Abstract
<jats:p>Lithium metal anodes offer significantly higher specific capacity than graphite anodes. However, inhomogeneous lithium deposition leads to dendritic microstructures that cause short circuits and limit industrial implementation. Characterization methods with high sensitivity to lithium metal are essential to understand growth mechanisms and develop mitigation strategies. Electron paramagnetic resonance (EPR) spectroscopy can monitor lithium dendrites, yet conventional cavity-resonator systems cannot selectively probe the deposited lithium, which is convoluted with the bulk lithium signal from the counter electrode. Here, we introduce an EPR-on-a-chip (EPRoC) approach that restricts the sensing volume to plated lithium on the negative electrode, simplifying spectral interpretation and enabling detection of the early onset of lithium plating. Using a physical model for conduction EPR (CEPR), the operando EPRoC signal of lithium plating on copper is utilized as a surrogate for anode-less batteries, enabling identification of charging stages associated with dendritic growth. Combined with a thin-film CEPR signal approximation and a triple-phase stochastic model optimized via deep learning, our approach maps EPR signals to dendritic growth characteristics and provides insight into lithium deposition behavior for specific cell conditions, offering a pathway toward a new generation of lithium-ion and anode-less batteries.</jats:p>