Critical assessment of nuclear sensitivity metrics for the r-process

TL;DR

This paper critically evaluates various nuclear sensitivity metrics used in r-process simulations, highlighting their differences, limitations, and proposing a normalized approach for more consistent comparisons.

Contribution

It assesses existing sensitivity metrics, introduces a new logarithmic measure, and recommends normalization techniques for more reliable sensitivity analysis in r-process studies.

Findings

Different metrics can lead to opposing sensitivity rankings for certain nuclides.

Normalization of abundances affects sensitivity factors and their comparability.

A minimized F statistic is recommended for better cross-study comparisons.

Abstract

Any simulation of the r-process is affected by uncertainties in our present knowledge of nuclear physics quantities and astrophysical conditions. It is common to quantify the impact of these uncertainties through a global sensitivity metric, which is then used to identify specific nuclides that would be most worthwhile to measure experimentally. Using descriptive statistics, we assess a set of metrics used in previous sensitivity studies, as well as a new logarithmic measure. For certain neutron-rich nuclides lying near the r-process path for the typical hot-wind scenario, we find opposing conclusions on their relative sensitivity implied by different metrics, although they all generally agree which ones are the most sensitive nuclei. The underlying reason is that sensitivity metrics which simply sum over variations in the r-process distribution depend on the scaling used in the baseline, which often varies between simulations. We show that normalization of the abundances causes changes in the reported sensitivity factors and recommend reporting a minimized F statistic in addition to a scale estimation for rough calibration to be used when comparing tables of sensitivity factors from different studies.