Phantom LAM and LLI: Resistance and Hysteresis Bias in Voltage-Curve Degradation Mode Analysis

Abstract: Degradation mode analysis (DMA) is widely used to decompose capacity fade into loss of lithium inventory (LLI) and loss of active material (LAM) from low-rate voltage-capacity data. Yet the measured trace is a pseudo-OCV (pOCV) that includes two non-degradation contributions: an SOC-dependent ohmic drop and intrinsic charge-discharge hysteresis, especially in graphite--silicon oxide (C/SiOx) negative electrodes. We show these can dominate attribution and generate Phantom LAM/LLI --apparent material loss created by curve registration, branch choice and voltage-windowing rather than true degradation. Using two commercial 21700 cells (LG M50T: higher resistance; Molicel P45B: lower resistance), we extract an SOC-dependent instantaneous resistance $R_Ω(\mathrm{SOC})$ from the first $\sim$50,ms pulse step and apply an IR correction to pOCV before fitting. In LG M50T, IR correction lifts the low-rate discharge pOCV by $+13$--$27$,mV with ageing; without it, PE-LAM is increasingly under-diagnosed (to $-8.80%$ relative error at late life) and LLI is suppressed (median $-3.07%$), with compensating inflation of apparent graphite loss. In P45B, on a branch-fair $3.0$--$4.2$,V window, end-of-life charge-branch DMA reports higher PE-LAM ($+3.42$,pp) and LLI ($+5.36$,pp), while the discharge branch recovers larger Si-LAM (discharge--charge difference to $+14.38$,pp). Raising the lower cutoff ($2.5$--$4.2 \rightarrow 3.0$--$4.2$,V) further under-reports Si-LAM by $13.61$,pp by removing the Si-sensitive low-voltage tail. We propose a practical protocol: correct only the instantaneous ohmic term, harmonize the voltage window, and base quantitative attribution on the discharge branch, treating anomalous/negative component LAMs on charge as allocation artefacts rather than recovery.