We introduce a new measure of the structure of a liquid which is the softness of the mean-field potential developed by us earlier. We find that this softness is sensitive to small changes in the structure. Then, we study its correlation with the supercooled liquid dynamics. The study involves a wide range of liquids (fragile, strong, attractive, repulsive, and active) and predicts some universal behaviors such as the softness being linearly proportional to the temperature and inversely proportional to the activation barrier of the dynamics with system dependent proportionality constants. We establish a master equation between the dynamics and the softness parameter and show that, indeed, the dynamics, when scaled by the temperature and system dependent parameters, show a data collapse when plotted against softness. The dynamics of fragile liquids show a strong softness dependence, whereas that of strong liquids show a much weaker softness dependence. We also connect the present study with the earlier studies of softness involving machine learning (ML), thus, providing a theoretical framework for understanding the ML results.

Foreign