Import muon pair production sampling table from Geant4

[amandalund]

This imports the table used for sampling the electron-positron pair energy from the Geant4 muon pair production model.

[amandalund]

One slight complication is that different Geant4 versions use different models for muon ionization...

[sethrj]

Is the newer model available in both? Or should we just default to the same models available in the selected version of G4?

[sethrj]

Ummmm....

155/265 Test #157: celeritas/ext/GeantImporter:FourSteelSlabs* .....................................***Exception: SegFault  0.39 sec
Celeritas version 0.5.0-dev.277+20f1e85
Note: Google Test filter = FourSteelSlabs*
[==========] Running 17 tests from 1 test suite.
[----------] Global test environment set-up.
[----------] 17 tests from FourSteelSlabsEmStandard
[ RUN      ] FourSteelSlabsEmStandard.em_particles

[sethrj]

Great! Nice work.

[sethrj]

Is the newer model available in both? Or should we just default to the same models available in the selected version of G4?

[amandalund]

Yeah I'm not sure what's up with that test, and I don't have Celeritas built with Geant4 11.2 yet...

[sethrj]

The good news is it's reproducible on my machine with 11.2.1; the bad news is that GetElement2DData returns garbage for Z=12... perhaps the semantics of that call have changed?

[amandalund]

Ah excellent, thanks @sethrj! I'll look into it...

[sethrj]

Wait wait wait, it's not too late, but we almost missed the opportunity to call that Element Table a MuPPET!

[amandalund]

Ok it looks like in v11.2+ G4ElementData can be constructed with the maximum number of elements (which should be greater than the largest Z). G4MuPairProductionModel now uses this constructor, but calls it with NZDATPAIR, which is the size of the Z array, which is 5. It looks like there should be some kind of error when initializing the data though for Z >= 5...

[sethrj]

@amandalund Yeah, so the ElementTable is designed to have the index be the element ID, but the MuPPET is fixed at size 5 (an implementation detail) and those are just indices (index of ZDATPAIR) rather than Z number (value in ZDATPAIR). There appears to be no way to get the size (the private data member maxNumElm). Also it seems to use the Z number to access the muppet in some places, and uses the index to set it. OK it looks like universally it just misuses the table, accessing by Z index rather than Z number. I don't know how this has worked before now ... but I do find the British slang appropriately ironic 😉 https://clip.cafe/lock-stock-two-smoking-barrels-1998/i-dont-wanna-know-you-use/

I think the solution is for the muppet code to hard-code Z = 5? I can submit a PR to add an accessor to the table limits.

[amandalund]

Haha lord, ok, using the index to set it rather than Z (and access it... sometimes?) is also new since v11.2. Thanks for looking into that more; I'll update the import for the newer versions and I guess just hardcode the size... 🐸 🐷

[sethrj]

Oh, so on older versions the table actually is indexed by element (and just has null for the ones that it's interpolating between?) That makes more sense and is a totally different implementation than the current one. See my update that it is actually using just entries 0 through 4; the current version does not expose the {1, 4, 13, 29, 92} element numbers in any way. Guess we will have to switch on version number... or reimplement the cross sections ourselves 🤪

[amandalund]

Yeah, the older versions used the table in a more intuitive way/how it was designed to be used.

or reimplement the cross sections ourselves 🤪

I'm tempted at this point ;)

[amandalund]

Also it seems to use the Z number to access the muppet in some places, and uses the index to set it. OK it looks like universally it just misuses the table, accessing by Z index rather than Z number.

It is still erroneously accessing the tables by the Z number in G4MuPairProductionModel::StoreTables(), right?

[sethrj]

Ah indeed, I got confused because I'd changed some lines but not all. Clearly no one used or tested that method...