AI Framework SCAN Accelerates Discovery of High-Performance Battery Electrolytes

Researchers led by Dr. Zhilong Wang and Prof. Fengqi You have developed a new AI framework designed to drastically accelerate the discovery of high-performance battery electrolytes. The system, known as SCAN (Shaping Conductivity Atlas for Non-aqueous electrolytes), provides a scalable method for predicting ionic conductivity in energy storage materials.

The SCAN framework utilizes a dynamic routing-guided, interpretable AI model to predict ionic conductivity based on the interaction between Li-salts, solvents, and specific environmental conditions. According to research published in Nature Computational Science (Volume 6, pages 271–284, 2026), the system allows researchers to navigate a vast salt-solvent design space more efficiently than previous methods. To support the scientific community, the team has released an open-source software package via GitHub (PEESEgroup/SCAN) and a corresponding web-based design platform. This software enables users to calculate descriptors for Li-salts and solvents, train their own models, and predict conductivity using either the core SCAN model or symbolic regression.

The Challenge of Electrolyte Design

Traditionally, the development of battery electrolytes has relied on a slow process of trial-and-error experimentation or computationally expensive molecular dynamics (MD) simulations. Because the number of possible combinations of salts and solvents is immense, identifying the optimal materials for next-generation batteries has remained a significant bottleneck. The search space is effectively an "underexplored universe," where the cost and time required for traditional simulation often outpace the need for rapid innovation in energy storage.

Implications for Energy Storage

By providing an interpretable AI map of this electrolyte universe, SCAN reduces the industry's reliance on costly physical experiments and resource-heavy simulations. This shift toward AI-driven material discovery is critical for the development of batteries with higher energy density, improved safety profiles, and faster charging capabilities. By targeting non-aqueous electrolytes specifically, the framework addresses the core components needed for advanced battery and energy storage applications, potentially shortening the timeline from theoretical design to commercial deployment.

Future Outlook

With the release of the open-source tools, the focus now shifts to how the broader research community integrates SCAN into their workflows to validate new material candidates. While the framework provides a powerful predictive tool, the next phase of development will likely involve the empirical testing of the high-conductivity candidates identified by the AI. Observers will be watching to see if this interpretable approach leads to a breakthrough in stable, high-capacity battery chemistries that can move beyond current lithium-ion limitations.