Bayesian Optimization of Direct-Drive Double-Shell Targets

Direct-drive inertial confinement fusion (ICF) offers higher energy coupling efficiency than indirect drive but suffers from higher sensitivities to asymmetries and laser conditions. The direct-drive-double-shell (DDDS) concept mitigates these issues by separating the compression and ignition roles between two shells. However, DDDS performance depends on a coupled parameter space, making design optimization challenging.

This work presents a Bayesian Optimization framework to efficiently identify optimal Direct Drive Double Shell designs tailored to facility-specific constraints. The optimization systematically explores the design space by varying parameters such as shell and material thickness to identify configurations that maximize neutron yield. The process begins by launching an initial batch of 1D xRAGE simulations in parallel using an initial set of inputs. As results return, a surrogate model is updated to inform the next batch of inputs, which are then automatically submitted for simulation. This approach enables rapid, data-driven design iteration and provides guidance on both target configurations and potential facility upgrades.