Impact of model defect and experimental uncertainties on evaluated output

• Published in: Nuclear instruments & methods in physics research. Section A, Accelerators, spectrometers, detectors and associated equipment Accelerators, spectrometers, detectors and associated equipment, 2013-09, Vol.723, p.163-172
• Main Authors: Neudecker, D., Capote, R., Leeb, H.
• Format: Article
• Language: English
• Subjects: 55Mn; Accelerators; Channels; Correlation; Correlation analysis; Cross sections; Defects; Experimental uncertainties; Mathematical analysis; Model defect uncertainties; Nuclear data evaluation; Small evaluated uncertainties; Uncertainty
• ISSN: 0168-9002; 1872-9576
• DOI: 10.1016/j.nima.2013.05.005

One of the current major problems in nuclear data evaluation is the unreasonably small evaluated uncertainties often obtained. These small uncertainties are partly attributed to missing correlations of experimental uncertainties as well as to deficiencies of the model employed for the prior information. In this article, both uncertainty sources are included in an evaluation of 55Mn cross-sections for incident neutrons. Their impact on the evaluated output is studied using a prior obtained by the Full Bayesian Evaluation Technique and a prior obtained by the nuclear model program EMPIRE. It is shown analytically and by means of an evaluation that unreasonably small evaluated uncertainties can be obtained not only if correlated systematic uncertainties of the experiment are neglected but also if prior uncertainties are smaller or about the same magnitude as the experimental ones. Furthermore, it is shown that including model defect uncertainties in the evaluation of 55Mn leads to larger evaluated uncertainties for channels where the model is deficient. It is concluded that including correlated experimental uncertainties is equally important as model defect uncertainties, if the model calculations deviate significantly from the measurements. •We study possible causes of unreasonably small evaluated nuclear data uncertainties.•Two different formulations of model defect uncertainties are presented and compared.•Smaller prior than experimental uncertainties cause too small evaluated ones.•Neglected correlations of experimental uncertainties cause too small evaluated ones.•Including model defect uncertainties in the prior improves the evaluated output.