Machine learning for nonadiabatic molecular dynamics: best practices and recent progress

2025/01/01 by Carolin Müller, Štěpán Sršeň, Brigitta Bachmair +7 · 1 voice · 17 citations
Biochemistry, Genetics and Molecular Biology · Chemistry · Materials Science · Physics and Astronomy · Psychology · #Chemistry #Cognitive science #Computational chemistry #Computer science #Data science #Dynamics (music) #Machine Learning in Materials Science #Materials science #Molecular dynamics #Nanotechnology #Physics #Protein Structure and Dynamics #Psychology #Spectroscopy and Quantum Chemical Studies #Statistical physics

paper · pdf · doi:10.1039/d5sc05579b

published in Chemical Science 16(38), 17542-17567 (Royal Society of Chemistry)

openalex publication_date 2025/01/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/08/06

Abstract

Exploring molecular excited states holds immense significance across organic chemistry, chemical biology, and materials science. Understanding the photophysical properties of molecular chromophores is crucial for designing nature-inspired functional molecules, with applications ranging from photosynthesis to pharmaceuticals. Non-adiabatic molecular dynamics simulations are powerful tools to investigate the photochemistry of molecules and materials, but demand extensive computing resources, especially for complex molecules and environments. To address these challenges, the integration of machine learning has emerged. Machine learning algorithms can be used to analyse vast datasets and accelerate discoveries by identifying relationships between geometrical features and ground as well as excited-state properties. However, challenges persist, including the acquisition of accurate excited-state data and managing the complexity of the data. This article provides an overview of recent and best practices in machine learning for non-adiabatic molecular dynamics, focusing on pre-processing, surface fitting, and post-processing of data.

Citations

• Gaussian approximation potentials: A brief tutorial introduction
• Constructing high‐dimensional neural network potentials: A tutorial review
• SchNet – A deep learning architecture for molecules and materials
• Less is more: Sampling chemical space with active learning
• CREST—A program for the exploration of low-energy molecular chemical space

Cited by

• XMCQDPT2-Fidelity Transfer-Learning Potentials and a Wavepacket Oscillation Model with Power-Law Decay for Ultrafast Photodynamics
• Analytical Excited-State Gradients and Derivative Couplings in TDDFT with Minimal Auxiliary Basis Set Approximation and GPU Acceleration
• Gradients not needed: ML-driven propagation of nonadiabatic molecular dynamics without reference gradients
• Gradients not needed: ML-driven propagation of nonadiabatic molecular dynamics without reference gradients. [europepmc]
• Controlling the fate of two triplet states: solid-state annihilator design for photon upconversion. [europepmc]
• The Newton-X platform for mixed quantum-classical dynamics. [europepmc]

Discussions

• Check out our new collaborative paper on machine learning for nonadiabatic molecular dynamics published in @chemicalscience.rsc.org. It provides an overview of the current state-of-the-art and best pr [bsky, 15 points, 0 comments]

Related

• Machine learning approaches for analyzing and enhancing molecular dynamics simulations
• Gradients not needed: ML-driven propagation ofnonadiabatic molecular dynamics without reference gradients
• Transferable Machine Learning Potential X-MACE for Excited States using Integrated DeepSets
• Molecular Dynamics with Neural-Network Potentials
• Charting electronic-state manifolds across molecules with multi-state learning and gap-driven dynamics via efficient and robust active learning