Efficient Quantum Vibrational Spectroscopy of Water with High-Order Path Integrals: From Bulk to Interfaces

Abstract

Vibrational spectroscopy is key in probing the interplay between the structure and dynamics of aqueous systems. To map different regions of experimental spectra to the microscopic structure of a system, it is important to combine them with first-principles atomistic simulations that incorporate the quantum nature of nuclei. Here we show that the large cost of calculating the quantum vibrational spectra of aqueous systems can be dramatically reduced compared with standard path integral methods by using approximate quantum dynamics based on high-order path integrals. Together with state-of-the-art machine-learned electronic properties, our approach gives an excellent description not only of the infrared and Raman spectra of bulk water but also of the 2D correlation and the more challenging sum-frequency generation spectra of the water−air interface. This paves the way for understanding complex interfaces such as water encapsulated between or in contact with hydrophobic and hydrophilic materials through robust and inexpensive surface-sensitive and multidimensional spectra with first-principles accuracy.

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

“…[38][39][40] Still, most of these methods exhibit an unfavorable scaling with system size. Second-generation potentials allow to address larger systems, and some vibrational studies for larger molecules, 41 clusters, 42 and condensed systems [43][44][45][46] have been reported. However, MLPs suitable for condensed systems have been typically constructed relying on density functional theory (DFT), which, although offering a good compromise between accuracy and efficiency for many systems, does not provide spectroscopic-quality vibrational frequencies.…”

On the Fresnel Factor Correction of Sum-Frequency Generation Spectra of Interfacial Water

“…We further suspect that the obtained 𝜒 𝑦𝑦𝑧 (2) spectra (or 𝜒′ 𝑦𝑦𝑧 (2) spectra with the interfacial dielectric profiles for different water species) can be used for the critical check of the simulated spectra. 18,25,31,[37][38][39][40][41][42][43] It is more robust if one can see the agreement on not only the spectral lineshape but also the absolute amplitude of the spectra. 17,44 Such an attempt is on the horizon.…”

Spectroscopy from Machine Learning by Accurately Representing the Atomic Polar Tensor

“…Molecular dipole moment vectors have been trained directly by symmetry-adapted Gaussian process regression to accurately calculate the IR spectrum of liquid ambient water . Using the same approach, it was shown that the molecular polarizability tensor can also directly be trained which enables one to calculate machine learned Raman − and SFG spectra from molecular dynamics simulations.…”

“…Besides such equivariant message passing Neural Networks, also Gaussian process regression can be used to create equivariant machine learning models . This approach was utilized recently to train molecular dipole moment vectors and polarizability tensors. − …”

“…The predicted APTs are then used to calculate the IR spectrum via eq and eq . Although the IR spectrum of liquid ambient water has already been machine learned by training molecular dipole moments, , even explicitly considering nuclear quantum effects, it is merely done here to demonstrate the applicability of the herein introduced technique. The introduced APTNN methodology and the training protocol is generally transferable to any other system, with and without periodic boundary conditions.…”