Assessing stray neutron dose variability across Monte Carlo codes in a proton therapy scenario

Monte Carlo (MC) simulations are widely used in proton therapy dosimetry, but neutron dose estimates vary significantly across different MC codes, nuclear models, and cross-section libraries. This study provides an overview of the expected differences in neutron fluence and dose equivalent calculations across five MC codes (PHITS, MCNP, FLUKA, GATE, and TOPAS) for a clinically-relevant proton beam model in a reference phantom.

A spread-out Bragg peak (SOBP) proton beam was simulated in a water phantom. Neutron fluence spectra and dose equivalents were evaluated at multiple out-of-field positions using each code's default nuclear models and cross-section libraries. Additional simulations explored the impact of alternative cross-section libraries (JENDL-5, JEFF-3.3, TENDL-2021) and nuclear models.

While proton dose distributions were consistent across codes (spatial variations < 2 mm), neutron fluences differed by up to 130 %, and dose equivalents varied by 88 % depending on nuclear data. Proton cross-sections had a greater influence on neutron spectra than neutron cross-sections. Codes using the same nuclear data libraries produced comparable results, highlighting cross-section selection as a key source of variability.

These results provide MC users with a reference for expected differences in neutron dose calculations. While default MC settings offer reasonable internal consistency, the observed variability underscores the need for benchmarking against experimental data to ensure accurate neutron dose predictions in proton therapy.