Monte Carlo Codes for Neutron Buildup Factors

Abstract

The point-kernel method is a widely used practical tool for gamma-ray shielding calculations. However, application of that method for neutron transport simulations is very limited. The accuracy of the method strongly depends on the accuracy of buildup factors used in the calculations. Buildup factors are usually obtained using appropriate computer codes, either based on discrete ordinates transport method or Monte Carlo approach. Since these codes put strong demands on computer resources, they are applied on a limited number of shielding configurations and an attempt is made to use these results and formulate an empirical expression for buildup factors estimation. Due to high physical complexity of neutron transport through shielding material it is very hard to perform parameterisation in order to establish adequate empirical formula. Existing formulas are very limited and are usually applicable to a narrow neutron energy range for few commonly used shielding materials, mostly in monolayer configuration. Recently, a new approach has been proposed for determination of gamma ray buildup factors for mono-layer, as well as multi-layer shielding configurations covering a wide gamma ray energy range. The new regression model is based on support vector machines learning technique, which has theoretical background in statistical learning theory. Development of named regression model required a large number of experimental data obtained by Monte Carlo computer code. More than 7000 Monte Carlo runs were required. Due to physical complexity neutron transport is likely to require even more experimental data in order to generate a model of reasonable accuracy. Therefore, the choice of appropriate Monte Carlo code is a very important question. One has to take into account the accuracy as well as the time required for input preparation and running the code. What also has to be considered is the possibility of the code to be incorporated in an algorithm for automated generation of experimental data. In this paper three Monte Carlo codes are analysed, namely SCALE4.4 code package (SAS3 sequence), SCALE6.0 code package (MAVRIC sequence), and MCNP5. Two simple experimental setups based on a point isotropic source in spherical and slab-like shield are modelled, and the codes are examined on previously mentioned issues. The comparison results show that each one of the examined codes has potential to be used for neutron buildup factor model generation. However, some aspects of their utilization require further analysis prior to final selection.