Study on iterative regularization method and application to neutron spectrum unfolding of multi-sphere spectrometer measurement

Abstract

An iterative regularization method combining the Lanczos bidiagonalization with Tikhonov regularization is proposed and applied to neutron spectrum unfolding of a newly-developed compact multi-sphere spectrometer. The unfolding is processed as a determined problem without any prior information, which is different from the traditional way. Comparison study of the two commonly-used unfolding methods, GRAVEL and MLEM with the iterative regularization method is carried out. The “semi-convergence” behavior appears for readings containing large uncertainties with the GRAVEL and MLEM methods, making the unfolded spectrums deviate from the true solutions. However, there is no semi-convergence behavior appearing with the iterative regularization method. The unfolded spectrums get close to the true solution as the number of iterations grows. Experimental measurement is carried out for measuring neutron spectrum of 241 Am–Be neutron source using the compact multi-sphere spectrometer. The measured data are also unfolded. Compared with the ISO reference spectrum, the unfolded results with GRAVEL and MLEM methods deviate from the reference spectrum due to the semi-convergence behavior, while the unfolded spectrum with the iterative regularization method agrees well with the reference spectrum, indicating that the iterative regularization method is adaptable to unfolding problems with both accurate measured readings and measured readings with large uncertainties.