By teaching a machine to learn some quantum tricks, physicists have discovered a strange new phase of hydrogen in solid form. While this discovery is purely theoretical for now, it could help us better understand the behavior of matter from the smallest of scales to the internal mechanics of the largest planets in the universe.

Discovered by an international team of researchers, this new phase of solid hydrogen followed the model’s demonstration of hydrogen molecules under extreme conditions: To use a food analogy, their shape transformed from spheres stacked like a stack of oranges to something much like eggs.

Hydrogen normally requires very low temperatures and very high pressures Solid form. Through a new machine learning study of this particular phase change, scientists have found the new molecular arrangement.

“We started with the not-so-ambitious goal of refining the theory of something we know,” He says Physicist Scott Jensen of the University of Illinois Urbana-Champaign.

“Unfortunately, or perhaps fortunately, it was more interesting than that. There was the emergence of this new behavior. In fact, this was the dominant behavior at higher temperatures and pressures, something that there was no hint of in the old theory.”

The updated machine learning algorithm played an important role in the research: it was able to model the actions of thousands of atoms rather than hundreds of numerous studies of quantum phenomena.

The researchers used an improved version of what is known as Quantum Monte Carlo QMC technique: Basically, it uses random sampling and probability mathematics to see how large groups of atoms behave collectively, groups that are difficult to study in an actual experiment.

A second computational method—one more capable of handling more atoms but without the precision—was used to verify the results. As the results are congruent, they indicate that the improved QMC technology is working as intended.

“Machine learning turns out to teach us a lot,” He says Physicist David Siberly of the University of Illinois Urbana-Champaign. “We’d seen signs of new behavior in previous simulations, but we didn’t trust them because we could only accommodate small numbers of atoms.”

“With our machine learning model, we can take full advantage of the most accurate methods and see what’s really going on.”

Simply put, the machine learning component has improved the accuracy and range of simulations that scientists can run, using existing data and past simulations to make future simulations more accurate in terms of their estimates.

Hydrogen is not only the most abundant element in the universe, but it is also the simplest of all in terms of its individual atoms: one proton and one electron. This means that new discoveries about hydrogen could affect almost everything else in physics.

Right now, it’s too early to tell what this new phase of solid hydrogen means, and more experiments and simulations are needed to look at it more closely. However, the study of hydrogen-filled planets such as Jupiter and Saturn is just one area where this additional understanding could be useful.

“We want to understand everything, so we have to start with the systems we can attack,” He says Cyberly. “Hydrogen is simple, so it’s worth knowing we can handle it.”

Research published in Physical review letters.