Researchers at Lawrence Livermore National Laboratory have developed machine learning techniques to study the phase behavior of superionic water - a phase in which hydrogen becomes liquid while oxygen remains solid-like on a crystalline lattice - with unprecedented resolving power.
It is thought that superionic water can exist at depths greater than approximately one third of the radii of Uranus and Neptune, as it makes up the majority of the mantles of these ice-giant planets.
Over three decades ago, superionic water was first proposed; however, it started to be measured recently and its optical properties (partly opaque) and oxygen lattices have been discovered recently.
Planetary science depends heavily on understanding its properties, which are difficult to determine experimentally or theoretically.
As a result of shorter simulation timeframes and small system sizes, superionic water simulations typically get a bit uncertain about the location of phase boundaries such as the melting line.
By using machine learning techniques to learn the atomic interactions from quantum mechanical calculations, Dr. Sebastien Hamel of Lawrence Livermore National Laboratory and colleagues made a leap forward in the ability to treat large systems and time scales.
To determine the phase boundaries, the researchers then used advanced free energy sampling methods, coupled with the machine-learned potential, to drive molecular dynamics.
We use machine learning and free energy methods to overcome the limitations of quantum mechanical simulations and analyze water phases, hydrogen diffusion and superionic transitions under extreme conditions, Dr. Hamel said.
The authors investigated what fractions of insulating ice, different superionic phases, and liquid water are present inside of ice giants by looking at phase boundaries that are consistent with existing experimental observations.
Dr. Hamel said, "Our quantitative understanding of superionic water sheds light on the interior structure, evolution, and magnetic fields of planets including Uranus, Neptune, and the increasing number of icy exoplanets."
The study was published in the journal Nature Physics.