varied in a wide range of
magnitude, findings described in this section are generally
applicable.

2.2. TLMs with electronic resistance

If the electronic resistance is large, the TLM shown in Fig. 1(b)

should be used. In contrast to the case of TLMs without electronic
resistance, it is challenging to extract the ionic conductivity from
the experimentally obtained Nyquist plot.
A theoretical solution for a TLM with electronic resistance is
given [21e23] as
Fig. 1. TLMs represented as ladder networks with (a) two rows, and (b) three rows of
elements with a traditional configuration of “ion-electron connection”, and with (c)  2  qffiffiffiffiffiffiffiffiffiffi
three rows of elements with “electron-electron connection”.
pffiffiffiffiffi zA þ z2C cosh L zAzþz C
þ 2zA zC
zA zC zB B
Z¼ Lþ $ qffiffiffiffiffiffiffiffiffiffi ;
zA þ zC 3
ðzA þ zC Þ2 sinh L zAzþzC
accompanied by charge-transfer resistance or is replaced by a B

constant-phase element (CPE), i.e., (10)

1 where zC is the value of the third row with a dimension of [U cm].

zB ¼ (8)
ju cdl þ r1ct Under the simple assumption that zA and zC are purely resistive and
zB is purely capacitive, substitution of Eqs. (5) and (6), and
for Fig. 2(b), and
1
zC ¼ re ¼ (11)
1 seff
e
zB ¼ (9)
ðjuÞj cdl
into Eq. (10) results in Fig. 3(a). In this case, two limit values of the
real part of the impedance are mathematically given [31] as
for Fig. 2(c) for use as zB in Eq. (4), where rct is the charge-transfer
resistance and j and cdl* are the exponent value and frequency- 1 1
independent parameter for the CPE, respectively. lim Z 0 ¼ ðr þ re ÞL ¼ ðRion þ Re Þ (12)
u/0 3 ion 3
The CPE is an empirical function introduced to describe the
behavior of distributed dielectric constant by Cole and Cole [25]. rion re Rion Re
Now it is widely used to describe non-ideal behavior of electric lim Z ¼ L¼ (13)
u/∞ rion þ re Rion þ Re
double layers, instead of a pure capacitance, in the EIS analysis of
various devices including electrochemical double layer capacitors A characteristic of this approach is that there is some non-zero
[26], lithium-ion batteries [27], and fuel cells [28]. Although some value of the high-frequency resistance (HFR), as shown in
potential origins of it, such as the roughness of the electrode j Fig. 3(a) and Eq. (13). Equations (12) and (13) show that, in prin-
electrolyte interface [29], have been proposed, its physical meaning ciple, the effective ionic and electronic resistance can be obtained
seems to be still uncertain. However, it is also true that adoption of from these two limit values. However, in an actual measurement,
this concept brings progress in the analysis of EIS. Similar to the some HFR value will always originate from the bulk electrolyte,
literature of the theory of TLMs [30], we also use CPE in our TLMs, as especially for the case of a symmetric cell as in Scheme 2. Even if a
discussed later in Results and Discussion. three-electrode cell as in Scheme 1 is used, an uncompensated
Fig. 2(b) and (c) demonstrate that the distortion from ideal resistance still remains unless the electrolyte has high conductivity.
behavior makes it difficult to use Eq. (7). However, if such a In such cases, it is difficult to use Eq. (13), since the parasitic
distortion is not significant, it is still possible to estimate Rion from resistance outside the TLM must be estimated exactly.
the transition point between the 45 line region and the Figure 3(a) shows that the “widths” of the curves (shown in the
218 Z. Siroma et al. / Journal of Power Sources 316 (2016) 215e223

Fig. 2. Nyquist plot of TLMs with electronic resistance with (a) capacitances, (b) parallel connections of resistance and capacitance, and (c) constant-phase elements as the second
row. Parameters are varied from a standard condition of L ¼ 0.5 mm, rion ¼ 10 kU cm, cdl ¼ 10 mF cm3, rct ¼ ∞, and j ¼ 1.

figure for the 5 kU cm case), which is the difference between two corresponds to a “Z-type connection” in Ref. [23]. This connection
values in Eqs. (12) and (13), are narrower than for the case of re ¼ 0. reflects an experimental condition with a three-electrode config-
Therefore, if the presence of electronic resistance is not considered, uration or a symmetric cell as in Scheme 2. Now, if the electronic
and therefore Eq. (7) is adopted, instead of Eqs. (12) and (13), to resistance is not zero, another type of connection becomes signif-
obtain rion from the “width” of a curve, an underestimation will icant, i.e., the impedance values between two terminals on both
result. In such a case, the ratio a of the obtained value to the actual sides of the electronic line as in Fig. 1(c). This type of connection
value is corresponds to a “T-type connection” in Ref. [23], and is hereafter
referred to as “electron-electron connection”. This connection re-
1 ðr þ re ÞL  rrionionþr
re
L flects an experimental condition between only two substrates as in
ion 1  K þ K2
a≡3 e
¼ ; (14) Scheme 3. An obvious advantage of this configuration is that the
1r
3 ion L
1þK ambiguity of the obtained HFR is minimized, as discussed below.
A theoretical solution for a TLM with electronic resistance with
where the electron-electron connection is given [23] as

re qffiffiffiffiffiffiffiffiffiffi
K¼ (15)
2 pffiffiffiffiffi zA þzC
rion z z 2z z B
cosh L zB  1
Zee ¼ A C Lþ C
$ q ffiffiffiffiffiffiffiffiffiffi (16)
This equation can be used as a criterion to evaluate whether or zA þ zC ðzA þ zC Þ2
3
sinh L zAzþz C
B
not the electronic resistance is small enough to use Eq. (7).
Both Eqs. (4) and (10) are impedance values between a terminal Impedance values for a TLM identical to Fig. 3(a), but which uses
of the ionic line on one side and another terminal of the electronic the electron-electron connection, are shown in Fig. 3(b). Although
line on the opposite side of each TLM. Hereafter, this type of it is self-evident that there is no value when re ¼ 0, a higher re value
connection is referred to as “ion-electron connection”, which gives a larger curve with a teardrop shape, i.e., a combination of a
 Z. Siroma et al. / Journal of Power Sources 316 (2016) 215e223 219

Fig. 3. Nyquist plot of TLMs with electronic resistance used for (a) the ion-electron connection and (b) the electron-electron connection, with L ¼ 0.5 mm, rion ¼ 10 kU cm, and
cdl ¼ 10 mF cm3, where re is varied in a range of 0e15 kU cm.

semicircle and a 45 line. Two limit values are mathematically due to the relatively high charge-transfer resistances of our elec-
obtained as trodes. The application of this analysis to an electrode with a low
charge-transfer resistance might require a further refinement of
lim Zee ¼ re L ¼ Re (17) the model described in this paper.
u/0 At any rate, even without the strict discussion presented above,
Fig. 4 demonstrates that the results obtained with the electron-
electron connection are less sensitive to deviation from an ideal
rion re Rion Re
lim Zee ¼ L¼ ; (18) condition. Here, it should be noted that the electron-electron
u/∞ rion þ re Rion þ Re
connection is effective when the electronic resistance is high,
which is opposite the case for the ion-electron connection. There-
where Eq. (18) is the same as Eq. (13). In this case, the low-
fore, the ion-electron connection and electron-electron connection
frequency limit value is also a real number, and is hereafter
can be complementary choices depending on the sample.
referred to as low-frequency resistance (LFR). Similar to the case of
In general, as for the ion-electron connection, the necessity to
Eqs. (12) and (13) for the ion-electron connection, the effective
consider a TLM depends on the relative magnitude of the ionic
ionic and electronic conductivities can be obtained from these two
resistance and/or the electronic resistance in comparison with the
limit values (HFR and LFR) for the electron-electron connection.
interfacial impedance. If the resistance is small, the effect is small
However, unlike with the ion-electron connection, measurements
and a simple equivalent circuit, such as the Randles-type model,
obtained with the electron-electron connection do not include a
can be used. On the other hand, under a condition that there is a
parasitic resistance that originates from the bulk electrolyte, and
certain electronic resistance, a certain value for LFR is guaranteed
therefore, they are easier to compare with theoretical estimations.
depending on it, and the ionic current induced in the TLM makes a
This is one of the greatest benefits of using the electron-electron
difference in LFR and HFR. This means that if the magnitude of the
connection.
LFR is experimentally detectable, some feature on the Nyquist plot
Similar to the situations shown in Fig. 2(b) and (c), deviation
originated from the TLM will surely be observed, unless the ionic
from a purely capacitive behavior of zB may also disturb the analysis
resistance is extremely large compared with the electronic
for the TLMs with electronic resistance. This is demonstrated in
resistance.
Fig. 4. Generally, less distortion is expected for the electron-
electron connection than for the ion-electron connection. As
shown in Fig. 4(d), the substitution of a CPE for a capacitance does 3. Experimental
not change the LFR, so Eqs. (17) and (18) are still applicable. Fig. 4(b)
demonstrates that the introduction of rct results in a small deviation As a Li2SeP2S5 solid electrolyte (SE), a 75Li2S,25P2S5 amorphous
of LFR. Strictly speaking, this is a model-derived error, and Eq. (17) glass powder (average particle size of 25.5 mm) was used. For the
is also essentially true in this case. The details are as follows. “graphite electrodes” and “NCM electrodes”, a graphite powder
A charge-transfer reaction means local charging/discharging of (average particle size of 20.5 mm) and a LiNi0.5Co0.2Mn0.3O2 powder
the active material. Since the capacity is finite, the reaction pro- (average particle size of 11.6 mm, covered by a LiNbO3 layer with a
duces a change in state-of-charge (SOC) and finally stops when DC thickness of ca. 5 nm), respectively, was mixed with the SE powder
(i.e., low frequency limit) is applied. This means that the LFR orig- in a mortar. The NCM powder was covered by niobium by spray
inates only from re. However, the present model does not reflect coating with an ethanol solution containing Nb(OEt)5 and LiOEt
this capacity limitation. To incorporate such processes in the model, using a tumbling fluidized bed granulating-coating machine, fol-
a finite-length Warburg impedance should be added to rct in series. lowed by heat treatment in an oxygen atmosphere at 400  C for
Theoretically, this produces another semicircle in the low- 30 min. The volume ratio of the active material to the SE was set in a
frequency region. However, such a detailed refinement of the range from 50:50 to 70:30, under the assumptions that the actual
model is beyond the scope of this paper, since our experimental densities of the graphite, NCM, and SE were 2.26, 4.72, and
results do not show such an additional semicircle. This is probably 1.90 g cm3, respectively.
220 Z. Siroma et al. / Journal of Power Sources 316 (2016) 215e223

Fig. 4. Nyquist plot of TLMs with electronic resistance with (a-b) parallel connections of resistance and capacitance, and (c-d) constant-phase elements as the second row, for (a,c)
the ion-electron connection, and (b,d) the electron-electron connection. Parameters are varied from a standard condition of L ¼ 0.5 mm, rion ¼ 10 kU cm, cdl ¼ 10 mF cm3, rct ¼ ∞,
j ¼ 1, and re ¼ 10 kU cm.

Table 1
DC conductivities of the electrode pellets. (DC conductivities of the graphite_60 and graphite_70 electrodes were not measured.)

Sample name Active material (vol %) SE (vol %) Thickness (mm) DC conductivity (S cm1)

graphite_50 50 50 586 9.5  101

NCM_50 50 50 392 3.2  105
NCM_60 60 40 342 9.7  105
NCM_70 70 30 331 1.8  104
NCM_100a 100 0 870 1.0  103
a
Not used for impedance measurements.

A setup shown in Scheme 3 was made for each electrode the current value observed after the application of a 10 mV DC
composition for the preliminary DC measurement and the imped- voltage for 10 s at room temperature. DC conductivity calculated
ance measurement with the electron-electron connection. For this based on the geometrical size of the pellet is also shown in Table 1.
setup, a sample pellet was made from ca. 100 mg of each mixture of A setup shown in Scheme 2 (a symmetric electrochemical cell) was
the electrode by pressing at 30 MPa for 1 min. Each pellet measured also made for each sample for the impedance measurements with
1 cm in diameter, and had a typical thicknesses of ca. 400 mm. the ion-electron connection. For this setup, lamination of the cen-
Details of the samples are summarized in Table 1. Each pellet was tral SE layer and two adjoining electrode layers was achieved by
set between two SUS plates as current collectors. Before the mea- sequential pressing at 10 MPa, followed by a final pressing at
surements, the setup (the sample and two SUS plates) was again 30 MPa for 1 min between two SUS plates. The thickness of the SE
subjected to ca. 30 MPa of pressure. DC resistance, which was used layer and the electrode layers was typically ca. 700 mm and 70 mm,
as a rough estimate of the electronic resistance, was obtained from respectively.
 Z. Siroma et al. / Journal of Power Sources 316 (2016) 215e223 221

The impedance of each cell was measured with an AC amplitude

of 10 mV in a frequency range of 10 mHze1 MHz, using a combi-
nation of a potentiostat (Solartron, SI 1287) and a gain-phase
analyzer (Solartron, SI 1260). The frequency was stepped logarith-
mically with 20 points per decade. The temperature of the cell was
set at 25  C. Parameter-fitting for the obtained impedance values
was performed using Microsoft Excel® Solver as follows: Based on
the assumed model, expected impedance value at each frequency
was calculated and compared with the experimental data. To
minimize the sum total of the squares of the relative deviations, all
parameters in the model (four or seven, depends on the model) are
fitted with the aid of Solver tool. In addition to the electrode
samples, the HFR of a pellet made from only SE was also measured
to obtain the ionic conductivity of the bulk SE.

4. Results and discussion

4.1. Graphite electrodes with the ion-electron connection

Fig. 6. The effective ionic conductivities of (a) the graphite electrodes and (b) the NCM
As shown in Table 1, DC resistivity measurements indicated that electrodes, and (c) the effective electronic conductivities of the NCM electrodes, rep-
our graphite electrodes had relatively high electronic conductiv- resented as functions of the volume ratio of the active material. The value for the 0%
sample (marked with *) is calculated from a HFR of a pellet made from only SE.
ities. These values can be assumed to be high compared to the ionic
conductivity, since the ionic conductivity of even the bulk SE was
found to be 5.5  104 S cm1. Therefore, we adopted the TLM becomes clear. Among the carbon electrodes tested, the highest
without electronic resistance (Fig. 1(a)) to interpret the impedance value of the ionic conductivity was 5.7  105 S cm1. Therefore, the
results. Fig. 5 shows the impedance of the graphte_50 sample ratio (K in Eqs. (14) and (15)) does not exceed 0.057, which corre-
measured with the configuration shown in Scheme 2. Except for sponds to an a of 0.90. This means that the error that originated
HFR that originated from the bulk electrolyte, the Nyquist plot is from ignoring of the electronic resistance is less than 10%.
qualitatively identical to the theoretical expectation shown in
Fig. 2(b) or (c). According to the theory, the real part of the inter-
section of two asymptotic lines can be used to estimate the effective 4.2. NCM electrodes with the electron-electron connection
ionic conductivity by Eq. (7), after subtraction of the HFR. In the
same manner, the ionic conductivity of each graphite electrode was In contrast to the case with graphite electrodes, NCM electrodes
calculated, and is represented as a function of the volume ratio of have low electronic conductivities (Table 1). Since the conditions
the active material (graphite) by line (a) in Fig. 6. might be beyond the limitation of the TLM without electronic
At this stage, the discussion of the effect of the electronic resistance, here we try to use the TLM with electronic resistance.
resistance on the accuracy of the obtained ionic conductivity

Fig. 7. Nyquist plot of the NCM_50 electrode measured with the configurations in (a)
Fig. 5. Nyquist plot of the graphite_50 electrode measured with the configuration in Scheme 2 (ion-electron connection of two electrodes with a thickness of 49 mm,
Scheme 2 (ion-electron connection). Since the cell is symmetric with two electrodes, divided by two to show the contribution of one electrode) and (b) Scheme 3 (electron-
the values are divided by two. electron connection) of an electrode with a thickness of 392 mm.
222 Z. Siroma et al. / Journal of Power Sources 316 (2016) 215e223

Fig. 7 shows the impedance of the NCM_50 sample. Line (a) shows it seems to be quite possible that the major part of the electronic
the results obtained with the ion-electron connection. Based on the resistance is due to the contact resistance between particles. Fitted
theoretical expectation shown in Fig. 4(a) or (c), it might be infor- results based on this model are shown in Fig. 8(b), where
mative if we could detect the transition point between the 45 line rion ¼ 13510 U cm, re,bulk ¼ 860 U cm, re,gb ¼ 31230 U cm. The other
of the high-frequency region and the perpendicular line (or its possibility is that the ionic path has a structure. Fitted results are
distortion) in the low frequency-region. Unfortunately, however, shown in Fig. 8(c), where rion, bulk ¼ 350 U cm, rion,gb ¼ 10720 U cm,
the experimental results are rather featureless. On the other hand, re ¼ 31530 U cm. In the same manner, the results with all the other
the results obtained with the electron-electron connection (Line (b) NCM electrodes were analyzed (See Supplementary information,
in Fig. 7) have characteristic features of two semicircles. At this Fig. S1 and Table S1). Notably, the total of the electronic or ionic
stage, we tentatively supposed that the semicircle of the lower resistance is almost the same among the three models. The dis-
frequency corresponds to the teardrop shape of the TLM with the crepancies among the models are at most 2% and 22% for electronic
electron-electron connection as described in Fig. 3(b) or Fig. 4(b, d). and ionic conductivity, respectively. Therefore, since the aim of the
Based on this assumption, the equivalent circuit shown in the insert analysis was to estimate of the ionic and electronic conductivity,
in Fig. 4(d) was used to explain the semicircle of the lower fre- the origin of the semicircle of the higher frequency can be left
quency, and the parameters were fitted with data points in the undetermined. Lines (b) and (c) in Fig. 6 are the ionic and electronic
range below the top of this semicircle (0.01 Hze0.35 Hz). The fitted conductivities obtained with the model shown in Fig. 8(a). As
result is shown in Fig. 8(a), where the obtained parameters are: shown, the carbon electrodes and NCM electrodes exhibit a similar
rion ¼ 11230 U cm, re ¼ 31490 U cm, jcpe ¼ 0.84, and dependency of the ionic conductivity on the SE volume ratio. This
cpedl ¼ 0.0914 F cm3 s0.16. The validity of this assumption strongly supports the validity of the assumption that the semicircle
regarding the assignment of the lower semicircle was left open at of the lower frequency originates from the TLM. In addition, the
this stage, and we went on to find explanations for the other behavior of the electronic conductivity is also reasonable, i.e., a
semicircles seen in the higher frequency region. Here we offer two higher SE ratio shows a lower electronic conductivity.
possible explanations. One possibility is that some of the electronic Although the validity of the assignment of the semicircle of the
path is accompanied by capacitance in parallel. A possible example lower frequency has been improved, the origin of the semicircle of
of such a situation is a grain boundary or contacts between parti- the higher frequency is not fixed. The models shown in Fig. 8(b) and
cles. Since the samples were made just by pressing of the powders, (c) are quite different, but the calculated results are almost

Fig. 8. Fitting results of the NCM_50 electrode measured with the electron-electron connection with TLMs with electronic resistance with (a) no structure, and structures of (b) the
electronic path or (c) the ionic path. Discrete points indicate experimental data shown in Fig. 7(b).
 Z. Siroma et al. / Journal of Power Sources 316 (2016) 215e223 223

identical. Various combinations of these tow models are also po- uncertainty is not fatal for the conductivity measurements, as long
tential candidates. Calculation results for two models identical to as the shape of the semicircle that originates from the TLM remains
those in Fig. 8(b) and (c), but with the ion-electron connection were clear.
also examined (See Supplementary information, Fig. S2). These
results are qualitatively similar to the featureless results shown in Acknowledgement
Fig. 7(a), which do not deny the validity of the models. However,
again, these two curves are similar to each other, and thus unlikely This work was partly supported by “Leading-edge, Innovative
to contribute to determining which model is correct. These calcu- Battery Material Evaluation Technology Development” project from
lations suggest that a judgment based only on impedance mea- the New Energy and Industrial Technology Department Organiza-
surements is limited, and some other experiment will be required tion (NEDO) in Japan.
to interpret the impedance results completely.
In this work, each NCM electrode was used as prepared, i.e., with
Appendix A. Supplementary data
the lowest SOC. Therefore, the charge-transfer resistance should be
high, and the analysis is expected to be relatively easy. However, as
Supplementary data related to this article can be found at http://
discussed for Figs. 2e4, a lower charge-transfer resistance makes
dx.doi.org/10.1016/j.jpowsour.2016.03.059.
the analysis more difficult. The model might need to be refined
further for the analysis during the charge-discharge process.
References
5. Conclusions
[1] K. Kanehori, K. Matsumoto, K. Miyauchi, T. Kudo, Solid State Ion. 9e10 (1983)
1445.
When the electronic conductivity in an electrode mixture layer [2] J.R. Akridge, H. Vourlis, Solid State Ion. 18e19 (1986) 1082.
is not sufficiently high, a TLM with electronic resistance is required [3] H. Ohtsuka, S. Okada, J. Yamaki, Solid State Ion. 40e41 (1990) 964.
[4] K. Tanaka, N. Aotani, K. Iwamoto, S. Kondo, Solid State Ion. 86e88 (1996) 877.
to analyze the impedance response. A theoretical consideration was [5] T. Minami, A. Hayashi, M. Tatsumisago, Solid State Ion. 136e137 (2000) 1015.
proposed based on the behaviors of TLMs with electronic resis- [6] T. Takeuchi, H. Kageyama, K. Nakanishi, T. Ohta, A. Sakuda, T. Sakai,
tance, not only with the traditional ion-electron connection but H. Kobayashi, H. Sakaebe, K. Tatsumi, Z. Ogumi, Solid State Ion. 262 (2014)
138.
also with the electron-electron connection. This approach revealed [7] A.M. Colclaure, R.J. Kee, Electrochim. Acta 55 (2010) 8960.
that the results with the electron-electron connection are infor- [8] J. Liu, M. Kunz, K. Chen, N. Tamura, T.J. Richardson, Phys. Chem. Lett. 1 (2010)
mative, especially for electrodes with relatively high electronic 2120.
[9] D.E. Stephenson, B.C. Walker, C.B. Skelton, E.P. Gorzkowski, D.J. Rowenhorst,
resistances, as can be seen in Fig. 3(b). A teardrop shape is char- D.R. Wheeler, J. Electrochem. Soc. 158 (2011) A781.
acteristic of the behavior of the electron-electron connection, for [10] N. Ogihara, S. Kawauchi, C. Okuda, Y. Itou, Y. Takeuchi, Y. Ukyo, J. Electrochem.
which the HFR and LFR can be used to obtain the ionic and elec- Soc. 159 (2012) A1034.
[11] R. De Levie, Electrochim. Acta 8 (1963) 751.
tronic conductivities by Eqs. (17) and (18).
[12] A. Lasia, J. Electroanal. Chem. 397 (1995) 27.
Adoption of the measurements with the ion-electron connec- [13] M.C. Lefebvre, R.B. Martin, P.G. Pickup, Electrochem. Solid-State Lett. 2 (1999)
tion should be sufficient to obtain the ionic conductivity in our 259.
graphite electrodes, since the electronic resistance is negligibly [14] R. Jiang, C.K. Mittelsteadt, C.S. Gittleman, J. Electrochem. Soc. 156 (2009)
B1440.
small and the results can be interpreted using a TLM without [15] J. Bisquert, G. Garcia-Belmonte, F. Fabregat-Santiago, N.S. Ferriols,
electronic resistance. However, it is difficult to apply the same P. Bogdanoff, E.C. Pereira, J. Phys. Chem. B 104 (2000) 2287.
analysis to our NCM electrodes since the electronic conductivity is [16] H.K. Song, J.H. Sung, Y.H. Jung, K.H. Lee, L.H. Dao, M.H. Kim, H.N. Kim,
J. Electrochem. Soc. 151 (2004) E102.
not high. In contrast, measurements with the electron-electron [17] S. Yoon, J.H. Jang, B.H. Ka, S.M. Oh, Electrochim. Acta 50 (2005) 2255.
connection and analysis by a TLM with electronic resistance are [18] Y. Liu, M.W. Murphy, D.R. Baker, W. Gu, C. Ji, J. Jorne, H.A. Gasteiger,
effective. Although two semicircles were observed, an analysis J. Electrochem. Soc. 156 (2009) B970.
[19] C.H. Chen, J. Liu, K. Amine, J. Power Sources 96 (2001) 321.
under the assumption that the semicircle of the lower frequency [20] K. Nakura, K. Ariyoshi, F. Ogaki, K. Takaoka, T. Ohzuku, J. Electrochem. Soc. 161
originates from the TLM was performed and the obtained ionic (2014) A841.
conductivity was compatible with our graphite electrodes. The [21] J. Bisquert, G. Garcia-Belmonte, F. Fabregat-Santiago, A. Compte, Electrochem.
Commun. 1 (1999) 429.
obtained electronic conductivity was also reasonable. These results €ltzsch, O. Kanoun, Electrochim. Acta 75 (2012) 347.
[22] U. Tro
strongly suggest that the ionic and electronic conductivities of an [23] Z. Siroma, N. Fujiwara, S. Yamazaki, M. Asahi, T. Nagai, T. Ioroi, Electrochim.
electrode mixture layer with a relatively low electronic conduc- Acta 160 (2015) 313.
[24] Z. Siroma, T. Ioroi, Electrochemistry 83 (2015) 425.
tivity can be obtained by the measurement with the electron-
[25] K.S. Cole, R.H. Cole, J. Chem. Phys. 9 (1941) 341.
electron connection. [26] J. Kang, J. Wen, S.H. Jayaram, A. Yu, X. Wang, Electrochim. Acta 115 (2014) 587.
The semicircle of the higher frequency is hypothetically [27] C.H. Chen, J. Liu, K. Amine, J. Power Sources 96 (2001) 321.
assigned to the grain boundaries of the ion path or electron path. [28] S. Cruz-Manzo, P. Rama, R. Chen, J. Electrochem. Soc. 157 (2010) B1865.
[29] W. Schelder, J. Phys. Chem. 79 (1975) 127.
Since each hypotheses almost explains the results, the attribution [30] J. Bisquert, Phys. Chem. Chem. Phys. 2 (2000) 4185.
of this semicircle will be left to future studies. Nevertheless, this [31] X. Ren, P.G. Pickup, J. Electroanal. Chem. 420 (1997) 251.