Point-kernel Method for Radiation Fields Simulation

Design and maintenance of nuclear cycle facilities mandatory imply implementation of efficient radiation protection. Radiation environment assessment and shielding optimization require a considerable amount of numerical calculations. Nowadays mathematical modelling methods and corresponding software are widely used for such kind of problems. All the software used for radiation fields modelling falls into two main groups. To the first belongs software based on Monte Carlo methods, such as MCNP [1], Geant [2], Penelope, Fluka, etc. Modern implementations of Monte Carlo method provide consistent accounting of radiation transport effects. This leads to perfect accuracy of the calculated values even for complex models. At the same time Monte Carlo calculations require considerable amount of computing time. Computation burden increases rapidly for complex geometries, multiple sources and thick shielding. Another group contains software implementing analytical methods. The examples of such programs are Microshield, QAD, Mercure. All this programs use point-kernel method for doze calculations. This method is much more less computationally intensive than Monte Carlo method. Series of calculations necessary for shielding optimization could be conducted at reliable time using point-kernel method. But due to macroscopic approach to radiation transport this methods lucks consistency. The main problem pointkernel method encounters is account for scattered radiation which is usually implemented through semiempirical approximation. Additional ”build-up” factor must be introduced as a multiplier to the attenuated doze. Determination of the appropriate buildup factor can be rather complex as it depends upon the energy, the thickness and type of material. Uncertainties in determining build-up factor essentially limit the accuracy of point-kernel method. In our paper we considered some aspects of pointkernel method and its implementation by Mercure-3 program code [4]. Results of doze calculations for some typical cases are presented. Also point-kernel method verification by the MCNP code [1] is discussed.