Title: Offline tau reconstruction overview
1Offline tau reconstruction overview
M. Pilar Casado (IFAE UAB)
- Introduction
- tauRec
- tau1P3P
- AOD and ESD
- Conclusion
2t Decays
35, but they belong to another story
- t decay modes
- Leptonic decay modes
- t ?nt ne e (17.4)
- t ? nt nm m (17.8)
- Hadronic decay modes
- 1 prong
- t ? nt p
(11.0) - t ? nt p p0
(25.4) - t ? nt p p0 p0
(10.8) - t ? nt p p0 p0 p0
(1.4) - t ? nt K np0
(1.6) - 3 prong
- t ? nt 3 p np0
(15.2)
How to identify them?
77
23
3Introduction
4Practical info
- tauRec
- Authors D. Cavalli, S. Resconi, F. Paige, S.
Rajagopalan, M. Heldmann. - Documentation
- Code http//atlassw1.phy.bnl.gov/lxr/source/atlas
/ - Reconstruction/tauRec/
- Wiki page https//uimon.cern.ch/twiki/bin/view/At
las/ - TopologicalClustering?topicTauRec
- Status in the offline releases Part of the
offline reconstruction. - tau1P3P
- Authors E. Richter-Was, T. Szymocha, L. Janyst.
- Documentation
- Code http//atlas-sw.cern.ch/cgi-bin/viewcvs-atla
s.cgi/users/erichter/ - Web page http//project-tau1p3p.web.cern.ch/proje
ct-tau1p3p/ESDtau1P3P/index.html - Status in the offline releases Still not
included.
5tauRec
- Reconstruct t-candidate
- Start from different objects CaloClusters
(default), Isolated tracks, - Associate tracks from the TrackParticle container
to the candidate - Build the set of variables for t-identification,
and then calculates Likelihood(s) - Calibrate candidates
- Apply set of the basic cuts for t-identification
6Some tau variables
- Some variables calculated in offline (and in the
trigger!)
7Variables used in tauRec
- Builds set of variables for t-identification
they are pT dependant - Calculate Likelihood from REM, FISO, Ntrack,
Strip Width, Nstrip, Charge, Impact - parameter, ET/pT(1sttrack)
REM
FISO
Charge
Ntrack
Nstrip
Impact parameter
Strip Width
ET/pT(1sttrack)
Signal A?tt?background QCD, 0ltpTlt44 full line and
134ltpTlt 334 dashed line
8 Tau1P3P (I)
- Motivation
- Tau1P3P dedicated to the tau identification in
searches for the light Higgs or soft SUSY t's
with ETvis 20-70 GeV. - Explores exclusive features of the t lepton
- hadronic t IS NOT a jet but 1 track ??0 or 3
tracks ??0 - Decay products well collimated in space, track
direction can provide precise estimate for
direction, if track sufficiently dominant. - energy scale defined from energy-flow and not
calorimetric. - Publications
- 1. E. Richter-Was, H. Przysiezniak and F.
Tarrade, ATL-PHYS-2004-030 - 2. E. Richter-Was, T Szymocha, ATL-PHYS-PUB-2005-0
05. - 3. D. Froidevaux et al., ATL-COM-PHYS-2005-024
- 4. E. Richter-Was, L. Janyst ATL-COM-PHYS-2005-008
9tau1P3P (II)
in real life we will probably start from
regions indicated by calorimetric clusters above
ETcut
identify good quality hadronic track, pT
gt 9 GeV, find nearby good quality tracks
with pT gt 2 GeV in DR lt 0.2
t3P
t1P
2 nearby tracks (h,f) from bary centre of
tracks S charge -1
no nearby-track (h,f) from track at vertex
build ETeflow, use DR lt 0.2
build ETeflow, use DR lt 0.2
build discrimination variables use DR lt 0.2
as a core, 0.2 lt DR lt 0.4 for isolation only
build discrimination variables use DR lt 0.2
as a core, 0.2 lt DR lt 0.4 for isolation only
- t1P - single-prong candidates
- t3P - three-prong candidates
10tau1P3P (III)
with D. Froidevaux, for more details see talk at
Nov. 2004 Physics Week
ETeflow ETemcl ETneuEM S pTtrack ...
used only EM cells within DR lt 0.2 around tau1P
direction
ETeflow/ETtruth
ETeflow/ETtruth
1 prong
3 prong
good resolution ETeflow/ETtruth t1P 65.2
in 0.9-1.1 88.3 in 0.8-1.2 t3P
73.4 in 0.9-1.1 93.8 in
0.8-1.2 good stability for ETtruth 20 70 GeV
ltgt1.011 s 0.0303
ltgt1.014 s 0.0663
ETeflow/ETtruth
ETeflow/ETtruth
1.2
1.2
0.8
0.8
11Comparison of performance
12Present AOD class for taus TauJet.h
- inherits from P4EEtaPhiM class,
- access to asociated tracks,
- access to associated CaloCluster,
- access to associated Vertex
- numTracks (use also for tau1p, tau3P)
- pEM 4-momenta of EM component
- other attributes set by the author
- authortau1P3P already implemented
13Present ESD class for taus TauObject.h
- trying to integrate different basic principles
of both algorithms - access to associated cluster, tracks, vertex can
be easily made - common
- several variables are very specific for tauRec,
- and other for tau1P3P
- possible approaches
- ? create author like for AOD ?
- ? inherit from a basic class, different
concrete classes ? - ? add more variables/methods ?
- Details being discussed at the moment in the
Tau group.
14Conclusion
- There are two algorithms for tau reconstruction
in the offline tauRec (cluster and track seeded
algorithm) and tau1P3P (track seeded algorithm). - There are many common variables EM radius, Strip
width and fracETR12. - Key differences
- H1 calibration vs energy flow.
- Likelihood vs PDE-RS.
- Energy resolution is similar for both packages.
Efficiency is better for tau1P3P for ET lt 25 GeV. - Whilst AOD is quite defined in the offline tau
reconstruction, the ESD class is under
discussion.