Electrochemical Impedance Spectroscopy for Battery Characterization

Physical Foundations, Equivalent Circuit Modeling, and Diagnostic Application to Li-Ion Cells

Electrochemical Impedance Spectroscopy (EIS) is the most information-dense single measurement available for non-destructive characterization of a Li-ion cell. A single spectrum collected across the frequency range of 100 kHz–10 mHz simultaneously resolves ohmic resistance, SEI ionic resistance, charge-transfer kinetics at both electrodes, double-layer capacitance, and solid-state Li⁺ diffusion — each process fingerprinted at its characteristic frequency. No DC electrochemical technique (GITT, HPPC, cyclic voltammetry) provides equivalent mechanistic separation in a single measurement. The challenge is not data collection — modern potentiostats acquire a full spectrum in minutes — but correct physical interpretation of the equivalent circuit parameters extracted from fitting. Misassignment of impedance features to the wrong physical mechanism is common in the battery literature and leads to incorrect conclusions about degradation state, rate-limiting processes, and cell design.

Physical Basis — What EIS Actually Measures

EIS applies a small-amplitude sinusoidal voltage perturbation to the cell at a series of discrete frequencies and measures the resulting current response. Because the perturbation amplitude is small (typically 5–10 mV), the system is assumed to operate in the linear regime — the current response is sinusoidal at the same frequency as the excitation, differing only in amplitude and phase. This linearity condition is the fundamental prerequisite for EIS validity; if it is violated (by using too large a perturbation amplitude, or by measuring a system far from steady state), the resulting spectrum is not physically interpretable.

The time-domain voltage and current signals are:

V(t) = V_0 e^{i\omega t

I(t) = I_0 e^{i(\omega t - \theta)

where $V_0$ and $I_0$ are the signal amplitudes, $\omega = 2\pi f$ is the angular frequency, and $\theta$ is the phase angle between voltage and current. The complex impedance is defined as:

Z(\omega) = \frac{V(t)}{I(t)} = \frac{V_0}{I_0}e^{i\theta} = Z_0(\cos\theta + i\sin\theta)

using Euler's relation $e^{i\theta} = \cos\theta + i\sin\theta$ . This yields the real and imaginary impedance components:

Z' = Z_0\cos\theta \quad \text{(real, resistive component)}

Z'' = Z_0\sin\theta \quad \text{(imaginary, reactive component)}

with the magnitude and phase recovered as:

|Z| = \sqrt{(Z')^2 + (Z'')^2}, \qquad \theta = \tan^{-1}\left(\frac{Z''}{Z'}\right)

The complete impedance $Z(\omega)$ is a frequency-dependent complex quantity. Plotting $-Z''$ vs. $Z'$ across the measured frequency range generates the Nyquist plot — the standard representation for battery EIS analysis. Plotting $|Z|$ and $\theta$ vs. $\log f$ generates the Bode plot, which is more sensitive to overlapping processes and is the preferred representation when time constants are close in value.

Dielectric Framework — Permittivity and Loss

For materials characterization applications (solid electrolytes, separator dielectric properties, electrode composite films), the impedance data is more naturally expressed in terms of complex permittivity:

\varepsilon^* = \varepsilon' - i\varepsilon''

where $\varepsilon'$ is the real permittivity (energy storage, in-phase response) and $\varepsilon''$ is the imaginary permittivity (dielectric loss, out-of-phase response). For a parallel-plate geometry:

\varepsilon' = \frac{Cd}{A\varepsilon_0}, \qquad \varepsilon'' = \varepsilon' \cdot D

where $C$ is measured capacitance, $d$ is electrode separation, $A$ is electrode area, $\varepsilon_0 = 8.854 \times 10^{-12}$ F/m is the permittivity of free space, and $D$ is the dielectric loss tangent ( $\tan\delta = \varepsilon''/\varepsilon'$ ).

The AC conductivity is related to the imaginary permittivity by:

\sigma_{AC} = 2\pi f \varepsilon_0 \varepsilon''

This relationship is used to extract ionic conductivity of solid electrolytes and separator-electrolyte composites from impedance measurements — a standard characterization route for solid-state battery electrolyte development.

Electric Modulus Representation

For systems where electrode polarization or conductivity relaxation processes dominate the permittivity spectrum and obscure weaker bulk relaxations, the electric modulus formalism is more appropriate:

M' = 2\pi f \varepsilon_0 \frac{A}{d} Z'', \qquad M'' = 2\pi f \varepsilon_0 \frac{A}{d} Z'

Electric modulus plots suppress electrode polarization contributions (which appear as low-frequency artefacts in permittivity plots) and resolve bulk ionic relaxation peaks that would otherwise be obscured. In the context of solid electrolyte characterization, $M''$ vs. frequency plots reveal the ionic hopping relaxation frequency, from which activation energy is extracted via the Arrhenius relation (see below).

Frequency Dispersion and Polarization Mechanisms

Each polarization mechanism contributing to the dielectric response of a material has a characteristic relaxation time ( $\tau$ ) — the timescale over which that mechanism responds to an applied field. Mechanisms with fast dynamics (short $\tau$ ) respond at high frequencies; slow mechanisms (long $\tau$ , large responding entities) drop out at high frequency.

Polarization Mechanism	Physical Origin	Characteristic Frequency
Electronic polarization	Electron cloud displacement relative to nucleus	~10¹⁵ Hz (UV/optical)
Atomic/ionic polarization	Relative displacement of ions in unit cell	~10¹²–10¹³ Hz (IR)
Dipole (orientational) relaxation	Reorientation of permanent dipoles	10⁶–10¹² Hz (microwave)
Ionic space charge relaxation	Migration of mobile ions to interfaces	10⁰–10⁶ Hz
Electrode polarization	Charge accumulation at electrode-electrolyte interface	10⁻³–10² Hz

For battery EIS specifically, the ionic space charge and electrode polarization mechanisms dominate the measurable spectrum (100 kHz–10 mHz range). Electronic and atomic polarization are not resolvable with standard EIS instrumentation — they require optical and IR spectroscopy respectively.

The Nyquist Plot of a Li-Ion Cell — Physical Assignment of Features

A full-range Nyquist plot of a commercial Li-ion cell exhibits several distinct regions, each mapping to a specific physical process within the cell. The frequency assignments below are representative — they shift with temperature, SOC, and degradation state.

-Z'' (Ω)
  |
  |    [L]   [R₀]      [R_SEI/C_SEI]    [R_ct/C_dl]        [W]
  |     *──────●────────(  )────────────(      )────────── /
  |                  ~100kHz-10kHz    ~1kHz-1Hz          <1Hz
  |
  └──────────────────────────────────────────────────── Z' (Ω)

Nyquist Feature	Frequency Range	Physical Origin	Extracted Parameter
Inductive tail (positive $Z''$ , high-f)	>100 kHz	Lead wire/cell tab inductance	Inductance $L$ (not intrinsic to cell)
High-frequency x-axis intercept	~100 kHz–10 kHz	Ohmic resistance: electrolyte + separator + current collectors + contacts	$R_0$ (internal ohmic resistance)
First semicircle (high-f)	~10 kHz–100 Hz	SEI ionic resistance + SEI capacitance	$R_{SEI}$ , $C_{SEI}$ (or $Q_{SEI}$ )
Second semicircle (mid-f)	~1 kHz–1 Hz	Charge-transfer resistance + double-layer capacitance	$R_{ct}$ , $C_{dl}$
Warburg tail (low-f oblique line)	<1 Hz	Semi-infinite solid-state Li⁺ diffusion in electrode particles	Warburg coefficient $\sigma_W$ , diffusion coefficient $D_{Li}$
Finite-length Warburg / low-f semicircle	<100 mHz	Finite diffusion length (thin-film or small particle)	$D_{Li}$ , diffusion length

Equivalent Circuit Models — Physical Justification

The mathematical equivalence between electrochemical interfaces and electronic circuits is not merely a fitting convenience — it reflects the physical reality that charge storage at an interface (double layer) behaves capacitively, and charge transfer across an interface (Faradaic reaction) behaves resistively, in the linear perturbation regime.

The Randles Circuit — Foundational Model

The Randles circuit is the simplest physically justified model for a single electrode-electrolyte interface:

Z_{Randles} = R_0 + \frac{1}{i\omega C_{dl} + \frac{1}{R_{ct} + Z_W}

Components:

$R_0$ : solution (electrolyte) resistance — real-axis intercept
$R_{ct}$ : charge-transfer resistance — kinetic barrier to the Faradaic reaction (Li⁺ desolvation and intercalation)
$C_{dl}$ : double-layer capacitance — non-Faradaic charge storage at electrode surface
$Z_W$ : Warburg impedance — semi-infinite diffusion

In Nyquist representation, the Randles circuit produces a single depressed semicircle at high-to-mid frequencies (from $R_{ct}$ and $C_{dl}$ in parallel) transitioning to a 45° Warburg tail at low frequencies.

Constant Phase Element (CPE/Q) — Why Ideal Capacitors Rarely Fit Battery Data

Real battery electrode interfaces rarely produce perfect semicircles — they produce depressed semicircles with the center below the real axis. This depression arises from surface heterogeneity: distributed reaction rates across a non-uniform SEI, roughness, porosity, and particle size distribution all produce a distribution of relaxation times rather than a single $RC$ time constant. The Constant Phase Element (CPE, symbol Q) captures this distribution:

Z_{CPE} = \frac{1}{Q_0(i\omega)^n}

where $Q_0$ is the CPE coefficient (F·s^{n-1}) and $n$ is the CPE exponent (0 ≤ n ≤ 1). When $n = 1$ , the CPE reduces to an ideal capacitor; when $n = 0$ , it reduces to a pure resistor; at intermediate $n$ , it represents a lossy capacitor (the "Q element" referenced in the original text) that produces the depressed semicircle observed in practice.

Complete Li-Ion Cell Equivalent Circuit

A physically motivated equivalent circuit for a full Li-ion cell (anode | separator | cathode, with SEI on anode) typically takes the form:

Z_{cell} = R_0 + Z_{SEI} + Z_{ct,anode} + Z_{ct,cathode} + Z_{diffusio

where each $Z_{SEI}$ and $Z_{ct}$ term is represented by a parallel R-CPE element. In practice, the anode and cathode charge-transfer semicircles often overlap in frequency, appearing as a single broadened semicircle — deconvolution requires either reference electrode measurements (three-electrode cell geometry) or operando temperature variation to separate the two contributions by their different activation energies.

Warburg Impedance and Li⁺ Diffusion Coefficient Extraction

The low-frequency oblique region of the Nyquist plot corresponds to semi-infinite linear diffusion of Li⁺ into the electrode particle bulk. The Warburg impedance is:

Z_W = \sigma_W \omega^{-1/2}(1 - i)

where $\sigma_W$ is the Warburg coefficient, extracted from the slope of the linear relationship between $Z'$ (or $Z''$ ) and $\omega^{-1/2}$ . The Li⁺ diffusion coefficient is then:

D_{Li} = \frac{1}{2}\left(\frac{RT}{n^2F^2Ac\sigma_W}\right)^2

where $R$ is the gas constant, $T$ is absolute temperature, $n$ is the number of electrons transferred per formula unit (n = 1 for Li⁺ intercalation), $F$ is the Faraday constant (96,485 C/mol), $A$ is the electrochemically active surface area (cm²), $c$ is the molar concentration of Li in the electrode (mol/cm³), and $\sigma_W$ is the Warburg coefficient (Ω·rad^{1/2}·s^{-1/2}).

Critical practical constraint: The active electrode area $A$ in this equation is the electrochemically active surface area — not the geometric electrode area. In a porous composite electrode, the true active area depends on tortuosity, particle size distribution, and carbon black coverage, and is difficult to determine independently. This introduces significant uncertainty in absolute $D_{Li}$ values extracted from EIS in composite electrodes. Reported $D_{Li}$ values in the battery literature therefore carry large uncertainty bands and should be treated as effective diffusion coefficients rather than intrinsic material properties, unless measured on model thin-film electrodes with known geometry.

Dielectric Relaxation and Activation Energy

In EIS spectra collected as a function of temperature, relaxation peaks in $\varepsilon''$ , $M''$ , or $Z''$ vs. frequency plots shift to higher frequency with increasing temperature — the relaxation process accelerates as thermal energy increases. The temperature dependence of the relaxation frequency $f_{max}$ (or equivalently, the relaxation time $\tau = 1/2\pi f_{max}$ ) follows the Arrhenius relation:

\tau = \tau_0 \exp\left(\frac{E_a}{k_B T}\right)

where $\tau_0$ is the pre-exponential factor (attempt frequency inverse), $E_a$ is the activation energy for the relaxation process, $k_B$ is Boltzmann's constant, and $T$ is absolute temperature. A plot of $\ln\tau$ vs. $1/T$ yields a straight line with slope $E_a/k_B$ .

Battery-relevant applications:

Activation energy for Li⁺ transport through the SEI layer — distinguishes different SEI compositions (inorganic Li₂CO₃-rich SEI: high $E_a$ ~0.5–0.8 eV; organic-rich SEI: lower $E_a$ )
Activation energy for charge-transfer reaction — tracks cathode surface passivation with cycling
Ionic conductivity activation energy in solid electrolytes — fundamental material parameter for solid-state battery design

EIS Frequency Range Requirements for Complete Battery Characterization

The full battery impedance spectrum spans approximately 6–7 decades of frequency. Different frequency windows resolve different physical processes:

Frequency Range	Physical Process	Measurement Time
1 MHz–100 kHz	Inductance, ohmic resistance	Seconds
100 kHz–1 kHz	SEI resistance and capacitance	Seconds
1 kHz–1 Hz	Charge-transfer resistance, double-layer capacitance	Seconds–minutes
1 Hz–10 mHz	Warburg diffusion	Minutes–hours
<10 mHz	Finite diffusion, intercalation capacitance	Hours

The practical lower frequency limit for routine battery EIS is typically 10–100 mHz — below this, measurement times become prohibitive (>15 minutes per spectrum) and the cell's SOC may change during measurement, violating the steady-state assumption. For high-accuracy diffusion coefficient determination, lower frequencies are required, with GITT (Galvanostatic Intermittent Titration Technique) used as a complementary or alternative approach.

Integrating EIS into a BMS enables real-time or periodic non-invasive assessment of:

BMS Application	EIS Parameter Used	Physical Basis
State of Charge (SOC) estimation	$R_0$ , OCV correlation; low-frequency $Z'$	$R_0$ and diffusion impedance are SOC-dependent
State of Health (SOH) monitoring	$R_{SEI}$ trend; $R_{ct}$ trend	SEI thickening → $R_{SEI}$ ↑; kinetic degradation → $R_{ct}$ ↑
Li plating detection	Low-frequency inductive feature; asymmetric relaxation	Plated Li introduces characteristic impedance signature
Temperature estimation	$R_0$ (strong T-dependence of electrolyte conductivity)	Can infer internal temperature independently of external sensors
End-of-life prediction	Rate of $R_{ct}$ increase per cycle; Warburg coefficient trend	Degradation rate extrapolation to capacity fade threshold

BMS-integrated EIS implementations inject the perturbation signal via the existing power electronics rather than a separate potentiostat — reducing hardware cost but limiting frequency range and amplitude control. Single-frequency impedance measurements (typically at one frequency in the charge-transfer semicircle range, ~100 Hz) are the most common BMS implementation, sacrificing spectral resolution for speed and hardware simplicity.

Data Quality Requirements — Validity Checks Before Fitting

Extracting physically meaningful parameters from EIS requires that the measurement satisfies three conditions, collectively assessed by the Kramers-Kronig (KK) relations:

Linearity — the perturbation amplitude must be small enough that the system response is linear. For Li-ion cells, 5–10 mV amplitude is standard; >20 mV amplitude risks nonlinear response, particularly at low SOC or on aged cells with high impedance.
Causality — the response must arise solely from the applied perturbation, not from independent cell processes (self-discharge, ongoing SEI growth during measurement). Cells must be at electrochemical steady state before measurement — typically requiring ≥2 hours of open-circuit rest after the last charge/discharge event.
Stationarity — the cell's impedance must not change during measurement. This limits the lower frequency bound for EIS on cycling cells and requires temperature control (±0.5°C) during measurement.

KK transformation of the measured $Z'(\omega)$ data should reproduce $Z''(\omega)$ (and vice versa) — systematic deviation between the measured and KK-transformed spectrum indicates a validity violation and means the spectrum should not be fitted or interpreted.

Battery Shortcut

Search This Blog