Investigation of variance reduction methods based on weight-windows for radiation transport simulation in deep penetration problems

Deep penetration problems have been studied in a very large set of applications; the radi- ation transport simulations have to be as reliable as possible. The results have to be reliable in order to carry out a series of successive calculations such as the dose rate assessment. Estimating dose rates to workers is a basic step of nuclear reactor licensing. The main issue encountered with these kinds of problems is that the simulated particles may not be able to reach the detector regions. This negatively contributes to the estimation of the particle flux, increasing the variance. How can we obtain reliable results quickly? Over the years, the evolution of Monte Carlo codes has been improved with method- ologies to reduce the variance, the uncertainty associated with the results. The Variance Reduction (VR) methods based on weight-windows are a choice that has been analyzed during this work. The VR methods use statistical artifices to improve the probability that the particles reach the detectors, without modifying the physics of the phenomenon. The work demonstrated that one method called MAGIC (Method of Auto- matic Generation of Importances by Calculation) can be considered for different studies without the use of deterministic and quasi-deterministic solutions coupled with MC codes.