Florian Pabst
(SISSA)

Abstract:
The rotational dynamics of molecules in the liquid phase can be routinely studied from the boiling point down to the glass transition temperature using dielectric spectroscopy, which spans 18 orders of magnitude in frequency. In recent years, depolarized light scattering has emerged as a competitive technique, offering comparable frequency range and resolution. Dielectric spectroscopy probes the permanent electric dipole moments of molecules, while light scattering is sensitive to polarizability tensors. Consequently, these methods provide complementary insights when their spectra are compared. In this talk, I will demonstrate how this dual approach can disentangle various dynamic contributions in ionic liquids, where conductivity complicates the interpretation of dielectric spectra ^[1]. Additionally, I will show that orientational cross-correlations between neighboring molecules dominate the dielectric response in hydrogen-bonded or highly polar substances but are largely absent from light scattering spectra ^[2].
Since both techniques access only the macroscopic response of the sample, microscopic interpretations of molecular dynamics are often ambiguous. Instead, ab initio simulations enable a detailed molecular-level understanding and simultaneously provide both dielectric and light scattering spectra. In the second part of my talk, I will present how deep neural networks, trained to reproduce interatomic forces, molecular dipoles, and polarizability tensors, yield spectra that align remarkably well with experimental results. This approach provides answers to questions such as: "Why does the dielectric absorption peak of water shift to higher frequencies with added salt, indicating faster dynamics, while increased viscosity suggests slower dynamics?" and "What causes the dramatic slowdown of molecular motions when approaching the glass transition temperature?" ^[3].

Zoom registration link:
https://zoom.us/meeting/register/lA5l9qH4ROKJ-97IiUUtLw