Computation Speeds and Memory Requirements of Mesh-Type ICRP Reference Computational Phantoms in Geant4, MCNP6, and PHITS.

Recently, Task Group 103 of the International Commission on Radiological Protection completed the development of new adult male and female mesh-type reference computational phantoms, which are planned for use in future International Commission on Radiological Protection dose coefficient calculations. In the present study, the performance of major Monte Carlo particle transport codes, i.e., Geant4, MCNP6, and PHITS, were investigated for the mesh-type reference computational phantoms by performing transport simulations of photons, electrons, neutrons, and helium ions for some external and internal exposures, and simultaneously measuring the memory usage, initialization time, and computation speed of the adult male mesh-type reference computational phantom in the codes. The measured results were then compared with the values measured with the current adult male voxel-type reference computational phantom in International Commission on Radiological Protection Publication 110 as well as five voxel phantoms produced from the adult male mesh-type reference computational phantom with different voxel resolutions, i.e., 0.1 × 0.1 × 0.1 mm, 0.6 × 0.6 × 0.6 mm, 1 × 1 × 1 mm, 2 × 2 × 2 mm, and 4 × 4 × 4 mm. From the results, it was found that in all of the codes, the memory usage of the mesh-type reference computational phantom is greater than that of the voxel-type reference computational phantom and the lowest resolution voxelized phantom, but it is sufficiently lower than the maximum memory, 64 GB, that can be installed in a personal computer. The required initialization time of the mesh-type reference computational phantom and of the voxel-type reference computational phantom and voxelized phantoms in resolutions lower than 0.6 × 0.6 × 0.6 mm was less than a few minutes in all of the codes. As for the computation speed among the codes, MCNP6 showed the worst performance for the mesh-type reference computational phantom, which was slower than that for the voxel-type reference computational phantom by up to ~50 times and slower than that for all of the voxelized phantoms by up to ~40 times. By contrast, PHITS showed the best performance for the mesh-type reference computational phantom, which was faster than that for the voxel-type reference computational phantom by up to ~3 times and faster than that for all of the voxelized phantoms by up to ~20 times. This high performance of PHITS is indeed encouraging considering that it is used nowadays by the International Commission on Radiological Protection for most dose coefficient calculations.