Machine learning spectroscopy to advance computation and analysis

Abstract

Spectroscopy, the exploration of matter through its interaction with electromagnetic radiation, is relevant in many diverse research fields, such as biology, materials science, medicine, and chemistry, and enables the qualitative and quantitative characterization of samples. Machine learning has revolutionized spectroscopy by enabling computationally efficient predictions of electronic properties, expanding libraries of synthetic data, and facilitating high-throughput screening. While machine learning has strengthened theoretical computational spectroscopy, its potential in processing experimental data has yet to be adequately explored. At the same time, automating structure and composition predictions from spectra remains a formidable challenge that requires theoretical simulations and expert knowledge. This review addresses the synergy between machine learning and spectroscopy, covering various techniques including optical, X-ray, nuclear magnetic resonance, and mass spectrometry. It outlines the fundamentals of machine learning, summarizes the techniques, and previews future developments to fully exploit the potential of machine learning and advance the field.