We present a robust protocol for affordable learning of electronic states to accelerate photophysical and photochemical molecular simulations.The protocol solves several issues precluding the widespread use of machine...We present a robust protocol for affordable learning of electronic states to accelerate photophysical and photochemical molecular simulations.The protocol solves several issues precluding the widespread use of machine learning(ML)in excited-state simulations.We introduce a novel physicsinformed multi-state ML model that can learn an arbitrary number of excited states across molecules,with accuracy better or similar to the accuracy of learning ground-state energies,where information on excited-state energies improves the quality of ground-state predictions.We also present gap-driven dynamics for accelerated sampling of the small-gap regions,which proves crucial for stable surfacehopping dynamics.Together,multi-state learning and gap-driven dynamics enable efficient active learning,furnishing robust models for surface-hopping simulations and helping to uncover long-timescale oscillations in cis-azobenzene photoisomerization.Our active-learning protocol includes sampling based on physics-informed uncertainty quantification,ensuring the quality of each adiabatic surface,low error in energy gaps,and precise calculation of the hopping probability.展开更多
基金the Polish Ministry of Education and Science for funding this research under the program “Perły Nauki,” grant number PN/01/0064/2022, amount of funding, and the total value of the project: 239 800,00 PLNas well as gratefully acknowledge Polish high-performance computing infrastructure PLGrid (HPC Centers: ACK Cyfronet AGH) for providing computer facilities and support within computational grant no. PLG/2024/017363+3 种基金M.B. thanks the funding provided by the European Research Council (ERC) Advanced grant SubNano (Grant agreement 832237)M.B. received support from the French government under the France 2030 as part of the initiative d’Excellence d’Aix-Marseille Université, A*MIDEX (AMX-22-IN1-48)P.O.D. acknowledges funding from the National Natural Science Foundation of China (via the Outstanding Youth Scholars (Overseas, 2021) project)via the Lab project of the State Key Laboratory of Physical Chemistry of Solid Surfaces. The computations were performed using the XACS cloud computing resources. The authors also acknowledge Max Pinheiro Jr, Prateek Goel, and Bao-Xin Xue for many non-published tests of not-so-successful protocols.
文摘We present a robust protocol for affordable learning of electronic states to accelerate photophysical and photochemical molecular simulations.The protocol solves several issues precluding the widespread use of machine learning(ML)in excited-state simulations.We introduce a novel physicsinformed multi-state ML model that can learn an arbitrary number of excited states across molecules,with accuracy better or similar to the accuracy of learning ground-state energies,where information on excited-state energies improves the quality of ground-state predictions.We also present gap-driven dynamics for accelerated sampling of the small-gap regions,which proves crucial for stable surfacehopping dynamics.Together,multi-state learning and gap-driven dynamics enable efficient active learning,furnishing robust models for surface-hopping simulations and helping to uncover long-timescale oscillations in cis-azobenzene photoisomerization.Our active-learning protocol includes sampling based on physics-informed uncertainty quantification,ensuring the quality of each adiabatic surface,low error in energy gaps,and precise calculation of the hopping probability.
