Bayesian Analysis of Inertial Confinement Fusion Experiments at the National Ignition Facility

TL;DR

This paper introduces a Bayesian inference method for analyzing complex high-energy-density experiments at NIF, enabling efficient parameter estimation and comparison with theoretical models using advanced statistical techniques.

Contribution

It develops a novel Bayesian approach incorporating nuisance parameters and prior knowledge, applied to NIF experiments for improved parameter inference and model validation.

Findings

Inclusion of prior knowledge improves model agreement.

Both MCMC and GA methods efficiently explore the posterior.

The approach reduces the need for extensive hydrodynamic simulations.

Abstract

We develop a Bayesian inference method that allows the efficient determination of several interesting parameters from complicated high-energy-density experiments performed on the National Ignition Facility (NIF). The model is based on an exploration of phase space using the hydrodynamic code HYDRA. A linear model is used to describe the effect of nuisance parameters on the analysis, allowing an analytic likelihood to be derived that can be determined from a small number of HYDRA runs and then used in existing advanced statistical analysis methods. This approach is applied to a recent experiment in order to determine the carbon opacity and X-ray drive; it is found that the inclusion of prior expert knowledge and fluctuations in capsule dimensions and chemical composition significantly improve the agreement between experiment and theoretical opacity calculations. A parameterisation of HYDRA results is used to test the application of both Markov chain Monte Carlo (MCMC) and genetic algorithm (GA) techniques to explore the posterior. These approaches have distinct advantages and we show that both can allow the efficient analysis of high energy density experiments.