Fully First-Principles Surface Spectroscopy with Machine Learning

Abstract

Our current understanding of the structure and dynamics of aqueous interfaces at the molecular level has grown substantially in the last few decades due to the continuous development of surface-specific spectroscopies, such as vibrational sum-frequency generation (VSFG). Similarly to what happens in other spectroscopies, to extract all of the information encoded in the VSFG spectra we must turn to atomistic simulations. The latter are conventionally based either on empirical force field models, which cannot describe bond breaking and formation or systems with a complex electronic structure, or on \textit{ab initio} calculations which are difficult to statistically converge due to their computational cost. These limitations ultimately hamper our understanding of aqueous interfaces. In this work, we overcome these constraints by combining two machine learning techniques, namely high-dimensional neural network interatomic potentials and symmetry-adapted Gaussian process regression, to simulate the SFG spectra of the water/air interface with \textit{ab initio} accuracy. Leveraging a data-driven local decomposition of atomic environments, we develop a simple scheme that allows us to obtain VSFG spectra in agreement with current experiments. Moreover, we identify the main sources of inaccuracy and establish a clear pathway towards the modelling of surface-sensitive spectroscopy of complex interfaces.

References in corpus (22)

• Perspective: How good is DFT for water?
• Machine Learning Molecular Dynamics for the Simulation of Infrared Spectra
• Nuclear quantum effects enter the mainstream
• Symmetry-Adapted Machine-Learning for Tensorial Properties of Atomistic Systems
• How to remove the spurious resonances from ring polymer molecular dynamics
• The first-principles phase diagram of monolayer nanoconfined water
• Polarization and Experimental Configuration Analysis of Sum Frequency Generation Vibrational Spectra of Air/Water Interface
• Incorporating long-range physics in atomic-scale machine learning
• Committee neural network potentials control generalization errors and enable active learning
• Deep neural network for the dielectric response of insulators
• Raman Spectrum and Polarizability of Liquid Water from Deep Neural Networks
• Self-Consistent Determination of Long-Range Electrostatics in Neural Network Potentials
• Both Inter- and Intramolecular Coupling of O-H Groups Determine the Vibrational Response of the Water/Air Interface
• Predicting molecular dipole moments by combining atomic partial charges and atomic dipoles
• Operators in Machine Learning: Response Properties in Chemical Space
• On the Consistency of Approximate Quantum Dynamics Simulation Methods for Vibrational Spectra in the Condensed Phase
• Fine Tuning Classical and Quantum Molecular Dynamics using a Generalized Langevin Equation
• Path-integral dynamics of water using curvilinear centroids
• Spectroscopy from Machine Learning by Accurately Representing the Atomic Polar Tensor
• Quantum dynamics using path integral coarse-graining
• A $\Delta$-Machine Learning Approach for Force Fields, Illustrated by a CCSD(T) 4-body Correction to the MB-pol Water Potential
• The Surface of Electrolyte Solutions is Stratified

Cited by in corpus (5)

• i-PI 3.0: a flexible and efficient framework for advanced atomistic simulations
• Revealing the molecular structures of a-Al2O3(0001)-water interface by machine learning based computational vibrational spectroscopy
• The Wetting of H${}_2$O by CO${}_2$
• Atomistic Simulations of Oxide-Water Interfaces using Machine Learning Potentials
• Efficient spectra from atomistic simulation: a generalized master equation study of the air-water interface