Water is a fundamental element of daily life, yet it remains a unique scientific phenomenon that defies complete explanation, as its behavior does not always follow the patterns typical of other liquids.

Perhaps the clearest example of this is the phenomenon of freezing; most materials become denser when they turn into a solid state, while water expands when it freezes, which is why ice floats on the surface, and frozen water causes pipes, rocks, and containers to crack. This behavior is related to how molecules are arranged at the microscopic level.

In this context, researchers from Osaka University turned to artificial intelligence to study this hidden molecular structure. According to the website "Scientific Russia," the study aimed to compare various scientific explanations for supercooled water and determine which best explains its unusual properties.

What is supercooled water?

Supercooled water is water that remains in its liquid state at a temperature below its normal freezing point, provided it is pure enough that there are no nucleation centers such as impurities or scratches on the container walls. This state is of great scientific importance because the unusual properties of water become more pronounced, giving researchers a better opportunity to study the molecular patterns responsible for its strange behavior.

Two competing structures

At the molecular level, water is formed by transient hydrogen bonds; they constantly form, break, and reorganize, making liquid water a dynamic network rather than a fixed structure.

The study focused on two competing structural forms within supercooled water: one is a low-density liquid structure in which molecules adopt a more open arrangement, and the other is a high-density liquid structure with a more compact arrangement. The balance between these two forms helps explain the unusual behavior of water, as the dominance of one over the other changes with temperature; as temperature rises, high-density structures become more common.

The role of artificial intelligence

Researchers had several methods to describe these molecular arrangements, known as structural indicators, but comparing them was difficult because they were created separately and relied on different criteria.

Artificial intelligence helped solve this problem; the researchers used molecular dynamics simulations to track molecular movements and interactions, then employed an artificial neural network to compare 16 structural indicators and evaluate each indicator's ability to distinguish between the high-density and low-density structures.

The role of artificial intelligence here was not to rediscover water, but to identify the most accurate mathematical descriptions that capture the important molecular differences within supercooled water.

Broader implications

The value of this work extends beyond a single experiment; understanding the microscopic structure of water could help researchers explain why it possesses exceptional physical properties, as well as highlight the ability of artificial intelligence to support fundamental science by organizing complex data and revealing the most meaningful measurements.

In conclusion, water, despite its apparent simplicity, involves a complex molecular competition that still hides deep scientific secrets, and artificial intelligence is becoming an advanced tool to uncover them.