PhD project: Machine learning accelerated studies of dense planetary matter

Project description

Extreme conditions of pressure and temperature can create exotic states of matter such as superionic ice, probably the most common form of water in the known universe. Deep inside icy planets' mantle regions, as Mbar pressures are approached, compression work becomes equal to bond energies and planetary matter is characterised by significant chemical changes, dissociation of molecules, phase transitions that might cross planetary adiabats, and uncertainties about miscibility limitations. This can lead to stable stratification inside planets with distinct electronic and thermal transport properties in different layers. In the absence of direct observational data, extracting these properties from computation is crucial for planetary modelling and our understanding of the solar system's formation and evolution. Accurate computational modelling of these materials usually relies on electronic structure calculations, but these are restricted in terms of system size, chemical complexity, and run time.

In this project, we will utilise machine learning approaches to train accurate and transferable interatomic potentials for matter at planetary conditions. We will create training sets for the machine learning process from random structure searching and molecular dynamics configurations. We will use those potentials to simulate specific mixtures within the H-C-N-O chemical space at conditions relevant inside icy planets, predict observables accessible to laboratory high-pressure experiments, and compare to astronomical observations. Calculations will, amongst others, utilise national resources such as the ARCHER2 supercomputer. The project will likely interface with experimental collaborators within Edinburgh's Centre for Science at Extreme Conditions.

Project supervisor

The project supervisor welcomes informal enquiries about this project.

Find out more about this research area

The links below summarise our research in the area(s) relevant to this project:

What next?

More PhD projects