Model Order Reduction for Quantum Molecular Dynamics
Please login to view abstract download link
Quantum molecular dynamics (QMD) provides a high-fidelity physical framework for modeling atomic and molecular phenomena from first principles. However, its mathematical formulation requires solving iterative, high-dimensional electronic structure problems at every physical state step. Within the broader data-driven computational mechanics paradigm, we present a projection-based model order reduction framework for electronic structure calculation in QMD. By discovering a low-dimensional manifold of the electronic solution space within Kohn-Sham density functional theory, our approach projects the governing variational formulation onto an intrinsically reduced subspace. This avoids full-space iterative wavefunction optimization while preserving rigorous physical symmetries and state consistency. We demonstrate this methodology on a water molecule, demonstrating strong agreement with high-fidelity reference solutions across structural trajectories and thermodynamic observables through canonical coordinate alignment and structured manifold sampling.
