Low‐loss STEM ‐ EELS analysis of beam‐sensitive lithium‐ion negative electrodes

Abstract

Silicon represents one of the most promising anode materials for next generation lithium‐ion batteries. However its colossal volume expansion (up to 300%) upon electrochemical reaction with lithium repeatedly exposes fresh surfaces to electrolyte solvent oxidation 1‐2 . This leads to very high irreversible capacities, compounded by the fact that parts of the silicon‐based electrodes are progressively disconnected from both electrical and ionic transport networks as the solid electrolyte interface (SEI) accumulates. Deeper insight into these degradation phenomena is critical to engineer adequate electrodes and/or electrolytes. Characterization of these electrodes has so far mostly focused on either bulk or surface analysis, both lacking spatial resolution. Little is known about the SEI's morphology in silicon nanoparticles (SiNPs) aggregates that make the electrode, or about the evolution of these nanoparticles themselves with cycling. Attempts to characterize this system through electron microscopy have been severely limited by the radiolysis and sputtering damage, respectively, undergone by the SEI and lithium‐silicon alloys (Li x Si). In this work we demonstrate the possibility to map major SEI and electrode components such as lithium carbonate (Li 2 CO 3 ), lithium fluoride (LiF) and lithium oxide (Li 2 O) as well as quantifying lithium‐silicon alloys compositions 4 and Si crystallinity from a single dataset by combining scanning transmission electron microscopy and low‐loss electron energy loss spectroscopy 5 (STEM‐EELS) (fig. 1). The low‐loss part of the EEL spectrum is considerably more intense than its high energy counterpart and contains both the Li K‐edge and plasmons. Fine tuning of the experimental parameters allows us to acquire low‐loss spectrum images with good signal‐noise ratios within timeframes compatible with minimal sample degradation. Plasmons can then either be used as unique molecular signatures for the SEI, or directly for quantification in the case of Li x Si compounds (fig. 2 inset). This can yield unique insight into electrode degradation phenomena through careful data processing (MLLS, Drude model fit…). Large spectrum images can be acquired within short timeframes (~10 ms/voxel), making this method a powerful and practical diagnostics tool for battery electrodes and other beam‐sensitive nanostructured systems. Results on electrodes disassembled from full cells at their 1 st , 10 th and 100 th charge and discharge, with a limited capacity of 1200 mAh/g, shed light on the SEI's deposition mechanism and morphological as well as chemical evolution along cycling for different electrolytes. Strong correlations were observed between the SEI's local chemistry and our nanoparticles cycling performance (fig. 2). Lithiation was also observed to proceed preferentially along grain boundaries, resulting in different behaviours between mono‐ and polycrystalline silicon powders.