🔬 materials science

Capturing the calendering U-shape in lithium-ion electrode thermal conductivity

This paper introduces a calendering-aware extension of the Zehner–Bauer–Schlunder model that successfully captures the non-monotonic, U-shaped evolution of through-plane thermal conductivity in lithium-ion electrodes by accounting for process-dependent microstructural changes, thereby reducing prediction error from 31.1% to 4.5% across various graphite and NMC formulations.

Original authors: Julius Störk

Published 2026-07-14

📖 6 min read🧠 Deep dive

CC BY 4.0

Original authors: Julius Störk

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine a lithium-ion battery as a bustling city made of tiny, porous neighborhoods. Heat is the traffic trying to get out of this city. For the city to stay safe and run smoothly, that traffic needs to flow easily through the layers of the battery. But here's the twist: the way we squeeze these layers during manufacturing—called "calendering"—changes the traffic rules in a weird, unexpected way.

For decades, scientists thought that if you squeezed a battery layer tighter, the heat would just flow better and better, like a crowded hallway getting wider as people push closer together. But new research shows that's not the whole story. Instead, the heat flow does a U-turn.

The U-Shape Mystery

When researchers squeezed a battery layer, the heat flow didn't just go up. It went down a little bit first, hit a low point, and then went up. It's like trying to walk through a hallway: at first, squeezing the walls together makes it harder to move because the floor gets slippery or the path gets blocked. But if you squeeze it even harder, the walls finally lock together, creating a solid bridge that lets you zip right through.

The paper's main discovery is a new mathematical "recipe" (a closure) that finally explains this U-shape. Before this, the best recipes could only predict a straight line going up. This new recipe adds a special ingredient: a contact term that changes as you squeeze. It accounts for the fact that squeezing first damages the tiny bridges between particles, then rebuilds them into stronger, interlocking connections.

The Two Suspects: Broken Bridges vs. Spinning Flakes

The researchers found two different reasons for this U-shape, depending on which part of the battery you look at:

The Cathode (The "Bridge" Story): In the positive side of the battery (the cathode), the particles are like round marbles. When you squeeze them, the tiny bridges holding them together get damaged first (making heat flow worse), but then they smash together and form new, stronger bridges (making heat flow better). The paper proves this is the main reason for the U-shape here.
The Anode (The "Flake" Story): In the negative side (the graphite anode), the particles are flat, like tiny pancakes or flakes of graphite. The paper suggests a clever second possibility: as you squeeze, these pancakes might rotate to lie flat against the floor. Since graphite conducts heat poorly when you try to push it through its thickness (like trying to push water through a stack of paper), this rotation actually lowers the heat flow before the squeezing gets strong enough to fix it.

Here is the catch: The paper says that looking at heat flow alone isn't enough to tell us for sure if the anode is doing the "bridge damage" thing or the "spinning flake" thing. Both stories fit the data almost perfectly. The authors suggest the "spinning flake" idea is a bit more elegant and fits the physics of graphite, but they admit they can't prove it without taking a special X-ray picture of the flakes. Until then, both ideas remain on the table as valid suspects.

The "Knudsen" Effect: The Gas That Disappears

There's another invisible player in the game. Inside the tiny pores of the battery, there is gas (or liquid). When the pores are super small, the gas molecules bounce off the walls more than they bump into each other. This is called the Knudsen effect.

The paper shows that this effect is huge. In the tiny pores of a separator (the wall between the battery layers), the gas loses 84% of its ability to carry heat because the pores are so small. If you ignore this, your math is way off. The new recipe includes this "Knudsen correction" to make sure the numbers are right.

What This Means for the Future

The authors didn't just write a theory; they built a tool that can run backwards. Because the math is "differentiable" (smooth and easy to calculate), you can plug in a heat measurement and instantly figure out how porous the battery layer is.

The Simulation: In computer simulations, this tool can guess the porosity of a battery layer with an error of only 0.008 (that's incredibly precise!).
The Real World: When they tried it on real data, it was a bit coarser (around 0.05 error), but still useful for catching bad batches.
The Goal: The ultimate dream is to put this on a factory line. Imagine a camera that measures heat in microseconds and tells the factory robot, "Hey, squeeze a little less, the bridges are breaking!" This could help make better batteries faster.

What the Paper Rules Out

The paper explicitly says that old models which only look at "how empty the space is" (porosity) are wrong. They can't explain the U-shape. They also rule out the idea that the graphite particles stay in the same orientation; the data suggests they must be moving or the bridges must be breaking.

How Sure Are They?

Proven: The U-shape exists, and the new recipe predicts it with an error of 4.5% on the data they tested. This is as good as the measurement tools themselves.
Highly Likely: The "contact damage and rebuild" story for the cathode is strongly supported.
Suggested but Not Proven: The "spinning flake" story for the anode is a strong candidate, but it's still a "degenerate" solution (meaning it's tied with the other story). The authors say, "We think this is it, but we need an X-ray to be 100% sure."
Simulated Only: The idea of using this for real-time factory control is a "feasibility study." It works in the computer, but the paper hasn't tested it on a live factory line yet.

In short, this paper gives us a new, smarter map for navigating the heat inside batteries. It shows us that squeezing isn't just about making things denser; it's a delicate dance of breaking and rebuilding tiny bridges, and sometimes, spinning the particles just right. And while we have a great map, the authors are honest: we still need to take a few more pictures to see exactly how the dancers are moving.

Technical Summary: Capturing the Calendering U-Shape in Lithium-Ion Electrode Thermal Conductivity

Problem Statement
Thermal management in lithium-ion batteries relies on accurate models of effective through-plane thermal conductivity ( $\lambda_{eff}$ ) for every porous layer. A critical gap exists in modeling the dependence of $\lambda_{eff}$ on the calendering process (roll-pressing). Experimental data reveals a non-monotonic, "u-shaped" dependence of conductivity on porosity during calendering: conductivity initially drops before rising again. Existing analytical closures, such as porosity-only models or static-contact effective-medium theories (e.g., Bruggeman, Maxwell-Eucken, standard Zehner–Bauer–Schlünder), fail to reproduce this u-shape, predicting only a monotonic rise. Furthermore, standard models often ignore the Knudsen effect (rarefied gas conduction in micrometer pores) and the evolution of particle contact networks or crystallographic reorientation during processing. This lack of an analytical closure that explains the u-shape introduces process-dependent errors that can approach an order of magnitude.

Methodology
The authors propose a calendering-aware closure built upon the Zehner–Bauer–Schlünder (ZBS) effective-medium theory, enhanced with three specific capabilities:

Knudsen-Corrected Base: The model incorporates the Smoluchowski/Knudsen correction for pore-gas conductivity, which is significant in micrometer-scale pores (especially under helium purge used in Laser Flash Analysis, LFA).
Calendering-Aware Contact Term: To address the u-shape, the authors introduce a compression-indexed contact term, $\phi(\Pi)$ , where $\Pi$ is the compression rate. This term models the evolution of the solid-solid bridge network. It is formulated as a quadratic function, $\phi(\Pi) = \max(0, \phi_0 + a\Pi + b\Pi^2)$ , to capture the initial damage to the binder network (shear, $a < 0$ ) followed by the recovery via particle interlocking ( $b \geq 0$ ) at higher compression.
Anisotropic Reorientation (Graphite Anode): For graphite anodes, the model includes a bounded reorientation mechanism. Graphite flakes are highly anisotropic ( $\lambda_a \gg \lambda_c$ ). The model posits that calendering aligns flakes in-plane, forcing the through-plane heat path to cross the weak $c$ -axis. This is modeled via a Hermans orientation factor $S(\Pi)$ that evolves with compression, reducing the effective solid conductivity $\lambda_s^\perp$ toward the $c$ -axis limit.

The model is implemented as a differentiable, low-dimensional closure using JAX, enabling both forward prediction and inverse inference (e.g., estimating porosity from conductivity). The calibration targets the apparent through-plane coating conductivity derived from LFA measurements after subtracting the current collector contribution.

Key Contributions

Primary Contribution: A calendering-aware closure that successfully reproduces the u-shaped dependence of $\lambda_{eff}$ on porosity across four electrode families (two graphite anodes, two NMC cathodes). This is the first low-dimensional variant found to capture this behavior, reducing the in-sample Mean Absolute Percentage Error (MAPE) from 31.1% (porosity-only) to 4.5%.
Secondary Contribution (Anode Specific): A bounded reorientation account for graphite anodes that explains the anomalously low fitted solid conductivity by aligning flakes toward the $c$ -axis. While conductivity data alone cannot uniquely distinguish this from a contact-only fit, the reorientation model is more parsimonious and physically bounded.
Differentiable Inverse: The implementation allows for the inversion of the closure to infer coating porosity or interface quality from thermal measurements, supporting inline quality control.
Identifiability Analysis: The paper rigorously addresses parameter identifiability given the small dataset (6–8 states per family). It reports grouped quantities (e.g., bridge conductance $G_0 = \phi_0 \lambda_b$ ) and profile-likelihood valleys rather than claiming unique point estimates for confounded parameters like $\lambda_s$ and $\phi_0$ .

Results

Accuracy: The calibrated closure achieves an in-sample MAPE of 4.5% across 27 calendering states, matching the estimated uncertainty of the reconstructed LFA target. The held-out "leave-one-state-out" error is 7.8%.
Mechanism Discrimination:
- Cathodes: The u-shape is driven entirely by contact-network evolution (damage and recovery). Reorientation is inert for quasi-isotropic NMC particles.
- Anodes: Both a contact-only model and a reorientation-coupled model fit the data equally well (MAPE ~1.8–5.4%). The reorientation model is favored for parsimony and its ability to explain the low $\lambda_s$ via alignment to the $c$ -axis, but it remains degenerate with the contact model until same-sheet XRD texture is measured.
External Consistency: The model passes several consistency checks:
- It predicts the correct sign of the temperature coefficient (negative, indicating solid-control) without fitting temperature data.
- It predicts in-plane conductivity ( $\lambda_{\parallel}$ ) consistent with independent photothermal measurements, despite being fitted only to through-plane data.
- It extrapolates to a commercial 18650 cell stack with ~15% error when interface resistance is inferred.
Inverse Feasibility: On synthetic data, the differentiable inverse recovers porosity with an RMSE of ~~0.008. On real data, the error is coarser (~~0.05) due to model-form residuals and metrology-state coupling, but it remains a viable tool for secondary checks.

Significance and Claims
The paper claims that the proposed closure is the first analytical, low-dimensional model to capture the calendering u-shape, bridging the gap between computationally expensive microstructure-resolved simulations and simple empirical fits. Its significance lies in:

Process Resolution: It explicitly links thermal conductivity to processing variables (compression), moving beyond static porosity-based models.
Mechanism Insight: It identifies that the u-shape arises from contact network evolution (damage then interlocking) and, for anodes, potentially from crystallographic reorientation.
Practical Utility: The differentiable nature of the closure enables its use as a "soft sensor" for inline quality control (porosity and interface monitoring) in continuous electrode manufacturing, a capability not offered by resolved simulations or purely data-driven regressions.
Methodological Scope: The authors are modest about the absolute values of fitted parameters, noting they are "method-scoped" to the LFA-derived target and shift significantly (1.8–2.4×) if recalibrated against guarded-hot-plate data. The robust findings are the grouped parameters, trend directions, and the ability to discriminate mechanisms via orthogonal measurements (XRD, adhesion, SEM).

The paper concludes that while the contact-versus-reorientation split for anodes remains unresolved by conductivity data alone, the closure provides a falsifiable framework. The proposed next step is a dedicated experimental campaign involving two-pressure gas measurements, XRD texture analysis, and paired interface metrology to definitively separate these mechanisms and validate the closure's physical assumptions.