Characterization of photoneutron fluxes emitted by electron accelerators in the 4-20 MeV range using Monte Carlo codes: A critical review
Résumé
Applications of electron accelerators range from nuclear waste package assay and security-related tasks to radiation therapy. Studies aiming at characterizing photoneutron fluxes generated by electron accelerators are usually based on Monte Carlo simulation. In this paper, we critically review the performance of Monte Carlo transport codes to simulate photoneutron fluxes emitted by electron accelerators operating between 4 and 20 MeV, typically the energy range of interest for the aforementioned applications. First, we go through the state of the art and lay the foundations of current theoretical knowledge on photoneutrons. By carrying out additional investigations, we show that contamination of photoneutron fluxes by electroneutrons is likely to lie between 0 and 2%. Second, we assess the characteristics of photoneutron fluxes emitted by tungsten or tantalum conversion targets and by heavy water or beryllium secondary targets. This characterization step is conducted with MCNP6.2, which is often considered as one of the reference Monte Carlo codes, and built around three parameters, i.e., photoneutron yield cross-sections, energy spectrum and angular distribution. In particular, we demonstrate that erroneous parameters in the nuclear data of MCNP6.2 lead to (γ, xn) cross-section threshold errors for two tungsten isotopes, i.e., $^{182}$W and $^{186}$W, inducing in turn a global underestimation of photoneutron production in tungsten. Furthermore, by taking an in-depth look at nuclear data libraries, we show that photoneutron yield cross-sections are sometimes poorly evaluated below 20 MeV, e.g., $^2$H, $^9$Be and $^{184}$W. Third, thanks to vanadium and aluminium foils, we benchmark MCNP against photoneutron activation measurements conducted in the vicinity of three different electron accelerators, including a medical one. Simulation of these measurements denote a systematic underestimation trend extending from a few percent to a factor ten. Recent findings reported in the literature proved that photoneutron kinematics is implemented in MCNP with erroneous equations related to neutron inelastic scattering, causing hardening of photoneutron energy spectrum and may explain in part the discrepancies encountered in this MCNP benchmark study. Finally, in light of the three main sources of errors that potentially lead to unreliable results when simulating photoneutron fluxes with Monte Carlo codes - implementation of nuclear data, modelling of photonuclear physics and fundamental knowledge of photoneutron yield cross-sections - we issue recommendations for the code developers and users. Until further progress is made in the field of photoneutron simulation, mastering the current limitations of Monte Carlo codes could be the first milestone for their users.