Water might secretly be a mix of 2 different liquids, scientists say

Researchers from the City University of Hong Kong have used artificial intelligence to find molecular evidence that water consists of two distinct liquid states—one denser than the other. Published June 4 in Nature Physics, the study suggests these states constantly switch, potentially explaining water’s unique density and viscosity behaviors.

How did AI prove the two-state hypothesis?

Physical chemist Xiao Cheng Zeng and postdoctoral researcher Liwen Li used “unsupervised deep learning” to identify patterns in molecular data. They employed the GROMACS simulation package to track hundreds of thousands of water molecules, generating tens of millions of data points.

The AI identified “reaction coordinates,” which are variables describing how a molecule shifts between denser and looser structures. According to Zeng, this AI-driven analysis took about a year and a half; he estimated the same work would have taken nearly a decade using traditional methods.

The team discovered two conversion paths. Most switches occur via a “semi-loop” with one energy barrier. However, near the 32 degrees Fahrenheit (zero degrees Celsius) boundary where ice and liquid coexist, molecules may take a “full-loop” path involving three separate barriers.

Did You Know? Most liquids become denser as they cool, but water becomes denser only until about 4 degrees Celsius before it starts to expand.

Why does water’s molecular structure matter for health?

This two-state model may offer a unifying explanation for water’s anomalies, such as why ice floats and why the liquid resists temperature changes better than similar substances. Understanding these molecular shifts is critical because most biological and pharmaceutical processes occur in water.

Zeng noted that a better understanding of this structure could clarify how drug molecules, proteins, and dissolved salts interact in a solution. According to the researcher, these interactions are vital for cell function and the development of injectable drugs.

Expert Insight: Samantha Carter suggests that while the molecular shift is a fundamental discovery, the real stakes lie in pharmaceutical precision. If the way drug molecules interact with water is better understood, it could eventually lead to more stable and effective injectable medications.

What happens next in the research?

Zeng and his team are currently building a more rigorous machine-learning model to confirm their results. They aim to connect these findings to specific properties like temperature, viscosity, and density.

Experimental confirmation in real water is a possible next step, though Zeng said this will likely require sensitive techniques. He mentioned that labs like the Pacific Northwest National Laboratory, which has found indirect spectroscopic evidence of this behavior, may be necessary for confirmation.

Once confirmed by experiment, Zeng believes this model could be used to understand how water interacts with nature. However, he cautioned that applying this knowledge to practical health uses is still a long way off.

Frequently Asked Questions

What is the two-state hypothesis?
It is the theory that water exists at the molecular level as two different liquids—one denser and one less-dense—that constantly switch places.

How did AI assist in this discovery?
Researchers used unsupervised deep learning to spot patterns in tens of millions of data points from molecular dynamics simulations, a process that reduced the analysis time from a potential decade to about 18 months.

How does this affect medicine?
Because cell functions and injectable drugs rely on interactions in water, understanding water’s molecular structure may eventually shed light on how proteins and drug molecules behave in solution.

Do you think AI will continue to solve long-standing mysteries in basic chemistry?