Julius Störk | Searcharxiv

arXiv2026

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

Calendering is a key manufacturing step in lithium-ion electrode production, increasing volumetric energy density by reducing electrode porosity. Its effect on through-plane effective thermal conductivity, however, can be non-monotonic: measurements of graphite-based anodes show an initial decrease in thermal conductivity during early calendering followed by recovery at higher compaction. Conventional porosity-based effective-medium closures cannot reproduce this U-shaped behaviour. We develop a calendering-aware extension of the Zehner--Bauer--Schlünder model that combines a Knudsen-corrected porous-medium baseline with a compression-indexed contact contribution. For graphite electrodes, the model represents the competing effects of increasing particle contact and calendering-induced reorientation of anisotropic graphite particles, which initially reduces favourable through-plane heat-transport pathways. For quasi-isotropic NMC cathodes, the observed response is instead captured through process-dependent contact-network evolution. Across 27 calendering states spanning thin and thick graphite anodes and NMC622 and NMC811 cathodes, the proposed closure reduces the mean absolute percentage error from 31.1% for the zero-fit reference model to 4.5%. The result shows that incorporating process-dependent microstructural evolution is necessary to capture the measured conductivity minimum. Validation across additional electrode formulations, thicknesses, and chemistries remains necessary to assess transferability.

cond-mat.mtrl-sci↗

arXiv2026

Interference and Retention in Continual Learning

Continual learning commonly relies on post-hoc mechanisms such as replay, elastic regularization, or distillation. This work argues that forgetting should instead be modeled directly as interference between tasks. In the frozen-feature regime, forgetting from learning a new task is exactly the interference energy induced on the old task. In deep networks, the same quantity is recovered through path-averaged curvature with minimal additional forward passes. When task supports are disjoint, forgetting can be eliminated structurally and when task supports overlap in conflicting directions, a non-zero distortion floor is unavoidable. The same geometry optimally merges models through task-aware orthogonalization. From this analysis we derive Interference-Gated Functional Allocation (IGFA), a replay-free, Fisher-free method that shares directions when tasks align and protects them when they conflict. Across benchmarks, IGFA achieves lossless retention when tasks are structurally separable and moves unavoidable cost from irreversible forgetting into deferred but recoverable plasticity when they are not. It matches the strongest replay-free structural baselines on dissimilar-task streams and improves on unconditional projection when similarity makes transfer worth preserving.

cs.LG↗