CLD fullsim for FCCee: muon->pion misidentification at 2-7 GeV using PandoraPF

[mcepeda]

Dear all,

I am working on tau reconstruction using full simulated samples with CLD, starting from Pandora PFOs.

I have noticed a sizeable contamination of muons in my analysis: these are muons originating from tau decays which are identified as pions by Pandora, concentrated between 2 and 7 GeV and at central Theta. This is using the "PandoraPFOs" sample (image attached) , and I see a similar tendency with "TightSelectedPandoraPFOs". Note in the image there are also muons below 2 GeV and a small leakage at high momentum at Theta=pi/2: those are not an issue.

@Zehvogel pointed me to a difference in muon algorithm starting at 7 GeV, which should explains this behaviour:

https://github.com/PandoraPFA/LCContent/blob/master/src/LCPlugins/LCParticleIdPlugins.cc#L434 (LCMuonId)
https://github.com/PandoraPFA/LCContent/blob/master/src/LCParticleId/MuonReconstructionAlgorithm.cc

Is there an easy way to tune the settings of the Muon ID in Pandora, eg changing the threshold to start at 5 GeV ? If there is any documentation about how to tune this I can do a check.

Alternatively, any further tweaks on how to improve the muon ID at low momentum would be great.

More plots are on the way (requested by @BrieucF ), I will update the ticket as soon as they are ready.

Regards,

María Cepeda

[andresailer]

https://github.com/PandoraPFA/LCContent/blob/aabb475ab87a538826befb36e79e27bc977e321f/src/LCParticleId/MuonReconstructionAlgorithm.cc#L28

    m_minTrackCandidateEnergy(7.f),

Add after

CLDConfig/CLDConfig/PandoraSettingsCLD/PandoraSettingsDefault.xml

Line 69 in f265f3d

<MinClusterOccupiedLayers>5</MinClusterOccupiedLayers>

<MinTrackCandidateEnergy> 5.0 </MinTrackCandidateEnergy>

[Zehvogel]

Just adding some old plots that I dug out showing muon PID efficiency and purity for the ILD detector. There the MuonReconstructionAlgorithm is enabled from 4 GeV but the efficiency is degraded in the barrel up to ~10 GeV. The LCMuonId running below 4 Gev produces many fake muons (at least for ILD) .

mu_eff_pandora.pdf
mu_pur_pandora.pdf

If the efficiency of the MuonReconstructionAlgorithm is too low for your purposes you can also try to relax the matching criteria of muon system clusters to tracks which should be these parameters.

    m_maxDistanceToTrack(200.f),
    m_minHelixClusterCosAngle(0.98f),

[mcepeda]

Thanks both and apologies for the delayed reply. Changing to 4.0 gave me a good enough performance to start the tau reconstruction in the decay channel to a single pion at 5 GeV, which should be Ok for now. The muon background is not a problem anymore in this channel. In this analysis low pt muons are not really used, the problem was just the muon->pion fake rate, so I did not tweak the Muon efficiency.

Before the fix:

After the fix (with smaller statistics, I'm processing a bigger sample as we speak):

[andresailer]

Great!

(Latex nitpick: $\cos\theta$ to have $\cos\theta$ and not $cos\theta$ 😄 )

[BrieucF]

Very Nice!

Do we have some kind of junction effect at $\cos\theta = 0$ which makes the pions more likely to punch trhough? Has anyone a plot of the CLD interaction length vs $\cos\theta$ including the calorimeters?

[andresailer]

From https://arxiv.org/pdf/1911.12230

Just from trigonometry there is less material particles have to go through at $\cos\theta = 0$

[andresailer]

PS: The result for muons is illustrated in Figure 41(a) as a function of energy and in Figure 41(b) as a
function of cos θ . The efficiency is above 99% for all energies, and generally flat as a function of cos θ .
The small dip around 90◦, which is also observed in CLICdet, is the subject of further investigations.

[BrieucF]

Thanks Andre, interesting!

The purely geometrical effect of "seeing" less material towards trajectories perpendicular to the barrel should have "smooth" consequences on pion punch-through. The interaction length plot makes me think this is not a material budget origin. Also seeing an effect in single muon particle gun efficiency suggests something more general.

[Zehvogel]

Could be an effect of the particle moving on exactly the border of the calorimeter cells at 90 degree's. Maybe causing hits on either/both sides of the border. That could throw off the muon reconstruction algorithm, it looks for a very particular shape of neatly lined up hits.

(Sorry for closing, fat fingers :()

[mcepeda]

Regarding the dip around 90º - I can add that those are not many, but populate higher momentum, and are very very narrowly focused in theta. You can see them in the first plot (2D) I circulated. They are not really an issue for the analysis since they are very easy to remove, but after the current deadline to try to get the tau result ready for the report I can investigate them further if you want.

(and thanks for the note on latex! indeed :) )