Study on the errors in numerical solution of the neutron slowing-down problem

In nuclear data files energy dependence of cross sections is presented as numerical values in points. For various cross sections and isotopes their own network of points is given, which optimally describes resonance structure of the cross sections. At numerical solution of the neutron slowing-down problem it becomes vecessary to create a unique calculational network, the density of which should be chosen based on the requirements of information storage and, finally, on the minimum of distortions in calculational functionals. These distortions essentially depend on the cross section interpolation law and the choice of quadrature formulae. While using nuclear data libraries available in the Centre of Nuclear Data (CND), a local error of calculating a neutron detailed spectrum of about 1% may be provided along with the error of calculating group constants averaged over this spectrum - of < or approximately 0.5% at the condition that: a) linear interpolation of cross sections in the log-log scale is taken; b) 3-4-fold multiplication of nodes in initial nuclear data files is performed; 3) while realizing the numerical solution of the slowing-down problem, the Simpson quadratures are used. On the basis of the study a working tape of cross sections has been compiled for a detailed calculation of reactor neutron spectra