Integrated computational and autonomous discovery for battery materials

Designing next-generation battery materials requires navigating an enormous design space defined by materials chemistry, physical properties, and structural variables. Our group addresses this challenge through an integrated computational framework that spans atomistic simulation, machine learning, and autonomous experimentation. At the atomistic level, we employ Density Functional Theory (DFT), Ab initio Molecular Dynamics (AIMD), and Machine-Learned Interatomic Potentials (MLIP) to investigate the fundamental properties of battery electrodes and electrolytes, from electronic structure and Li⁺ migration barriers to large-scale ionic transport and long-time structural evolution. To bridge the gap between atomic-scale understanding and materials-level performance, we develop machine learning models that predict key properties including lifetime, electrochemical performance, and optimal electrolyte composition, enabling high-throughput screening across vast compositional spaces. Ultimately, these computational insights are seamlessly integrated into a Self-Driving Laboratory (SDL) platform, where automated electrolyte synthesis, real-time electrochemical data acquisition, and closed-loop AI-driven optimization work in concert to accelerate the discovery and validation of superior battery materials. By connecting first-principles physics to autonomous experimentation, our group aims to fundamentally transform how battery materials are designed and discovered.

Multiscale simulation of cathode and solid electrolyte materials

Our group employs a hierarchical simulation approach integrating DFT, AIMD, and MLIP to investigate the fundamental properties of battery materials from first principles. For layered oxide cathodes, we investigate key atomistic processes including oxygen dimerization pathways via NEB(Nudged elastic band) calculations, defect formation energies, and Li/Ni mixing energies, providing mechanistic understanding of structural degradation and redox activity at the atomic scale. For solid electrolytes, we perform systematic DFT calculations of binding energies, structural stability, and mechanical properties across various dopant configurations to establish design principles for superionic conductors. Extending beyond DFT, AIMD directly yields Li⁺ diffusivity and transport mechanisms at finite temperature. Finally, MLIP trained on DFT datasets retains first-principles accuracy while dramatically reducing computational cost, enabling simulations over extended length and time scales that are otherwise inaccessible.