Utilization of MCNP®6 implicit-capture simulations for quantification of systematic uncertainties from experimental environments

摘要

The high statistical precision that can be achieved using modern data collection techniques implies that systematic uncertainties that were ignored or not considered for past experiments must now be explored. The folding of Monte Carlo simulations into experimental data analysis, which is becoming common practice, requires an estimate of the systematic uncertainty present in the experimental result from uncertainties in the library of cross sections used during the simulation, but this source of uncertainty is typically not quantified. Here we describe a method to estimate this systematic uncertainty by varying the cross sections used during simulation runtime. As opposed carrying out hundreds of sequential analog Monte Carlo simulations, this task was accomplished through a post-processing calculation of output from a single simulation that was run using the implicit-capture formalism available with MCNP®6.2 and the PTRAC output format. This variation procedure allows for the quantification of the spread in counts expected in each bin of a simulated data histogram as well as the correlations between all bins of the same histogram, both resulting from variations in the investigated cross section. This method is demonstrated using simulations of the Chi-Nu experiment through variations of the ~6Li(n, t)α reaction in simulations involving Li-glass detectors. Although the cross-section sampling procedure can be carried out ignoring correlations between the values of the varied cross section at different energies, we also demonstrate the importance of including these correlations in this procedure by considering a full assessment of the systematic uncertainty present in simulated counts from the Chi-Nu experiment both with and without these correlations included. Despite the fact that the ~6Li(n, t)α cross section is a standard below 1 MeV, a significant systematic uncertainty is obtained for counts from the simulations discussed in this work, though a nearly negligible uncertainty would have been determined if correlations in the assumed cross-section library were ignored.