Osaka University Researchers Use AI to Crack Water’s Mystery, Distinguishing Supercooled Water’s Two Liquid States

July 12, 2026

ScienceDaily / The University of Osaka Japan

Overview

Researchers at the University of Osaka have deployed an AI model, trained on computer simulations, to effectively distinguish between the two competing liquid states of supercooled water. By evaluating 16 structural descriptors, the AI identified key indicators, offering a unified framework for studying water’s unusual behaviors. This breakthrough clarifies the microscopic structural changes in water, deepening understanding of this fundamental substance.

In Depth

Key Findings

Researchers at the University of Osaka have made a significant breakthrough in understanding one of water’s greatest mysteries: distinguishing between its two competing liquid states in supercooled conditions. Utilizing an AI model trained on extensive computer simulations, the team successfully evaluated 16 different structural descriptors for supercooled water, identifying the most effective metrics to differentiate these elusive states. This AI-driven approach provides a clearer, unified framework for investigating water’s complex and unusual behaviors.

Technical / Clinical Details

The research involved training an AI model with vast datasets generated from high-fidelity molecular dynamics simulations. The AI analyzed various structural descriptors of water, including hydrogen bond network patterns, intermolecular distances, and local density fluctuations. Through machine learning algorithms, the AI pinpointed the most sensitive indicators for distinguishing between the ‘high-density liquid’ (HDL) and ‘low-density liquid’ (LDL) states of supercooled water. This methodology allows for a more consistent and comprehensive analysis of the microscopic structural changes and their correlation with the macroscopic anomalies of water, which was previously challenging with conventional methods. This advanced understanding is expected to shed light on unresolved phenomena such as the glass transition and liquid-liquid critical point of supercooled water.

Background & Context

Water, essential for life on Earth, exhibits numerous anomalous physical properties compared to most other liquids. Particularly, supercooled water—water cooled below its freezing point without solidifying—displays highly complex behaviors, including anomalous volume expansion and a sharp increase in specific heat. These anomalies have been attributed to the hypothesis that water exists in two distinct liquid states, but direct microscopic evidence has been exceedingly difficult to obtain. The Osaka University study demonstrates AI’s potent capability to uncover hidden patterns and relationships within complex physical systems, opening new avenues for AI application in physical chemistry.

Strategic Significance & Outlook

This AI-based analytical framework not only paves new ground for supercooled water research but also holds potential for application in other complex liquid systems and phase transition phenomena. Researchers anticipate that this unified methodology will facilitate deeper exploration into the fundamental mechanisms driving water’s anomalies, contributing to a better understanding of its behavior in biological processes and materials science. In the long term, this AI model could lead to practical applications such as controlling ice formation, developing more efficient water treatment technologies, or engineering novel water-based functional materials, signifying a profound impact beyond theoretical physics.

Source: https://www.sciencedaily.com/releases/2026/07/260707025047.htm

Get our weekly technology intelligence — free

Receive an infographic that lets you judge at a glance whether each field’s analysis report is worth reading.

Subscribe Free — Weekly Tech Intelligence

By subscribing, you’ll receive Troy-Technical’s weekly technology intelligence newsletter.