Using Deep Learning to Investigate Laboratory Astrophysics Experiments Through Collective Thomson Scattering Analysis*

As we begin to enter a new paradigm of massive data collection and availability for high energy density science, AI and machine learning have become great candidates for the large scale data analysis needed in these experiments.

Here we present our work using a deep neural network (DNN) surrogate model to analyze the ion acoustic wave (IAW) feature from a Thomson scattering (TS) image for a control shot in a laboratory astrophysics campaign at the OMEGA Laser Facility. To train the DNN, a large dataset of Thomson scattered light spectra is generated from a multi-species 3-Maxwellian plasma model for a variety of plasma conditions using the open-source code PlasmaPy. We show the DNN predictions are comparable to results from two popular analysis methods; a 1D hybrid (kinetic ions and fluid electrons) Particle-In-Cell simulation using the code CHICAGO, and a Markov Chain Monte Carlo (MCMC) analysis of the TS data.