ACS Publications: OMol25-Trained MLIPs Enable High-Accuracy, High-Throughput Na-ion Battery Electrolyte Solvation Structure Prediction with Experimental Verification

August 17, 2026

ACS Publications USA

Overview

An ACS Publications paper demonstrates that OMol25-trained machine learning interatomic potentials (MLIPs) provide an efficient and accurate route to predictive, high-throughput electrolyte simulations for next-generation Na-ion batteries. Specifically, UMA-OMol, a particular MLIP, offers near-DFT accuracy at significantly reduced computational cost compared to direct DFT molecular dynamics. This breakthrough holds broad promise for modeling complex electrochemical systems with chemical transferability, facilitating a molecular-level understanding of electrolyte solvation structure crucial for battery chemistry.

In Depth

Key Findings

A paper published in ACS Publications has demonstrated that machine learning interatomic potentials (MLIPs) trained with the OMol25 dataset offer an efficient and highly accurate pathway for predictive, high-throughput simulations of electrolytes for next-generation sodium-ion (Na-ion) batteries. Notably, the specific MLIP dubbed UMA-OMol achieves near-Density Functional Theory (DFT) accuracy while drastically reducing computational costs compared to direct DFT molecular dynamics simulations.

Technical / Clinical Details

Understanding electrolyte behavior, particularly its solvation structure, is paramount for determining battery performance. However, atomic-level electrolyte simulations using high-accuracy quantum chemical methods like DFT demand immense computational resources and time. This research resolves this bottleneck with MLIPs. MLIPs are machine learning models that construct interatomic interaction potentials from a small number of DFT calculation data points. Once trained, they can simulate large atomic systems much faster than DFT while maintaining comparable accuracy. UMA-OMol, an MLIP trained on the extensive OMol25 molecular database, exhibits high “chemical transferability” across various chemical environments, making it versatile for application to unknown electrolyte systems. Experimental verification showed that UMA-OMol could predict the crucial solvation structure of electrolytes—how solvent molecules surround Na ions—with accuracy that closely matches experimental results. This enables researchers to screen numerous electrolyte candidates and evaluate their properties in a short time, accelerating battery material exploration by orders of magnitude compared to conventional simulation methods.

Background & Context

Sodium-ion batteries are gaining attention as a next-generation energy storage system due to concerns over resource scarcity and rising costs of lithium-ion batteries. The commercialization of Na-ion batteries critically depends on the development of high-performance and safe electrolytes. However, their molecular-level behavior is complex, making it challenging to design optimal electrolytes based solely on empirical rules. AI and materials informatics offer powerful solutions to this challenge. The use of high-accuracy MLIPs efficiently predicts electrolyte stability, ion conductivity, and interfacial properties, thereby accelerating the design process. This plays a crucial role in driving innovation in battery technology and contributing to the realization of a more sustainable energy society.

Strategic Significance & Outlook

The success of OMol25-trained MLIPs like UMA-OMol has the potential to fundamentally transform the paradigm of electrolyte design in next-generation battery chemistry. This methodology will be broadly applicable not only to Na-ion batteries but also to other energy storage systems such as lithium-ion and all-solid-state batteries, as well as general complex electrochemical systems like fuel cells and electrocatalysts. By integrating AI with high-accuracy simulation tools, the R&D cycle in materials science will be dramatically shortened, and molecular-level understanding of phenomena previously only theoretically predicted will deepen. In the future, these MLIPs are expected to function as part of “self-driving laboratories,” where AI autonomously manages the entire process of electrolyte design, synthesis, characterization, and the generation of new insights from data, achieving “closed-loop optimization.” This will lead to the rapid market introduction of high-performance next-generation batteries, accelerating the transition to a sustainable society.

Source: https://pubs.acs.org/jpclcd/article/17/31/9073/5231863/Prediction-and-Experimental-Verification-of

Get our weekly technology intelligence — free

Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.

Subscribe Free — Weekly Tech Intelligence

By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.