Dresden 2026 – scientific programme
Parts | Days | Selection | Search | Updates | Downloads | Help
O: Fachverband Oberflächenphysik
O 47: New methods: Theory
O 47.4: Talk
Tuesday, March 10, 2026, 15:15–15:30, HSZ/0201
AI-Driven Optimization Techniques for Density Functional Theory Calculations — Álvaro Fraile-Carmena1, Damián Sánchez-Maqueda1, Cristian Ramírez-Atencia1, and •María Camarasa-Gómez2 — 1Universidad Politécnica de Madrid, C/ Alan Turing, s/n, Madrid 28031, Spain — 2Centro de Física de Materiales CFM/MPC (CSIC-UPV/EHU), Paseo de Manuel Lardizábal 5, Donostia-San Sebastián 20018, Spain
Accurately determining optoelectronic properties of molecules and solids is an ongoing challenge in first-principles methods. This objective has gained even more relevance, as research on quantum materials evolves rapidly. So far, density functional theory (DFT) has remained the primary computational tool for this task. However, reaching with DFT accuracy levels comparable to many-body perturbation theory still requires significant computational effort [1]. Here we introduce an approach that integrates AI-driven optimization techniques into DFT workflows [2]. Embedding these optimization methods directly in ab initio software enhances both accuracy and efficiency. We present results obtained with surrogate models and state-of-the-art open source optimization libraries [3], showing their potential to accelerate and improve electronic structure simulations. [1] M. Camarasa-Gómez, S. E. Gant, et al., npj Comput. Mater. 10, 288 (2024) [2] Á. Fraile-Carmona, D. Sánchez-Maqueda, C. Ramírez-Atencia, and M. Camarasa-Gómez (2025) [in preparation] [3] J. Blank, and K. Deb, IEEE Access 8, 89497 (2020)
Keywords: optoelectronic properties; artificial intelligence; optimization; DFT; surrogate models