Efficient Analysis of High-Resolution Imaging Data in High-Energy Density Physics

Efficient Analysis of High-Resolution Imaging Data in High-Energy Density Physics

This project will center around observing properties of warm dense matter. Warm dense matter is a state of matter that is too hot to be considered a solid, but too dense to be considered a plasma. It occurs in planetary interiors, as well as during inertial confinement fusion experiments. In order to model and understand these complex environments, certain properties of warm dense matter must be experimentally measured, such as its thermal conductivity. This knowledge could aid in the modelling of planetary evolution, as well as in the design of fusion capsules.
The main goal of the project is to observe the thermal conductivity of various metal-plastic interfaces. A novel technique will be developed and employed in the process of analyzing the X-ray images. This measurement of thermal conductivity will be done using data already collected at the Linac Coherent Light Source, or LCLS, which is an X-ray Free Electron Laser (XFEL). The experiment also utilized an optical laser. The optical laser was used to heat a plastic-coated metal wire to warm dense matter conditions, where it was then imaged by the XFEL. By repeating the experiment with various different time delays between the two lasers, the wire's expansion can be observed through time. Simulating the dynamics of the target can give information about its thermal conductivity.
Observing the evolution of warm dense matter systems is a difficult process. When creating these systems in a lab, measurements are limited by the small size and short duration in which the system we are observing exists. Therefore, few experimental values of the thermal conductivity of warm dense matter exist, and any new measurements would be valuable for benchmarking theoretical models. During this process of extracting a thermal conductivity measurement, a novel technique involving machine learning will be utilized. Currently, a forward model exists that can simulate a diffraction pattern created from a given density profile. The inverse of this process is much harder. We propose using an untrained neural network in combination with the forward model as a tool to extrapolate the density profile
given its diffraction pattern. The neural network will be trained to output a density profile, which will then be run through a diffraction code and compared to the experimental results. In this way, the neural network can be trained to minimize the difference between the simulated image and the experimental data. Finally, the density profiles can be matched to hydrodynamic simulations, which would allow for the measurement of thermal conductivity.
This project falls within the EPSRC Plasma and Lasers research area. It involves the investigation of a high density plasma, created by a short pulse laser. The experimental measurement of the thermal conductivity of warm dense matter will bolster the design of inertial confinement fusion experiments.

Collaborators
SLAC
Tom White, UNR
Matthew Oliver, STFC
Dan Eakins, University of Oxford