Flash time too wide for multiple intime flash matching

[bear-is-asleep]

Description

The flash time window is currently 8 us, motivated by the cumulative light from an interaction being collected in that time period being maximal (see SBND PDS paper). However, if there are multiple flashes in a time window, such as an intime cosmic and a neutrino, or more commonly >1 neutrinos, the light from both of them will be merged. This can cause inefficiencies in any flash matcher, and only one of these interactions/slices can be matched (per TPC).

See docdb 40918 for more info.

Checklist

• I've added at least 1 label under Labels
• [na] Assigned a reviewer under Assignees

Possible Solution (optional)

A very short term solution may be to do topological cuts for analyses first. So if you have 2 neutrinos, you can possibly store the one most similar to your signal definition.

One option is to just use the fast component of the light to construct a flash for flash matching. We know the ratio of fast and slow light components, so we can use this as a scaling for our hypothesis construction for a fast flash. This will allow us to significantly reduce the time window (maybe to 100 ns).

[fjnicolas]

@bear-is-asleep thanks for the issue, just dropping here a few ideas.

I agree disentangling light from pile-up interactions would be a great enhancement to the OpFlash algorithm. It is is not trivial though, and it would require a dedicated study to evaluate its performance and ensure it doesn't degrade our current capabilities (i.e. I do not think it can be done in a rush for an imminent production).

• Reducing the flash integration window to 100 ns could result in a very large number of reconstructed OpFlashes, which may also negatively affect flash-matching. A significant amount of slow light still arrives after the first 100 ns, potentially triggering additional OpFlashes (after the 'prompt' OpFlash). While increasing the threshold could help suppress these slow-light OpFlashes, it would also reduce sensitivity to actual low energy physics.
• Instead, I propose keeping the current OpFlash finder setup with its 8 µs integration window. We could then implement a subroutine to search for a secondary peak within that original 8 µs window. For this, using waveforms instead of OpHits might be a good idea. To avoid retriggering on slow light, we can apply a more stringent threshold for detecting this second peak so we make sure we truly identify a second in-time interaction (or possibly a Michel electron 😉) instead of slow light garbage.