Monte Carlo simulations usually rely on simulating a large number of particles in parallel. In nuclear simulations, the number of particles has to be controlled, so that the system does not die out or grow beyond the limits of computer memory. This population control has the unintended consequence of “clustering” particles in specific spatial regions as the simulation progresses. In turn, clustering renders calculations based on simulation output unreliable.
Understanding when clustering begins to occur is critical in accurate Monte Carlo simulations, and is a problem that extends far beyond reactor physics. In his talk, Eric Dumonteil likened the effect to the classical “gambler’s ruin” problem. In turn, Oliver Kelsey Tough presented results on the closely related Fleming–Viot process, and thus was able to utilise results from populations genetics
