Water has puzzled scientists for decades. Over the past 30 years or so, they have theorized that when cooled to extremely low temperatures like -100 degrees Celsius, water might separate into two liquid phases of different densities. Like oil and water, these phases don’t mix and may help explain some of water’s other odd behaviors, such as how it becomes less dense as it cools.
However, studying this phenomenon in the laboratory is nearly impossible because water crystallizes into ice so quickly at such low temperatures. Now, new Georgia Tech research uses machine learning models to better understand phase transitions in water, opening up even more avenues to better theoretically understand a variety of substances. With this technique, the researchers found strong computational evidence in support of water’s liquid-to-liquid transition that can be applied to real-world systems that operate using water.
“We’re doing this with very detailed quantum chemical calculations that try to get as close as possible to the real physics and physical chemistry of water,” said Thomas Gartner, an assistant professor in Georgia Tech’s School of Chemical and Biomolecular Engineering. “This is the first time anyone has been able to study this transition with this level of accuracy.”
The research is published in the journal “Liquid-Liquid Transitions in Water from First Principles” Physical Review Letterswith co-authors at Princeton University.
To better understand how water interacts, the researchers ran molecular simulations on a supercomputer, which Gartner compared to a virtual microscope.
“If you had an infinitely powerful microscope, you could zoom all the way down to the level of individual molecules and watch them move and interact in real time,” he said. “That’s what we’ve done by making what’s almost a computational movie.”
The researchers analyzed how molecules move and the structural characteristics of the liquid at different water temperatures and pressures, simulating phase separation between high-density and low-density liquids. They collected vast amounts of data—running some simulations for up to a year—and continued to fine-tune their algorithms for more accurate results.
Running such long and detailed simulations would have been impossible even a decade ago, but today’s machine learning offers a shortcut. The researchers used a machine learning algorithm to calculate the energy of how water molecules interact. The model performs calculations significantly faster than conventional techniques, allowing simulations to be performed more efficiently.
Machine learning is not perfect, so these long simulations also improve the accuracy of predictions. The researchers were careful to test their predictions with different types of simulation algorithms. If multiple simulations give similar results, then it validates their accuracy.
“One of the challenges of this work is that there isn’t a lot of data for us to compare, because it’s a problem that’s nearly impossible to study experimentally,” Gartner said. “We’re really pushing the boundaries here, so that’s another reason why it’s so important that we try to do this using a number of different computing technologies.”
Some of the extreme conditions the researchers tested may not exist directly on Earth, but could exist in various aqueous environments in the solar system, from Europa’s ocean to the water at the center of a comet. However, the findings could also help researchers better explain and predict water’s strange and complex physical chemistry, inform water use in industrial processes, develop better climate models, and more.
According to Gartner, this work is even more pervasive. Water is a well-studied research area, but this approach can be extended to other difficult-to-simulate materials, such as polymers, or complex phenomena, such as chemical reactions.
“Water is essential to life and industry, so the particular question of whether water can undergo this phase transition has been a long-standing question, and it will be important if we can find an answer,” he said. “But now we have this very powerful new computing technology, but we don’t yet know what the boundaries are, and there’s a lot of room for growth in this field.”