Machine Learning Enabled Lineshape Analysis in Optical Two-Dimensional Coherent Spectroscopy
Optical two-dimensional (2D) coherent spectroscopy excels in studying coupling and dynamics in complex systems. The dynamical information can be learned from lineshape analysis to extract the corresponding linewidth. However, it is usually challenging to fit a 2D spectrum, especially when the homogeneous and inhomogeneous linewidths are comparable. We implemented a machine learning algorithm to analyze 2D spectra to retrieve homogeneous and inhomogeneous linewidths. The algorithm was trained using simulated 2D spectra with known linewidth values. The trained algorithm can analyze both simulated (not used in training) and experimental spectra to extract the homogeneous and inhomogeneous linewidths. This approach can be potentially applied to 2D spectra with more sophisticated spectral features.