Improving variance convergence rate in monte carlo eigenvalue simulations via delayed neutrons

Abstract

This paper investigates a new approach to Monte Carlo eigenvalue simulations in an effort to reduce the impact of correlations and improve the associated variance convergence rate in the active generations. For a given active generation size, larger inactive generations are simulated and split into two banks, neutrons in one bank are transported in every active generation (i.e. prompt bank) and contribute to tallies, while neutrons in the other bank (i.e. delayed bank) are carried forward and injected in the prompt bank at later generations. This process mimics delayed neutrons and provides better variance behavior at the cost of more work in the inactive generations which are usually less costly due to the absence of tallies. A predictive model for the delayed scheme is also developed to help identify an optimal run strategy and provide an estimate of the variance correction prior to running the model.