Add NR cut to file loader

[PavelKavrigin]

Simple implementation of NR selections from ROOT files using a modified cut string. An additional parameter in the configuration ('nr_only') is used to turn on/off this cut. This PR can be used as a basis for other cuts or a modification of this cut.

[PavelKavrigin]

Here's a basic test using 20 files with AmBe simulations. These are 'e_dep' histograms, with NR-only selection via cut string (first plot) and without the cut (second plot). The output of this branch is compared to the master branch of epix. NR selection in the master branch was done by selecting only 'recoil==0' rows from the output.

[zhut19]

Hi @PavelKavrigin @ramirezdiego , it is better to keep the non-neutron / ion energy deposition in the same event, no?
So the nr selection cannot apply at the root to array level, but a bit trickier after that line.
Here is some example code I am trying:
mask = ((interactions['type'] == "neutron") & (interactions['edproc'] == "hadElastic")) | (interactions['edproc'] == "ionIoni")
interactions = interactions[ak.num(interactions[mask])['x'] > 0]

[ramirezdiego]

@zhut19 Agreed (and @PavelKavrigin too, as we discussed this fix last week again). The idea is to propagate this cut, but keep the entire event level information, rather than this step in particular.

[ramirezdiego]

@shenyangshi Do you think you can jump on this, following the syntax by Pavel and the above suggestion by @zhut19?

[shenyangshi]

Hi @ramirezdiego, the NR selection is implemented. I select events with summed ER components smaller than 32kev and having NR components larger than 0kev. Here is the graph showing the NR components considering the ER effect or not.

[WenzDaniel]

Just for the record: This means that we are removing events with an inelastic component a 40 keV and 100 keV. This might be important for any TPC based neutron-veto.

[PavelKavrigin]

Is it safe to ignore the timing and the location of the energy depositions? There could be cases where NR and ER do not occur close to each other. We will throw away all of these events based on the total energy.

[shenyangshi]

Maybe yes, as NR and ER particles are all nearly to the speed of light, so both NR and ER will contribute to the S1, but for S2 if the distance is far maybe one of them won't be in the same event window, but this kind of event won't be in the neutron band anyway.

[WenzDaniel]

Based on how we build events the S2 should be within the event as long as the second interaction happens in the sensitive volume.

[WenzDaniel]

In any case a suggestion. I think it is worthwhile to store a second epix file containing all "rejected events".

[ramirezdiego]

Just for the record: This means that we are removing events with an inelastic component a 40 keV and 100 keV. This might be important for any TPC based neutron-veto.

@WenzDaniel Not sure I get your comment. This cut should only be applied when we want to estimate the pure NR SS, so disabled by default.

[ramirezdiego]

@shenyangshi @PavelKavrigin If this is not adding a significan delay to the code, I would suggest that we always store the e_dep_er and e_dep_nr branches per event, and use the nr_cut flag to simply select events according to them (this can also help @WenzDaniel with his current checks, maybe?).

On the other side, if the output file can't be modified now (because this would require doing the same with WFSim), let's just leave these branches floating and fix that in future commits.

[ramirezdiego]

@shenyangshi Can you refresh my memory on the 32 keV upper bound for ERs? I was assuming we'd use a lower one. Anything to do with Kr83m?

[shenyangshi]

No, it just comes to me that the ER energy in NR events is well above the NR energy scale, so selecting 10kev or 32kev wouldn't make a difference to the results. The upper bound for Ed_nr*Lindhard factor is around 32kev, so I selected this value.

[ramirezdiego]

@shenyangshi Could you please add the NR cut as a keyword argument in bin/run_epix?

[shenyangshi]

Sure, it's done, sorry I didn't notice this file before.

[shenyangshi]

This NR cut essentially finds out NR events (labeled by g4id, shared by interactions within) that contain ER components (beta and gamma) smaller than 10kev.

For example, the following whole event is removed because there are two huge ER energy depositions:

But the following event is preserved because only NR is involved: