Spectroscopy from Machine Learning by Accurately Representing the Atomic Polar Tensor

Abstract

Vibrational spectroscopy is a key technique to elucidate microscopic structure and dynamics. Without the aid of theoretical approaches, it is, however, often difficult to understand such spectra at a microscopic level. Ab initio molecular dynamics has repeatedly proved to be suitable for this purpose; however, the computational cost can be daunting. Here, the E(3)-equivariant neural network is used to fit the atomic polar tensor of liquid water a posteriori on top of existing molecular dynamics simulations. Notably, the introduced methodology is general and thus transferable to any other system as well. The target property is most fundamental and gives access to the IR spectrum, and more importantly, it is a highly powerful tool to directly assign IR spectral features to nuclear motiona connection which has been pursued in the past but only using severe approximations due to the prohibitive computational cost. The herein introduced methodology overcomes this bottleneck. To benchmark the machine learning model, the IR spectrum of liquid water is calculated, indeed showing excellent agreement with the explicit reference calculation. In conclusion, the presented methodology gives a new route to calculate accurate IR spectra from molecular dynamics simulations and will facilitate the understanding of such spectra on a microscopic level.

First-principles spectroscopy of aqueous interfaces using machine-learned electronic and quantum nuclear effects

“…We exploit equivariant (invariant and covariant to the SO(3) group symmetry operations) atom-centred ML models 72,73 that bypass niteelectric eld electronic structure calculation and respect the physical symmetries of the rank-1 polarization and rank-2 polarizability tensors. We, amongst others, have combined equivariant ML of total polarization and polarizability tensors with classical or PI quantum dynamics methods to predict liquid water IR 17,74,75 and Raman spectra. 17,18,22 A hurdle towards rst-principles SFG modelling is that its computable expressions are written in terms of molecular polarization and polarizability tensors, which are ill-dened in rst-principles modelling.…”

Computation of CCSD(T)-Quality NMR Chemical Shifts via Δ-Machine Learning from DFT

“…The field of machine learning (ML) within computational chemistry has grown very rapidly in the past decade. 8 Application examples are found in various areas, such as molecular property prediction, 9 crystal structure determination via NMR chemical shifts, 10 prediction of vibrational spectra, 11 and a complete general density functional approximation (DM21). 12 Since quantum-chemistry-based methods for predicting NMR parameters can have serious drawbacks, such as insufficient accuracy or high computational cost (depending on the systems of interest), the use of artificial neural networks (ANN) and machine learning for this purpose has become increasingly popular in recent years.…”

Integrating Explainability into Graph Neural Network Models for the Prediction of X-ray Absorption Spectra

“…Machine learning (ML) techniques are being increasingly applied to various areas of theoretical and computational chemistry given their ability to infer structure–property relationships on the basis of large amounts of data. − Among those ML techniques, graph neural networks (GNN) and deep neural networks (DNN) are promising candidates to predict the properties of matter, such as the electronic structure, at a higher computational speed, already making them favorable for high-throughput calculations in materials design and drug discovery. , Thus, the ability to perform efficient computations with high accuracy has demonstrated that ML techniques are advantageous in domains such as various types of spectroscopy, including vibrational and optical. ,− …”