Machine learning sheds new light on the opacity of dense plasmas

8 June 2020

Research led by Oxford Physicists shows how a novel functional discovery approach can be used to address a longstanding problem in the study of dense plasma opacities.

The work was performed at the FLASH free-electron laser in Hamburg, where femtosecond XUV pulses were focused down to micron spots to create solid density plasmas at temperatures exceeding 200,000 K.

One of the major challenges in plasma opacity experiments is that gradients in temperature in the sample make it all but impossible to measure the plasma absorption coefficient (which is temperature dependent), and indeed temperature measurements themselves in these plasmas are notoriously difficult to determine. In this work the team showed how to sidestep this problem by employing functional optimization to use the gradients themselves to help understand the absorption process. From the oversampled transmission data they were able to learn the form of the absorption coefficient as a function of electron temperature and density, yet without ever needing to measure the electron temperature, density or ionization explicitly.

The work led by Sam Vinko’s group in Atomic and Laser Physics was performed as part of a long-standing international collaboration including Charles University Prague, the Institute of Physics, Academy of Sciences and ELI Beamlines in the Czech Republic; Chalmers University of Technology Göteborg and Uppsala University in Sweden; Sandia National Laboratories and SLAC National Accelerator Laboratory in the USA; and DESY, CFEL, European XFEL and Münster University in Germany.

Time-Resolved XUV Opacity Measurements of Warm Dense Aluminum
Vinko et al., Phys. Rev. Lett. 124, 225002 (2020).