AI Helps Decode the Century-Old Mystery of Why Water Expands When Freezing
Researchers at Osaka University utilized machine learning to evaluate 16 molecular structure descriptors, identifying the most effective one to capture the microscopic changes of water molecules. This establishes a unified framework to understand water's anomalous expansion and its behavior in a supercooled state. This research represents a significant application of AI in physical chemistry, solving a long-standing thermodynamic mystery about water. By providing a systematic method to analyze liquid-liquid transitions, it could advance our understanding of anomalous liquids and materials science. The study reveals that water's anomalous properties stem from the structural competition between High-Density Liquid (HDL) and Low-Density Liquid (LDL) states in supercooled water. As temperature rises, the collapsed HDL structures begin to outnumber the open, tetrahedral-like LDL structures.
## BACKGROUND
Unlike most liquids that contract when cooling, water expands when freezing due to its unique hydrogen-bonded network. Supercooled water refers to liquid water that remains unfrozen below its standard freezing point. Scientists hypothesize that a liquid-liquid phase transition between two distinct phases (HDL and LDL) explains many of water's thermodynamic anomalies.