betaFeedback

WorksPhysics and Astronomy

A theory of inductive loops in electrochemical impedance spectroscopy

Abstract

We demonstrate that failure of time-invariance assumption in the modeling of electrochemical systems by equivalent circuits can lead to the formation of low frequency "inductive loops" that manifest themselves as positive imaginary parts of the impedance function. Assuming that the properties of the equivalent circuits change slowly in time we perform an asymptotic analysis and obtain a new integral representation of the impedance function that reduces to the standard one at high frequencies, while exhibiting inductive loops at low frequencies.

Introduction

Electrochemical impedance spectroscopy (EIS) is an indispensable tool to describe complex electrochemical systems in a unified and graphical way. In this approach any system is described by a single complex-valued impedance function Z(ω)Z(\omega). The basic theory of EIS [3, 4] [19] says that this function possesses special analytic properties that are most concisely expressed by the representation

Z(ω)=1iC0ω+∫0∞dσ(τ)1+iωτ,∫0∞dσ(τ)1+τ<+∞,0<C0≤∞.(1)Z(\omega) = \frac{1}{iC_0\omega} + \int_0^\infty\frac{d\sigma(\tau)}{1+i\omega\tau}, \qquad\int_0^\infty\frac{d\sigma(\tau)}{1+\tau} < +\infty,\quad0<C_0\leq\infty. \tag*{(1)}

Here σ\sigma is a positive Borel-regular measure on [0,+∞)[0,+\infty). It is often convenient to approximate such measures by linear combinations of delta-functions, resulting in rational approximations of Z(ω)Z(\omega) [1]. Such approximations can be interpreted as impedances of electrical circuits made of resistors and capacitors only. They are called the equivalent circuit models (ECM). One easily verifiable feature of the representation (1) is the negativity of the imaginary part of the impedance function.

The experimentally observed arcs in a Nyquist plot of the impedance with positive imaginary part in the low frequency part of the spectrum received the unfortunate moniker of an “inductive loop” [8] [12, 23, 21, 10]. It is generally agreed that magnetic effects play no significant role in the functioning of most electrochemical systems and devices, and even when they do, the inductive effects manifest themselves at the high end of the frequency spectrum [7, 6] [16]. The explanation of inductive loops in the current literature [6] [16] by means of ECMs with negative resistances and capacitances creates more problems than it solves. If one permits negative resistances and capacitances, then arbitrary “experimental data” can be matched by such an impedance function with any degree of precision due to the Riesz theorem (see e.g. [17, 14, 15]). This is not surprising, since it is the passivity of the system [22, 9, 20], violated by the the negative resistances and capacitances in ECMs, that are responsible for the possibility of stable reconstruction of the impedance function.

A far more reasonable explanation of the experimental appearance of low frequency inductive loops is the violation of the time-invariance assumption at time scales commensurate with the inverses of low frequencies. This explanation is semi-explicit in [21, 16, 18], where specific electrochemical processes altering the properties of the system on slow time scales are identified.

When the time-invariance assumption is violated the input-output behavior of the system can no longer be mathematically described by a well-defined impedance function at low frequencies. In this paper we will show that if one applies a particular impedance measurement procedure, delivering the correct impedance for any time-invariant system, the inductive loops could be observed in non time-invariant systems.

The goal of this paper is to extend the classical EIS/ECM approach to electrochemical systems with slowly “drifting” properties. In § 3 we demonstrate that a single Voigt element with drifting properties may exhibit an inductive loop. In § 5 we use asymptotic analysis to extend the EIS/ECM approach to non time-invariant systems with slowly varying properties. The resulting impedance representation formula, though more complicated, may still be used to reconstruct the impedance function values within the frequency band containing experimental measurements by performing the classical Kramers-Kronig analysis [2, 5] [11, 20] for high frequency data, while fitting the additional “drift” parameters using the inductive loop low frequency data.

The paper is organized as follows. In § 2 we discuss a measurement technique that gives a good approximation of the true impedance function for time-invariant systems. We then show in § 3 that the same measurement technique for non time-invariant Voigt elements can produce “inductive loop” data. Representing a general non time-invariant electrochemical system by an ECM with non time-invariant Voigt elements connected in series, gives us a mathematical model that can be analyzed. An asymptotic analysis applied to each non time-invariant Voigt element leads us in § 5 to a generalization of the EIS theory that reduces to classical formulas at high frequencies and is capable of modeling low frequency inductive loops.

Impedance measurements

Let us assume that the input signal is the current

I(t)=I0sin⁡(ωt),t∈[0,T(ω)],T(ω)=2πn(ω)ω,(2)I(t) = I_0\sin(\omega t), \quad t \in[0,T(\omega)], \quad T(\omega) = \frac{2\pi n(\omega)}{\omega}, \tag*{(2)}

where n(ω)∈Nn(\omega) \in\mathbb{N} should be as large as possible, so that the time T(ω)T(\omega) it takes to make a measurement is still acceptable to whoever makes the measurements. Of course, this is an issue only for low frequencies ω\omega. Thus, we can measure the system response only for frequencies ω≥ωmin⁡=2π/Tmax⁡\omega\ge\omega_{\min} = 2\pi/T_{\max}, where Tmax⁡T_{\max} is the maximal admissible time for making low frequency measurements. Hence, in the low frequency regime we always choose n(ω)=1n(\omega) = 1. At high frequencies we use a different measurement strategy. We choose a fixed measurement time T0T_0 and set n(ω)=round⁡(T0ω/2π)n(\omega) = \operatorname{round}(T_0\omega/2\pi). Hence, for all ω≥ωmin⁡\omega\ge\omega_{\min} we define

n(ω)=max⁡{1,round⁡(T0ω2π)},ω≥ωmin⁡.(3)n(\omega) = \max\left\{1,\operatorname{round}\left(\frac{T_0\omega}{2\pi}\right)\right\}, \quad\omega\ge\omega_{\min}. \tag*{(3)}

The Fourier transform of I(t)I(t) can be computed explicitly, but is an unwieldy expression, except at the frequency ω\omega of the input sinusoid:

I^(ω)=I0∫0T(ω)sin⁡(ωt)e−iωt dt=πI0niω=I0T(ω)2i.(4)\widehat{I}(\omega) = I_0\int_0^{T(\omega)}\sin(\omega t)e^{-i\omega t}\,\mathrm{d}t = \frac{\pi I_0 n}{i\omega} = \frac{I_0T(\omega)}{2i}. \tag*{(4)}

For linear, time-invariant (LTI) systems the output voltage U(t)U(t) must satisfy U^(ω)=Z(ω)I^(ω)\widehat{U}(\omega) = Z(\omega)\widehat{I}(\omega), where the complex factor Z(ω)Z(\omega) is called the impedance. Theoretically, when I(t)=0I(t) = 0 for t>T(ω)t > T(\omega), the output voltage is still non-zero. However, in most cases it decays exponentially fast and can therefore be neglected. In this case we have the approximation

U^(ω)≈∫0T(ω)U(t)e−iωt dt.(5)\widehat{U}(\omega) \approx\int_0^{T(\omega)}U(t)e^{-i\omega t}\,\mathrm{d}t. \tag*{(5)}

Since U(t)U(t) is measured, formula (5) can be used to estimate U^(ω)\widehat{U}(\omega) and thus, we have the formula for estimating the impedance from the experimental data:

Z0(ω)=2iT(ω)∫0T(ω)U(t)e−iωt dt.(6)Z_0(\omega) = \frac{2i}{T(\omega)}\int_0^{T(\omega)}U(t)e^{-i\omega t}\,\mathrm{d}t. \tag*{(6)}

If we apply formula (6) to an elementary Voigt circuit, consisting of a resistor RR and a capacitor CC connected in parallel, we will not get the correct answer

ZV(ω)=Riθω+1,θ=RC.(7)Z_V(\omega) = \frac{R}{i\theta\omega+ 1}, \quad\theta= RC. \tag*{(7)}

Figure 1 shows that Z0Z_0 is not a very good approximation for ZVZ_V, especially in the intermediate frequency range. The remedy is to understand this discrepancy and then devise a way to correct for it. Using the explicit expression for Z0(ω)Z_0(\omega) (which we don’t display here), we discover that for an elementary Voigt circuit we have

Z0(ω)=ZV(ω)−(1Tωℑ(ZV(ω))+iTZV′(ω))(1−e−T/θ),(8)Z_0(\omega)=Z_V(\omega)-\left(\frac{1}{T\omega}\Im(Z_V(\omega))+\frac{i}{T}Z'_V(\omega)\right)\left(1-e^{-T/\theta}\right), \tag*{(8)}

where ZV′(ω)Z'_V(\omega) denotes the derivative of ZV(ω)Z_V(\omega) with respect to ω\omega. Since relation (8) between the measured and the true impedance function of an elementary Voigt circuit is linear, it extends to all LTI systems.

We recall that in an LTI system the output (voltage U(t)U(t)) depends on the input (current I(t)I(t)) via

U(t)=ρ0I(t)+∫−∞tI(τ)K(t−τ) dτ,(9)U(t)=\rho_0 I(t)+\int_{-\infty}^{t}I(\tau)K(t-\tau)\,\mathrm{d}\tau, \tag*{(9)}

where the function K(s)K(s) is called a memory kernel.

Theorem 2.1. Suppose that the memory kernel decays exponentially:

∣K(s)∣≤R0θe−s/θ,s>T0.(10)\lvert K(s)\rvert\leq\frac{R_0}{\theta}e^{-s/\theta},\qquad s>T_0. \tag*{(10)}

for some T0<T(ω)T_0<T(\omega) for all ω\omega. Then

∣Z0(ω)−Z(ω)+1Tωℑ(Z(ω))+iTZ′(ω)∣≤R0(1+θT)e−T/θ,(11)\left\lvert Z_0(\omega)-Z(\omega)+\frac{1}{T\omega}\Im(Z(\omega))+\frac{i}{T}Z'(\omega)\right\rvert\leq R_0\left(1+\frac{\theta}{T}\right)e^{-T/\theta}, \tag*{(11)}

where

Z(ω)=ρ0+∫0∞K(s)e−iωs ds(12)Z(\omega)=\rho_0+\int_{0}^{\infty}K(s)e^{-i\omega s}\,\mathrm{d}s \tag*{(12)}

is the true impedance of the system.

The proof can be found in Appendix § 7

Thus, we are lead to a simple method for estimating the true impedance by making measurements over a longer time T2=2πn2(ω)/ω>T1=2πn1(ω)/ωT_2 = 2\pi n_2(\omega)/\omega> T_1 = 2\pi n_1(\omega)/\omega. Then, up to an exponentially small error we must have

Z(ω)≈Zexp⁡(ω)=T2Z0(ω;T2)−T1Z0(ω;T1)T2−T1.(13)Z(\omega) \approx Z_{\exp}(\omega) = \frac{T_2 Z_0(\omega;T_2)-T_1 Z_0(\omega;T_1)}{T_2-T_1}. \tag*{(13)}

We still want to require that at high frequencies the time it takes to make a an impedance measurement is fixed. Let us call it T0′T'_0, since we still want n1(ω)n_1(\omega) to be given by (3). In that case

n2min⁡(ω)=n1(ω)+1≤n2(ω)≤max⁡{2,round⁡(T0′ω2π)}=n2max⁡(ω).(14)n_2^{\min}(\omega) = n_1(\omega)+1 \le n_2(\omega) \le\max\left\{2,\operatorname{round}\left(\frac{T'_0\omega}{2\pi}\right)\right\} = n_2^{\max}(\omega). \tag*{(14)}

Thus, for measuring very low frequencies we set n1(ω)=1n_1(\omega)=1 and n2(ω)=2n_2(\omega)=2. When ω=4π/T0\omega=4\pi/T_0 we have n1(ω)=2n_1(\omega)=2, in which case we must require that n2(ω)≥3n_2(\omega)\ge3. It follows that T0′≥3T0/2T'_0\ge3T_0/2, and therefore, n2max⁡(ω)>n1(ω)n_2^{\max}(\omega)>n_1(\omega). In practice, the experimenter can choose any integer n2(ω)n_2(\omega) satisfying n2min⁡(ω)≤n2(ω)≤n2max⁡(ω)n_2^{\min}(\omega)\le n_2(\omega)\le n_2^{\max}(\omega), once the constant T0′≥3T0/2T'_0\ge3T_0/2 has been set.

The correction method (13) can be interpreted as a directive to ignore the transient response over the time interval [0,T1][0,T_1], effectively taking data during time interval [T1,T2][T_1,T_2]. Indeed, using formula (6), we have

Zexp⁡(ω)=2iT2(ω)−T1(ω)∫T1(ω)T2(ω)U(t)e−iωt dt,T1,2(ω)=2πn1,2(ω)ω.(15)Z_{\exp}(\omega)=\frac{2i}{T_2(\omega)-T_1(\omega)}\int_{T_1(\omega)}^{T_2(\omega)}U(t)e^{-i\omega t}\,dt,\qquad T_{1,2}(\omega)=\frac{2\pi n_{1,2}(\omega)}{\omega}. \tag*{(15)}

One convenient choice is n1=n(ω)n_1=n(\omega) and n2=2n(ω)n_2=2n(\omega), where n(ω)n(\omega) is given by (3), so that the length of the time interval over which the data is taken is still T(ω)=2πn(ω)/ωT(\omega)=2\pi n(\omega)/\omega.

Theorem 2.2. Under assumptions of Theorem 2.1 we have the estimate

∣Zexp⁡(ω)−Z(ω)∣≤T1T2−T1R0(1+θT1)e−T1/θ+R0e−T2/θ.(16)\left|Z_{\exp}(\omega)-Z(\omega)\right|\le\frac{T_1}{T_2-T_1}R_0\left(1+\frac{\theta}{T_1}\right)e^{-T_1/\theta}+R_0e^{-T_2/\theta}. \tag*{(16)}

The proof can be found is in Appendix § 7.

To see how formula (15) improves the evaluation of the impedance we take the same Voigt circuit with R=2kΩR=2\mathrm{k}\Omega, C=200μFC=200\mu\mathrm{F}, shown in Figure 1, and compare Zexp⁡(ω)Z_{\exp}(\omega) and Z(ω)=ZV(ω)Z(\omega)=Z_V(\omega). The difference between the Nyquist plots of the two functions can no longer be visualized as in Figure 1, since it is less that 0.01%0.01\%, according to numerics, when we choose n1=n(ω)n_1=n(\omega) and n2=2n(ω)n_2=2n(\omega).

Image not extracted

Figure 1. Comparison of Z0(ω)Z_0(\omega) and Z(ω)Z(\omega) for an elementary Voigt circuit with R=2 kΩR = 2\,\mathrm{k}\Omega, C=200 μFC = 200\,\mu\mathrm{F}.

Non time-invariant elementary Voigt circuits

In this section we examine elementary Voigt circuits made of a resistor and a capacitor connected in parallel. Our main assumption is that the parameters RR and CC of the circuit do not stay constant, but slowly change in time, instead. We will show that if we apply the impedance measurement recipe from the previous section to such non time-invariant systems we may obtain impedance curves with inductive loops.

General theory

Let us start by considering an elementary Voigt circuit consisting of a resistor R=R(t)R=R(t) and a capacitor C=C(t)C=C(t) connected in parallel. In such a circuit we have the Ohm’s laws for each of the elements

IR(t)=U(t)R(t),IC(t)=C(t)U˙(t).I_R(t)=\frac{U(t)}{R(t)},\qquad I_C(t)=C(t)\dot{U}(t).

Combining this with the Kirchhoff’s law I(t)=IR(t)+IC(t)I(t)=I_R(t)+I_C(t) we obtain the constitutive relation (i.e. dependence of I(t)I(t) on U(t)U(t)) in the form

I(t)=U(t)R(t)+C(t)U˙(t).(17)I(t)=\frac{U(t)}{R(t)}+C(t)\dot{U}(t). \tag*{(17)}

Since we use the current as the input we need to solve (17) for U(t)U(t):

U(t)=∫−∞tI(s)C(s)exp⁡{−∫stdxθ(x)}ds,θ=RC.(18)U(t)=\int_{-\infty}^{t}\frac{I(s)}{C(s)}\exp\left\{-\int_s^t\frac{\mathrm{d}x}{\theta(x)}\right\}\mathrm{d}s,\qquad\theta=RC. \tag*{(18)}

The quantity θ\theta is called the relaxation time of the Voigt circuit. Formula (15) gives the experimentally measured “impedance” of a simple Voigt circuit with non time-invariant elements:

Zexp(ω)=2iΔT∫T1T2e−iωt∫0tsin⁡(ωs)C(s)exp⁡{−∫stdxθ(x)}dsdt,ΔT=T2−T1.(19)Z_{\mathrm{exp}}(\omega)=\frac{2i}{\Delta T}\int_{T_1}^{T_2}e^{-i\omega t}\int_0^t\frac{\sin(\omega s)}{C(s)}\exp\left\{-\int_s^t\frac{\mathrm{d}x}{\theta(x)}\right\}\mathrm{d}s\mathrm{d}t,\qquad\Delta T=T_2-T_1. \tag*{(19)}

For future reference we also have the following formula for the impedance of a non time-invariant capacitor and resistor

ZexpC(ω)=2iΔT∫T1T2e−iωt∫0tsin⁡(ωs)C(s)dsdt=2ΔTω∫T1T2(e−iωt−1)sin⁡(ωt)C(t)dt,(20)Z_{\mathrm{exp}}^C(\omega)=\frac{2i}{\Delta T}\int_{T_1}^{T_2}e^{-i\omega t}\int_0^t\frac{\sin(\omega s)}{C(s)}\mathrm{d}s\mathrm{d}t=\frac{2}{\Delta T\omega}\int_{T_1}^{T_2}\left(e^{-i\omega t}-1\right)\frac{\sin(\omega t)}{C(t)}\mathrm{d}t, \tag*{(20)}
ZexpR(ω)=2iΔT∫T1T2e−iωtR(t)sin⁡(ωt)dt.(21)Z_{\mathrm{exp}}^R(\omega)=\frac{2i}{\Delta T}\int_{T_1}^{T_2}e^{-i\omega t}R(t)\sin(\omega t)\mathrm{d}t. \tag*{(21)}

Explicit non time-invariant models

Let us now examine a particular model of the time dependence of R(t)R(t) and C(t)C(t). We assume that RR and CC undergo an “exponential drift” from R−R_- (resp. C−C_-) at t=−∞t=-\infty to R+R_+ (resp. C+C_+) at t=+∞t=+\infty:

R(t)=aR−+R+et/τa+et/τ,C(t)=bC−+C+et/τb+et/τ,a,b>0.(22)R(t)=\frac{aR_-+R_+e^{t/\tau}}{a+e^{t/\tau}},\qquad C(t)=\frac{bC_-+C_+e^{t/\tau}}{b+e^{t/\tau}},\qquad a,b>0. \tag*{(22)}

The evolution law for R(t)R(t) is governed by two parameters: the time scale τ\tau and the “current position” a>0a>0 that tells us how far along R(t)R(t) is on the way from R−R_- to R+R_+. It is in 1-1 correspondence with R(0)R(0). The value a=1a = 1 means R(0)R(0) is exactly half-way between R+R_{+} and R−R_{-}. Exact same comments apply to C(t)C(t). It is easy to calculate explicitly

−∫stdxθ(x)=τθ+(s−tτ+κCln⁡(ρCe−s/τ+1ρCe−t/τ+1)+κRln⁡(ρRe−s/τ+1ρRe−t/τ+1))(23)-\int_s^t \frac{dx}{\theta(x)} = \frac{\tau}{\theta_{+}}\left(\frac{s-t}{\tau}+\kappa_C\ln\left(\frac{\rho_C e^{-s/\tau}+1}{\rho_C e^{-t/\tau}+1}\right)+\kappa_R\ln\left(\frac{\rho_R e^{-s/\tau}+1}{\rho_R e^{-t/\tau}+1}\right)\right) \tag*{(23)}

where

κC=(a−ρC)(b−ρC)ρC(ρC−ρR),κR=(a−ρR)(b−ρR)ρR(ρR−ρC),ρR=aR−R+,ρC=bC−C+,ρR≠ρC.\kappa_C=\frac{(a-\rho_C)(b-\rho_C)}{\rho_C(\rho_C-\rho_R)},\qquad\kappa_R=\frac{(a-\rho_R)(b-\rho_R)}{\rho_R(\rho_R-\rho_C)},\qquad\rho_R=\frac{aR_{-}}{R_{+}},\qquad\rho_C=\frac{bC_{-}}{C_{+}},\qquad\rho_R\ne\rho_C.

We note a relation

κC+κR=1−θ+θ−,θ±=C±R±.(24)\kappa_C+\kappa_R=1-\frac{\theta_{+}}{\theta_{-}},\qquad\theta_{\pm}=C_{\pm}R_{\pm}. \tag*{(24)}

If ρR=ρC=ρ\rho_R=\rho_C=\rho, then

−∫stdxθ(x)=τθ+(s−tτ+αln⁡(ρe−s/τ+1ρe−t/τ+1)+β(e−s/τ−e−t/τ)(ρe−s/τ+1)(ρe−t/τ+1))(25)-\int_s^t \frac{dx}{\theta(x)} = \frac{\tau}{\theta_{+}}\left(\frac{s-t}{\tau}+\alpha\ln\left(\frac{\rho e^{-s/\tau}+1}{\rho e^{-t/\tau}+1}\right)+\frac{\beta(e^{-s/\tau}-e^{-t/\tau})}{(\rho e^{-s/\tau}+1)(\rho e^{-t/\tau}+1)}\right) \tag*{(25)}

where

α=1−θ+θ−,β=(a−ρ)(b−ρ)ρ.\alpha=1-\frac{\theta_{+}}{\theta_{-}},\qquad\beta=\frac{(a-\rho)(b-\rho)}{\rho}.

A different drift model

R(t)=R0e−t/τ+R∞(1−e−t/τ),C(t)=C0e−t/τ+C∞(1−e−t/τ)(26)R(t)=R_{0}e^{-t/\tau}+R_{\infty}(1-e^{-t/\tau}),\qquad C(t)=C_{0}e^{-t/\tau}+C_{\infty}(1-e^{-t/\tau}) \tag*{(26)}

is a limiting case of our model (22) with R+=R∞R_{+}=R_{\infty}, C+=C∞C_{+}=C_{\infty}, and

a,b→0,R−,C−→∞,aR−→R0−R∞,bC−→C0−C∞.a,b\to0,\qquad R_{-},C_{-}\to\infty,\qquad aR_{-}\to R_{0}-R_{\infty},\qquad bC_{-}\to C_{0}-C_{\infty}.

In this limit, however, the restriction that R−R_{-} and C−C_{-} are positive are no longer required, since R(t)R(t) and C(t)C(t) are required to be positive only for t≥0t\ge0. In this drift model we can still use (23) with

ρR→R0R∞−1,ρC→C0C∞−1,κR→ρRρR−ρC,κC→ρCρC−ρR,\rho_R\to\frac{R_{0}}{R_{\infty}}-1,\qquad\rho_C\to\frac{C_{0}}{C_{\infty}}-1,\qquad\kappa_R\to\frac{\rho_R}{\rho_R-\rho_C},\qquad\kappa_C\to\frac{\rho_C}{\rho_C-\rho_R},

provided ρR≠ρC\rho_R\ne\rho_C. If ρR=ρC=ρ\rho_R=\rho_C=\rho, then we use formula (25) with α=1\alpha=1 and β=ρ\beta=\rho.

Once, ∫stθ(x)−1dx\int_s^t\theta(x)^{-1}dx has been evaluated, we compute

U(t)=I0C+∫0tsin⁡(ωs)1+be−s/τρCe−s/τ+1exp⁡{−∫stdxθ(x)}ds.(27)U(t)=\frac{I_0}{C_{+}}\int_0^t\sin(\omega s)\frac{1+be^{-s/\tau}}{\rho_Ce^{-s/\tau}+1}\exp\left\{-\int_s^t\frac{dx}{\theta(x)}\right\}ds. \tag*{(27)}

The experimentally measured impedance Zexp⁡(ω)Z_{\exp}(\omega) is then computed by means of formula (15). The numerical evaluation of Zexp⁡(ω)Z_{\exp}(\omega) presents several challenges, since parameter τ\tau must be very large, relaxation times θ±\theta_{\pm} are typically very small and ω\omega can range from 10−6Hz10^{-6}\mathrm{Hz} to 106Hz10^{6}\mathrm{Hz}, making some integrands highly oscillatory. These issues are addressed in Appendix B. Figure 2 shows an inductive loop in the Nyquist impedance plot of a non time-invariant Voigt element, with R(t)R(t) and C(t)C(t) given by (22), where R+=0.4kΩR_{+}=0.4k\Omega, R−=2kΩR_{-}=2k\Omega, a=10a=10, C+=1.6mFC_{+}=1.6mF, C−=0.8mFC_{-}=0.8mF, b=6b=6, τ=600s\tau=600s. The figure shows that at high frequencies the impedance is very well captured by the impedance of the time-invariant Voigt element with R=R(0)R=R(0), and C=C(0)C=C(0).

Slow parameter drift asymptotics

Assume that 1/τ=ϵ1/\tau=\epsilon is a small parameter and that ω=ϵΩ\omega=\epsilon\Omega, where the rescaled frequency Ω\Omega is assumed to be fixed. We need to find the asymptotics of Z(ϵΩ)Z(\epsilon\Omega) as ϵ→0\epsilon\to0 and see if Im⁡(Z(ϵΩ))\operatorname{Im}(Z(\epsilon\Omega)) can indeed be positive. We now assume that R(t)=R0(ϵt)R(t)=R_0(\epsilon t) and C(t)=C0(ϵt)C(t)=C_0(\epsilon t), where R0(x)R_0(x) and C0(x)C_0(x) are strictly positive, bounded smooth functions with bounded first derivatives on R\mathbb{R}. Our goal is not only to obtain the asymptotics of Z(ω)Z(\omega), defined by (6), as ϵ→0\epsilon\to0, but estimate the error between the actual impedance and its asymptotic approximation. We obtain for the elementary Voigt circuit, using (18)

Zϵ(ω)=2iΔT∫T1T2e−iωt∫0tsin⁡(ωs)C0(ϵs)exp⁡{−∫stdτθ(ϵτ)} ds dt,ΔT=T2−T1.Z_\epsilon(\omega)=\frac{2i}{\Delta T}\int_{T_1}^{T_2}e^{-i\omega t}\int_0^t\frac{\sin(\omega s)}{C_0(\epsilon s)}\exp\left\{-\int_s^t\frac{d\tau}{\theta(\epsilon\tau)}\right\}\,ds\,dt,\qquad\Delta T=T_2-T_1.

When the drift time scale τ=1/ϵ\tau=1/\epsilon is large, the formula for Zϵ(ω)Z_\epsilon(\omega) above can be simplified:

Zϵ(ω)=iΩπΔn(ω)∫2πn1(ω)/Ω2πn2(ω)/Ωe−iΩηIm⁡(R0(η)eiΩη1+iωθ0(η))dη+O(ϵ),(28)Z_\epsilon(\omega)=\frac{i\Omega}{\pi\Delta n(\omega)}\int_{2\pi n_1(\omega)/\Omega}^{2\pi n_2(\omega)/\Omega}e^{-i\Omega\eta}\operatorname{Im}\left(\frac{R_0(\eta)e^{i\Omega\eta}}{1+i\omega\theta_0(\eta)}\right)d\eta+O(\epsilon), \tag*{(28)}

where the order ϵ\epsilon error O(ϵ)O(\epsilon) is uniform over the entire frequency spectrum.

Formula (28) shows that if the functions θ0(η)\theta_0(\eta) and R0(η)R_0(\eta) are constants, then the approximation becomes exact, as it reduces to the classical impedance of the elementary Voigt circuit. One more simplification is possible in the asymptotics ϵ→0\epsilon\to0. Specifically, we can replace θ0(η)\theta_0(\eta) by θ(0)\theta(0) in (28). Our final asymptotics is

Zϵ∗(ω)=Z(ω,Ω)=iΩπΔn(ω)∫2πn1(ω)/Ω2πn2(ω)/Ωe−iΩηIm⁡(R0(η)eiΩη1+iωθ0(0))dη.(29)Z_\epsilon^*(\omega)=Z(\omega,\Omega)=\frac{i\Omega}{\pi\Delta n(\omega)}\int_{2\pi n_1(\omega)/\Omega}^{2\pi n_2(\omega)/\Omega}e^{-i\Omega\eta}\operatorname{Im}\left(\frac{R_0(\eta)e^{i\Omega\eta}}{1+i\omega\theta_0(0)}\right)d\eta. \tag*{(29)}

For numerical evaluation of (29) we choose n1=n(ω)n_1=n(\omega), n2=2n(ω)n_2=2n(\omega) and change variables of integration η=sn/Ω\eta=sn/\Omega. Then, using the representation Im⁡(z)=(z−zˉ)/2i\operatorname{Im}(z)=(z-\bar{z})/2i we obtain the formula we use in the Matlab code to make Figure 3:

Image not extracted

Figure 3. The quality of the asymptotics of an inductive loop.

Zϵ∗(ω)=12π∫2π4πR0(snΩ)1+iωθ0(0) ds−12π∫2π4πR0(snΩ)e−2ins1−iωθ0(0) ds.(30)Z_\epsilon^*(\omega)=\frac{1}{2\pi}\int_{2\pi}^{4\pi}\frac{R_0\left(\frac{sn}{\Omega}\right)}{1+i\omega\theta_0(0)}\,ds-\frac{1}{2\pi}\int_{2\pi}^{4\pi}\frac{R_0\left(\frac{sn}{\Omega}\right)e^{-2ins}}{1-i\omega\theta_0(0)}\,ds. \tag*{(30)}

For the model (22) we compute

12π∫2π4πR0(snΩ) ds=R++(R+−R−)Ω2πnln⁡(1+a(e−4πnΩ−e−2πnΩ)1+ae−2πnΩ).\frac{1}{2\pi}\int_{2\pi}^{4\pi}R_0\left(\frac{sn}{\Omega}\right)\,ds=R_++\frac{(R_+-R_-)\Omega}{2\pi n}\ln\left(1+\frac{a\left(e^{-\frac{4\pi n}{\Omega}}-e^{-\frac{2\pi n}{\Omega}}\right)}{1+ae^{-\frac{2\pi n}{\Omega}}}\right).

In the high frequency regime

R0(snΩ)=R0(ϵsT02π).R_0\left(\frac{sn}{\Omega}\right)=R_0\left(\frac{\epsilon sT_0}{2\pi}\right).

We can therefore, use a linear approximation R0(x)≈bR+mRxR_0(x)\approx b_R+m_Rx, when x∈[ϵT0,2ϵT0]x\in[\epsilon T_0,2\epsilon T_0], where the parameters bRb_R and mRm_R are found from the least squares fit. In that case

12π∫2π4πR0(snΩ)e−2ins ds≈imR2Ω.\frac{1}{2\pi}\int_{2\pi}^{4\pi}R_0\left(\frac{sn}{\Omega}\right)e^{-2ins}\,ds\approx\frac{im_R}{2\Omega}.

Figure 3 shows the Nyquist plot of Zϵ∗(ω)Z_\epsilon^*(\omega) superimposed on the computed “measured” impedance corresponding to the model (22). The difference between the two graphs is not detectable at the scale of the entire Nyquist plot, shown in the left panel of the figure. The right panel shows the blown-up portion of the plot containing the inductive loop. The overall relative error of the asymptotic approximation (29) is about 0.1%. In the figure we used (22) with the same parameter values as in Figure 2.

Now, if we have a general electrochemical system exhibiting parameter drift, then the measured impedance can be represented as

Zexp⁡(ω)=iΩπ(n2(ω)−n1(ω))∫2πn1(ω)Ω2πn2(ω)Ωe−iΩηℑ(Z(ω,η)eiΩη) dη+O(ϵ).(31)Z_{\exp}(\omega)=\frac{i\Omega}{\pi(n_2(\omega)-n_1(\omega))}\int_{\frac{2\pi n_1(\omega)}{\Omega}}^{\frac{2\pi n_2(\omega)}{\Omega}}e^{-i\Omega\eta}\Im\left(Z(\omega,\eta)e^{i\Omega\eta}\right)\,d\eta+O(\epsilon). \tag*{(31)}

where for each η∈R\eta\in\mathbb{R} the function ω↦Z(ω,η)\omega\mapsto Z(\omega,\eta) is a classical (instantaneous) impedance function, provided there is no leading capacitance in the ECM. In the presence of the leading capacitance, the term

Zexp⁡C(ω)=iΩπ(n2(ω)−n2(ω))∫2πn1(ω)Ω2πn2(ω)Ω(e−iΩη−1)sin⁡(Ωη)iωC0(η) dηZ_{\exp}^{C}(\omega)=\frac{i\Omega}{\pi(n_2(\omega)-n_2(\omega))}\int_{\frac{2\pi n_1(\omega)}{\Omega}}^{\frac{2\pi n_2(\omega)}{\Omega}}\left(e^{-i\Omega\eta}-1\right)\frac{\sin(\Omega\eta)}{i\omega C_0(\eta)}\,d\eta

must be added to the right-hand side of (31).

If we approximate the instantaneous impedance function Z(ω,ϵt)Z(\omega,\epsilon t) by a finite Voigt circuit ECM [1]

Z(ω,η)=R0(η)+∑j=1NRj(η)1+iωθ(η),Z(\omega,\eta)=R_0(\eta)+\sum_{j=1}^{N}\frac{R_j(\eta)}{1+i\omega\theta(\eta)},

then, up to a uniform, over the entire spectrum, order ϵ\epsilon error, we have

Zexp(ω)=iΩπ(n2(ω)−n1(ω))∫2πn1(ω)/Ω2πn2(ω)/Ωe−iΩηIm⁡m(Z0(ω,η)eiΩη) dη+O(ϵ),(32)Z_{\mathrm{exp}}(\omega)=\frac{i\Omega}{\pi(n_2(\omega)-n_1(\omega))}\int_{2\pi n_1(\omega)/\Omega}^{2\pi n_2(\omega)/\Omega}e^{-i\Omega\eta}\operatorname{Im}_m\left(Z^0(\omega,\eta)e^{i\Omega\eta}\right)\,d\eta+O(\epsilon), \tag*{(32)}

where

Z0(ω,η)=R0(η)+∑j=1NRj(η)1+iωθ(0).(33)Z^0(\omega,\eta)=R_0(\eta)+\sum_{j=1}^{N}\frac{R_j(\eta)}{1+i\omega\theta(0)}. \tag*{(33)}

If Rj(η)R_j(\eta) in (33) are modeled according to (26),

Rj(η)=Rj(0)+ΔRj(1−e−μjη),j=0,…,N,(34)R_j(\eta)=R_j(0)+\Delta R_j\left(1-e^{-\mu_j\eta}\right),\qquad j=0,\ldots,N, \tag*{(34)}

then the integral in (32) can be computed explicitly:

Zexp(ω)=∑j=0N{Rj(0)+ΔRj1+iωθj(0)+ΔRjF(ϵμjω,ωθj(0))}+O(ϵ),(35)Z_{\mathrm{exp}}(\omega)=\sum_{j=0}^{N}\left\{\frac{R_j(0)+\Delta R_j}{1+i\omega\theta_j(0)}+\Delta R_jF\left(\frac{\epsilon\mu_j}{\omega},\omega\theta_j(0)\right)\right\}+O(\epsilon), \tag*{(35)}

where θ0=0\theta_0=0, and

F(W,v)=ie−2πn1W−e−2πn2Wπ(n2−n1)Wv(W+i)−1(1+v2)(W+2i).F(W,v)=i\frac{e^{-2\pi n_1W}-e^{-2\pi n_2W}}{\pi(n_2-n_1)W}\frac{v(W+i)-1}{(1+v^2)(W+2i)}.

Typically, one would choose n1(ω)=n(ω)n_1(\omega)=n(\omega) and n2(ω)=2n(ω)n_2(\omega)=2n(\omega), where n(ω)n(\omega) is given by (3). At low frequencies n(ω)=1n(\omega)=1, while at high frequencies 2πn(ω)/ω2\pi n(\omega)/\omega can be replaced with a constant T0T_0.

Conclusions and discussion

In this paper we have proposed an explanation of the inductive loop phenomena observed in experiments. The hypothesis is that the act of the impedance measurement of an electrochemical system might speed up slow processes, such as corrosion and charge diffusion, altering the properties of the system on time scales commensurate with the inverses of the frequencies at which the impedance is measured. The modeling and analysis of these processes can lead to specific drift laws that can be used to validate our general theory.

A natural question is whether our theory makes it apparent how the time-dependent nature of the system causes the occurrence of inductive loops. By way of the answer we refer to formula (30) and set ω=0\omega=0, n(ω)=1n(\omega)=1. Then we obtain

Im⁡(Z(ϵΩ))≈12π∫2π4πR0(sΩ)sin⁡(2s) ds(36)\operatorname{Im}(Z(\epsilon\Omega))\approx\frac{1}{2\pi}\int_{2\pi}^{4\pi}R_0\left(\frac{s}{\Omega}\right)\sin(2s)\,ds \tag*{(36)}

While R0(u)R_0(u) is a strictly positive, smooth real function, the sign of the integral in (36) is the outcome of balancing positive and negative contributions of the oscillatory integrand. For example, if R0(u)R_0(u) is a decreasing function, then positive contributions will always be larger than the negative ones and inductive loops will be present. Conversely, if R0(u)R_0(u) is an increasing function, no inductive loops will be produced. If the slowly oscillating sinusoidal current input causes a corresponding in-phase oscillation of R0R_0, the inductive loop effect could be several times stronger.

One open question, not directly addressed in the paper is capturing the time scale τ\tau over which the system’s properties change. A crude estimate would be the inverse frequency at which the imaginary part of the impedance hits zero. For example, in the simulation corresponding to Figure 2 the time scale τ=600s\tau= 600\mathrm{s}, while 2π/ω0=610s2\pi/\omega_0 = 610\mathrm{s}, where Im⁡(Z(ω0))=0\operatorname{Im}(Z(\omega_0)) = 0. Our simulation of the time-dependent elementary Voigt circuit shows that the deviations from the classical model start at frequencies ω\omega as large as two orders of magnitude over τ\tau. They become vividly pronounced at frequencies on the order of 1/τ1/\tau and persist over frequencies an order of magnitude lower.

An inductive loop in a non time-invariant Voigt element

Figure 2. An inductive loop in a non time-invariant Voigt element.

Finally, whether or not the inductive loop data can give additional information about the system beyond the time scale τ\tau depends on whether the drift model (34) is acceptable. In this case one should be able to use the algorithm in [13] applied to the high frequency data to compute parameters Rj(0)R_j(0), θj(0)\theta_j(0) and NN. The intermediate and low frequency data can then be used in a non-linear least squares fit to to estimate drift parameters μj\mu_j and ΔRj\Delta R_j. Future research into the inductive loops should address these questions.

Acknowledgments. This material is based upon work supported by the National Science Foundation under Grant No. DMS-2005538.

Mathematical proofs

Proof of Theorem 2.1 and 2.2

For I(t)=χ[0,T(ω)](t)sin⁡(ωt)I(t) = \chi_{[0,T(\omega)]}(t)\sin(\omega t) we have, making a change of variables s=t−τs = t - \tau,

U(t)=ρ0sin⁡(ωt)+∫0tK(s)sin⁡(ω(t−s)) ds,0≤t≤T(ω).U(t) = \rho_0 \sin(\omega t) + \int_0^t K(s)\sin(\omega(t-s))\,\mathrm{d}s,\qquad0 \le t \le T(\omega).

We can then write

U(t)=ρ0sin⁡(ωt)+∫0∞χ(s,+∞)(t)K(s)sin⁡(ω(t−s)) ds.U(t) = \rho_0 \sin(\omega t) + \int_0^\infty\chi_{(s,+\infty)}(t)K(s)\sin(\omega(t-s))\,\mathrm{d}s.

so that we can substitute this into (6) and switch the order of integration. We obtain

Z0(ω)=ρ0+2iT(ω)∫0T(ω)K(s)(∫sT(ω)sin⁡(ω(t−s))e−iωt dt)ds.Z_0(\omega) = \rho_0 + \frac{2i}{T(\omega)}\int_0^{T(\omega)} K(s)\left(\int_s^{T(\omega)}\sin(\omega(t-s))e^{-i\omega t}\,\mathrm{d}t\right)\mathrm{d}s.

Computing the inner integral we obtain

Z0(ω)=ρ0+∫0T(ω)K(s)(e−iωs−ωse−iωs−sin⁡(ωs)2πn)ds.Z_0(\omega) = \rho_0 + \int_0^{T(\omega)} K(s)\left(e^{-i\omega s} - \frac{\omega s e^{-i\omega s} - \sin(\omega s)}{2\pi n}\right)\mathrm{d}s.

Using formula (12) for the true impedance, we compute

Z(ω)−12πn(ω)ℑ(Z(ω))−iω2πn(ω)Z′(ω)=ρ0+∫0∞K(s)(e−iωs+sin⁡(ωs)−ωse−iωs2πn) ds.Z(\omega)-\frac{1}{2\pi n(\omega)}\Im(Z(\omega))-\frac{i\omega}{2\pi n(\omega)}Z'(\omega)=\rho_0+\int_{0}^{\infty}K(s)\left(e^{-i\omega s}+\frac{\sin(\omega s)-\omega s e^{-i\omega s}}{2\pi n}\right)\,\mathrm{d}s.

Therefore

Z(ω)−ℑ(Z(ω))ωT(ω)−iT(ω)Z′(ω)−Z0(ω)=∫T(ω)∞K(s)(e−iωs+sin⁡(ωs)−ωse−iωs2πn) ds.(37)Z(\omega)-\frac{\Im(Z(\omega))}{\omega T(\omega)}-\frac{i}{T(\omega)}Z'(\omega)-Z_0(\omega)=\int_{T(\omega)}^{\infty}K(s)\left(e^{-i\omega s}+\frac{\sin(\omega s)-\omega s e^{-i\omega s}}{2\pi n}\right)\,\mathrm{d}s. \tag*{(37)}

For s≥T(ω)=2πn/ωs\ge T(\omega)=2\pi n/\omega we have ωs≥2πn\omega s\ge2\pi n. Denoting x=ωsx=\omega s and A=1/(2πn)A=1/(2\pi n), we estimate

∣e−ix+A(sin⁡x−xe−ix)∣≤∣1−Ax∣+A=Ax−1+A≤Ax,\left|e^{-ix}+A(\sin x-xe^{-ix})\right|\le|1-Ax|+A=Ax-1+A\le Ax,

since Ax≥1Ax\ge1 and A≤1/(2π)<1A\le1/(2\pi)<1. Hence,

∣Z0(ω)−Z(ω)+ℑ(Z(ω))+iωZ′(ω)2πn(ω)∣≤1T(ω)∫T(ω)∞s∣K(s)∣ ds.\left|Z_0(\omega)-Z(\omega)+\frac{\Im(Z(\omega))+i\omega Z'(\omega)}{2\pi n(\omega)}\right|\le\frac{1}{T(\omega)}\int_{T(\omega)}^{\infty}s|K(s)|\,\mathrm{d}s.

Using the exponential decay (10) of the memory kernel we obtain the estimate

∣Z0(ω)−Z(ω)+ℑ(Z(ω))+iωZ′(ω)2πn(ω)∣≤R0(1+θT(ω))e−T(ω)/θ,(38)\left|Z_0(\omega)-Z(\omega)+\frac{\Im(Z(\omega))+i\omega Z'(\omega)}{2\pi n(\omega)}\right|\le R_0\left(1+\frac{\theta}{T(\omega)}\right)e^{-T(\omega)/\theta}, \tag*{(38)}

proving (11).

Let us now prove Theorem 2.2. We compute,

Zexp⁡(ω)−Z(ω)=T1T2−T1∫T1T2K(s)(e−iωs+sin⁡(ωs)−ωse−iωs2πn1) ds−∫T2∞K(s)e−iωs ds.Z_{\exp}(\omega)-Z(\omega)=\frac{T_1}{T_2-T_1}\int_{T_1}^{T_2}K(s)\left(e^{-i\omega s}+\frac{\sin(\omega s)-\omega s e^{-i\omega s}}{2\pi n_1}\right)\,\mathrm{d}s-\int_{T_2}^{\infty}K(s)e^{-i\omega s}\,\mathrm{d}s.

Using estimate (38) we obtain

∣Zexp⁡(ω)−Z(ω)∣≤T1(ω)T2(ω)−T1(ω)R0(1+θT1(ω))e−T1(ω)/θ+R0e−T2(ω)/θ.|Z_{\exp}(\omega)-Z(\omega)|\le\frac{T_1(\omega)}{T_2(\omega)-T_1(\omega)}R_0\left(1+\frac{\theta}{T_1(\omega)}\right)e^{-T_1(\omega)/\theta}+R_0e^{-T_2(\omega)/\theta}.

Proof of the asymptotic foormula (31)

Let us analyze the asymptotic behavior of Zε(ω)Z_{\varepsilon}(\omega). We first change variables in the innermost integral σ=ετ\sigma=\varepsilon\tau:

Zε(ω)=2iΔT∫T1T2e−iωt∫0tsin⁡(ωs)C0(εs)exp⁡{−1ε∫εsεtdσθ(σ)} ds dt.Z_{\varepsilon}(\omega)=\frac{2i}{\Delta T}\int_{T_1}^{T_2}e^{-i\omega t}\int_{0}^{t}\frac{\sin(\omega s)}{C_0(\varepsilon s)}\exp\left\{-\frac{1}{\varepsilon}\int_{\varepsilon s}^{\varepsilon t}\frac{\mathrm{d}\sigma}{\theta(\sigma)}\right\}\,\mathrm{d}s\,\mathrm{d}t.

Next we change variables ξ=εs\xi=\varepsilon s in the integral with respect to the ss variable:

Zε(ω)=2iΔTε∫T1T2e−iωt∫0εtsin⁡(ωξ/ε)C0(ξ)exp⁡{−1ε∫ξεtdσθ(σ)} dξ dt.Z_{\varepsilon}(\omega)=\frac{2i}{\Delta T\varepsilon}\int_{T_1}^{T_2}e^{-i\omega t}\int_{0}^{\varepsilon t}\frac{\sin(\omega\xi/\varepsilon)}{C_0(\xi)}\exp\left\{-\frac{1}{\varepsilon}\int_{\xi}^{\varepsilon t}\frac{\mathrm{d}\sigma}{\theta(\sigma)}\right\}\,\mathrm{d}\xi\,\mathrm{d}t.

Finally, we change variables in the outermost integral η=ϵt\eta=\epsilon t:

Zϵ(ω)=2iΔTϵ2∫T1ϵT2ϵe−iΩη∫0ηsin⁡(Ωξ)C0(ξ)exp⁡{−1ϵ∫ξηdσθ(σ)}dξdη,Z_{\epsilon}(\omega)=\frac{2i}{\Delta T\epsilon^{2}}\int_{T_{1}\epsilon}^{T_{2}\epsilon}e^{-i\Omega\eta}\int_{0}^{\eta}\frac{\sin(\Omega\xi)}{C_{0}(\xi)}\exp\left\{-\frac{1}{\epsilon}\int_{\xi}^{\eta}\frac{d\sigma}{\theta(\sigma)}\right\}d\xi d\eta,

where Ω=ω/ϵ\Omega=\omega/\epsilon could be large, when ω\omega is not very small, but could also be of order 1, when ω\omega is of order ϵ\epsilon.

The main approximation idea is to realize that

Eϵ(ξ,η)=exp⁡{−1ϵ∫ξηdσθ(σ)}E_{\epsilon}(\xi,\eta)=\exp\left\{-\frac{1}{\epsilon}\int_{\xi}^{\eta}\frac{d\sigma}{\theta(\sigma)}\right\}

is exponentially small when η−ξ\eta-\xi is not very small. However, when η≈ξ\eta\approx\xi, then θ(σ)≈θ(η)\theta(\sigma)\approx\theta(\eta), which means that

Eϵ(ξ,η)≈Eϵ0(ξ,η)=exp⁡{−η−ξθ(η)ϵ}.E_{\epsilon}(\xi,\eta)\approx E_{\epsilon}^{0}(\xi,\eta)=\exp\left\{-\frac{\eta-\xi}{\theta(\eta)\epsilon}\right\}.

To make this approximation quantitative we use the inequality

ex2−ex1x2−x1<emax⁡{x1,x2},\frac{e^{x_{2}}-e^{x_{1}}}{x_{2}-x_{1}}<e^{\max\{x_{1},x_{2}\}},

which is a consequence of convexity of the exponential function. Hence,

∣Eϵ(ξ,η)−Eϵ0(ξ,η)∣≤Lθ−1(η−ξ)22ϵexp⁡{−η−ξMθϵ}=ϵLB(η−ξϵ),\left|E_{\epsilon}(\xi,\eta)-E_{\epsilon}^{0}(\xi,\eta)\right|\leq\frac{L_{\theta^{-1}}(\eta-\xi)^{2}}{2\epsilon}\exp\left\{-\frac{\eta-\xi}{M_{\theta}\epsilon}\right\}=\epsilon L B\left(\frac{\eta-\xi}{\epsilon}\right),

where

Mθ=max⁡t∈Rθ(t),Lθ−1=max⁡t∈R∣(1θ(t))′∣,B(x)=x22exp⁡{−xMθ}.M_{\theta}=\max_{t\in\mathbb{R}}\theta(t),\qquad L_{\theta^{-1}}=\max_{t\in\mathbb{R}}\left|\left(\frac{1}{\theta(t)}\right)'\right|,\qquad B(x)=\frac{x^{2}}{2}\exp\left\{-\frac{x}{M_{\theta}}\right\}.

Thus, we can replace Zϵ(ω)Z_{\epsilon}(\omega) with its approximation

Zϵ(1)(ω)=2iΔTϵ2∫T1ϵT2ϵe−iΩη∫0ηsin⁡(Ωξ)C0(ξ)Eϵ0(ξ,η)dξdη.Z_{\epsilon}^{(1)}(\omega)=\frac{2i}{\Delta T\epsilon^{2}}\int_{T_{1}\epsilon}^{T_{2}\epsilon}e^{-i\Omega\eta}\int_{0}^{\eta}\frac{\sin(\Omega\xi)}{C_{0}(\xi)}E_{\epsilon}^{0}(\xi,\eta)d\xi d\eta.

Moreover, we also have

∣Zϵ(ω)−Zϵ(1)(ω)∣≤2Lθ−1mCΔTϵ∫T1ϵT2ϵ∫0ηB(η−ξϵ)dξdη,\left|Z_{\epsilon}(\omega)-Z_{\epsilon}^{(1)}(\omega)\right|\leq\frac{2L_{\theta^{-1}}}{m_{C}\Delta T\epsilon}\int_{T_{1}\epsilon}^{T_{2}\epsilon}\int_{0}^{\eta}B\left(\frac{\eta-\xi}{\epsilon}\right)d\xi d\eta,

where

mC=min⁡t∈RC(t).m_{C}=\min_{t\in\mathbb{R}}C(t).

Changing variables in the inner integral x=(η−ξ)/ϵx=(\eta-\xi)/\epsilon we obtain

∣Zϵ(ω)−Zϵ(1)(ω)∣≤2Lθ−1ϵmC∫0∞B(x)dx=ϵMθ3Lθ−1mC.\left|Z_{\epsilon}(\omega)-Z_{\epsilon}^{(1)}(\omega)\right|\leq\frac{2L_{\theta^{-1}}\epsilon}{m_{C}}\int_{0}^{\infty}B(x)dx=\frac{\epsilon M_{\theta}^{3}L_{\theta^{-1}}}{m_{C}}.

The function Eε0(ξ,η)E_{\varepsilon}^{0}(\xi,\eta) is exponentially small when ξ\xi is not very close to η\eta. However, when ξ\xi is very close to η\eta we can replace C0(ξ)C_{0}(\xi) with C0(η)C_{0}(\eta). Hence we have

Zε(1)(ω)≈Zε(2)(ω)=2iΔTε2∫T1εT2εe−iΩηC0(η)∫0ηsin⁡(Ωξ)Eε0(ξ,η) dξdη.Z_{\varepsilon}^{(1)}(\omega) \approx Z_{\varepsilon}^{(2)}(\omega)=\frac{2i}{\Delta T\varepsilon^{2}}\int_{T_{1}\varepsilon}^{T_{2}\varepsilon}\frac{e^{-i\Omega\eta}}{C_{0}(\eta)}\int_{0}^{\eta}\sin(\Omega\xi)E_{\varepsilon}^{0}(\xi,\eta)\,d\xi d\eta.

Moreover,

∣Zε(1)(ω)−Zε(2)(ω)∣≤2LC−1ΔTε2∫T1εT2ε∫0η(η−ξ)exp⁡{−η−ξεMθ} dξdη,\left|Z_{\varepsilon}^{(1)}(\omega)-Z_{\varepsilon}^{(2)}(\omega)\right|\leq\frac{2L_{C^{-1}}}{\Delta T\varepsilon^{2}}\int_{T_{1}\varepsilon}^{T_{2}\varepsilon}\int_{0}^{\eta}(\eta-\xi)\exp\left\{-\frac{\eta-\xi}{\varepsilon M_{\theta}}\right\}\,d\xi d\eta,

where

LC−1=max⁡t∈R∣(1C(t))′∣.L_{C^{-1}}=\max_{t\in\mathbb{R}}\left|\left(\frac{1}{C(t)}\right)'\right|.

Changing variables x=(η−ξ)/εx=(\eta-\xi)/\varepsilon in the inner integral we obtain the bound

∣Zε(1)(ω)−Zε(2)(ω)∣≤2LC−1ΔT∫T1εT2ε∫0∞xexp⁡{−xMθ} dxdη=2LC−1Mθ2ε.\left|Z_{\varepsilon}^{(1)}(\omega)-Z_{\varepsilon}^{(2)}(\omega)\right|\leq\frac{2L_{C^{-1}}}{\Delta T}\int_{T_{1}\varepsilon}^{T_{2}\varepsilon}\int_{0}^{\infty}x\exp\left\{-\frac{x}{M_{\theta}}\right\}\,dx d\eta=2L_{C^{-1}}M_{\theta}^{2}\varepsilon.

Now, the inner integral in Zε(2)(ω)Z_{\varepsilon}^{(2)}(\omega) can be evaluated explicitly:

∫0ηsin⁡(Ωξ)exp⁡{−η−ξθ(η)ε} dξ=εθsin⁡(Ωη)−ε2θ2Ωcos⁡(Ωη)+ε2θ2Ωe−η/(εθ)1+ε2θ2Ω2.\int_{0}^{\eta}\sin(\Omega\xi)\exp\left\{-\frac{\eta-\xi}{\theta(\eta)\varepsilon}\right\}\,d\xi=\frac{\varepsilon\theta\sin(\Omega\eta)-\varepsilon^{2}\theta^{2}\Omega\cos(\Omega\eta)+\varepsilon^{2}\theta^{2}\Omega e^{-\eta/(\varepsilon\theta)}}{1+\varepsilon^{2}\theta^{2}\Omega^{2}}.

We observe that the first two terms can be combined nicely, and

Zε(2)(ω)=2iΔTε∫T1εT2εe−iΩηIm⁡(R0(η)eiΩη1+iωθ0(η)) dη+Δε(ω)=Zε(3)(ω)+Δε(ω),Z_{\varepsilon}^{(2)}(\omega)=\frac{2i}{\Delta T\varepsilon}\int_{T_{1}\varepsilon}^{T_{2}\varepsilon}e^{-i\Omega\eta}\operatorname{Im}\left(\frac{R_{0}(\eta)e^{i\Omega\eta}}{1+i\omega\theta_{0}(\eta)}\right)\,d\eta+\Delta_{\varepsilon}(\omega)=Z_{\varepsilon}^{(3)}(\omega)+\Delta_{\varepsilon}(\omega),

where

Δε(ω)=2iΩΔT∫T1εT2εe−iΩηθ2e−η/(εθ)C0(η)(1+ε2θ2Ω2) dη.\Delta_{\varepsilon}(\omega)=\frac{2i\Omega}{\Delta T}\int_{T_{1}\varepsilon}^{T_{2}\varepsilon}e^{-i\Omega\eta}\frac{\theta^{2}e^{-\eta/(\varepsilon\theta)}}{C_{0}(\eta)(1+\varepsilon^{2}\theta^{2}\Omega^{2})}\,d\eta.

We estimate

∣Δε(ω)∣≤2Ωεmax⁡η∈[T1ε,T2ε]θ(η)2e−η/(εθ(η))C0(η)(1+ε2θ(η)2Ω2)≤MRe−T1/Mθ.\left|\Delta_{\varepsilon}(\omega)\right|\leq2\Omega\varepsilon\max_{\eta\in[T_{1}\varepsilon,T_{2}\varepsilon]}\frac{\theta(\eta)^{2}e^{-\eta/(\varepsilon\theta(\eta))}}{C_{0}(\eta)(1+\varepsilon^{2}\theta(\eta)^{2}\Omega^{2})}\leq M_{R}e^{-T_{1}/M_{\theta}}.

Zε(3)(ω)Z_{\varepsilon}^{(3)}(\omega) can be written as a two-scale impedance function (28). In order to prove a simplified formula (29) we estimate

∣Zε(3)(ω)−Zε∗(ω)∣≤ΩπΔn(ω)MRLθ1+mθ2ω2∫2πn1(ω)/Ω2πn2(ω)/Ωωη dη=2πε(n1+n2)MRLθ1+mθ2ω2.\left|Z_{\varepsilon}^{(3)}(\omega)-Z_{\varepsilon}^{*}(\omega)\right|\leq\frac{\Omega}{\pi\Delta n(\omega)}\frac{M_{R}L_{\theta}}{1+m_{\theta}^{2}\omega^{2}}\int_{2\pi n_{1}(\omega)/\Omega}^{2\pi n_{2}(\omega)/\Omega}\omega\eta\,d\eta=\frac{2\pi\varepsilon(n_{1}+n_{2})M_{R}L_{\theta}}{1+m_{\theta}^{2}\omega^{2}}.

If we choose n1=nn_{1}=n and n2=2nn_{2}=2n, then in the regime n(ω)=1n(\omega)=1 we get the bound

∣Zε(3)(ω)−Zε∗(ω)∣≤6πεMRLθ.\left|Z_{\varepsilon}^{(3)}(\omega)-Z_{\varepsilon}^{*}(\omega)\right|\leq6\pi\varepsilon M_{R}L_{\theta}.

If n(ω)>1n(\omega)>1, then n(ω)=T0ω/2πn(\omega)=T_0\omega/2\pi, and we obtain

∣Zε(3)(ω)−Zε∗(ω)∣≤3εT0ωMRLθ1+mθ2ω2≤3εT0MRLθ2mθ.\left|Z_{\varepsilon}^{(3)}(\omega)-Z_{\varepsilon}^{*}(\omega)\right|\leq\frac{3\varepsilon T_0\omega M_RL_\theta}{1+m_\theta^2\omega^2}\leq\frac{3\varepsilon T_0M_RL_\theta}{2m_\theta}.

Hence, we obtain a uniform in the entire spectrum bound

∣Zε(ω)−Zε∗(ω)∣≤Mε.\left|Z_\varepsilon(\omega)-Z_\varepsilon^*(\omega)\right|\leq M\varepsilon.

for some constant MM that depends on the parameter functions R0(s)R_0(s) and C0(s)C_0(s), and that scales like RθR\theta.

Numerical evaluation of the impedance in explicit drift models

The key to the effective numerical computation of Zexp⁡(ω)Z_{\exp}(\omega) for explicit drift models (22) and (26) is a non-dimensionalization of all quantities. We therefore begin by rescaling the variable of integration in the definition (19) of Z(ω)Z(\omega), t=nt^/ωt=n\hat{t}/\omega.

Z(ω)=iI0π∫2π4πU(nt^ω)e−int^ dt^.Z(\omega)=\frac{i}{I_0\pi}\int_{2\pi}^{4\pi}U\left(\frac{n\hat{t}}{\omega}\right)e^{-in\hat{t}}\,d\hat{t}.

where we have chosen n1=n(ω)n_1=n(\omega), n2=2n(ω)n_2=2n(\omega), and where n(ω)n(\omega) is given by (3). We then observe that U(t)U(t) has the form

U(t)=I0C+∫0tsin⁡(ωs)f(sτ)eτθ+[g(sτ)−g(tτ)] ds(39)U(t)=\frac{I_0}{C_+}\int_0^t\sin(\omega s)f\left(\frac{s}{\tau}\right)e^{\frac{\tau}{\theta_+}\left[g\left(\frac{s}{\tau}\right)-g\left(\frac{t}{\tau}\right)\right]}\,ds \tag*{(39)}

where

f(u)=be−u+1ρCe−u+1,g(u)=u+κRln⁡(ρRe−u+1)+κCln⁡(ρCe−u+1),f(u)=\frac{be^{-u}+1}{\rho_Ce^{-u}+1},\qquad g(u)=u+\kappa_R\ln(\rho_Re^{-u}+1)+\kappa_C\ln(\rho_Ce^{-u}+1),

provided ρC≠ρR\rho_C\ne\rho_R. If ρC=ρR=ρ\rho_C=\rho_R=\rho, then we use

g(u)=u+αln⁡(ρe−u+1)−βρ(ρe−u+1).g(u)=u+\alpha\ln(\rho e^{-u}+1)-\frac{\beta}{\rho(\rho e^{-u}+1)}.

We therefore, rescale the variable of integration s=ns^/ωs=n\hat{s}/\omega in (39):

U(nt^ω)=I0nC+ω∫0t^sin⁡(ns^)f(ns^ωτ)eτθ+[g(ns^ωτ)−g(nt^ωτ)] ds^.U\left(\frac{n\hat{t}}{\omega}\right)=\frac{I_0n}{C_+\omega}\int_0^{\hat{t}}\sin(n\hat{s})f\left(\frac{n\hat{s}}{\omega\tau}\right)e^{\frac{\tau}{\theta_+}\left[g\left(\frac{n\hat{s}}{\omega\tau}\right)-g\left(\frac{n\hat{t}}{\omega\tau}\right)\right]}\,d\hat{s}.

Hence, we obtain the formula we use in our Matlab code

Z(ω)=inπC++ω∫2π4πe−int^(∫0t^sin⁡(ns^)f(ns^ωτ)eτθ+[g(ns^ωτ)−g(nt^ωτ)] ds^)dt^.(40)Z(\omega)=\frac{in}{\pi C_++\omega}\int_{2\pi}^{4\pi}e^{-in\hat{t}}\left(\int_0^{\hat{t}}\sin(n\hat{s})f\left(\frac{n\hat{s}}{\omega\tau}\right)e^{\frac{\tau}{\theta_+}\left[g\left(\frac{n\hat{s}}{\omega\tau}\right)-g\left(\frac{n\hat{t}}{\omega\tau}\right)\right]}\,d\hat{s}\right)d\hat{t}. \tag*{(40)}

In the frequency band where n(ω)/ω≈T0n(\omega)/\omega\approx T_0, we approximate, assuming that τ\tau is large,

f(ns^ωτ)≈bf+mfns^ωτ,g(ns^ωτ)≈bg+mgns^ωτ.f\left(\frac{n\hat{s}}{\omega\tau}\right) \approx b_f + m_f\frac{n\hat{s}}{\omega\tau}, \qquad g\left(\frac{n\hat{s}}{\omega\tau}\right) \approx b_g + m_g\frac{n\hat{s}}{\omega\tau}.

where the slopes mfm_f and mgm_g, and intercepts bfb_f and bgb_g are found from linear least squares fits of f(x)f(x) and g(x)g(x), x∈[T0/τ,2T0/τ]x \in[T_0/\tau, 2T_0/\tau]. Then, all integrals in (40) can be computed explicitly.

Up to exponentially small terms

Z(ω)≈bfR+mg+iωθ++12ωτmfR+mg+iωθ+(3T0ω+3i−4img2mg2+ω2θ+2),n(ω)>1.Z(\omega) \approx\frac{b_fR_+}{m_g+i\omega\theta_+}+\frac{1}{2\omega\tau}\frac{m_fR_+}{m_g+i\omega\theta_+}\left(3T_0\omega+3i-\frac{4im_g^2}{m_g^2+\omega^2\theta_+^2}\right), \qquad n(\omega)>1.

where in the final expression we replaced n(ω)n(\omega) with T0ω/(2π)T_0\omega/(2\pi), which is valid exactly in the high frequency regime.

References

  1. [1]P. Agarwal, M. E. Orazem, and L. H. Garcia-Rubio. Measurement models for electrochemical impedance spectroscopy: I. demonstration of applicability. Journal of the Electrochemical Society, 139(7):1917, 1992.DOI
  2. [2]P. Agarwal, M. E. Orazem, and L. H. Garcia-Rubio. Application of measurement models to impedance spectroscopy: III. Evaluation of consistency with the Kramers-Kronig relations. Journal of the Electrochemical Society, 142(12):4159, 1995.DOI
  3. [3]A. Bard and L. Faulkner. Electrochemical Methods; Fundamentals and Applications. Wiley Interscience, Hoboken, NJ, 2nd edition, 2000.
  4. [4]E. Barsoukov and J. R. Macdonald, editors. Impedance spectroscopy: theory, experiment, and applications. John Wiley & Sons Inc., 2nd edition, 2005.
  5. [5]B. A. Boukamp. A linear Kronig-Kramers transform test for immittance data validation. Journal of the electrochemical society, 142(6):1885, 1995.DOI
  6. [6]B. A. Boukamp. Distribution (function) of relaxation times, successor to complex nonlinear least squares analysis of electrochemical impedance spectroscopy? Journal of Physics: Energy, 2(4):042001, 2020.DOI
  7. [7]H. Brandstätter, I. Hanzu, and M. Wilkening. Myth and reality about the origin of inductive loops in impedance spectra of lithium-ion electrodes—a critical experimental approach. Electrochimica acta, 207:218–223, 2016.DOI
  8. [8]H.-M. Cho, Y. J. Park, and H.-C. Shin. Semiempirical analysis of time-dependent elementary polarizations in electrochemical cells. Journal of The Electrochemical Society, 157(1):A8–A18, 2009.DOI

Paper details

vlnc:W_hde87k6z
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 0

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents