Sandwiched between the electrolyte and electrodes in a lithium-ion battery is a remarkably thin layer of material that strongly dictates battery performance and durability: the interphase. Although typically only a few to tens of nanometers thick, interphases are among the least understood components of an operating battery cell. Their complex and constantly evolving structures make it challenging to determine how their atomic features and microscopic variations translate into macroscopic battery performance.
In a recent study, published in Advanced Energy Materials, researchers at Lawrence Livermore National Laboratory (LLNL) used large-scale molecular dynamics simulations accelerated by machine learning, combined with data-driven analysis, to tackle this challenge. Their approach disentangles the dynamically evolving chemistry and structure at battery interphases and reveals how this complexity impacts battery functionality.
"Typically, we expect these spontaneously formed interphases to be detrimental to battery performance. They are one of the main causes of capacity fade and performance degradation in batteries," said LLNL scientist and author Sabrina Wan. "But if we can understand and control their functionality through targeted design, they could instead be beneficial."
Interphases are thin layers that form when electrodes and electrolytes chemically and electrochemically react during battery fabrication and operation. Characterized by highly variable and constantly changing structures, interphases develop their own compositions and properties - often markedly different from those of the bulk materials on either side.
"Unless we can understand and control the interphase, we cannot identify tractable strategies to improve cell performance, safety and longevity, all of which are critical for economic considerations and safe operation of batteries," said Wan.
Experimentally probing these buried interphase structures is extremely challenging, requiring exceptionally high spatial and temporal resolution as well as sensitivity to their complex chemistry.
Addressing these complexities computationally has been a longstanding challenge in the battery field as well. Quantum-mechanical simulations can accurately describe chemical reactions and interactions among different materials, but their high computational cost limits simulations to relatively small systems - typically only a few hundred atoms. On the other hand, classical approaches can model much larger systems and extend into the relevant nanometer regime, but they cannot capture the chemical reactions that drive interphase formation and its continued evolution during battery operation.
This creates a fundamental computational challenge: how can researchers simultaneously capture the chemical reactions within an interphase structure and the length and time scales needed to understand how it works?
"Enabled by recent advances in machine-learning methodologies, we now have the tools to address this grand challenge and bridge these scales with high fidelity," said Wan.
By training a highly scalable and accurate machine-learning model on quantum-mechanical data, the researchers extended atomistic simulations to the length and time scales needed to resolve the structure-property relationships governing interphases. They were able to examine the actual size scale of the interphase with millions of atoms at the nanosecond timescale of chemical reactions.
"Especially when coupled with data-driven post-analysis approaches, we can establish statistically meaningful correlations between the variation within these interphase structures and their performance," said Wan.
The authors demonstrated the approach on a lithium fluoride/lithium carbonate system that has long posed a puzzle in the battery research community. Bulk lithium fluoride has intrinsically poor conductivity, but, counterintuitively, interphases rich in that material remain functional during battery operation. The new simulations reveal that mixtures of lithium fluoride and other chemical species within the interphase create practical pathways for ion transport.
These simulation-derived insights directly translate into actionable strategies for the design and optimization of interphase structures with targeted properties and performance.
Beyond lithium-ion batteries, the simulation platform provides a broadly applicable framework for studying complex, reactive interphases in other areas of materials science and energy technology, including catalysis, fusion materials and weapons science, where evolving interface chemistry plays a critical role in material performance and degradation.
"That's really the value of this - it isn't limited to batteries," said Wan.
