Optimization of Direct-Drive Inertial Fusion Implosions Through Predictive Statistical Modeling

Robust predictive models are essential to the design of high-performance inertial confinement fusion (ICF) implosions. Despite progress in modeling, radiation−hydrodynamics codes do not predict a priori the results of ICF experiments with enough accuracy to enable the implementation of iterative design methodologies. The lack of an accurate predictive capability is likely a major obstacle in the quest for ignition. This talk describes a successful attempt to transform inaccurate code outputs into accurate predictive tools using statistical mapping onto the experimental database. The fundamental principle behind this new method is that even though the codes are imperfect, the experimental observables are expected to be correlated to a combination of code output variables because both the experiment and the code take the same input. Remarkably, the correlation between experimental observables and code output exists even if the codes are 1-D and the real implosions are distorted in 3-D, as long as the seeds of the nonuniformities are systematic. This technique only fails if the experiments are dominated by random effects, leading to large shot-to-shot variations, which is not the case for OMEGA implosions. This method has been successfully used to increase yields above 10¹⁴, areal densities to 150 mg/cm2, and convergence ratios to 17 with the goal of finding the optimum implosion that can be fielded on the OMEGA laser.