The Finite Difference Time Domain (FDTD) method is widely adopted as a suitable numerical technique to efficiently solve the Maxwell equations both in free space and in presence of lossy media, like a human being. In non ionizing electromagnetic fields radioprotection, it is well known that the computer simulation is the only feasible approach to describe the detailed SAR distribution inside the whole human body or selected districts of it. Spatial resolutions on the order of 1-2 mm are now easily achievable regarding the tissues complex permittivity maps [1], [2]. Therefore to describe a human body, a complete 3-dimensional FDTD grid requires a large number of cubic cells, typically on the order of 107-108, which cause an unmanageable request of computer RAM and long running time on a single processor. To overcome these drawbacks we propose a parallel MPI (Message Passing Interface) implementation [3] of the subgridding algorithm as described in [4] with the aim of sharing the RAM request and spatial iterations of the FDTD grid among a number of processors. As practical example here we show the MPI parallelization of a large FDTD domain, in which five human bodies are simultaneously frontally illuminated by a 1800 MHz DCS antenna in the near field situation.
On the Combined Use of MPI Code and a Subgridding Algorithm to Solve Maxwell s Equations in Large and Complex FDTD Domains
Pontalti, Rolando;Vaccari, Alessandro
2009-01-01
Abstract
The Finite Difference Time Domain (FDTD) method is widely adopted as a suitable numerical technique to efficiently solve the Maxwell equations both in free space and in presence of lossy media, like a human being. In non ionizing electromagnetic fields radioprotection, it is well known that the computer simulation is the only feasible approach to describe the detailed SAR distribution inside the whole human body or selected districts of it. Spatial resolutions on the order of 1-2 mm are now easily achievable regarding the tissues complex permittivity maps [1], [2]. Therefore to describe a human body, a complete 3-dimensional FDTD grid requires a large number of cubic cells, typically on the order of 107-108, which cause an unmanageable request of computer RAM and long running time on a single processor. To overcome these drawbacks we propose a parallel MPI (Message Passing Interface) implementation [3] of the subgridding algorithm as described in [4] with the aim of sharing the RAM request and spatial iterations of the FDTD grid among a number of processors. As practical example here we show the MPI parallelization of a large FDTD domain, in which five human bodies are simultaneously frontally illuminated by a 1800 MHz DCS antenna in the near field situation.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
