Using MCNP for fusion neutronics

Abstract

Any fusion reactor using tritium-deuterium fusion will be a prolific source of 14 MeV neutrons. In fact, 80% of the fusion energy will be carried away by these neutrons. Thus it is essential to calculate what will happen to them, so that such quantities as the tritium breeding ratio, the neutron wall loading, heat deposition, various kinds of material damage and biological dose rates can be determined. Monte Carlo programs, in particular the widely-used MCNP, are the preferred tools for this. The International Fusion Materials Irradiation Facility (IFMIF), intended to test materials in intense neutron fields with a spectrum similar to that prevailing in fusion reactors, also requires neutronics calculations, with similar methods. In some cases these calculations can be very difficult. In particular shielding calculations - such as those needed to determine the heating of the superconducting field coils of ITER or the dose rate, during operation or after shutdown, outside ITER or in the space above the test cell of IFMIF - are very challenging. The thick shielding reduces the neutron flux by many orders of magnitude, so that analog calculations are impracticable and heavy variance reduction is needed, mainly importances or weight windows. On the other hand, the shields contain penetrations through which neutrons may stream. If the importances are much higher or the weight windows much lower at the outer end of such a penetration than at the inner end, this may lead to an excessive proliferation of tracks, which may even make the calculation break down. This dissertation describes the author's work in fusion neutronics, with the main emphasis on attempts to develop improved methods of performing such calculations. Two main approaches are described: trying to determine near-optimal importances or weight windows, and testing the "tally source" method suggested by John Hendricks as a way of biasing the neutron flux in angle.