Journal of The

Electrochemical Society

OPEN ACCESS You may also like

- Nonlinear Electrochemical Impedance
Extending Newman's Pseudo-Two-Dimensional Spectroscopy of Lithium-Ion Batteries:
Experimental Approach, Analysis, and
Lithium-Ion Battery Impedance Simulation Initial Findings
Matthew D. Murbach, Victor W. Hu and
Daniel T. Schwartz
Approach to Include the Nonlinear Harmonic
- Second-Harmonic Nonlinear
Response Electrochemical Impedance Spectroscopy:
Part I. Analytical Theory and Equivalent
Circuit Representations for Planar and
Porous Electrodes
To cite this article: Matthew D. Murbach and Daniel T. Schwartz 2017 J. Electrochem. Soc. 164 Yuefan Ji and Daniel T. Schwartz
E3311
- Nonlinear Electrochemical Impedance
Spectroscopy for Lithium-Ion Battery
Model Parameterization
Toby L. Kirk, Adam Lewis-Douglas, David
Howey et al.
View the article online for updates and enhancements.

This content was downloaded from IP address 103.118.50.122 on 26/03/2025 at 11:25

 Journal of The Electrochemical Society, 164 (11) E3311-E3320 (2017) E3311

JES FOCUS ISSUE ON MATHEMATICAL MODELING OF ELECTROCHEMICAL SYSTEMS AT MULTIPLE SCALES IN HONOR OF JOHN NEWMAN
Extending Newman’s Pseudo-Two-Dimensional Lithium-Ion
Battery Impedance Simulation Approach to Include the Nonlinear
Harmonic Response
Matthew D. Murbach∗ and Daniel T. Schwartz∗∗,z
Department of Chemical Engineering and Clean Energy Institute, University of Washington, Seattle, Washington
98105-1750, USA

The pseudo-two-dimensional (P2D) model of lithium-ion batteries couples a volume-averaged treatment of transport, reaction,
and thermodynamics to solid-state lithium diffusion in electrode particles. Here we harness the linear and nonlinear physics of
the P2D model to evaluate the fundamental (linear) and higher harmonic (nonlinear) response of a LiCoO2 |LiC6 cell subject to
moderate-amplitude sinusoidal current modulations. An analytic-numeric approach allows the evaluation of the linearized frequency
dispersion function that represents electrochemical impedance spectroscopy (EIS) and the higher harmonic dispersion functions we
call nonlinear electrochemical impedance spectroscopy (NLEIS). Base case simulations show, for the first time, the full spectrum
second and third harmonic NLEIS response. The effect of kinetic, mass-transport, and thermodynamic parameters are explored.
The nonlinear interactions that drive the harmonic response break some of the degeneracy found in linearized models. We show
that the second harmonic is sensitive to the symmetry of the charge transfer reactions in the electrodes, whereas EIS is not. At low
frequencies, NLEIS probes aspects of the cell thermodynamics that are not accessible with EIS. In short, NLEIS has the potential
to increase the number of physicochemical parameters that can be assessed in experiments similar in complexity to standard EIS
measurements.
© The Author(s) 2017. Published by ECS. This is an open access article distributed under the terms of the Creative Commons
Attribution 4.0 License (CC BY, http://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse of the work in any
medium, provided the original work is properly cited. [DOI: 10.1149/2.0301711jes] All rights reserved.

Manuscript submitted April 3, 2017; revised manuscript received May 16, 2017. Published June 2, 2017. This paper is part of the
JES Focus Issue on Mathematical Modeling of Electrochemical Systems at Multiple Scales in Honor of John Newman.

Electrochemical impedance spectroscopy (EIS) is a commonly lenge of maintaining a stationary state during EIS measurements in a
utilized tool for the noninvasive analysis of a wide range of electro- sealed cell.
chemical systems.1,2 Using a sufficiently small perturbation ampli- An approach that breaks some of the degeneracy of linearized
tude simplifies the interpretation of the impedance response because EIS without violating the requirement for stationarity is the use of
the system under analysis can be approximated as linear. The con- moderate-amplitude current modulations around a zero-mean current.
dition of linearity and ease of fitting equivalent circuit analogs has Moderate-amplitude modulations are used to drive a weakly non-
led to the widespread adoption of EIS for both characterization3–5 linear response, as evidenced by higher harmonics in the frequency
and prognostic6–8 applications within lithium-ion battery research. In response of the system. Typically, higher harmonics in impedance
particular, EIS has been used to garner insight into important physico- experiments have been treated as noise, to be eliminated as a way of
chemical processes including the growth of the solid electrolyte inter- ensuring the linearity of EIS measurements.18–21 However, measuring
phase (SEI) layer,9 mass transfer and kinetics in electrode materials,10 the nonlinear harmonic response of an electrochemical system has
as well as other degradative and capacity loss mechanisms.11 The been used to study corrosion,22–24 fuel cells,12,25,26 and mass transport
challenge with EIS, however, is that the analysis is restricted to prob- controlled hydrodynamic systems27–29 as well as in chemical sens-
ing the linear regime, despite the inherently nonlinear nature of the ing applications.30–32 The increased informational power of nonlinear
electrochemical system. Fundamentally, linearization reduces the in- EIS (NLEIS) measurements enabled Wilson et al. to distinguish be-
formation content of a given EIS measurement. The consequence of tween physical models which were degenerate in the linear regime,12
linearization is model degeneracy, in that, one EIS dataset can be while Nakata et al. used the sensitivity of the nonlinear response to
represented equally well by different physical models12 and different distinguish between different chemical targets for improved sensor
circuit analogs,13 because of the loss of information inherent with selectivity.31
linearization. This degeneracy and loss of information may be an es- Extending EIS to probe the full-spectrum nonlinear response re-
pecially important limitation for the study of lithium-ion batteries if quires a more sophisticated approach than fitting lumped-parameter
we hope to carefully dissect degradative interfacial processes.14,15 equivalent circuit analogs. Physics-based models of the full electro-
The issue of linear EIS model degeneracy and information loss is chemical system offer the flexibility and interpretability required for
particularly challenging for sealed batteries. Analytical solutions for a meaningful analysis. Here we extend the P2D impedance model
the pseudo-two-dimensional (P2D) model impedance response under developed by Newman et al.33,34 to include the linear and nonlinearly-
stationary conditions show that linearized EIS is unable to uniquely de- generated higher order harmonic response generated by a lithium-ion
termine many of the model parameters needed for simulating charge- battery system. As with EIS measurements, NLEIS is a “local” probe
discharge curves.16,17 For example, the anodic and cathodic transfer of the battery around a single state of charge, so EIS and NLEIS
coefficients, αa and αc , respectively, cannot be uniquely determined; experiments will typically involve a sequence of measurements at
only the sum (αa + αc ) appears in the linearized solution. Changes to various open circuit potentials (states of charge) to parameterize the
the battery potential (state of charge) do not alter the inaccessibility full battery behavior. Later papers show that the extra information
of unique transfer coefficients. Alternative ways to measure unique available in an NLEIS experiment comes at a modest increase in
transfer coefficients using EIS, such as passing a mean current that experimental difficulty. To the best of our knowledge, full-spectrum
drives the electrode kinetics into the Tafel regime, introduce the chal- nonlinear EIS has not been applied to the study of lithium-ion batter-
ies. Yet, the ability to distinguish between different physical models,
noninvasively probe nonlinear internal states, and investigate the sym-
∗ Electrochemical Society Student Member. metry and reversibility of degradative reactions make NLEIS a can-
∗∗ Electrochemical Society Fellow. didate for providing information-rich insights that can move the field
z
E-mail: [email protected] forward.
E3312 Journal of The Electrochemical Society, 164 (11) E3311-E3320 (2017)

The Mathematical Model Current only flows in the solution- and solid-phases at
the electrode-current collector and separator-electrode interfaces,
Pseudo-two-dimensional battery impedance model.—The widely
respectively,
adopted P2D battery model has been a powerful tool for
understanding,35 optimizing,36,37 and controlling38 lithium-ion bat- ∂φs
teries for many decades. The P2D model is a physics-based repre- i2 = 0 and − σe f f = −I at x = 0 and x = L , [11-12]
∂x
sentation of the coupled nonlinear mass transport, thermodynamic,
and reaction processes inside a lithium-ion battery. Porous electrode ∂φs
theory treatment of the electrodes requires volume averaging over i2 = −I and − σe f f = 0 at x = lneg and x = lneg + lsep
the particles and electrolyte pores leading to two macrohomogeneous ∂x
[13-14]
solid- and solution-phase continua superimposed in the x direction and the solution-phase potential is set to zero (φ = 0) at the negative
across the cell sandwich thickness. The x direction equations are cou- current collector as a reference. The equations in the separator are
pled to solid-phase diffusion of lithium inside the electrode particles, given by Equations 6–8 with the interfacial flux set to zero ( j = 0).
governed by, Continuity of the fluxes across the electrode/separator interfaces is
 
∂cs 1 ∂ ∂cs maintained. In all cases, effective transport coefficients (diffusivities,
= 2 Ds r 2 [1] conductivities) are modifications of the intrinsic transport coefficients
∂t r ∂r ∂r
using a Bruggeman-type tortuosity factor to account for the porous
where r is the radial distance from the center of the particle. Here we character of the medium. For example, De f f = D Br ugg for effective
take the solid-phase diffusivity, Ds , to be independent of concentration diffusivity, where D is the intrinsic diffusivity, and Br ugg is the
in the solid, cs . Bruggeman coefficient.
The concentration of lithium in the electrode particles is coupled
to the interfacial flux from faradaic charge transfer by the boundary Frequency domain formulation.—The governing equations de-
condition at the solid-solution interface, r = R p , scribed above are both coupled and nonlinear. If we assume the
∂cs system of equations is stable to finite amplitude current perturba-
− Ds= jf [2] tions, then nonlinear theory suggests a pure single frequency sinu-
∂r
soidal input perturbation will generate an output response with spec-
where we assume Butler-Volmer kinetics, tral components at the excitation frequency (fundamental or 1st har-
    
i0 αa F −αc F monic) and integer higher harmonics (2nd , 3rd , and 4th harmonic of
jf = exp η − exp η . [3] the fundamental frequency). We have shown previously that a fast-
F RT RT
computing analytic-numeric method can be used for analyzing the
The total interfacial flux, j, also has a non-faradaic Helmholtz linear and nonlinear response of many different electrochemical and
double-layer capacitive contribution, jdl , transport-reaction systems when they are stable to finite amplitude
j = j f + jdl [4] perturbations.12,28,29,39,40 Here we apply these methods to the analysis
of the nonlinear impedance in a lithium-ion battery system using the
where above P2D equations. The key step in the fast analytic-numeric formu-
Cdl ∂ (φs − φ) lation is to expand all dependent variables in the governing equations
jdl = . [5] and boundary condition as a complex Fourier series
F ∂t
∞
Here we follow the convention of Doyle et al.33 for a simple model
of interfacial processes where the double-layer capacitance, Cdl , is X (x, t; I, ω1 ) = Re X̃ k (x; I, ω1 ) exp ( jkω1 t) [15]
dominated by cation rearrangement, and both Cdl and the exchange- k=0

current density, i 0 , are independent of concentration. where X denotes any of the dependent variables (cs , c, i2 , φs , and φ),
The current density in the electrolyte, i2 , is related to interfacial X̃ k are the complex Fourier coefficients, X̃ k = X̃ k − j X̃ k , for the
processes by Faraday’s law, respective dependent variables, x is the vector of spatial independent
∂i2 variables (x, r), t is time, I is the current modulation amplitude, ω1
= aF j [6] is the fundamental frequency of the perturbation, and k is an integer
∂x
that indicates the harmonic number. The notation of X (a; b, c, . . .) is
where a is the specific interfacial area of the electrode. Binary elec- used to show X is the dependent variable, a is an independent variable,
trolyte theory describes the concentration in the electrolyte, c, and b, c, . . . are parameters.
 
∂c ∂ ∂c   For the weakly nonlinear, time-periodic (stationary) system studied
 = De f f + a 1 − t+0 j [7] here, the complex Fourier coefficients can be expressed by a Volterra
∂t ∂x ∂x
series expansion in powers of the current modulation amplitude,
and solution-phase potential, φ, ∞
∂φ 2κe f f RT   ∂ ln c X̃ k (x; I, ω1 ) = I k+2r X̃ k,k+2r (x; ω1 ) . [16]
− κe f f = i2 − 1 − t+0 [8]
∂x F ∂x r =0

where  is the void fraction, De f f is the effective diffusivity, t+0 is the The resulting double-indexed complex coefficients X̃ k, p (x; ω1 )
transference number of Li+ in the electrolyte, and κe f f is the effective within the summation of Equation 16 are the key frequency disper-
solution-phase conductivity. The potential in the solid, φs , is governed sion coefficients for the Fourier coefficients at harmonic number, k,
by Ohms’ law, and order of nonlinear coupling, p. When k = p = 1, the governing
∂φs equations and boundary conditions represent the linear response of
− σe f f = −I − i2 [9] the system to infinitesimal current modulations (e.g., this is the con-
∂x dition Doyle et al. computed33 to determine the traditional linear EIS
where σe f f is the effective conductivity in the solid-phase and I is the of a half-cell). When k = p = 2, the governing equations and bound-
total current density, −I = i1 + i2 . ary conditions represent the leading order 2nd harmonic response of
Zero flux boundary conditions are taken at the positive and negative the system of equations to a finite perturbation, and so on. Solutions
current-collectors, for p > k, represent corrections to the leading order k = p solutions.
∂c ∂φ Here, we ignore all correction terms and focus on the leading order
= = 0 at x = 0 and x = L [10] (k = p) linear and nonlinear harmonics.
∂x ∂x
 Journal of The Electrochemical Society, 164 (11) E3311-E3320 (2017) E3313

Computing the leading order linear and nonlinear spectra.—The

resulting governing equations are shown in the Appendix. Table AI
contains the governing equations and boundary conditions for simu-
lating the linear impedance (k = p = 1) response. Tables AII and AIII
contain the sets of equations for computing the second (k = p = 2)
and third (k = p = 3) harmonic responses. The higher harmonic equa-
tions contain terms which depend on the lower harmonic solutions in
a hierarchical manner such that the harmonics are solved sequentially
(not simultaneously). The full hierarchical set of ordinary differential
equations for the first three leading order impedance harmonics was
solved using the generic Coefficient Form PDE physics module in
COMSOL v4.4. We chose to mesh the solid-phase equations (A1.1,
A2.1, and A3.1) instead of using an analytical solution in order to
have a general approach for any arbitrary concentration dependent
diffusivity parameters. The mesh nodes were spaced using a geomet-
ric ratio such that there existed a high concentration of nodes at the
interfaces between the solid- and solution-phases (at r = R p,i ) and at
the electrode/separator interfaces (at x = lneg and x = lneg + lsep ).
Mesh refinement in the interfacial regions with high gradients was
used so that the solution converged for all parameter and frequency
combinations resulting in approximately 9,000 total elements. Further
mesh refinement produced negligible changes in the computed results.
The frequency parameter, ω1 , was varied logarithmically from 10−3
to 105 Hz with 10 points per decade. Each full spectrum (with 81
frequencies) required a computational time of around 25 minutes on
a Dell Precision T1500 with an Intel Core i7 CPU @ 2.80 GHz and
8 GB RAM using a Windows 7 Professional 64-bit operating system.
Parameters were updated and the results were saved to disk using the
LiveLink for MATLAB version R2013b.
The leading order whole-cell impedance for the first harmonic
(k = p = 1), the second harmonic (k = p = 2), and the third
harmonic (k = p = 3) can be directly computed from the simulations
as

Z̃ k, p (ω1 ) = ϕ̃k,
s
p (x = L; ω1 ) − ϕ̃k, p (x = 0; ω1 )
s
[17]

where ϕ̃k, s
p (x = L; ω1 ) is the computed Fourier coefficient for the
solid-state potential evaluated at the boundary of the positive cur-
rent collector (x = L) and ϕ̃k, s
p (x = 0; ω1 ) is the computed potential
at the negative current collector (x = 0). These impedance coeffi-
cients are complex with both a real and an imaginary component such
 
that Z̃ k, p (ω1 ) = Z̃ k, p (ω1 ) + j Z̃ k, p (ω1 ). The double indexed coeffi-
cient Z̃ 1,1 (ω1 ) is identically equal to the normal (linear) electrochem-
ical impedance of the system. Z̃ 2,2 (ω1 ) is the leading order nonlinear
electrochemical impedance derived from the second harmonic of the
voltage response to a pure cosine input current and Z̃ 3,3 (ω1 ) is the
leading order nonlinear electrochemical impedance derived from the
third harmonic response.

Results and Discussion

Simulations of a LiCoO2 |LiC6 cell.—Figure 1 shows Nyquist Figure 1. Nyquist representations of the (a) linear impedance, (b) second
representations of simulated impedance spectra for a typical harmonic, and (c) third harmonic spectra for the base parameters given in
LiCoO2 |LiC6 cell using the parameters in Table I. The spectra cover Table I. The high frequency kinetics arc and low frequency mass transport
a wide range of frequencies from 1 mHz to 100 kHz logarithmically tail are characteristic of a lithium-ion battery. The lack of a second harmonic
spaced with 10 points per decade. The linear impedance response, response (b) in frequencies dominated by the reaction kinetics (1 Hz to 1 kHz)
Z̃ 1,1 (ω1 ), displays the expected characteristics of a lithium-ion battery is related to the typical assumption of symmetric transfer coefficients, αa and
with a purely real ohmic resistance at high frequencies, a depressed αc .
mid-frequency arc linked to kinetic processes, and a low-frequency
mass transport and thermodynamic dominated tail. The high frequency given by,
resistance is given by the sum of the ohmic resistances in each region
of the cell, 1 (αa + αc ) i 0,i F 1
ω RC,i = = [19]
Rct,i Cdl,i RT Cdl,i
     
 pos sep neg which evaluate to 1284.5 H z and 1428.5 H z for the negative
Rohmic = + +
σe f f, pos + κe f f, pos κe f f,sep σe f f,neg + κe f f,neg and positive electrodes, respectively. The two electrodes have a
[18] similar contribution to this kinetic arc for the base parameters
which evaluates to 0.467 m − m 2 for the parameters in Table I. The (Rct,neg = 7.79 m − m 2 and Rct, pos = 7.0 m − m 2 ). The charac-
typical characteristic frequencies for the kinetics of each electrode are teristic frequencies for the diffusional processes in the solution- and
E3314 Journal of The Electrochemical Society, 164 (11) E3311-E3320 (2017)

Table I. Base parameters for LiCoO2 |LiC6 simulations.43

Symbol Parameter Negative Electrode Separator Positive Electrode Units

Geometric Parameters
ai Specific interfacial area 723600 − 885000 m2 /m3
Br ugg Bruggeman coefficient 4 4 4
li Region thickness 88 25 80 μm
Ri Particle radius 2 − 2 μm
i Void fraction 0.485 0.724 0.385
 f,i Filler fraction 0.0326 – 0.025
Physicochemical Parameters
c0 Initial electrolyte 1000 1000 1000 mol/m3
concentration
Cdl,i Double-layer 10 − 10 μF/cm2
capacitance
D Electrolyte diffusivity 7.5 × 10−10 7.5 × 10−10 7.5 × 10−10 m2 /s
Ds,i Solid-phase diffusivity 3.9 × 10−14 − 1.0 × 10−14 m2 /s
i 0,i Exchange-current 3.30 − 3.67 A/m2
density
t+0 Transference number 0.364 0.364 0.364
αa,i Anodic transfer 0.5 − 0.5
coefficient
κ Electrolyte conductivity 0.2047 0.2047 0.2047 S/m
σi Solid-phase 100 − 100 S/m
conductivity
dU (cs )i
dcs |cs =c0 Derivative of open −3.21038 − −11.6724 V · cm3 /mol
s

circuit potential
d 2 U (cs )i
dcs 2
|cs =c0s Second derivative of −281.56 − 643.371 V · cm6 /mol2
open circuit potential
d 3 U (cs )i
dcs 3
|cs =c0s Third derivative of open 2222.82 − −177808 V · cm9 /mol3
circuit potential
Constants
R Gas Constant 8.314 J/mol · K
F Faraday’s Constant 96485 C/mol
T Temperature 298.15 K
Open Circuit Potentials
crs =R p ,neg
U (cs )neg 0.7222 + 0.1387θn + 0.029θn 0.5 − 0.0172θn −1 + 0.0019θn −1.5 + 0.2808exp(0.90 − 15θn ) θn = s
cmax,neg
−0.7984exp(0.4465θn − 0.4108)
−4.656+88.669θ p 2 −401.119θ p 4 +342.909θ p 6 −462.471θ p 8 +433.434θ p 10 crs =R p , pos
U (cs ) pos . θp = s
−1.0+18.933θ p 2 −79.532θ p 4 +37.311θ p 6 −73.083θ p 8 +95.96θ p 10 cmax, pos

D D
solid-phases are given by ω D,i = e f f,i
li 2
and ω Ds ,i = R s,i2 , re- scenario (symmetric transfer coefficients on both electrodes); under
p,i
more general conditions, the second harmonic is indeed sensitive to
spectively. For the base parameters, the fastest transport process
charge transfer.
(solution-phase diffusion in the separator) has a characteristic fre-
The base case third harmonic spectrum, shown in Figure 1c, ex-
quency of ω D,sep = 0.33 Hz, and the slowest process (solid-phase
hibits signal in the low-frequency mass-transport and thermodynamic
diffusion in the positive electrode) has a characteristic frequency of
regime as well as the mid-frequency regime dominated by the reac-
ω Ds , pos = 2.5 mHz, with characteristic transport frequencies in the
tion kinetics, much like the standard linear electrochemical impedance
negative electrode falling between them.
spectrum shown in Figure 1a. For both the second and third har-
If the governing equations and boundary conditions representing
monic, the nonlinear impedance asymptotes to zero in the limit of
the physics of a battery were linear, then the standard electrochemical
high frequencies. At high frequencies, the response of the battery
impedance shown in Figure 1a would be the only response of the sys-
is dominated by ohmic processes (that are governed by intrinsically
tem to a sinusoidal current perturbation of any magnitude. However,
linear physics). As a result, the nonlinear harmonics vanish at high
because the P2D model is a coupled nonlinear set, it also produces a
frequencies whereas the linear impedance remains finite.
frequency-dependent higher harmonic response that we can compute.
The five characteristic transport and reaction timescales, and the high
frequency ohmic limit, described above, can be used to help interpret Effect of charge transfer symmetry.—Figure 2a shows that chang-
the higher order harmonic response. For instance, the second harmonic ing the positive electrode anodic transfer coefficients, αa, pos , in Equa-
impedance, Z̃ 2,2 (ω1 ), shown in Figure 1b, only exhibits an apprecia- tion 4, while keeping the sum αa + αc = 1, has no effect on the linear
ble signal in the characteristic frequency range for mass-transport and impedance, Z̃ 1,1 (ω1 ), response; all five curves lie directly on top of
thermodynamics (ω1 < 1 Hz). For the base case parameters used here, each other. The transfer coefficients, αa and αc , for a reaction are
there is no appreciable second harmonic response at frequencies in a measure of charge transfer symmetry on that electrode. The base
the reaction kinetics regime (1 Hz − 10 kHz). As we will show in the case values of 0.5 for αa and αc used in Figure 1 for both electrodes
next section, this lack of kinetically-driven second harmonic response means there is complete symmetry in oxidation and reduction on both
is an artifact of the particular literature values used in the base case electrodes; i.e. the magnitude of the kinetic overpotential is identical
 Journal of The Electrochemical Society, 164 (11) E3311-E3320 (2017) E3315

insensitive to charge transfer symmetry and the overlapping spectra

in Figure 2a.
Figure 2b shows that symmetry in charge transfer is clearly im-
portant for the second harmonic response. The solid curve in Figure
2b is the same case as shown Figure 1b, with symmetric transfer
coefficients in the battery (αa = αc = 0.5 for both electrodes).
Mathematically, symmetric transfer coefficients nullify terms in the
kinetic contributions to the second harmonic (k = p = 2) equations
in the Appendix. When we break the symmetry of charge transfer in
Figure 2b — making the positive electrode have more facile oxidation
(αa > 0.5) or more facile reduction (αa < 0.5) — we see a second
harmonic arc arise at frequencies that correspond to kinetic processes.
Moreover, the direction of the kinetic arc depends on the direction
of the charge transfer asymmetry. When the transfer coefficients pro-
mote more facile battery discharge (αa, pos < 0.5 and αa,neg > 0.5),
the second harmonic kinetic arc moves toward the first quadrant in the
Nyquist plot. In contrast, if the charge transfer asymmetry makes bat-
tery charging more facile (αa, pos > 0.5 and αa,neg < 0.5), the second
harmonic response shifts to the third quadrant.
The mathematical origins of these second harmonic simulation
results are given by the governing equations for k = p = 2. The real
and imaginary oscillations in the second harmonic flux density at the
solid/electrolyte interface are given by Equation A2.6r and A2.6i in
the appendix,
 
 i 0 (αa + αc )  i 0 αa 2 − αc 2 F  2
η1,1 − η1,1
2
j f 2,2 = η2,2 +
RT 4 R2 T 2
[A2.6r]
and
 
i 0 (αa + αc )  i 0 αa 2 − αc 2 F    
j f 2,2 = η2,2 + η1,1 η1,1 . [A2.6i]
RT 2 R2 T 2
We find that terms involving oscillations in the overpotential at the
fundamental frequency, η1,1 , are significantly larger than the oscillat-
ing overpotential at the second harmonic, η2,2 . As a result, the second
term on the right-hand side (RHS) of both equations dominates the
harmonic response in the second order flux density. The second term
of the RHS disappears if αa = αc . This sensitivity to reaction asym-
metry has been demonstrated experimentally by Xu and Riley41 who
studied the ferri–ferrocyanide redox couple as well as Heubner et al.42
who used second harmonic measurements to determine the transfer
coefficients in the charge transfer kinetics of LiFePO4 .
The third harmonic shown in Figure 2c is also sensitive to charge
transfer symmetry, but in an inverse sense from the second harmonic.
The largest third harmonic kinetic loop is observed when there is
symmetry, and then the kinetic loop shrinks as the symmetry is broken.
It is also noteworthy that the shift in third harmonic is insensitive to
whether the charge transfer asymmetry favors oxidation or reduction.
Thus, Figure 2c produces overlapping curves for αa = 0.45 and
αa = 0.55, and a different set of overlapping curves for αa = 0.35
and αa = 0.65.
The results of Figure 2 show that there is new information to be
Figure 2. Nyquist representations of the (a) linear impedance, (b) second gleaned by measuring the second and third harmonics. In the case of
harmonic, and (c) third harmonic spectra for varying values of the positive the transfer coefficients, the standard linear EIS spectra (Figure 2a)
electrode transfer coefficient, αa, pos , while keeping the sum αa + αc = 1. are insensitive to the specific values, whereas the second harmonic
The linear response is insensitive to the transfer coefficients with all five provides information about the magnitude of the asymmetry and di-
curves superimposed upon one another. The second harmonic response is very rection (charge or discharge) that is kinetically more facile. The third
sensitive to the asymmetry of the transfer coefficients with the presence of
harmonic just gives insights into the magnitude of the asymmetry.
a kinetic arc only when αa = 0.5. While only the reaction symmetry of
the positive electrode was varied here, a similar dependence on the negative
electrode charge transfer symmetry also exists. The , , and  symbols Effect of kinetic parameters.—Figure 3 shows the effect of the
indicate 1 mHz, 1 Hz, and 1 kHz, respectively. exchange current densities, i 0,i , and double-layer capacitances, Cdl,i ,
for the positive and negative electrodes. An asymmetric set of transfer
during charge and discharge. There is no intrinsic reason to assume coefficients (αa,neg = 0.45 and αa, pos = 0.55) has been assumed to
charge transfer symmetry on either electrode. A difference in αa and ensure there is a kinetic arc in the second harmonic. The remaining
αc for either electrode would reflect asymmetry in the charge transfer discussion of individual parameter effects on the linear and nonlinear
processes on that electrode. harmonic response will focus on only the first and second harmonic,
Mathematically, only the sum (αa + αc ) appears as a parameter for clarity.
in the linear k = p = 1 equations in the Appendix, and that sum is The linear impedance spectra in Figures 3a and 3c show the sen-
unity in all cases simulated, resulting in the linear impedance being sitivity of the kinetic arc width to the changes in exchange current
E3316 Journal of The Electrochemical Society, 164 (11) E3311-E3320 (2017)

Figure 4. Nyquist representations of the first (a, c, e) and second (b, d, f)

harmonic impedance spectra for different values of solution- and solid-phase
diffusion coefficients, D and Ds,i . The second harmonic shows a more exagger-
ated response to changes in the mass-transport parameters. All other parameters
were held constant at the values given in Table I except αa,neg = 0.45 and
αa, pos = 0.55. The , , and  symbols indicate 1 mHz, 1 Hz, and 1 kHz,
respectively.

separates the timescales for each electrode kinetic process and intro-
duces distinct arcs for the two electrodes. In many ways, the linear and
Figure 3. Nyquist representations of the first (a, c, e, g) and second (b, d, second harmonic show similar behaviors, namely, they go from being
f, h) harmonic impedance spectra for different values of exchange current one distinctive arc to two clear arcs. Because the timescales between
density, i 0 , and double-layer capacitance, Cdl . The characteristic timescales the kinetic, transport, and thermodynamic regimes are well separated
of the system result in changes to these kinetic parameters only affecting for the base case parameters, changing either the exchange-current
the mid-frequency arc in both the first and second harmonic response. All
densities or the double-layer capacitances has little effect on the low
other parameters were held constant at the values given in Table I except
αa,neg = 0.45 and αa, pos = 0.55. The , , and  symbols indicate 1 mHz, frequency response of any of the harmonics for the values shown here.
1 Hz, and 1 kHz, respectively.
Effect of mass transport.—In equivalent circuit approaches to
impedance analysis, the low frequency tail of the linear EIS response
density on each electrode. The increasing size of the kinetic arc is is often used to extract information about the diffusional processes.
normally interpreted through the charge transfer resistances, Rct , with In their original description of the P2D impedance model on which
smaller exchange current densities resulting in larger Rct . Similar to this work builds, Doyle et al.33 noted that lithium-ion batteries have
the linear response, the second harmonic spectra in Figures 3b and 3d many different diffusional impedances, and for simple analysis, the
show a growing kinetic arc (at frequencies in excess of 1 Hz) as the low frequency spectrum must be dominated by a transport process in
exchange-current density is decreased. The width of the second har- order for Warburg plots to be accurate. The overlapping diffusional
monic kinetic arc is dependent on both the asymmetry of the charge time constants remain a challenge for extracting individual coefficients
transfer as well as on the value of the exchange current density. from impedance spectra. Here we explore if the nonlinear second
Varying the double-layer capacitance, Cdl , changes the charac- harmonic spectra may offer more sensitivity.
teristic frequency for each electrode without changing the overall Figure 4 shows the effects of varying the solution- and solid-phase
charge transfer resistance of the electrode. Figures 3e–3h show that diffusion coefficients over several orders of magnitude. The charac-
independently varying the capacitances by four orders of magnitude teristic “wagging” of the low frequency linear impedance tail shown
 Journal of The Electrochemical Society, 164 (11) E3311-E3320 (2017) E3317

in Figures 4a, 4c, 4e makes clear the difficulty in extracting a single

diffusion coefficient from a Warburg analysis, since all three diffu-
sional impedance responses manifest similarly. Figures 4b, 4d, 4f
shows the effect of the mass-transport parameters on the second har-
monic response. Similarly to the linear impedance response, the high
frequency arc remains constant due to the well separated timescales
between the kinetic and diffusion regimes. Figures 4b and 4d show
that varying either the negative or positive solid-phase diffusion co-
efficients has the effect of changing the magnitude and phase of the
low frequency response like the linear impedance. For particularly
low values of the solution-phase diffusivity, however, Figure 4f shows
that the structure of the low frequency second harmonic rearranges as
this diffusional impedance dominates the response. In nearly all cases,
the low frequency changes in second harmonic impedance are more
dramatic than in the linear response, suggesting a full physics analysis
of experimental linear and second harmonic data may result in a much
greater ability to fit low frequency processes. We are exploring this
currently.

Effect of thermodynamic parameters.—The thermodynamics of

the cell determine the impedance response as the frequency goes to
zero. The important thermodynamic parameters in the low frequency
range are the derivatives of the open circuit potential, U (cs ), with re-
spect to the concentration of lithium in the electrodes. Mathematically,
dU
the linear response shows a sensitivity to only the first derivative, dc s ,
d2U
while the second and higher harmonics depend on the second, dcs 2 ,
and higher order derivatives, respectively. Each of these derivatives
are evaluated at the lithium concentration in the solid present at the
particular state of charge being evaluated, c0s .
The sensitivity to second and higher order derivatives could be use-
ful for applications such as estimation of state of charge or probing
fundamental thermodynamic models predictions for lithium intercala-
tion. The sign of the second derivative (the curvature of U (cs )) changes
in between different phase-change plateaus in the open circuit poten-
tial. Figure 5 shows the dependence of these open circuit potential
derivatives around the base case parameters on the linear and second
harmonic spectra. As with the mass transfer case, one sees similar
effects between the linear and second harmonic impedance spectra,
though with exaggerated effects in the second harmonic, when the
first derivative of the open circuit potential is varied (Figures 4a–4d).
In contrast, the first harmonic is completely insensitive to the second
derivative of open circuit potential, whereas the second harmonic is
sensitive at low frequencies.

Conclusions and Future Implications

Physics-based impedance models are an important component of Figure 5. Nyquist representations of the first (a, c, e, g) and second (b, d,
understanding and interpreting impedance spectra due to their flexibil- f, h) harmonic impedance spectra for different values of the first and second
ity and physically meaningful parameters. Here, we presented an ex- dU
open-circuit derivatives, dc d2U
s and . As the first derivative is varied (a)-(d),
tension of Newman’s P2D impedance model for a lithium-ion battery dcs 2
a similar effect is seen in the linear and second harmonic response, with a
cell sandwich and demonstrated the simulated linear and harmonic re- slight exaggeration in the second harmonic response. The linear impedance is
sponse for a LiCoO2 |LiC6 cell. The strong dependence of the second insensitive to the changing second derivative. All other parameters were held
harmonic on the symmetry of the reaction kinetics for each electrode constant at the values given in Table I except αa,neg = 0.45 and αa, pos = 0.55.
was demonstrated and the individual effects of many kinetic, mass The , , and  symbols indicate 1 mHz, 1 Hz, and 1 kHz, respectively.
transport, and thermodynamic parameters were shown.
Introducing higher harmonic spectra into the analysis of an elec-
trochemical system provides additional informational content over the
linear response alone. Expanding the model to introduce additional cells using standard potentiostats and frequency response analyzers.
nonlinear contributions to the physics (concentration dependence of These results will be explored in the near future.
parameters like the diffusion coefficients, exchange-current densities,
and double-layer capacitances, for example) could lead to the gen-
eration of additional insight into the P2D model and its parameters. Acknowledgments
Quantifying the sensitivity of the model to the many different physic-
ochemical and geometric parameters could provide insight into the Partial support of this research is provided by the U. S. Depart-
experimentally measurable information in the linear and harmonic ment of Education grant P200A120023, an NSF IGERT grant DGE-
spectra. Preliminary experimental data indicate that these higher har- 1258485, and the Boeing Endowment for Excellence. The authors
monic signatures are easily measurable for commercial lithium-ion thank Hannaneh Hajishirzi for the helpful discussions.
E3318 Journal of The Electrochemical Society, 164 (11) E3311-E3320 (2017)

Appendix Table AII. Governing equations for the leading order second
harmonic response of the P2D model.

Second Harmonic (k = p = 2)
Solid-Phase Material Balance

s
 1 ∂ ∂c2,2
s
Real: 2ω1 c2,2 = r 2 ∂r
(Ds r 2 ∂r ) [A2.1r]
Table AI. Governing equations for the linear impedance response 
s 1 ∂ ∂cs
of the P2D model. Imaginary: −2ω1 c2,2 = r 2 ∂r
(Ds r 2 ∂r2,2 ) [A2.1i]

∂cs
BC at r = R p : −Ds ∂r2,2 = j f 2,2
Linear (k = p = 1) s 
∂c2,2
Solid-Phase Material Balance −Ds ∂r = j f 2,2
∂c s  
s  1 ∂ s
∂c2,2 s
∂c2,2
Real: ω1 c1,1 = r 2 ∂r
(Ds r 2 ∂r1,1 ) [A1.1r] BC at r = 0: =0 =0
s  ∂r ∂r
 1 ∂ ∂c1,1
Imaginary: −ω1 c1,1
s
= r 2 ∂r
(Ds r 2 ∂r ) [A1.1i] Solution-Phase Material Balance
s
∂c1,1 ∂c
BC at r = R p : −Ds ∂r = j f 1,1 Real: ∂∂x (De f f ∂2,2
x )= [A2.2r]
  − a(1 − t 0 )( j  + j  )
∂cs 2ω1 c2,2 +
= j f 1,1
f 2,2 dl 2,2
−Ds ∂r1,1 ∂c
s
∂c1,1
 s
∂c1,1

Imaginary: ∂∂x (De f f ∂2,2
x )= [A2.2i]
BC at r = 0: =0 =0  − a(1 − t 0 )( j  + j  )
∂r ∂r −2ω1 c2,2 + f 2,2 dl 2,2
Solution-Phase Material Balance BC at x = 0 and x = lneg + lsep + l pos :

∂c2,2 ∂c
∂c
Real: ∂∂x (De f f ∂1,1 = 0 ∂2,2
x =0
x )= [A1.2r] ∂x
 − a(1 − t 0 )( j  + j  )
ω1 c1,1 Faraday’s Law
+ f 1,1 dl 1,1
∂c ∂i2 2,2
Imaginary: ∂∂x (De f f ∂1,1 = a F( j f 2,2 + jdl 2,2 )
x )= [A1.2i] Real: ∂x [A2.3r]
 − a(1 − t 0 )( j  + j  )
−ω1 c1,1 ∂i2 2,2
+ f 1,1 dl 1,1 Imaginary: ∂x + = a F( j f 2,2 jdl 2,2 ) [A2.3i]
BC at x = 0 and x = lneg + lsep + l pos : BC at x = 0 and = x = L: i2 2,2 0 i2 2,2 = 0

∂c1,1 
∂c1,1
∂x =0 =0 ∂x BC at x = lneg and x = lneg + lsep :
Faraday’s Law i2 2,2 = 0 i2 2,2 = 0
∂i2 1,1
Real: ∂x = a F( j f 1,1 + jdl 1,1 ) [A1.3r] Solid-Phase Potential
∂i2 1,1 
Imaginary: = a F( j f 1,1 + jdl 1,1 ) [A1.3i] s
∂ϕ2,2
∂x Real: −σe f f = −i2 2,2 [A2.4r]
BC at x = 0 and x = L: i2 1,1 = 0 i2 1,1 = 0 ∂x

∂ϕs
BC at x = lneg and x = lneg + lsep : i2 1,1 = −1 Imaginary: −σe f f ∂2,2 = −i2 2,2 [A2.4i]
x
i2 1,1 = 0 ∂ϕs s
∂ϕ2,2

BC at x = 0 and x = x = 0 ∂x
L: ∂ 2,2 =0
Solid-Phase Potential BC at x = lneg and x = lneg + lsep :
s   
∂ϕ1,1 s
∂ϕ2,2 ∂ϕs
Real: −σe f f ∂x = −1 − i2 1,1 [A1.4r] ∂x = 0 ∂2,2
x =0
s 
∂ϕ1,1 Solution-Phase Potential
Imaginary: −σe f f ∂x = −i2 1,1 [A1.4i] 
∂ϕ2,2
∂ϕs

Real: −κe f f ∂x = i2 2,2 − [A2.5r]
BC at x = 0 and x = L: σe f f ∂ 1,1 =1 ∂c   
x κe f f RT
(1 − t+0 )(2c0 ∂2,2  ∂c1,1  ∂c1,1 )
x − c1,1 ∂ x +
s 
∂ϕ1,1 c1,1
c0 2 F ∂x
∂x =0 ∂ϕ 
s  Imaginary: −κe f f ∂ 2,2 
∂ϕ1,1 x = i2 2,2 − [A2.5i]
BC at x = lneg and x = lneg + lsep : ∂x =0 ∂c   
κe f f RT
(1 − t+0 )(2c0 ∂2,2  ∂c1,1  ∂c1,1 )
s 
∂ϕ1,1 c0 2 F x − c1,1 ∂ x − c1,1 ∂x
∂x =0
BC at x = 0: 
ϕ2,2 =0 
ϕ2,2 =0
Solution-Phase Potential 
∂ϕ2,2
Real:  [A1.5r] BC at x = lneg + lsep + l pos : ∂x =0
 
∂ϕ2,2
∂ϕ  2κe f f RT 0 ∂c1,1
x = i2 1,1 −
−κe f f ∂ 1,1 c0 F (1 − t+ ) ∂ x ∂x = 0
Imaginary: [A1.5i] Faradaic Flux
∂ϕ  2κ RT

∂c1,1
−κe f f ∂ 1,1
x = i2 1,1 − c0 F (1 − t+ )
ef f 0
∂x
  Real: j f 2,2 = [A2.6r]
BC at x = 0: ϕ1,1 = 0 ϕ1,1 = 0
i 0 (αa +αc )  i 0 (αa 2 −αc 2 )F  2

∂ϕ1,1 η2,2 + (η1,1 − η1,1 2 )
BC at x = lneg + lsep + l pos : ∂x =0 RT 4 R2 T 2

∂ϕ1,1 Imaginary: j f 2,2 = [A2.6i]
∂x =0 i 0 (αa +αc )  i 0 (αa 2 −αc 2 )F 
RT η2,2 + 2 R2 T 2
(η1,1 η1,1 )
Faradaic Flux Double-layer Flux
a +αc ) 
Real: j f 1,1 = i0 (αRT η1,1 [A1.6r] s 
 a +αc ) 
Real: jdl 2,2 = 2ω1 CFdl (ϕ2,2  )
− ϕ2,2 [A2.7r]
Imaginary: j f 1,1 = i0 (αRT η1,1 [A1.6i] 
Imaginary: jdl 2,2 = 2ω1 CFdl (ϕ2,2
 − ϕs )
2,2 [A2.7i]
Double-layer Flux
Overpotential
Real: jdl 1,1 = ω1 CFdl (ϕ1,1
s  )
− ϕ1,1

[A1.7r] Real: η2,2 = [A2.8r]
   2  2
 − ϕs  )
Imaginary: jdl 1,1 = ω1 CFdl (ϕ1,1 ϕ2,2  − dU cs − 1 d U (cs
− ϕ2,2
2
4 dcs 2 1,1 − c1,1 )
s s
1,1 [A1.7i] dcs 2,2

Imaginary: η2,2 = [A2.8i]
Overpotential   − dU s  1 d 2 U s  s 
s  − dU cs  ϕ2,2
s
− ϕ2,2 dcs c2,2 − 2 dcs 2 (c1,1 c1,1 )
Real: η1,1 = ϕ1,1 − ϕ1,1 dcs 1,1 [A1.8r]
 s   − dU cs 
Imaginary: η1,1 = ϕ1,1 − ϕ1,1 dcs 1,1 [A1.8i]
 Journal of The Electrochemical Society, 164 (11) E3311-E3320 (2017) E3319

Table AIII. Governing equations for the leading order third harmonic response of the P2D model.

Third Harmonic (k = p = 3)
Solid-Phase Material Balance
s
∂c3,3
 1 ∂
s
Real: 3ω1 c3,3 = r 2 ∂r
(Ds r 2 ∂r ) [A3.1r]
∂c s 
s 1 ∂
Imaginary: −3ω1 c3,3 = r 2 ∂r
(Ds r 2 ∂r3,3 ) [A3.1i]
 
∂cs ∂cs
BC at r = R p : −Ds ∂r3,3 = j f 3,3 −Ds ∂r3,3 = j f 3,3
 
∂cs ∂cs
BC at r = 0: ∂r3,3 = 0 ∂r3,3 =0

Solution-Phase Material Balance


∂c3,3
∂   
∂ x (De f f ∂ x ) = 3ω1 c3,3 − a(1 − t+ )( j f 3,3 + jdl 3,3 )
Real: 0 [A3.2r]

∂c3,3
∂ 

Imaginary: ∂ x (De f f ∂ x ) = −3ω1 c3,3 − a(1 − t+ )( j f 3,3 +
0 jdl 3,3 ) [A3.2i]
∂c 
∂c3,3
BC at x = 0 and x = lneg + lsep + l pos : ∂3,3 x = 0 ∂x = 0

Faraday’s Law
∂i2 3,3
Real: ∂x = a F( j f 3,3 + jdl 3,3 ) [A3.3r]
∂i2 3,3
Imaginary: = a F( j f 3,3 + jdl 3,3 )
∂x [A3.3i]
BC at x = 0 and x = L: i2 3,3 = 0 i2 3,3 = 0
BC at x = lneg and x = lneg + lsep : i2 3,3 = 0 i2 3,3 = 0

Solid-Phase Potential
s 
∂ϕ3,3
Real: −σe f f ∂x = −i2 3,3 [A3.4r]

∂ϕs
Imaginary: = −i2 3,3
−σe f f ∂3,3
x [A3.4i]
 s 
∂ϕs ∂ϕ3,3
BC at x = 0 and x = L: ∂ 3,3
x = 0 ∂x  = 0
∂ϕs s 
∂ϕ3,3
BC at x = lneg and x = lneg + lsep : ∂ 3,3
x = 0 ∂x =0

Solution-Phase Potential

∂ϕ3,3 ∂c    
Real: −κe f f
κe f f RT
= i2 3,3 −
(1 − t+0 )(4c02 ∂3,3  ∂c2,2  ∂c2,2  ∂c1,1  ∂c1,1
∂x 2c03 F x + 2c0 (c1,1 ∂ x − c1,1 ∂ x + c2,2 ∂ x − c2,2 ∂ x )
∂c  2  2 

 ∂c1,1
x (c1,1 − c1,1 ) − 2c1,1 c1,1 ∂ x )
+ ∂1,1 [A3.5r]
∂ϕ  κe f f RT

2 ∂c3,3

 ∂c2,2

 ∂c2,2

 ∂c1,1

 ∂c1,1
Imaginary: −κe f f ∂ 3,3
x = i2 3,3 − 2c0 3 F (1 − t+ )(4c0 ∂ x − 2c0 (c1,1 ∂ x + c1,1 ∂ x + c2,2 ∂ x + c2,2 ∂ x )
0

∂c  2  2 

 ∂c1,1
x (c1,1 − c1,1 ) + 2c1,1 c1,1 ∂ x )
+ ∂1,1 [A3.5i]
 
BC at x = 0: ϕ3,3 = 0 ϕ3,3 = 0
∂ϕ 
∂ϕ3,3
BC at x = lneg + lsep + l pos : ∂ 3,3 x = 0 ∂x = 0

Faradaic Flux
i 0 (αa +αc ) 
η3,3 + i0 (α2a R−α a +αc )F
2 2 3 3 2
Real: j f 3,3 = RT 2T 2
c )F
(η1,1 η2,2 − η1,1 η2,2 ) + i0 (α24 R3 T 3
(η1,1 3 − 3η1,1 η1,1 2 ) [A3.6r]
 i 0 (αa +αc )  i 0 (αa 2 −αc 2 )F     i 0 (αa 3 +αc 3 )F 2
Imaginary: j f 3,3 = RT η3,3 + 2 R2 T 2
(η1,1 η2,2 + η1,1 η2,2 ) + 24 R 3 T 3
(3η1,1 η1,1 2 − η1,1 3 ) [A3.6i]

Double-layer Flux
s 
Real: jdl 3,3 = 3ω1 CFdl (ϕ3,3  )
− ϕ3,3 [A3.7r]
  − ϕs  )
Imaginary: jdl 3,3 = 3ω1 CFdl (ϕ3,3 [A3.7i]
3,3

Overpotential
 dU s  1 d2U s s s  s  1 d3U s 3 s  s  2
Real: η3,3 = ϕ3,3
s  −
− ϕ3,3 dcs c3,3 − 2 dcs 2 (c1,1 c2,2 − c1,1 c2,2 ) − 24 dcs 3 (c1,1 − 3c1,1 c1,1 ) [A3.8r]
  − dU cs  − 1 d 2 U (cs  cs  + cs  cs  ) − 1 d 3 U (3cs  cs  2 − cs  3 )
Imaginary: η3,3 = ϕ3,3
s
− ϕ3,3 dcs 3,3 2 dcs 2 1,1 2,2 1,1 2,2 24 dcs 3 1,1 1,1 1,1 [A3.8i]

References 8. B. Saha, K. Goebel, and J. Christophersen, Trans. Inst. Meas. Control, 31, 293 (2009).
9. E. Barsoukov, J. H. Kim, J. H. Kim, C. O. Yoon, and H. Lee, J. Electrochem. Soc.,
1. E. Barsoukov and J. R. Macdonald, Eds., Impedance spectroscopy: theory, experi- 145, 2711 (1998).
ment, and applications, 2nd ed., p. 595, Wiley-Interscience, Hoboken, N.J, (2005). 10. P. Yu, B. N. Popov, J. A. Ritter, and R. E. White, J. Electrochem. Soc., 146, 8
2. M. E. Orazem and B. Tribollet, Electrochemical impedance spectroscopy, p. 523, (1999).
Wiley, Hoboken, N.J, (2008). 11. U. Tröltzsch, O. Kanoun, and H. -R. Tränkler, Electrochimica Acta, 51, 1664 (2006).
3. M. Thomas, P. G. Bruce, and J. B. Goodenough, J. Electrochem. Soc., 132, 1521 12. J. R. Wilson, D. T. Schwartz, and S. B. Adler, Electrochimica Acta, 51, 1389 (2006).
(1985). 13. S. Fletcher, J. Electrochem. Soc., 141, 1823 (1994).
4. D. Andre, M. Meiler, K. Steiner, C. Wimmer, T. Soczka-Guth, and D. U. Sauer, J. 14. R. Fathi, J. C. Burns, D. A. Stevens, H. Ye, C. Hu, G. Jain, E. Scott, C. Schmidt, and
Power Sources, 196, 5334 (2011). J. R. Dahn, J. Electrochem. Soc., 161, A1572 (2014).
5. J. Landesfeind, J. Hattendorff, A. Ehrl, W. A. Wall, and H. A. Gasteiger, J. Elec- 15. S. J. An, J. Li, C. Daniel, D. Mohanty, S. Nagpure, and D. L. Wood, Carbon, 105,
trochem. Soc., 163, A1373 (2016). 52 (2016).
6. P. Singh, R. Vinjamuri, X. Wang, and D. Reisner, Electrochimica Acta, 51, 1673 16. G. Sikha and R. E. White, J. Electrochem. Soc., 154, A43 (2007).
(2006). 17. G. Sikha and R. E. White, J. Electrochem. Soc., 155, A893 (2008).
7. B. Saha, K. Goebel, S. Poll, and J. Christophersen, IEEE Trans. Instrum. Meas., 58, 18. K. Darowicki, Electrochimica Acta, 42, 1781 (1997).
291 (2009). 19. M. Kiel, O. Bohlen, and D. U. Sauer, Electrochimica Acta, 53, 7367 (2008).
E3320 Journal of The Electrochemical Society, 164 (11) E3311-E3320 (2017)

20. B. Hirschorn, I. Ibrahim, M. E. Orazem, H. Takenouti, and B. Tribollet, in, p. 81, 32. S. Nakata, K. Takemura, and K. Neya, Anal. Sci., 17, 365 (2001).
ECS (2008) http://ecst.ecsdl.org/cgi/doi/10.1149/1.3004031. 33. M. Doyle, J. P. Meyers, and J. Newman, J. Electrochem. Soc., 147, 99 (2000).
21. J. J. Giner-Sanz, E. M. Ortega, and V. Pérez-Herranz, Electrochimica Acta, 186, 598 34. J. P. Meyers, M. Doyle, R. M. Darling, and J. Newman, J. Electrochem. Soc., 147,
(2015). 2930 (2000).
22. K. Darowicki, Corros. Sci., 37, 913 (1995). 35. M. Doyle, J. Newman, A. S. Gozdz, C. N. Schmutz, and J. -M. Tarascon, J. Elec-
23. R. -W. Bosch and W. F. Bogaerts, Corrosion, 52, 204 (1996). trochem. Soc., 143, 1890 (1996).
24. R. W. Bosch, J. Hubrecht, W. F. Bogaerts, and B. C. Syrett, Corrosion, 57, 60 (2001). 36. M. Doyle, T. F. Fuller, and J. Newman, J. Electrochem. Soc., 140, 1526 (1993).
25. J. R. Wilson, M. Sase, T. Kawada, and S. B. Adler, Electrochem. Solid-State Lett., 37. T. F. Fuller, M. Doyle, and J. Newman, J. Electrochem. Soc., 141, 1 (1994).
10, B81 (2007). 38. M. Pathak, D. Sonawane, S. Santhanagopalan, R. D. Braatz, and V. R. Subramanian,
26. T. Kadyk, R. Hanke-Rauschenbach, and K. Sundmacher, Int. J. Hydrog. Energy, 37, ECS Trans., 75, 51 (2017).
7689 (2012). 39. D. T. Schwartz, Proc. R. Soc. Lond. Math. Phys. Eng. Sci., 442, 397 (1993).
27. D. T. Schwartz, P. Stroeve, and B. G. Higgins, J. Electrochem. Soc., 136, 1755 (1989). 40. J. R. Wilson, S. B. Adler, and D. T. Schwartz, Phys. Fluids, 17, 63601 (2005).
28. J. A. Medina and D. T. Schwartz, Phys. Fluids, 8, 2895 (1996). 41. N. Xu and D. J. Riley, Electrochimica Acta, 94, 206 (2013).
29. J. A. Medina and D. T. Schwartz, J. Electrochem. Soc., 144, 155 (1997). 42. C. Heubner, M. Schneider, and A. Michaelis, J. Power Sources, 288, 115 (2015).
30. S. Nakata, K. Yoshikawa, and T. Matsuda, Biophys. Chem., 42, 213 (1992). 43. P. W. Northrop, V. Ramadesigan, S. De, and V. R. Subramanian, J. Electrochem. Soc.,
31. S. Nakata, R. Takitani, and Y. Hirata, Anal. Chem., 70, 4304 (1998). 158, A1461 (2011).