Lithium-Ion Battery Impedance Spectroscopy: Testing Methods and Analysis Techniques That Actually Work

If you are serious about understanding what is happening inside a lithium-ion cell, voltage curves and capacity numbers only tell part of the story. Electrochemical impedance spectroscopy, or EIS, cuts deeper. It separates overlapping processes that happen at different speeds, lets you quantify resistance at every interface, and gives you a window into degradation mechanisms that show up nowhere else.

This guide covers how EIS testing actually works in lithium-ion battery research, what each part of the spectrum means, and how to extract real numbers from the data.

What EIS Actually Measures and Why It Matters

The core idea is simple. You hit the battery with a tiny sinusoidal signal, usually 5 millivolts or less, sweep across a wide frequency range from microhertz up to megahertz, and measure how the cell responds. The ratio of voltage to current at each frequency gives you the complex impedance. Real part tells you about resistive losses. Imaginary part tells you about capacitive and inductive behavior.

What makes EIS powerful is time constant separation. Fast processes like ohmic resistance respond at high frequencies. Slower processes like solid-state diffusion only show up at low frequencies. A single scan gives you a layered map of everything happening inside the cell simultaneously.

The measurement relies on three assumptions: linearity, causality, and stability. The perturbation must be small enough that the system stays linear. The response must be caused by the input, not external noise. And the cell state must not drift during the scan. Violate any of these and the data becomes unreliable. Kramers-Kronig transforms can validate whether your data even meets these criteria.

How to Run EIS Tests on Lithium-Ion Cells

Setting Up the Test Correctly

Get the setup wrong and the data is worthless. Use a proper electrochemical workstation with frequency response analysis capability. Connect the leads carefully, red to red, black to black, and keep the shielding box closed throughout the entire measurement. Wear plastic gloves when handling cells. Keep phones and other RF sources away from the setup.

For a standard lithium-ion cell, the frequency range typically spans from 0.1 hertz to 1 megahertz, though broadband EIS can extend down to 0.01 hertz or even microhertz to capture slow diffusion processes. Set the amplitude to 5 millivolts. Go higher and you start pushing the cell out of its linear regime. Go lower and you lose signal-to-noise ratio.

The cell must be at a stable state of charge before testing. Charge or discharge to the target SOC using constant current constant voltage, then let it rest for at least 30 minutes. Some protocols use open circuit voltage relaxation until the voltage drift drops below a few millivolts per hour. Skipping this step is the most common mistake in EIS testing.

Classical versus Dynamic EIS

Classical EIS, also called single-sine sweep, steps through each frequency one at a time, measures the steady-state response, and moves to the next. It gives the cleanest data with the best signal-to-noise ratio. The downside is speed. A full scan from millihertz to kilohertz can take minutes to tens of minutes, and the cell must stay perfectly still the whole time.

Dynamic EIS solves the speed problem by overlaying a multi-sine or broadband excitation signal on top of a DC bias. You can run it during charge or discharge, which means you get impedance data at every SOC point in a single pass. The trade-off is that the frequency resolution drops, and you need to account for SOC drift during the scan. Two common approaches handle this: either introduce a time axis alongside the real and imaginary axes and map impedance back to specific SOC points, or use multiple single-frequency measurements stitched together across the full SOC range.

Nonlinear EIS takes a different angle. Instead of keeping the signal tiny, you deliberately use a larger amplitude to push the cell into nonlinear behavior. Then you run fast Fourier transforms on the distorted response to extract harmonic components. This exposes nonlinear dynamics that classical EIS completely misses, though the interpretation gets significantly harder.

Reading the Nyquist Plot and Bode Plot

What Each Region of the Spectrum Tells You

A typical lithium-ion cell Nyquist plot shows three to five distinct features as you move from high to low frequency.

The high-frequency intercept on the real axis is the ohmic resistance, often labeled Rs or Rb. This includes electrolyte ionic resistance, separator resistance, and contact resistance between electrodes and current collectors. It barely changes with SOC or temperature, which makes it a useful baseline reference.

The first semicircle, usually appearing between 1 kilohertz and 100 hertz, represents the solid electrolyte interphase, or SEI. The resistance here is Rsei and it pairs with a capacitance Csei in the equivalent circuit. As the cell ages, this semicircle grows because the SEI layer thickens. Monitoring Rsei over cycles is one of the most direct ways to track calendar aging.

The second semicircle, sitting between 100 hertz and 1 hertz, is the charge transfer resistance, Rct, in parallel with the double-layer capacitance, Cdl. This is the kinetic bottleneck of the electrochemical reaction itself. Rct is extremely sensitive to SOC and temperature. At very low or very high SOC, Rct spikes because there are fewer active sites or the lithium concentration gradient becomes steep. In the middle SOC range, typically 20 to 80 percent, Rct hits its minimum.

The low-frequency tail, below 1 hertz, shows up as a 45-degree line. This is the Warburg impedance, representing solid-state diffusion of lithium ions inside the active material particles. The steeper the line, the more diffusion-limited the cell is. At extreme SOC values, this tail becomes nearly vertical, indicating severe diffusion restriction.

At very low frequencies, below 0.01 hertz, you may see another semicircle related to crystal structure changes or new phase formation, plus a vertical line from lithium accumulation and consumption. These features are hard to capture in practice because the test would take hours.

Bode Plot: The Complementary View

The Bode plot shows impedance magnitude and phase angle versus frequency on logarithmic scales. It does not separate individual processes as cleanly as the Nyquist plot, but it makes frequency-dependent trends much easier to spot. A sharp drop in impedance magnitude at a specific frequency tells you exactly which process dominates at that timescale. The phase angle peak corresponds to the characteristic frequency of each semicircle in the Nyquist plot.

Equivalent Circuit Modeling and Data Fitting

Building the Right Circuit

The standard equivalent circuit for a lithium-ion cell starts with Rs in series, followed by a parallel Rsei-Csei branch for the SEI, then a parallel Rct-Cdl branch for charge transfer, and finally a Warburg element for diffusion. In practice, pure capacitors rarely fit well because real electrode surfaces are rough and heterogeneous. Replace them with constant phase elements, or CPEs, which have an exponent n that quantifies the deviation from ideal capacitive behavior. When n equals 1, you have a perfect capacitor. When n is 0.5, you are looking at pure diffusion.

Software tools like ZView, ZSimpWin, EIS300, or Nova fit the model to your data using nonlinear least squares algorithms, typically Levenberg-Marquardt. The fitting quality is judged by the chi-squared error. Keep it below 0.001 for publication-quality data. Above 0.1 and you need to rethink your circuit topology or your data quality.

Common Fitting Pitfalls

The biggest trap is over-fitting. Adding more R-CPE pairs will always reduce the error, but it does not mean the extra elements are physically real. Different circuit topologies can produce nearly identical fits, a problem known as model degeneracy. Always validate your fit against Kramers-Kronig transforms and cross-check with physical intuition. If a fitted parameter makes no sense, like a negative resistance or a capacitance of 10 farads, the model is wrong regardless of how low the chi-squared value is.

Another issue is initial parameter selection. Nonlinear fitting is sensitive to starting values. Pick initial guesses based on what you see in the raw Nyquist plot. The high-frequency intercept gives you Rs. The diameter of the first semicircle gives you Rsei. The second semicircle diameter gives you Rct. Use these as starting points and let the algorithm refine from there.

What EIS Reveals About Cell State and Health

Tracking SOC and Temperature Effects

Charge transfer resistance follows a U-shaped curve versus SOC. It is lowest around 50 percent SOC and climbs sharply below 20 percent and above 80 percent. Diffusion impedance behaves similarly, with the worst performance at the extremes. Ohmic resistance stays nearly flat across the entire SOC range, which is why the 1000 hertz real impedance value works well as a stable reference point.

Temperature affects every part of the spectrum, but not equally. At low temperatures, the low-frequency diffusion tail dominates the impedance increase. At high temperatures, the charge transfer resistance drops because reaction kinetics speed up, but the SEI may start breaking down, causing Rsei to drift upward over time. This is why high-temperature EIS data often shows oscillations in the mid-frequency region, a signature of unstable interface chemistry.

Detecting Internal Short Circuits

One of the most practical applications of EIS is fault detection. When a cell develops an internal short, the ohmic resistance measured at high frequency, around 1000 hertz, increases noticeably. This parameter is stable across SOC and temperature in a healthy cell, so any drift is a red flag. Researchers have shown that monitoring the 1000 hertz real impedance value during cycling can catch over-discharge-induced internal shorts before any thermal event occurs. The method is fast, non-invasive, and works at any SOC.

Cycle Life and Degradation Tracking

Over hundreds of cycles, Rsei grows steadily as the SEI thickens. Rct may increase or decrease depending on whether the dominant degradation mode is loss of active lithium or loss of active material. The Warburg coefficient shifts as the diffusion pathways degrade. By fitting EIS data collected at regular intervals, you can separate these mechanisms and predict remaining useful life with far more confidence than capacity tracking alone.