Coupled electron/photon S(N) calculation in lattice geometry

Abstract

The capabilities of the coupled charged/neutral particle transport S(N) code SMARTEPANTS (Simulating Many Accumulative Rutherford Trajectories Electron Photon and Neutral Transport Solver) have been extended from x-y-z geometry to x-y-z geometry with embedded cylinders. A new method called the super-cell algorithm was applied to accommodate cylindrical shapes using a rectangular mesh. The super-cell is defined as a rectangular mesh cell containing one or more material interfaces. Each material region within a super-cell constitutes a sub-cell. To model cylindrical shapes, curved sub-cell interfaces were used. The critical aspect of the super-cell method was to determine the angular fluxes in a sub-cell within the super-cell. To do this, the super-cells were divided into two major categories, Type-1 and Type-2. The cylinder's radius compared to the super-cell's mesh size was used as a basis to distinguish the super-cell's type. Each type was divided into several sub-cases depending on the direction cosine of the angular flux when entering the super-cell. The super-cell method was integrated into SMARTEPANTS and then used to calculate the energy deposition for a variety of test problems to check the method's sensitivity to its parameters. For a block of galium-arsenide (Ga-As) with an embedded gold cylinder, it was found that an S₈ quadrature set with a five energy groups is both time efficient and yields satisfactory results. The effect of cylinder radius compared to the mesh size in Type-2 super-cells was found to be minimum for an optimum mesh size. Several benchmark problems were performed to compare the super-cells results with coupled electron/photon Monte Carlo code (ITS). The total energy deposition in the peak energy cell was selected to facilitate the comparison. Peak energy cell is the cell with the maximum energy deposition. For an isotropic electron source in a Ga-As block embedded with Type-1 and Type-2 gold cylinders the results were within 3% and 6% respectively and SMARTEPANTS results in the non-super-cells were more symmetric than Monte Carlo. Super-cell also demonstrated better computer efficiency both in CPU time and memory when compared with the Monte Carlo method on the same machine.