SUMMARY

A thermal neutron scattering kernel data evaluation framework for computation of model-dependent predictions and their uncertainties is outlined. In this framework, model parameters are fitted to double-differential cross section measurements and their uncertainties. For convenience, the initial implementation of this framework uses the molecular dynamics (MD) model implemented in the NAMD code. It is applied to light water using the TIP3P and TIP4P interaction models. The MD trajectories computed by the NAMD simulation are then processed using the nMOLDYN code to compute the thermal neutron scattering kernel. Double-differential cross sections computed from the scattering kernel are then fitted to double-differential scattering data measured at the Spallation Neutron Source detector at Oak Ridge National Laboratory. The fitting procedure is designed to yield optimized model-parameters and their uncertainties in the form of a covariance matrix, from which new evaluations of thermal neutron scattering kernel will be generated. The fitting procedures will be done using the Generalized Least Squares method. The purpose is to show how the computation of sensitivities of the scattering kernel with respect to model parameters could be computed by a novel application of the adjoint method to MD simulations.