Constructing the initial guess spectrum for neutron unfolding without a priori knowledge of the measured field

Accurate neutron spectrum measurement is vital for nuclear energy and nuclear medicine applications. While iterative unfolding methods are favored for their efficiency and accuracy, their precision critically depends on the initial guess spectrum. Without prior information of the measured field, a non-informative “all-ones flat spectrum” is typically used, limiting further accuracy improvements. This paper proposes a method to construct preset spectra based on analytical neutron spectrum models (Watt, Maxwell-Boltzmann, moderation, evaporation spectra), incorporating physical characteristics to provide more instructive initial values for iterative unfolding. Numerical validation using IAEA-403's PTB multi-sphere response functions and 251 radiation-protection scenario spectra shows that the moderation-model-based preset spectrum reduces average relative deviation by ∼20% compared to the flat preset spectrum, achieving superior accuracy in 80.1% of scenarios and comparable accuracy in 17.1%. The method also suppresses spurious structures from uneven energy-group division, enhancing spectral restoration. Cf-252 neutron source experiments confirmed a 26.4% reduction in deviation and improved restoration using the moderation preset spectrum. The proposed method offers a clear, versatile approach compatible with various unfolding algorithms, supporting high-accuracy broad-energy neutron spectrometry. © 2001 Elsevier Science. All rights reserved.

Duke Scholars

Author Sifan Liu Statistical Science