Screening of Li-Based Solid Electrolytes Using Bond-Valence Methods and Graph Neural Networks

Stephen R Xie; Shreyas J Honrao; John W Lawson

NASA NTRS · 20240002676

Screening of Li-Based Solid Electrolytes Using Bond-Valence Methods and Graph Neural Networks

Stephen R Xie·Shreyas J Honrao·John W Lawson

Abstract

Li-based solid-state electrolyte (Li-SSE) materials enable safer, all-solid-state batteries but the computational search for candidates with favorable stability and Li-ion conductivity is challenging due to the size of the search space and the cost of evaluating transport properties with ab initio methods. We present a high-throughput screening approach for Li-SSE materials using a combination of bond-valence methods and graph neural networks. We demonstrate the screening approach with a dataset containing tens of thousands of Li-containing compounds. Furthermore, we combine the machine-learning screening procedure with an isovalent substitution scheme to generate and screen additional Li SSE candidates beyond existing databases. Finally, we discuss relative importances of geometric and bond-valence quantities in the training of graph neural networks, providing insight for future modeling of ionic conductivity in Li-SSE materials.

Explore related subjects

Materials discovery Solid state batteries

Keep this discovery

Explore connections, maps & timelines

BibTeX RIS

Stephen R Xie, Shreyas J Honrao, John W Lawson. Screening of Li-Based Solid Electrolytes Using Bond-Valence Methods and Graph Neural Networks. https://ntrs.nasa.gov/citations/20240002676

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related reports

DOE OSTI2025

Pressure-driven density match nucleates metastable r8 phases from amorphous Si and Ge

The pressure–temperature phase behavior of covalent disordered solids such as amorphous silicon and germanium is complex. Questions remain on possible glass transitions, on polyamorphism via amorphous–amorphous transitions, on connections with liquid–liquid transitions, on structure-behavior relationships, and on their potential as precursor for novel methods for material discovery. Here we demonstrate experimentally the nucleation of a metastable, four-fold coordinated rhombohedral r8 phase from pure amorphous silicon and germanium upon room temperature compression at pressures below 10 GPa. Accompanying theory reveals a strong pressure-driven distortion of the bond angle transforming the starting tetrahedral low-density amorphous network to a distorted four-fold coordinated medium-density state. This state is of lower density than metallic high-density networks, resembles the crystalline r8 phase and initiates its nucleation. Our finding shows that polyamorphism is not the only possible transformation mode for these amorphous solids and that instead nucleation of interesting functional phases at potentially useful pressures is possible. Such novel access modes to metastable structures are critical for future exploitability and could be useful for other tetrahedral materials including carbon, where the related (bc8) post-diamond phase remains elusive. Our observed density match between an amorphous and a metastable crystalline phase clearly allows for a new phase transition pathway, while corresponding theory demonstrates how carefully validated atomistic simulations can guide prediction, discovery and synthesis of novel material structures.

Materials discovery↗

NASA NTRSDate not supplied

Interpretable Tree-Based and Graph Neural Network Approaches for Novel Solid State Electrolyte Design

All-solid-state batteries with Li metal anode can address the safety issues surrounding traditional Li-ion batteries as well as the demand for higher energy densities. However, the development of solid electrolytes simultaneously possessing high ionic conductivity and good chemical and electrochemical stabilities has proven to be a challenge. I will present our informatics approach to explore the Li compound space for promising solid electrolytes using high-throughput multi-property screening and interpretable machine learning. This is accomplished through the generation of a large database of battery-related materials properties of Li compounds. We use tree-based ensemble learning methods and graph neural network approaches to accurately learn relationships between crystal structures and corresponding thermodynamic and kinetic properties, with interpretability being a major focus. Our models give us the ability to enable rapid discovery and design of novel solid-state battery chemistries.

Materials discovery↗

NASA NTRSDate not supplied

Hierarchical screening for Li-based solid electrolytes using fast, interpretable machine-learned potentials

Li-based solid-state electrolyte materials enable safer, all-solid-state batteries but the computational search for candidates with favorable stability and Li-ion conductivity is challenging due to the size of the search space and the cost of evaluating transport properties with ab initio methods. The prohibitive cost of high-throughput screening with DFT has lead to the development of surrogate models using geometric analysis, empirical potentials, and descriptors for ionic transport. Here, I will discuss a hierarchical screening approach for identifying promising materials using a combination of density functional theory, bond-valence methods, and machine learning potentials generated with the Ultra-Fast Force Fields (UF3) framework. We show how the inexpensive bond-valence method can be used to guide the generation of training samples for machine learning, in addition to filtering candidates. Finally, we apply the hierarchical workflow to screen for ionic conductivity across a database of Li-containing compounds.

Materials discovery↗