Abstract
Vibrational spectroscopy, which encompasses Raman and infrared (IR) spectroscopy, serves as a crucial instrument for investigating vibrational properties and the structure of condensed systems. However, experimental data often encompass a myriad of overlapping effects, and isolating the contributions of individual molecular interactions is non-trivial. In this study, first-principles Raman and infrared spectra of water and hydrocarbons were calculated under various extreme conditions, including extreme pressure-temperature conditions and nanoconfinement, which pose challenges for experimental approaches. Raman spectra were employed to extract structural information by analyzing the pressure and temperature dependence of the characteristic Raman peaks for methane, ethane, and propane.
Furthermore, a novel method within density functional theory was developed with the aim of calculating molecular polarizability in condensed phases, compatible with hybrid functionals. This method can be used for molecules, ions, and nanoparticles. Derived from the finite field method and extending beyond linear response, this technique is implemented within the first-principles molecular dynamics (FPMD) code.Consequently, it enables the calculation of molecular polarizability alongside MD simulations, facilitating the consideration of temperature effects.
Then, a neural network dipole model was established to provide dielectric propxiii erties of water across an extensive pressure-temperature range. The findings reveal that the static dielectric constant may vary by an order of magnitude in Earth’s upper mantle, indicating considerable alterations in solvation properties of water at varying depths. Additionally, the study evaluated the frequency-dependent dielectric constant
of water and determined that temperature has a higher impact on dielectric absorption than pressure.
Lastly, machine learning models were leveraged to predict the sum frequency generation (SFG) spectrum of the water-air interface. By training models on ab initio molecular dynamics data, we accurately reproduced the SFG spectrum. This enabled large-scale simulations of over 16,000 water molecules, resolving prior disagreements between theory and experiments. Our analysis suggests a tiny SFG peak around 2000 cm−1 (we used D2O instead of H2O) arises due to long-range intermolecular interactions at the water-air interface.