Table of Contents
Metallic hydrogen can be tricky to understand. Things are intense beneath the clouds of Jupiter. In these extreme situations, hydrogen itself doesn’t exist the usual form. The covalent bonds of the hydrogen atoms break down and the material becomes a lot more like a metallic solid. The field of astronomy seeks to understand the physical nature of gas giants and other worlds, so a highly detailed understanding of hydrogen’s this metallization process is important.
Getting to the bottom of things
To gain a little more insight on the situation, researchers and scientists have to examine hydrogen’s phase diagram. A phase diagram is what shows exactly how a material’s physical properties will vary with pressure and temperature. Hydrogen’s less exotic states are already well understood. There is, however, a knowledge-gap when it comes to hydrogen’s metallic form. The thing is, it’s really hard to make those kinds of pressures in a lab.
Recently, scientists have been able to use computer-based simulations to understand metallic hydrogen. They can closely model hydrogen’s atoms under simulated extreme conditions. These models showed that when under an increasing pressure, hydrogen suddenly undergoes a first-order phase-transition. It transitions to an atomic state from a molecular state. This process was then applied to a gas giant’s interior. Doing that pointed towards the possibility of there being distinct boundaries between conductive and insulating layers within the planet.
The problem is, however, that these scientists have faced quite a serious limitationthat has ended up casting a bit of doubt on their theories. Bingqing Cheng explains at the University of Cambridge, ‘Even state-of-the-art methods can only model hundreds of hydrogen atoms. Since we were previously restricted to using such small system sizes, the first-order transition was assumed.’
Using computers to learn about atoms
The fundamental issue with simulations is one of scale. The fact that as you add more atoms, complexity increases dramaticly very fast due to quantum interactions that are at play. Running larger models would require resources that are far beyond the capabilities of today’s most powerful supercomputers. In a new study, Cheng and his colleagues in Switzerland and the UK approached the with machine learning because machine learning can build models of highly complex systems by training itself on real data.
Machine learning drastically reduced the computing power required to simulate metalic hydrogren. This enabled the team to simulate more than 1700 atoms across a wide range of pressures and temperatures and about a nanosecond. ‘In a nutshell, we exploited machine learning techniques to learn the atomic interactions from quantum mechanics,’ Cheng further explained, ‘The inputs to the neural network were the positions of each atom, and the outputs were the energy and forces they experience.’
The right tools for the job
With such tools at their disposal, Cheng’s research team was able to discover something interesting. They saw that hydrogen’s molecular-to-atomic transitions don’t happen quit as cleanly as the practice had previously thought. As it turns out, hydrogen undergoes a smooth transition as the pressure rises. This phenomenon only occurs past a certain ‘critical point’ in the phase diagram. It can only happen where the pressures and temperatures are high enough. That’s when the physical distinction between metal and liquid impossible. Conventional hydrogen’s transformation into metallic hydrogen occurs because hydrogen’s conductive and insulating phases happen both continuously and gradually.
Revealing this ‘supercritical’ property of hydrogen could have some serious implications for humanities current understanding of hydrogen’s phase diagram. It’s important to know how hydrogen operates at very extreme temperatures pressures. As Cheng describes, the discovery his team made provides a drasticly clearer picture on the nature of hydrogen deep within gas giants. ‘Inside these bodies, the insulating and the metallic layers have a smooth density profile between them, instead of an abrupt change,’ he explained.
Machine learning shows great promise
It’s obvious that this discovery has huge implications for the planetary sciences. The team behind the project thinks that their results highlight a promising potential for artificial intelligence and machine learning. According to the team, these techniques can effectively explore hydrogen’s phase diagram. Cheng said, ‘We believe machine learning will not only change the way high pressure hydrogen is modelled, but will be in the standard toolbox of computational physicists. With machine learning, we can run nanosecond-scale simulations with thousands of atoms, with first principle accuracy. This was beyond our wildest dreams just ten years ago.’
John Proctor, a University of Salford researcher involved in the project hinted that exploring this work for planetary interiors will not be a simple task. John elaborated, ‘We propose that there is no first order liquid-liquid transition in hydrogen, and that planetary models assuming this therefore need to be revised. Since the actual interiors of Jupiter and Saturn do not consist of pure hydrogen – they are fluid mixtures predominantly composed of hydrogen and helium – a lot of further work is required to fully explore the implications of this work for planetary science.’
In other space-related news, NASA uses Artificial Intelligence to predict and increase in a hurricane’s (or typhoon’s) intensity.