Antares/KM3NeT group meeting - reconstruction




Join Zoom Meeting

Meeting ID: 958 9367 1004
Passcode: 745504




For MC-hits, I use Poisson hit probability, within 80 m from the track.
For real hits, use both hit and no-hit information.

I use two tests to see whether there is a Bremsstrahlung shower.
Poisson-test shows 5 Bremsstrahlung shower candidates in 21 events.

If I scan over a range of shower distances, the likelihood contains a lot of local minima.
I'm investigating whether the scipy-optimizer finds the global minimum correctly here.

If I fix the distance, but vary the energy, the likelihood looks much better.

For the real hits, the time-residuals are not very well approximated by the PDF.
Usually, I see less hits than I would expect around a time-residual of 0.


Ronald: Can you look at whether the extra spread in the time residuals derives from the amount of misreconstruction?

Ronald: Are you constructing the PDF from the reco-tracks or MC-truth tracks?

--> reco-tracks.

Ronald: Using the reco-tracks may bias the PDF.

Maarten: The reconstruction looks for the optimal track, so you don't expect a perfect match here. Additionally, your hypothesis may be more perfect than reality here. The events you show seem not very well reconstructed and maybe also do not contain as perfect a shower as ...
If your procedure can be used to pick out events with a huge stochastic energy loss, this would be extremely cool and we can use your tool to
Would love to see identification of events with 2 or more showers and plot the energy ratio of the first and the last shower. There should be a bias there.

Ronald: You can actually fit the energy loss.


A small addition to what I showed yesterday:
Deviation around x-position of vertex gives very sharp likelihood minimum.

Maarten: What do you expect? Shower fit angular resolution is several dozen of degrees. So an improvement here is probably not expected.

Jordan: You can maybe estimate the resolution from the width of the minimum and compare to the angular resolution you expect.

Ronald: The x-, y- and z-axis are fixed to the detector geometry here, so these plots are sometimes difficult to interpret.

Maarten: At these energies, the only information that depends on the shower direction is the hit- no-hit info.
Can we do a signal+background and background only hypothesis and take the difference?
Then the number of no-hit PMTs included should not matter a lot anymore.

Ronald: One of the issues here is that we may be including a large amount of hits also from the hadronic shower in the nue-CC events. These should be removed to make a clear comparison.

Thijs: In energy-deviation right-side plot, we see the minimum located at -10 GeV. If you would have a hadronic shower that was not taken into account, we would expect the minimum to be at positive energy here.

Ronald: That depends upon how the deviation is defined here.

Jordan: Might also be a good idea to include energy-elongation.

There are minutes attached to this event. Show them.
    • 14:00 14:10
      Brían 10m
      Speaker: Mr Brían Ó Fearraigh (Nikhef)
    • 14:10 14:20
      Thijs 10m
      Speaker: Thijs Juan van Eeden
    • 14:20 14:30
      Ronald 10m
      Speaker: Ronald Bruijn
    • 14:30 14:40
      Jordan 10m
      Speaker: Jordan Seneca
    • 14:40 14:50
      Aart 10m
      Speaker: Aart Heijboer
    • 14:50 15:00
      Maarten 10m
      Speaker: Maarten de Jong
    • 15:00 15:10
      Bouke 10m
      Speaker: Bouke Jisse Jung
    • 15:10 15:25
      Valentin 15m
      Speaker: Valentin Pestel
    • 15:25 15:40
      Timon 15m
      Speaker: Timon van Dieren