OSAKA (Web Desk) – Researchers have used artificial intelligence to gain a deeper understanding of the microscopic structure of water, providing new insights into one of nature’s most unusual substances. The study offers a new method for analyzing water’s molecular behavior and could improve scientific understanding of everything from climate systems to advanced materials.
Water is one of the most familiar substances on Earth, yet it continues to puzzle scientists because of its unique physical properties. Unlike most liquids, water expands when it freezes, making ice less dense than liquid water. This unusual characteristic allows ice to float, helping preserve aquatic ecosystems during winter and playing a vital role in Earth’s climate.
Scientists have long known that water’s unusual behavior results from the way its molecules continuously form and break hydrogen bonds. However, understanding how these microscopic structures change under different temperatures and pressures has remained a major scientific challenge.
A research team from Osaka University has now developed an artificial intelligence-based system capable of comparing different methods used to describe water’s molecular structure. Their findings were published in the scientific journal Communications Chemistry.
The researchers focused on supercooled water, a state in which water remains liquid even after being cooled below its normal freezing point. This phenomenon occurs when water is stored in extremely clean, smooth containers without impurities that normally trigger ice crystal formation.
Supercooled water exhibits several unusual properties that scientists believe are linked to the existence of two competing molecular arrangements known as high-density liquid (HDL) and low-density liquid (LDL). The balance between these two structures changes as temperature varies, influencing water’s physical behavior.
At the molecular level, hydrogen bonds connect water molecules into an ever-changing network. Warmer temperatures generally favor the denser HDL arrangement, while colder conditions encourage the formation of more open LDL structures.
Over the years, researchers have developed numerous mathematical descriptors to represent these microscopic structures. These include measurements of local density, bond angles and tetrahedral arrangements. However, because these methods were developed independently, comparing their effectiveness has proven difficult.
To solve this problem, the Osaka University team designed a neural network capable of evaluating multiple structural descriptors simultaneously. The artificial intelligence system analyzed data generated through molecular dynamics simulations, enabling it to identify hidden relationships between different molecular configurations.
Rather than relying on a single measurement, the AI examined sixteen different structural descriptors and assessed how effectively each one distinguished between HDL and LDL states across a wide range of temperatures.
The machine-learning model gradually improved its performance by identifying recurring patterns in large datasets. This approach allowed researchers to evaluate which descriptors contained the most meaningful structural information while eliminating less useful measurements.
According to the research team, the unified AI framework provides a more systematic way to study water’s microscopic behavior than previous methods.
Scientists believe the findings could improve understanding of water’s thermodynamic properties, including why it behaves differently from nearly every other liquid under changing environmental conditions.
The research may also help scientists develop more accurate computer simulations for studying natural processes involving water, including cloud formation, ice development, biological systems and chemical reactions.
Beyond water itself, the approach demonstrates how artificial intelligence is becoming an increasingly valuable tool in scientific research. Machine-learning models are now being used across chemistry, biology, physics and materials science to analyze complex datasets that would be difficult to interpret using conventional techniques.
Researchers say AI allows scientists to recognize subtle patterns that may otherwise remain hidden, accelerating discoveries and improving the accuracy of scientific models.
The team believes their framework could eventually be adapted to investigate other liquids and complex molecular systems, broadening its usefulness beyond water research.
Although further studies will be needed to refine the model and validate its findings under additional experimental conditions, scientists say the work represents an important advance in understanding one of the world’s most essential substances.
As artificial intelligence continues transforming scientific research, studies like this highlight its growing role in solving long-standing mysteries and expanding knowledge about the fundamental behavior of matter.
