Toward Spectroscopic Accuracy in Quantum Dynamics Simulations with Fourth-Generation High-Dimensional Committee Neural Network Potentials

Abstract: Predictive simulation of vibrational spectra of complex condensed phases and interfaces with thousands of degrees of freedom has long been a challenging task of modern condensed matter theory. In this work, fourth-generation high-dimensional committee neural network potentials (4G-HDCNNPs) are developed for the first time, using active learning and query-by-committee, and introduced to the essential nuclear quantum effects (NQEs) as well as conformational entropy and anharmonicities from path integral (PI) molecular dynamics simulations. Using representative bulk water and air-water interface systems, we demonstrate the accuracy of the developed framework in infrared and vibrational sum frequency generation spectral simulations. Specifically, by seamlessly integrating non-local charge transfer interactions from 4G-HDCNNPs with the NQEs from PI methods, our introduced methodology yields accurate infrared spectra using predicted charges from the 4G-HDCNNP architecture without explicit training of dipole moments. Our vibrational sum frequency generation spectra also illustrate high accuracy for the considered air-water interface system. The introduced framework in this work is general and offers a benchmark tool and a simple, practical paradigm for predictive spectral simulations of complex condensed phases and interfaces, free from empirical parameterizations and/or ad hoc fittings.