Tau Studies in MAIA

Thursday 9 Oct 2025, 15:00 → 16:00 Europe/Berlin

Gregory Penn (Yale University)

Zoom link:
https://yale.zoom.us/j/91200672837

15:00 → 15:20 Updates from Yale Group

Speakers: Ethan Martinez (Yale University), Gregory Penn (Yale University)

Minutes:

Ethan has started to submit BIB samples in chunks of 5 events. The 5 are taking > 20 hours. We will look into whether the OSG is using CPUs or GPUs, and explore options with GPUs (possibly UW-Madison's cluster) if time is a constraint. Giacomo brought up the tracking config as a place where we can speed up reconstruction. We'll compare the 3 TeV and 10 TeV tracking scripts to see if there are improvements we can make to shorten the runtime. 


Intro:

• BIB samples: There were slides shown (not yet uploaded) in last weeks' MAIA general meeting that highlighted asymmetries in the simulated samples. These samples are being used for our current BIB submission. We'll have to follow this as it develops.
• This meeting:
  • Organize BIB submission. Is everyone able to access the open science grid? Any more technical challenges?
  • Define "money plots" (without BIB) to produce soon. 
    • Efficiency for all decay modes overlaid (such as below)
      • How should we define charged pions? What message do we want to communicate?
    • Rejection / efficiency on jets, electrons, muons vs. pT
    • Tau energy resolution vs. pT. Example plots shown below (ATLAS tau reconstruction note)
    • H / Z / mumu invariant mass distributions (such as below, already sufficient for current definitions)

 

15:20 → 15:40 Updates from University of Wisconsin

Speakers: Cyrus Kianian (UW-Madison), Moses Glassman (University of Wisconsin-Madison)

CK - TauFinder 10_9 - MAIA.pdf

TauFinder Update 9-29-25.pdf

Cyrus has looked at the effect of merging tau candidates:

• Showed that only a few percent of truth taus are split into multiple reconstructed taus.
• Showed that increasing the signal cone size catches nearly all of these fail cases.

Q&A:

• Clarification the purpose of tau merging. It is done to see if the change of a tau axis, through adding additional tracks, now moves another track within the signal cone. This is a product of seeding the tau on tracks. Nothing like this exists in ATLAS / CMS due to seeding taus in the calorimeter. 
• Asked about 0-prong taus. Cyrus will look into what particles they are compose of (presumably only neutrals), and how TauFinder can seed on tracks yet return a tau without charged particles.
• Noted that we could remove merging given its small effect, even while keeping the signal cone the same. We will see how TauFinder must be changed with BIB added, then re-evaluate.

 

Moses has looked into using the yoke for muon veto:

• 3% of taus leave energy in the yoke, 59% of muons leave energy in the yoke.
• Cutting on yoke energy indeed lowers the efficiency and rejection by this amount.

 

Q&A:

• Requested slides explaining what the yoke variable is. The definition is somewhere in the code. Greg and Moses will follow up offline.
• Muon veto in the future will probably take form as muon reconstruction is explored. At the moment, vetoing based on the yoke variable is best if it is done without any hit to tau efficiency.

15:40 → 16:00 Updates from LIP

Speaker: Kevin Dewyspelaere (MuColl)

LIP Muon collider 09_10.pdf

Tau energy corrections.pdf

Kevin:

• Explored isolation energy requirements on truth taus in various processes and for light-quarks and b-quark backgrounds. Looked at both the total isolation energy and the fraction of isolation energy over the leading tau track pT. Found that they both provide very strong rejection power with only a reduction in tau efficiency of a few percent.
• Showed tau energy correction plots. Demonstrated that the energy correction is minimal with the dynamic cone, and worsens significantly with its removal.
  
  

Q&A:

• The cut of relative isolation energy to leading tau track pT is more robust than the absolute isolation energy, as it roughly removes tau pT dependence.
  • It could be considered to change the leading tau track pT to tau pT. Kevin will compare these two.
• The process dependence of the tau isolation energy is interesting. From the isolation energy vs. tau track pT curve, it appears that there is additional dependence beyond pT. This is probably a dR(tau,tau) effect. 
• Discussed turning the table of jet efficiencies into a plot vs truth pt (or eta). There are some caveats in how to define the denominator, as we don't have a solid jet definition (no FastJet in the reco config used). We settled on the denominator as # of truth quarks, and the numerator being the total # of reconstructed taus in the sample. Some caveats will exist with the denominator definition.
• Discussed turning the energy correction plots into an energy resolution plot. Kevin will generate these.
• Discussed how to handle taus with very poor resolution, particularly due to misclassification of its decay mode. These can impact the energy correction (the example on slide 7 is a 22 GeV correction, which is dramatic at low pT). Decided that it is reasonable to throw away very poor taus for the energy correction, but we will revisit this once we see the energy residual distributions.
  • Kevin mentioned that the energy correction is ~ the same when 0 - 20 GeV taus are excluded.