The Sandwich Method for Determining Source Convergence in Monte Carlo Calculation

In slow source convergence problems, it is often difficult to ascertain whether the source iteration has converged or not. In order to solve this problem, a new "sandwich method" has been proposed. The essence of this method is that a finally converged eigenvalue keff is approached starting from two kinds of initial source guesses which give higher and lower neutron multiplication factors. It is especially important for evaluating nuclear criticality safety to know how to choose a biasing source to obtain an upper limit for keff. In this paper, (1) an example is shown to explain the difficulties in ascertaining the source convergence, (2) a method is proposed to obtain the upper and lower limit curves for keff by biasing the initial source distribution, (3) the sandwich method is applied to four benchmark problems proposed by the source convergence group of the OECD/NEA Working Party on Nuclear Criticality Safety. Our calculation results show that the sandwich method is an effective means to confirm source convergence in such slow convergence problems. Appendix is prepared to support the method theoretically.