XGPT: Extending Monte Carlo Generalized Perturbation Theory capabilities to continuous-energy sensitivity functions

The XGPT method extends the Generalized Perturbation Theory capabilities of Monte Carlo codes to continuous-energy sensitivity functions. In this work, this method is proposed as a new approach to nuclear data uncertainty propagation. XGPT overcomes some of the limitations of legacy perturbation-based approaches. In particular, it allows the nuclear data uncertainty propagation to be performed adopting continuous energy covariance matrices, instead of discretized (multi-group) data. The XGPT capabilities are demonstrated in three simple fast criticality benchmarks for 239Pu and 208Pb cross section uncertainties. The new method is also applied in selected cases to estimate higher moments of the keff distribution, starting from TENDL random evaluations. The XGPT estimates, when compared against reference Total Monte Carlo (TMC) results, show a good agreement and a significant reduction in computational requirements with respect to the TMC approach. Finally, the capabilities for uncertainty propagation involving adjoint-weighted response functions are demonstrated.