Vibrational Spectroscopy Predictions Using Machine Learning

K. Chen, S. Luber

Vibrational spectroscopy, including infrared (IR) and Raman spectroscopy, is a powerful tool for investigating molecular structure, bonding, and dynamics. However, accurately and efficiently predicting spectra remains challenging due to the complexity of vibrational interactions, anharmonic effects. Traditional computational methods, such as ones based on density functional theory (DFT), can offer accurate predictions but are often limited by their high computational cost, especially for large and complex systems[1]. Machine learning (ML) has emerged as a powerful alternative, enabling efficient and scalable spectra prediction[2]. In this work, we introduce an ML-based approach, which can accurately predict IR spectra and extends to systems subjected to external electric fields. Furthermore, we explore Raman spectra prediction under external electric fields, demonstrating the model`s ability to capture field-dependent vibrational responses. The developed approach offers a fast and accurate solution for vibrational spectra prediction, bridging the gap between efficiency and precision in vibrational spectroscopic simulations.

1. E. Ditler; S, Luber, WIREs Comput. Mol. Sci. 2022 12, e1605.
2. R. Han; R. Ketkaew; S. Luber, J. Phys. Chem. 2022 126, 6, 801-812.