UC Berkeley

The accuracy and precision of nuclear data is of great importance to many fields, including nuclear energy, nuclear medicine, non-proliferation, stockpile stewardship and fundamental physics. New Generation-IV nuclear reactor designs, which are a vital part of the solution to current climate change crisis, rely on nuclear data for their simulations. The nuclear energy community has developed sophisticated methods and codes to propagate uncertainties through the simulation and design process. There is still work to be done, however, on the nuclear data uncertainties that these methods aim to incorporate. In the newest release of the Evaluated Nuclear Data File (ENDF) library, the majority of the 557 individual isotopic evaluations do not even have associated uncertainties. For many isotopes, experimental data are available and the evaluations and their uncertainties are in part based on these data. In this dissertation, methods for improving the accuracy and consistency of the uncertainties on the experimental data are presented. Improving the quality of the reported experimental uncertainties is the first step in the vital process of improving the evaluated uncertainties in libraries such as ENDF.

In order to achieve this goal, templates of measurement uncertainties were created for total and capture cross section measurements. These templates can be used by evaluators to ensure that the uncertainties reported by experimentalists are complete and realistic. Templates are provided for total cross section measurements (utilizing transmission) and for capture cross section measurements (utilizing total absorption spectroscopy, total energy detection, activation analysis, partial gamma detection, or accelerator mass spectrometry). A large-scale literature review for the creation of the measurement uncertainties templates is introduced, which will help to ensure consistency between the estimated uncertainties and other data sets. One common and significant source of uncertainty---the efficiency of an HPGe detector---is studied in detail. The proper method for calculating a data covariance matrix is explained, and new intensity correlation matrices are presented which will allow for more realistic correlations between the measured data points. A new method for determining curve fitting uncertainties is developed, and future work that will allow for proper interpolation uncertainties is explored, with applications in many fields of the physical sciences. Finally, a new method to combine experiment and theory for partial gamma measurements is presented. This method focuses on the partial gamma cross sections that are the least discrepant with the experimental data to infer the total reaction cross section, and includes a simple method for putting uncertainties on the deduced reaction cross section. This uncertainty is a realistic measure of the discrepancies between the experimental data and the calculation. Using this method, an accurate \isotope[238]{U} inelastic scattering cross section, which is critical for accurately modeling some fast reactor systems, is calculated from a data set which has significant issues in the measurement of the strongest gamma. Together, these methods improve the experimental uncertainties that the evaluations rely on.