No abstract

Work is reported here on the process and present results of an international evaluation of the neutron cross section standards. The evaluations include the H(n,n), 3He(n,p), 6Li(n,t), 10B(n,α), 10B(n,α1γ), 197Au(n,γ), 235U(n,f), and 238U(n,f) standard reactions as well as the 238U(n,γ) and 239Pu(n,f) reactions. This evaluation was performed to include new experiments on the standards that have been made since the ENDF/B-VI evaluation was completed and to improve the evaluation process. Evaluations have been completed for the 6Li(n,t), 197Au(n,γ) and 238U(n,γ) cross sections. Also below 20 MeV the H(n,n), 235U(n,f), 238U(n,f) and 239Pu(n,f) cross sections are completed. Many of the cross sections being evaluated are used in neutron dosimetry for fluence determination. The general trend observed for the evaluations is an increase in the cross sections for most of the reactions from fractions of a percent to several percent compared with the ENDF/B-VI results.

The underestimation of the errors in the recommended neutron cross sections calculated on the basis of an analysis of measurement results using strict statistical methods is discussed. The basic reasons for the underestimation of the errors in the cross sections are described. A statistical model is proposed -a constant bias model -for taking into account effectively the component of the measurement error that is unknown to the experimentor. The results of a statistical analysis of the measurements of the ratio of the fission cross sections of 238 U and 235 U on the basis of the constant bias model are presented. It is noted that the experimental errors are much higher (on the average by a factor of 2) than the declared errors. The consequences of using the generally accepted method for correcting the errors in the evaluated cross sections are examined for an exactly solvable model example. It is shown that such a procedure does reconstruct the real errors in the evaluated cross sections.There is a crisis of confidence in the results obtained by using classical statistical methods for analyzing experimental data consisting of a collection of several sets of measurements, including mutually contradictory measurements. Classical statistical methods are taken to mean the least-squares method, the Bayesian approach, and the maximum-likelihood method. The reason is that the errors in the physical dependences calculated on the basis of these methods are exceedingly small. An instructive example are the errors in the neutron standards which are widely used in experiments and are obtained after statistical analysis of more than 10000 measurements on the basis of the generalized least-squares method [1]. Computational results which contain no explicit errors have been deemed by experts as not deserving of confidence, and the errors in the evaluated cross sections recommended by experts are 2-19 times greater than the computed errors [1].With time such results were incorporated into a single continuous series. This gave rise to a crisis of confidence in the classical statistical methods and actually led to complete rejection of the errors in the evaluated data in the National Neutron Data Libraries (Russian BROND-2 [2], American ENDF/B VI [3], and Japanese JENDL-3.1 [4]). The existing errors for many isotopes and dependences are, with rare exceptions, a product of an "expert" assessment.Examples of the underestimation of errors in the neutron cross sections calculated using strict statistical methods are presented in [5]. The examples examined are cases where the individual components of the systematic part of the total experimental error declared by the experimenters were excluded from statistical analysis. However, it turned out that other sources make the main contribution to the underestimation of the errors in the neutron cross sections. The present paper is devoted to a description of these sources. A statistical model which makes it possible to analyze a collection of mutually contradictory sets of measurements is a...

No abstract

No abstract