Genetic algorithm-based optimization of a target for the production of atmospheric-like neutrons via 100 MeV proton beam

Abstract

This paper presents the optimization of a white neutron target for 100 MeV proton beam to produce a fast neutron spectrum shape similar to that of the atmospheric neutron at ground level. The neutron target consists of 0.5 mm-thick discs consecutively placed in front of a cooling water channel. A sensitivity analysis is performed with different materials to assimilate resulting neutron spectrum with the ground-level cosmic ray-induced neutron spectrum by applying the genetic algorithm (GA). Optimization of the spectrum is conducted in consideration of neutron impact to semiconductor devices. To reconstruct a target-generated neutron spectrum, parametrization on neutron generation and stopping power by target material and position is configured. The loss function in the GA is defined to evaluate similarity between the target neutron spectrum and the atmospheric neutron spectrum. The GA is modeled with a package of consecutive discs as a gene for optimization, and the length and material of target for each disc are decided throughout generations. As a result, the target is configured with six different materials, with approximately half of the weighted deviation of atmospheric neutrons being compared with any other single-solid neutron target. The GA-based optimization process shown here suggests further use for designing targets with other incident proton energies or neutron spectra for various applications.