Learning Time-Dependent Density Functional Theory via Geometry and Physics Aware Latent Evolution
Keywords: AI for science, Density functional theory, Real-time TDDFT, Neural PDE solver
Abstract: We consider using machine learning to simulate time-dependent density functional theory (TDDFT) to predict physical properties of molecules and materials beyond their ground states. In particular, by simulating the electronic response of the system under an external electromagnetic field, the optical absorption spectrum can be calculated using real-time TDDFT (RT-TDDFT), which provides physical information about the excited states and dipole strength function. However, RT-TDDFT simulation requires the direct propagation of electronic wavefunctions of all valence electrons for extended periods, making the process very time-consuming. In this work, we model electron density as volumetric data and train neural networks to map between coarse time steps. To make the model aware of the atomistic environment, we incorporate 3D message passing into the model architecture. Additionally, we use latent evolution to regularize the model towards learning the underlying physics. Our method is termed TDDFTNet. To evaluate our approach, we generate datasets using molecules from the MD17 dataset. Results show that TDDFTNet can learn the time propagation of electron densities accurately and efficiently.
Primary Area: applications to physical sciences (physics, chemistry, biology, etc.)
Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics.
Submission Guidelines: I certify that this submission complies with the submission instructions as described on https://iclr.cc/Conferences/2025/AuthorGuide.
Reciprocal Reviewing: I understand the reciprocal reviewing requirement as described on https://iclr.cc/Conferences/2025/CallForPapers. If none of the authors are registered as a reviewer, it may result in a desk rejection at the discretion of the program chairs. To request an exception, please complete this form at https://forms.gle/Huojr6VjkFxiQsUp6.
Anonymous Url: I certify that there is no URL (e.g., github page) that could be used to find authors’ identity.
No Acknowledgement Section: I certify that there is no acknowledgement section in this submission for double blind review.
Submission Number: 12913
Loading