Monte Carlo simulation of migration of fusion plasma neutrons - optimization and traps

Abstract

We recall some of the sensitive points and stages in random walk of neutron, frequent solutions and their consequences on quality and duration of MC neutronic and photonic simulations. We present some unconventional approaches we developed to precisely meet the double paramount MC goal: maximal probing at minimal variance, whence minimum CPU time. Instead of the traditional point observation, enhanced probing is used to limit collected random scatter dispersion. Vector probing by shower (through ”nuclear reaction‑channel” space) and drizzle (improved sampling throughout space involving even the deepest parts, with region fragmentation allowed) drastically reduces the collected variance. By treating analogically close‑collision flux-at-a-point tallies, the unphysical pole discontinuity at-the-detecting-point is avoided - this allows the study of even within-detector collisions. For deep shielding treatment the use of the two-step Cascade Monte Carlo is recommended as it reproduces from physical considerations the mathematical approach. Making sure to distinguish volumetric versus local destructive effects, the latter requiring the use of "statistics of extreme values".