A Method to Optimize the Geometry of Radiation Detectors by Exploiting the Multiphysics Nature of the Simulations

Monte Carlo simulations of radiation detectors invoke computationally-intensive methods, to model a broad range of dynamical physical processes. In this paper, we examine the applicability of a black-box optimization technique, multiple fidelity sequential kriging optimization, as a framework for studying design optimization in systems modeled by multiphysics. The proposed optimization technique is able to combine the results of simulations with varying degree of multiphysical detail in order to reduce the overall computational cost. In this preliminary case study, geometrical parameters of a prototypical HPGe detector are adjusted to maximize the detection efficiency. Microphysics of secondary effects are approximated, rather than explicitly calculated, to generate fast, lower-fidelity simulations. Low-cost and high-cost data are integrated in the optimization method which sequentially searches the parameter space on variable fidelity levels.