Bad conversion of a MCNP SSW using ssw2mcpl?

[s0menz]

Hello,

I converted an SSW file created in MCNP6 (version 6.1.0) to an MCPL (version 2.2.4) file for use in Geant4. The original source is a box containing thermal neutrons and gamma rays, mainly prompt gammas from graphite. After the conversion to MCPL, I split the source into its two components: neutrons and gammas.

The issue is that the gamma source is missing some peaks that we would expect from the activation of graphite (around 1.8 MeV and 3 MeV). This does not happen for all gamma peaks; most of them are present. I verified this by comparing a plot of the MCPL gamma source with a tallied gamma current in MCNP using a box slightly smaller than the SSW. In that MCNP simulation, particles are killed when they enter the tally box, and they travel only 1 mm in air.

Furthermore, when I use the two MCPL sources (neutrons and gammas) in Geant4 and reproduce the same tally as in MCNP, I observe a gamma current 100 times higher and a neutron current 10 times higher.

An additional note: I used different importances when building the SSW in MCNP. Could this issue be related to a wrong conversion of particle weights using ssw2mcpl? What puzzles me the most is that some gammas are entirely missing in the MCPL file.

Any help would be greatly appreciated. Thanks!

[tkittel]

It is really hard to say what might be the problem here. But I am strongly suspecting that the error lies on the MCNP side of things, since the conversion from ssw to mcpl of each particle is always based on more or less the same unpacking of fortran records of 11 (IIRC) double precision floats. Also, it is hard to imagine why only some gamma peaks would be missing.

Of course, hard to say for sure. Perhaps you can first try with a simpler MCNP setup where there are less complications (importances, funny tallies, etc.) and see if the problem persists.

[tkittel]

Sorry @s0menz I notice that you had closed your question again. If this was in error, let me know. If the question has instead been resolved, we would appreciate a comment here about what was wrong, so others might benefit in the future :-)

[s0menz]

Thanks @tkittel for your reply. I'm sorry — I had already closed the discussion before seeing your message, as I realized I had made some mistakes in evaluating the issue. The conversion with ssw2mcpl went well.

1. I discovered that the two peaks were also missing in the original MCNP source. I think I probably saw them in the output due to coarse energy binning in the tally, which made some small, broadened peaks appear. I initially believed those peaks were supposed to be present because they're listed in the NIST databases. However, after refining the binning in the MCNP6 tally, I saw that those peaks are actually buried within the gamma background noise.

2. The weights were correctly converted. The issue came from two mistakes I found in the Geant4 setup:

• First, I was using the Primary Vertex in the PrimaryGeneratorAction, which altered (I still don't know why ) the weights as they passed to the SteppingAction, breaking the proportionality.
• Second, I wasn’t applying the weights when filling the histograms from the ntuples — so they weren’t being accounted for properly.

Now that I’ve fixed these issues, the two sources match perfectly. The only remaining difference is that the MCNP6 source appears to be about 1.2 times higher than the MCPL one. Could this be due to ssw2mcpl having recorded more NPS events than the original MCNP source? In the mcpl file I see 19% more of particles than the original mcnp source

Thanks again for your help!

[tkittel]

Ok, thanks for the details! I can't really say that I can explain the 19% difference unfortunately. My best guess is that this 19% difference is already present in the ssw file written by MCNP, since all ssw2mcpl does is more or less copy the particles found in the ssw file into the new mcpl file.