Neutron spectrum reconstruction for liquid organic scintillators in low information scenarios via genetic algorithm

Abstract

Construction of neutron energy spectrum is of interest in various scientific fields such as nuclear power, nuclear security, industrial applications of nuclear and fundamental physics. A genetic neutron spectrum unfolding method is proposed to generate neutron energy spectrum giving light output data from a liquid organic scintillation detector. The method presented attempts to unfold given minimum a priori data, specifically it does not require an initial guess spectrum to be supplied. Two response matrices corresponding to an EJ-309 organic liquid scintillator are constructed for testing, with one matrix using the traditional discretisation and the other using an alternate discretisation based on the energy-to-light conversion process. Test cases include MCNPX-PoliMi simulated $^{252}$Cf, AmLi and AmBe spectra with corresponding detector responses. The genetic method can perform a coarse unfolding of the test spectra, potentially enough to perform an initial categorisation of the spectra. Promising results are obtained when genetic method is used in conjunction with the existing MAXED unfolding code, which operates on the maximum entropy principle. In this scenario, accuracy appears to exceed MAXED using its built-in default a priori spectrum. The $^{252}$Cf test case saw a reduction in the unfolded spectrum error from 5.08$\times 10^{-3}$ to 9.42$\times 10^{-5}$. The AmLi and AmBe test cases could not be unfolded by MAXED using its default spectrum; however, when supplied with the genetic method result, MAXED could complete a successful unfold.