High-dimensional neural network potentials for accurate vibrational frequencies: The formic acid dimer benchmark

TL;DR

This paper demonstrates that high-dimensional neural network potentials can accurately reproduce vibrational frequencies of the formic acid dimer, validated against high-level quantum calculations and experimental data, highlighting their reliability for complex molecular systems.

Contribution

The study introduces a benchmark approach using vibrational frequencies to validate neural network potentials for complex molecular systems.

Findings

Neural network potentials can match coupled cluster and experimental vibrational frequencies.

High-dimensional neural network models provide reliable potential-energy surfaces.

Validation method enhances confidence in machine learning potentials for spectroscopy.

Abstract

In recent years, machine learning potentials (MLP) for atomistic simulations have attracted a lot of attention in chemistry and materials science. Many new approaches have been developed with the primary aim to transfer the accuracy of electronic structure calculations to large condensed systems containing thousands of atoms. In spite of these advances, the reliability of modern MLPs in reproducing the subtle details of the multi-dimensional potential-energy surface is still difficult to assess for such systems. On the other hand, moderately sized systems enabling the application of tools for thorough and systematic quality-control are nowadays rarely investigated. In this work we use benchmark-quality harmonic and anharmonic vibrational frequencies as a sensitive probe for the validation of high-dimensional neural network potentials. For the case of the formic acid dimer, a frequently studied model system for which stringent spectroscopic data became recently available, we show that high-quality frequencies can be obtained from state-of-the-art calculations in excellent agreement with coupled cluster theory and experimental data.