Multi-pass Monte Carlo simulation method in nuclear transmutations

Abstract

Monte Carlo methods, in their direct brute simulation incarnation, bring realistic results if the involved probabilities, be they geometrical or otherwise, remain constant for the duration of the simulation.However, there are physical setups where the evolution of the simulation represents a modification of the simulated system itself.Chief among such evolving simulated systems are the activation/transmutation setups. That is, the simulation starts with a given set of probabilities, which are determined by the geometry of the system, the components and by the microscopic interaction cross-sections. However, the relative weight of the components of the system changes along with the steps of the simulation. A natural measure would be adjusting probabilities after every step of the simulation.On the other hand, the physical system has typically a number of components of the order of Avogadro's number, usually 1025 or 1026 members. A simulation step changes the characteristics for just a few of these members; a probability will therefore shift by a quantity of 1/1025. Such a change cannot be accounted for within a simulation, because then the simulation should have then a number of at least 1028 steps in order to have some significance. This is not feasible, of course. For our computing devices, a simulation of one million steps is comfortable, but a further order of magnitude becomes too big a stretch for the computing resources.We propose here a method of dealing with the changing probabilities, leading to the increasing of the precision. This method is intended as a fast approximating approach, and also as a simple introduction (for the benefit of students) in the very branched subject of Monte Carlo simulations vis-à-vis nuclear reactors.