Optimal space-energy splitting in MCNP with the DSA

The DSA theory is based on the possibility to obtain exact explicit expressions for the dependence of the second moment and calculation time on the splitting parameters. This allows the automatic optimization of the splitting parameters by "learning" the problem's bulk parameters from which the problem depended coefficients of the quality function (second moment ^* time) are constructed. The above procedure was exploited to implement an automatic optimization of the splitting parameters in the MCNP code. This was done in a number of steps. Firstly only spatial surface splitting was considered. In this step, the major obstacle has been the truncation of an infinite series of "products" of "surface path's" leading from the source to the detector. The encouraging results of the first phase led to the inclusion of full space/energy phase space splitting.