Dear all,

Iam trying to compare the analytical solution for transmission of a neutron beam (fast neutrons, from 60 keV to 20 MeV, mean energy of 2.3 MeV) with the solution given by MCNP6 (version 6.1).

Some words to the analytical solution:

I have used the NJOY software to obtain grouped cross sections, for every neutron energy bin. Dividing the sum of the product between the neutron flux and the grouped cross section in each bin by the integral neutron flux in the total considered neutron energy bin rage (60 keV to 20 MeV) yields an integral microscopic cross section (ENDF MT=1). Based on this cross section the macroscopic cross section is calculated analytically as usual and the mean free path length equals the reciproke of it.

For aluminum I find a total macroscopic cross section of 0.189 cm^-1 and hence a mean free path length of 5.304 cm in the analytical solution. However, in MCNP6 I find a value of 5.7560 cm^-1 (mfp value in the print table 126 in the output file) for the case of a 2.5 x 2.5 cm^2 Al sample, completely irradiated by a parallel and uniformaly-distributed neutron beam. In both solutions I use exactly the same cross section libraries (ENDF/B-VIII.0). For other samples I found even higher or sometimes lower discrepancies, so the observed difference depends on the sample.

The analytical model is simple and well established, it should be correct. I can not understand from where the discrepancy comes, for certain simulations with monoenergetic neutron beams of discrete energy I find a nearly perfect agreement. Effects such as a multiple scattering and so on should be negligible in a sample of 3 mm or less.

What I have also found out:

Comparing the results for a single energy bin, the solutions agree well if the respective cross-section curve has no gradient but is constant. When it has a slope or resonances, the discrepancy becomes large. Also, if I substitute each energy bin by the mean energy and perform a simulation with a distribution of discrete energies, then the results rather agree. The problem/discrepancy arises once I simulate an energy distribution of neutrons.

Did anyone of you face the same issue with MCNP? I would be very grateful if someone could give me a hint what might be the problem and how I could resolve the observed discrepancy.

Kind regards,

Niklas