Machine learning of two-dimensional spectroscopic data

Abstract

Two-dimensional electronic spectroscopy has become one of the main experimental tools for analyzing the dynamics of excitonic energy transfer in large molecular complexes. Simplified theoretical models are usually employed to extract model parameters from the experimental spectral data. Here we show that computationally expensive but exact theoretical methods encoded into a neural network can be used to extract model parameters and infer structural information such as dipole orientation from two dimensional electronic spectra (2DES) or reversely, to produce 2DES from model parameters. We propose to use machine learning as a tool to predict unknown parameters in the models underlying recorded spectra and as a way to encode computationally expensive numerical methods into efficient prediction tools. We showcase the use of a trained neural network to efficiently compute disordered averaged spectra and demonstrate that disorder averaging has non-trivial effects for polarization controlled 2DES.

Molecular excited states through a machine learning lens

“…A wide range of electronic spectroscopic methods exist, which measure various physicochemical phenomena, 9 and ML has been already applied to assist different spectroscopic techniques, including steady-state absorption (in both optical 26,32,34,35,125,126 and X-ray 127-129 domains), emission, 28 and multidimensional time-resolved spectroscopy. 34,126,130 Some of these studies [26][27][28] were the earliest works using ML for excited-state research, and we are currently experiencing a resurgence in interest in such methods.…”

“…One step towards the direct simulation of spectra with ML was made for the simulation of twodimensional electronic spectra of a light-harvesting complex. 130 In that study, calculated pigment-site energies were used as descriptors, rather than the more conventional structural descriptors. An approach to directly simulate spectrum for a given structure was suggested recently for the prediction of K-edge X-ray absorption near-edge structure spectra of diverse molecules from the QM9 dataset 132 and solid materials.…”

Effect of disorder and polarization sequences on two-dimensional spectra of light-harvesting complexes

“…This is in contrast to the allparallel polarization sequence, where ESA has only negative components and SE positive values, which largely stay in place. Rodríguez and Kramer (2019) for the real part of the rephasing signal. Corresponding measured spectra by Thyrhaug et al (2018) are shown in their Fig.…”

“…Numerically, we compute the disorder average from 5000 realizations of disorder added to the site energies with standard deviation 30 cm −1 and 50 cm −1 . The resulting 2DES are efficiently encoded in a neural network representation following Rodríguez and Kramer (2019). This encoding allows us to study various disorder realizations, shown in Fig.…”

Machine learning for excitation energy transfer dynamics

“…Machine learning techniques have been used to represent and solve quantum systems such as in a work by Carleo and Troyer [21] where the authors introduce an ansatz capable of both finding the ground state and describing the unitary time evolution of complex interacting quantum systems. More specifically and in a bid to investigate open quantum systems further, the applications of supervised machine learning that we are interested in are related to forms of approximating solutions to open quantum system dynamics such as where multilayer perceptrons have been used to obtain the exciton dynamics of large photosynthetic complexes [22] and to better understand the relationship between the structure of light-harvesting systems and their excitation energy transfer properties [23]; where recurrent neural networks were used to model quantum systems interacting with an unknown environment [24] and where convolutional neural networks were used to predict long-time dynamics of an open quantum system [25].…”