Physics informed reduced electron evolution models from rapid target heating based on measurements

We present a method of mixing modern machine learning methods and physics-based reduced order models using parameterized Boltzmann distribution functions for modelling the spatio-temporal evolution of plasma plumes generated from rapid target heating by 20 MeV electron beams. The generation of plasma plumes during such heating processes can significantly impact various high-energy physics experiments, making accurate prediction of their behavior crucial for experimental design and analysis. Experiments were conducted using pulsed bunches of approximately 10¹⁵ electrons to heat a range of thin targets to temperatures above 1 eV. The shadowgraph and interferometer measurements provide spatially resolved snapshots of the expanding plume's electron density and temperature distributions at discrete time points. From these experimental data sets, we use both physics-informed and physics-agnostic based parameterizations along with gradient-based and gradient-free optimization in order to train our parameterized reduced models for predicting the electron density and temperature evolution. Initial validation tests for a 100-micrometer-thick Aluminum target show promise in the predictive capability of these reduced models to capture the complex physics at a fraction of the computational cost when compared to radiation hydrodynamics simulations. Furthermore, we measure the generalizability of these reduced models with respect to different target materials and compare to radiation hydrodynamics simulations.