Speaker
Description
Both neutron nuclear reaction data and fusion reaction data of $^4$He system play an important role in astrophysics and energy fields. Especially the microscopic nuclear cross section of the fusion reaction D(d,n)$^3$He is the core data among them. However, there are obvious differences among current experimental data for the incident energy from 0.2 to 1 MeV, and there is no experimental data available above 6 MeV. The relevant evaluation value has not been published in the main nuclear data evaluation library. "Generalized reduced R-matrix theory", with the corresponding program RAC, analyzes all experimental data about $^4$He system simultaneously, which includes the database of n+$^3$He, d+D, and p+T, and provides a complete set of self-consistent evaluation values for the incident energy between threshold and 30 MeV. RAC calculation outputs the integral cross sections and corresponding covariance matrices. In particular, D(d,n)$^3$He and $^3$He(n,d)D are inverse cross sections of each other, and they achieve precise matching to ensure the reliability of the final evaluation values.
The main feature of RAC is to adopt the $\chi^2$ expression of "Conventional Least-Squares" or "General Least-Squares", instead of "Approximate Least-Squares" which being widely used. In "Approximate Least-Squares", only the diagonal elements of the experimental data covariance matrix are considered, in essence, systematic errors are treated as statistical errors, and the correlation between the statistical errors and the experimental data is reduced, this makes the data fitting lose the sensitivity to obtain the optimal normalization coefficient. In "Conventional Least-Squares" or "General Least-Squares", all elements of the covariance matrix of experimental data are considered, that is, the correlation between statistical errors and experimental data is considered completely. In order to avoid the occurrence of "PPP", it will force the evaluator to carefully normalize the experimental data, so that the experimental data set can achieve a high degree of internal consistency, so as to obtain the most reliable evaluation value.