Modeling of a Spherical Tokamak as an Extended Neutron Source Using ASTRA and MCNP

Fast neutrons can perform two functions that are critical to the nuclear power industry: breed fissile material from fertile material and burn minor actinides in spent fuel. The fusion reaction in a fusion reactor between deuterium and tritium nuclei yields an alpha particle and a neutron with 14.5-MeV energy. These neutrons can easily convert fertile material (such as 238U and 232Th) into fissile material (such as 239Pu and 233U) and destroy minor actinides (Pu, Np, Am, and Cm) present in the spent nuclear fuel and significantly reduce its radiotoxicity. However, each neutron generated implies one tritium consumed and tritium is generated by nuclear reactions between a neutron and lithium nuclei. Therefore, neutron economy becomes a crucial issue and an adequate neutronic modeling of the neutron source becomes critical. In this article, the Automatic System for TRansport Analysis (ASTRA) and Monte Carlo N-Particle (MCNP) codes are used in conjunction to characterize the neutron flux coming out of a low-aspect-ratio tokamak machine. ASTRA is first used to determine temperature and density profiles within the plasma volume and MCNP is used to evaluate the neutron flux. Since the simulation of an extended neutron source with complex geometry such as the confined plasma inside the tokamak is difficult, steps are taken to simplify the geometry by defining two simple sources, a cylindrical wall and a disk, which should have similar neutron emission distributions to the volumetric source. This reduction is done using a preprocessing Monte Carlo algorithm to establish probability density functions of neutron emission for the cylindrical wall as a function of the height and for the disk as a function of the radius, based on the emission characteristics of the extended source. The results that compare the neutron flux leaving a volume for four different configurations of neutron sources (isotropic source at the geometrical origin, ring source at the plasma geometrical center, cylindrical wall source, and distributed multipoint source) are presented. It is found that the processing of the two simple sources has a higher computational cost in terms of both processing time and random number consumption.