Built-in physics models and proton-induced nuclear data validation using MCNP, PHITS, and FLUKA – Impact on the shielding design for proton accelerator facilities

The MCNP, PHITS, and FLUKA are general-purpose Monte Carlo (MC) radiation transport codes that are widely used for many real-world shielding problems at accelerator facilities around the world. Nuclear interactions are described in these codes by either built-in physics models, or tables with evaluated cross sections and secondary energy-angular distributions, or a combination of both. Over the decades, many code validation efforts have been made, owing to the availability of shielding benchmarks to test the physics models and nuclear data and to verify the accuracy of simulation codes.

For high beam energy and high beam current accelerator applications, neutron emission through the vacuum pipe along the reverse direction of incident proton beam is an important factor for a shielding design in order to correctly assess the dose rates for workers and the structural materials of the accelerator and handle with the waste activated by the backscattered neutron fluxes. In this work, neutron-production cross sections and thick-target yield predictions from MC codes relying on physics models and nuclear data libraries are benchmarked against the experimental data, in order to assess their accuracy in predicting neutron emission and furthermore to assess the corresponding impact on shielding design.

The results of this study demonstrate that the nuclear data libraries and physics models, which are not expected to give good results at lower energies (< 150 MeV) but are used anyhow when there is no nuclear data available or above the energy range where the data tables end in the so-called “mix-and-match” strategy, need further improvements. Among the investigated proton induced nuclear data libraries, JENDL-4.0/HE produces the most satisfactory agreement to experimental data for all target materials, but may still benefit from refinement. Concerning the physics models of the codes, FLUKA V4-4.0 has the best performance in terms of reproducibility of the experimental values. It is also shown that all discrepancies between the calculations and the experiments for the energy range < 10 MeV (which is dominant on the dose rates through the shield thickness), are up to factor of two. This might be considered as an acceptable figure as it is equivalent to a normal safety margin (x2) considered in shielding calculations of accelerator facilities around the world.