Workshop I: Quantum Chemistry and Electronic Structure

Electronic structure methods underpin predictive simulations in chemistry and materials science and provide the reference data used to train machine learning models. AI creates opportunities to improve these methods while placing additional emphasis on consistency across chemical environments, reliable derivatives, and automated calculations at scale. These requirements are particularly demanding for transition metals, excited states, strongly correlated systems, and extended materials, where reference calculations can remain costly or unreliable.
The workshop will bring together quantum chemists, materials scientists, applied mathematicians, and ML researchers to examine both how AI can advance electronic structure calculations and how electronic structure theory can provide reliable foundations for AI. Topics include exchange–correlation functionals in density functional theory (DFT), coupled-cluster and multireference methods, Green’s-function approaches, tensor networks, and quantum Monte Carlo, including neural representations of electronic states. When are conventional DFT approximations sufficient, and when are higher levels of theory necessary? How can physical constraints and mathematical structure guide learned functionals, electronic states, and algorithms? Mathematical questions include variational consistency, convergence and stability, error estimation, and certification.

A second theme is the generation of high-quality reference data. How should datasets balance accuracy, chemical diversity, and computational cost? Numerical errors must be distinguished from errors in the underlying electronic structure approximation, and both must be considered when assessing models trained on these data. The workshop will examine how systematic errors propagate into learned predictions, how reference calculations should be selected and validated, and how data provenance, reproducible workflows, and open datasets can support reliable large-scale learning.

Particular attention will be given to bridging quantum chemistry and materials science. Many recent AI developments have focused on molecules, while high-accuracy benchmarks for periodic materials remain scarce. Transferring methods between these settings requires understanding electronic screening, delocalization, finite-size effects, and the limits of locality. How can reliable reference calculations be established when experimental validation is limited or comparisons involve additional physical approximations? The aim is to connect advances in electronic structure theory and its mathematical foundations to the reliable training and assessment of the reduced-order and machine learning models explored in subsequent workshops.