Comparison of Bonner sphere responses calculated by different Monte Carlo codes at energies between 1 MeV and 1 GeV – Potential impact on neutron dosimetry at energies higher than 20 MeV

Bonner Spheres Spectrometry in its high-energy extended version is an established method to quantify neutrons at a wide energy range from several meV up to more than 1 GeV. In order to allow for quantitative measurements, the responses of the various spheres used in a Bonner Sphere Spectrometer (BSS) are usually simulated by Monte Carlo (MC) codes over the neutron energy range of interest. Because above 20 MeV experimental cross section data are scarce, intra-nuclear cascade (INC) and evaporation models are applied in these MC codes. It was suspected that this lack of data above 20 MeV may translate to differences in simulated BSS response functions depending on the MC code and nuclear models used, which in turn may add to the uncertainty involved in Bonner Sphere Spectrometry, in particular for neutron energies above 20 MeV. In order to investigate this issue in a systematic way, EURADOS (European Radiation Dosimetry Group) initiated an exercise where six groups having experience in neutron transport calculations with the MC codes (MCNP, MCNPX, FLUKA, PHITS, MARS, or GEANT4) calculated the responses of a bare He-3 proportional counter, a He-3 proportional counter embedded in the middle of a 9 inch polyethylene sphere, and a He-3 proportional counter centred in a 9 inch polyethylene sphere containing a lead shell, at neutron energies of 1, 10, 100, and 1000 MeV. In general, calculated responses agreed very well for neutron energies below 20 MeV, whatever MC code used. At higher energies, however, certain differences were observed among the different calculations, which may mainly be attributed to the application of different INC models and their parameters. It was found that up to 1 GeV most of the results ranged between calculations previously published that were obtained with MCNP/LAHET using the Bertini INC model and GEANT4 using the Binary and Bertini INC models. These results indicate that use of different MC codes and INC models for the calculation of BSS response functions may result in an uncertainty of unfolded neutron fluences above 20 MeV of about 20%.