Accurate Surface Chemistry Enabled by Neural NeTworks

Harnessing light to drive chemical reactions offers a selective alternative to traditional thermal activation, with transformative potential for industrial catalysis. Light-driven activation of challenging molecules, such as CO2, can enable sustainable pathways for chemical production. Modern nanocatalysts can be designed to combine efficient light absorbers (plasmonic metals) with catalytically active components to funnel light energy into chemical reactions via electronic excitations. At the same time, ultrafast spectroscopic techniques are beginning to reveal atomic-scale details of these processes, opening unprecedented opportunities for rational catalyst design.Computational chemistry calculations are essential for understanding reaction mechanisms and providing the fundamental insights needed for rational design of catalysts and materials. Unfortunately, current simulation methods fall short in describing excited-state dynamics in realistic heterogeneous catalysts. Multiconfigurational wavefunction methods, adept at describing electronic excitations, are not suitable for surface reaction simulations due to their steep computational scaling.ASCENNT aims to overcome this challenge by developing a computational framework that can accurately simulate electronic excitations at surfaces and their subsequent dynamics at the atomic scale by integrating machine learning, multiscale modelling, density functional theory, and multiconfigurational wavefunction theory. These advances will unlock new understanding across photochemistry, electrochemistry, spectroscopy, and materials science.The fellowship will merge the researcher’s expertise in multiconfigurational methods with the host’s leadership in data-driven approaches, creating a unique interdisciplinary profile. ASCENNT will deliver both conceptual breakthroughs and practical tools while preparing the researcher to launch an independent career as an innovative leader in computational material chemistry.