Uncertainty band evaluation of optical potentials and differential cross-sections. Application to $^8$Li + $^{58}$Ni elastic scattering

Abstract

A statistical method is presented to evaluate the uncertainty bands in the optical nucleus-nucleus potential and in differential cross sections. The starting point is the least square fit of a set of experimental values of elastic differential cross sections, varying the relevant optical potential parameters. This is done using standard $\chi^2$ minimization codes, that provide the covariance matrix of the parameters. A maximum likelihood exploration of the $\chi^2$ surface in parameter space allows to determine the covariance matrix of the parameters associated to a contour of a given $\chi^2$ value. Bayes theorem allows to assign probabilities (p-values) to the regions in parameter space, characterized by $\chi^2$ contours. The method allows to obtain uncertainty bands of an arbitrary observables associated to a given p-value using two approaches. The general approach determines the extremes of the observables calculated in the region of parameter space associated to that p-value. This requires an adequate sampling of parameter space, and explicit calculations of the observables on all sampling points. The simplified approach considers uncertainty propagation of the observable in terms of the optical model parameters. This involves the least-square covariance matrix, given by $\chi^2$ minimization codes, and analytically calculated enhancement factors for each p-value. The method, in the general and simplified approaches, is applied to recent measurements of the elastic differential cross sections of $^8$Li + $^{58}$Ni. $1\sigma$ and $2\sigma$ uncertainty bands are obtained for the optical potentials as a function of the distance, and the differential cross sections as a function of the angle. The general and simplified approaches are very similar in this case. The application of the procedure to determine uncertainty bands of complex scattering calculations is discussed.