Atomistic simulation methods are undergoing a significant transformation due to the integration of machine learning (ML) techniques, which allow for highly accurate short-range interatomic potentials in molecular dynamics (MD) simulations. Despite this advancement, there are still critical limitations that hinder their broader application. Short-range potentials fail to capture essential long-range electrostatic interactions, charge relaxation processes, and important non-local quantum-mechanical (QM) effects. These shortcomings become particularly evident in simulations involving polarizable complex chemical systems where reactive dynamics, such as bond-breaking, bond-formation, charge transfer, and ionic fragmentation, play a crucial role. To address these limitations, a new class of atomistic models is being developed that combines ML approaches with explicit electronic structure calculations. These hybrid approaches, including flexible charge models as well as semi-empirical quantum chemistry methods, aim to bridge the gap between classical atomistic simulations and high-level ab-initio methods, enabling more accurate predictions in systems where long-range interactions and non-local effects are crucial.

This workshop will bring together experts from various disciplines to advance the development and application of innovative atomistic models, encouraging collaboration to address the challenges in simulating complex and reactive chemical systems. The workshop will cover a range of topics, including the creation of novel reduced complexity approaches, ML for coarse-grained flexible charge models and semi-empirical QM methods, data analysis of large complex systems, e.g. with the integration of quantum response data to capture non-local quantum effects, the design of electronic structure solvers for AI-accelerated hardware, and strategies to enhance the efficiency of existing methods in tackling quantum chemistry problems.