Predicting hydration layers on surfaces using deep learning

Abstract

Characterisation of the nanoscale interface formed between minerals and water is essential to the understanding of natural processes, such as biomineralization, and to develop new technologies where function is dominated...

Molecular Features of Hydration Layers: Insights from Simulation, Microscopy, and Spectroscopy

“…These simulated configurations were used to train the artif icial neural network (ANN) model, which accurately predicted the density profiles of hydration layers from the structures of the calcite surfaces. 46 When trained properly against the data set generated by (classical or ab initio) MD simulations, such an ANN model can instantly and accurately predict the molecular structure of a hydration layer without having to run lengthy MD simulations. If such an ML model is trained against the structures obtained from the AFM and VSFG experiments, an ML model might predict the hydration structures even better than the state-of-the-art molecular simulations.…”

“…Machine learning is recently proposed as a tool for translating the AFM and VSFG signals to the underlying molecular structure. 46 Regarding the molecular dynamics of a hydration layer, experimental work has been scarce. Instead, the molecular dynamics of reorientation and diffusion have been studied by using MD simulation.…”

“…By using MD simulation, Ranawat et al created the configurations of the hydration layers formed on different faces of calcite. These simulated configurations were used to train the artificial neural network (ANN) model, which accurately predicted the density profiles of hydration layers from the structures of the calcite surfaces …”

“…The ANN-predicted density profile matches the simulated one (bottom right). Reprinted with permission from ref . Copyright 2021 Royal Society of Chemistry.…”

“…In both AFM and VSFG, a molecular simulation proves to be very useful in clarifying exactly and quantitatively how the molecular structure manifests as the experimental signals. Machine learning is recently proposed as a tool for translating the AFM and VSFG signals to the underlying molecular structure . Regarding the molecular dynamics of a hydration layer, experimental work has been scarce.…”