An alternative Monte Carlo method for the κ-eigenfunction?

Major Monte Carlo codes such as MCNP (1) typically use variations on the power iteration [1,2] method to obtain the fundamental mode k-eigenvalue and eigenfunction. This paper describes an alternative to the power iteration method. The new method builds the eigenfunction, one source particle at a time, by starting each new source particle at a location chosen so that the source density tends to better approximate the fundamental eigenfunction. That is, the method adds new source points to regions where the empirical source density appears too low to satisfy the eigenfunction equation.