<p>Water inherently contains trace amounts of various salts, yet the microscopic processes by which salts influence some of its physical properties remain elusive. Notably, the mechanisms that reduce the dielectric constant of water upon salt addition are still debated. The primary absorption peak for electromagnetic radiation-commonly used in microwave heating-shifts towards higher frequencies in saline solutions, suggesting faster water molecular dynamics. This observation, however, contrasts with the simultaneous increase in viscosity and experimental reports that ionic solutes would slow down water molecular motion. In this work, we use molecular dynamics (MD) simulations with deep-neural-network models trained on high-quality quantum mechanical data to mimic interatomic forces and molecular dipoles, to compute the dielectric spectra of perchlorate water saline solution, which may be relevant to the recent discovery of liquid water beneath the thick ice crust at Mars’s south pole. Our results reveal that both the reduction in the dielectric constant and the absorption peak shift can be attributed to ion-induced changes in the orientational ordering of water molecules. Additionally, we demonstrate that the self-part of the molecular dipole-dipole correlation function reveals clear signatures of the slowing dynamics within the first cationic solvation shell, consistent with the experimentally observed increase in viscosity.</p> <p>This record contains the DeepMD neural network potential and the underlying training data as well as the neural network for prediciting Wannier centroids and the underlying training data. </p>