Preconditioning and acceleration of the power method for reactor calculations

New preconditioning and acceleration techniques are applied to the power method in order to facilitate the solution of reactor problems. First, the drawbacks of an approximate inversion are mitigated by fitting an error recovery term to the power method. Secondly, the authors introduce a new overrelaxation technique based on a stationarity principle. Unlike the usual acceleration approaches, no estimate of the dominance ratio is required to calculate an optimal overrelaxation parameter. Validation results are reported for the IAEA-2D benchmark with two forms of approximate inversion, the second comprising an incomplete factorization with no fill-in. These new techniques also appear to be valuable tools for solving the 3D diffusion equation