Uncertainty Quantification for Optical Model Parameters

TL;DR

This paper applies statistical uncertainty quantification methods to optical model parameters in nuclear reaction theory, analyzing how parameter uncertainties affect cross section predictions and emphasizing the importance of parameter correlations.

Contribution

It introduces a systematic approach to quantify uncertainties in optical model parameters and propagates these to reaction cross sections, highlighting the role of parameter correlations.

Findings

Correlated chi-squared functions provide more natural parameterizations.

Uncertainty bands are broader when parameter correlations are included.

Systematic trends identified across multiple nuclear reactions.

Abstract

Although uncertainty quantification has been making its way into nuclear theory, these methods have yet to be explored in the context of reaction theory. For example, it is well known that different parameterizations of the optical potential can result in different cross sections, but these differences have not been systematically studied and quantified. The purpose of this work is to investigate the uncertainties in nuclear reactions that result from fitting a given model to elastic-scattering data, as well as to study how these uncertainties propagate to the inelastic and transfer channels. We use statistical methods to determine a best fit and create corresponding 95\% confidence bands. A simple model of the process is fit to elastic-scattering data and used to predict either inelastic or transfer cross sections. In this initial work, we assume that our model is correct, and the only uncertainties come from the variation of the fit parameters. We study a number of reactions involving neutron and deuteron projectiles with energies in the range of 5-25 MeV/u, on targets with mass $A$=12-208. We investigate the correlations between the parameters in the fit. The case of deuterons on $^{12}$C is discussed in detail: the elastic-scattering fit and the prediction of $^{12}$C(d,p)$^{13}$C transfer angular distributions, using both uncorrelated and correlated $\chi^2$ minimization functions. The general features for all cases are compiled in a systematic manner to identify trends. Our work shows that, in many cases, the correlated $\chi ^2$ functions (in comparison to the uncorrelated $\chi^2$ functions) provide a more natural parameterization of the process. These correlated functions do, however, produce broader confidence bands. Further optimization may require improvement in the models themselves and/or more information included in the fit.