1 / 28

L’étiquetage des jets b dans ATLAS

Understand the crucial role of b-tagging in signal extraction from backgrounds in various LHC channels like top physics, Higgs physics, SUSY, and Exotics. Explore requirements and techniques like spatial tagging, soft-lepton tagging, likelihood ratios, and secondary vertex tagging. Delve into main taggers, performances, impact of material on b-tagging, and track classifications for improved analysis at ATLAS. Sessions from Atelier Physique Atlas France in Seignosse 2007 presented by Laurent Vacavant, CPPM Marseille.

yongr
Download Presentation

L’étiquetage des jets b dans ATLAS

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. L’étiquetage des jets bdans ATLAS Introduction Performances Préparation aux premières données Laurent Vacavant / CPPM Marseille, Atelier Physique Atlas France 2007 – Seignosse - 11/09/2007

  2. Motivation b-tagging crucial to extract signal from backgrounds in a large variety of channels of interest at LHC: • top physics (pair or single top) • Higgs physics (Hbb) for discovery and/or property measurements • SUSY (hbb, bbH/A) • Exotics (little Higgs, hidden valley models,….) • unexpected… Requirements vary from channel to channel: • ttbar: emphasis on background rejection for clean measurements, efficiency not a problem (prochaines présentations) • ttH(bb): large background as well, requires 4 btags, high efficiency given low signal Xsection (session Higgs demain) Atelier Physique Atlas France 2007 - Seignosse

  3. Soft lepton Jet axis a0>0 a0<0 B Primary vertex Secondary Vertex y x B-tagging basics Lifetime of B hadrons: c ~ 470 m (mixture B+/B0/Bs), ~ 390 m (b) for E(B) ~ 50 GeV, flight length ~ 5 mm, d0 ~ 500 m •  Spatial tagging: • Signed impact parameter • of tracks (or significance) • Secondary vertex • (see note ATL-PHYS-2004-06: • S.Corréard, V.Kostioukhine, • J. Lévêque, A. Rozanov, • J.B. de Vivie) •  Soft-lepton tagging: • Low pT electron from B (D) • Low pT muon from B (D) Atelier Physique Atlas France 2007 - Seignosse

  4. 1-2 GeV/c 2-5 GeV/c 5-10 GeV/c >50 GeV/c Key ingredients CSC-01-00-00 • tracking • newTracking/ iPatRec algo. • impact parameter resolution • σ(d0) ~ 10 (+) 150/pT√sinθm • tracking in (dense) jets • specific quality requirements • esp. N(B-layer) > 0 • ~ - 3% absolute eff. • partionning track coll. • primary vertex • mostly needed along Z • @ high-luminosity • lepton ID (soft leptons) • jets (axis) DC3-02 Atelier Physique Atlas France 2007 - Seignosse

  5. Likelihood Ratio & IP2D Tagger • Smoothed & normalized distributions of significances for the two • b | u hypotheses: b(S), u(S) • For each track: ratio b(S)/u(S) • Advantages: • more powerful • allows to combine non-correlated • variables just by adding the weights • Drawbacks: • requires p.d.f. for b and u • hypotheses: may be difficult to • measure in data Atelier Physique Atlas France 2007 - Seignosse

  6. Secondary Vertex Search • Track selection • Search for all 2-track vertices in the jet (VKalVrt) • Remove V0s, identified interaction with material,… • Common (inclusive) vertex for remaining tracks (NB: reconstruction of all vertices also being studied) beam-pipe, pixel layers  conv. Ks Λ Atelier Physique Atlas France 2007 - Seignosse

  7. Secondary Vertex (SV) Tagger • Variables used: low correlation with IP (Lxy not used) • Use multiD distributions  correlations light jets light jets SVX mass b-jets b-jets E(SVX)/E(jet) M=SVX mass F=E(SVX)/E(jet) N=# 2-track v. Combined with IP: Atelier Physique Atlas France 2007 - Seignosse

  8. List of Main Taggers Atelier Physique Atlas France 2007 - Seignosse

  9. Performances

  10. G.Aad, JB de Vivie Generic performances with release 12 • CSC samples • CSC-01-02-00 geo. • NewTracking • old (CSC11) calib. • noise, 3% pixel ineff. • no pile-up • cone 0.4 jets • standard purification (b) = 60% • careful w/ (pT,η) shape (right plot) • ttbar events: R ~ 200 for (b) = 60% R ~ 800 for (b) = 50% •  ttbar events w/o purification: • R ~ 150 for (b) = 60% • R ~ 400 for (b) = 50% Atelier Physique Atlas France 2007 - Seignosse

  11. JB de Vivie, R. Zaidan, LV Impact of material with CSC geometries Material increase: d0 resolution: CSC-01-00-00 DC3-02 1-2 GeV/c 2-5 GeV/c 5-10 GeV/c >50 GeV/c • More realistic description of material • Biggest change is from DC3-02 to CSC-01-00-00 • Additional degradation with CSC-01-02-00 (and a bit more to come) G. Gorfine Impact on b-tagging:rejection /~1.5 for fixed ε(b), for R(u)=100 <~5% absolute on ε(b) • Understanding: • rescaling in DC3 of IP • resolution to match CSC • removal of G4 secondaries • Explain 60-70% of losses (more in central region, ~100%) • Need to understand the missing part & address recoverability secondaries ! resolution ! Atelier Physique Atlas France 2007 - Seignosse

  12. LV Track classification: example of shared hits Use dedicated calibration for several track classes e.g. tracks with shared hits (≥1 Pix || ≥2 Sct) (NB: default treatment since release 11) IP3D, 50% sh. rejected sh. calibrated  Good understanding/monitoring of tracking (beyond resol.) is important !  More track categories are being studied (0B-layer,fragmentation, pT, η) Atelier Physique Atlas France 2007 - Seignosse

  13. JetFitter: a whole new algorithm • Main idea:try to constrain all tracks from the • B/D hadron decays to intersect the same flight • axis (instead of the common vertex approach) • Technique: a new dedicated Kalman filter fitting: • Treatment of incomplete/1-prong topologies: • on average 1.9 (1.7) OK track from D (B) decays • recover « 1 track from B + 1 track from D  » (9%) • can improve purity • Based on likelihood: • categories for topology (0,1,2+ vertices) • vertex mass, energy fraction, significances • of {disti} • can also use charm p.d.f for improving • b/c separation •  Available in release 13 Atelier Physique Atlas France 2007 - Seignosse

  14. Towards first data

  15. JB. de Vivie, LV JetProb Tagger • Tagger à la Aleph/Lep: • Compatibility of the tracks with the primary vertex • Fit negative side of Signed impact parameter w/ resolution func (gaus+2exp) • But easier to understand/calibrate in early data  Performances poorer than IP2D: x0.5 (but less optimized)  Should give results early (performances to be quantified) Atelier Physique Atlas France 2007 - Seignosse

  16. H.Bachacou Soft Muons • Br(bX)+Br(bcX) = 11% + 10% • Independent of vertex tagging • Muon reconstruction: currently Staco+Mutag (high-efficiency down to 4 GeV/c pT, low fakes) • 2 steps: • association muon-jet (cone ∆R < 0.5) • Likelihood ratios (1D, 2D) for b/light hypothesis using pT, pTrel. • Rejection of 200 for 10% b-tagging efficiency (incl. Br. & -ID) • Fake muons vs pile-up/cavern bckgd being checked Atelier Physique Atlas France 2007 - Seignosse

  17. @60% e(b) DC3 R(u) = 130 CSC R(u) = 133 R(u) @50% e(b) DC3 R(u) = 165 CSC R(u) = 190 Rejection of u-jets (60% a.e.  ~ 8%) DC1- no pileup R(u) = 151 (166) ± 11 (7.8.0) DC1- low lumi R(u) = 136 (144) ± 9 (7.8.0) DC1-high lumi R(u) = 104 (123) ± 11 (7.8.0) Rome R(u) = 145 ± 6 (11.0.3) DC3 R(u) = 130 (12.0.6) CSC R(u) = 133 (12.0.6) Rejection of c-jets DC1-no pileup R(c) = 35 (35) ± 2 (7.8.0) DC1-low lumi R(c) = 36 (35) ± 2 (7.8.0) DC1-high lumi R(c) = 33 (33) ± 2 (7.8.0) e(alg-b) – efficiency of the b-tagging algorithm F. Derue Soft Electrons • Soft Electron reconstruction: • starts from ID track • associate cluster • Soft Electron ID: likelihood • tracks (TRT, …) • shower shape NB: derniers résultats avec géométrie CSC-01-02: R(u)~80 Atelier Physique Atlas France 2007 - Seignosse

  18. Muon jet   Muon jet SVT tag jet L.Feligioni Measuring b-tagging efficiency with di-jets • pTrel templates • Non-linear system (System8 à la D0): • 2 samples • Muon Jets (n) • MJ + other tag opposite jet Jet (p) • 2 different b/l fractions • 2 non-correlated taggers • Tracks (impact param./SVT) • Soft Muon (pTrel) • system can be solved analytically • Results: (for ~10k MJ)  definition of Muon+Jet trigger in progress Atelier Physique Atlas France 2007 - Seignosse

  19. W jets: ET>40 GeV, 20 b-veto (weight < 3) Hadronic side b-jet: ET>40 GeV b-tag (weight>3) Lepton: ET>20 GeV Leptonic side b-jet: ET>20 GeV No tag requirement ETmiss> 20 GeV Selecting a pure b-jet sample in ttbar events Leptonic top mass:  bkgd shape from mjjb sideband • b-tag on hadronic side • predict (kinematics) b-jet • on leptonic side (unbiased) 100 pb-1 Atelier Physique Atlas France 2007 - Seignosse

  20. H. Bachacou Counting method for measuring b-tag efficiency in ttbar events Atelier Physique Atlas France 2007 - Seignosse

  21. Activities and CSC notes for b-tagging CPPM, LAL Algorithms and generic performances: • BT0: performances and parametrization [Ed: L. V.] • BT1: vertexing (primary, secondary, beam-lines) [Ed: V. Kostioukhin] • BT2: alignment and b-tagging [Ed: G. Gorfine] • BT3: soft muons [Ed: H. Bachacou] • BT4: soft electrons [Eds: F. Derue, A. Kaczmarska] • BT5: HLT b-tagging [Ed: C. Schiavi] Specific performance studies: • BT6: SUSY [Ed: J.B. de Vivie] • BT7: ttH [Ed: A. Rozanov] • BT9: bbH/A->tautau [Ed: E. Staneckla] Calibration & performance estimation from data: • BT10: with di-jet events [Ed: G. Watts, F. Rizatdinova] • BT8: with ttbar [Eds: H. Bachacou, R. Hawkings] • BT11: 900 GeV run [Ed: F. Rizatdinova] (now merged with BT10) Saclay LPNHE LAL CPPM CPPM  Saclay  Atelier Physique Atlas France 2007 - Seignosse

  22. Conclusion • High-performances b-taggers have been developped • R>100 for (b)=60% achieved, not so far from (b)=70% • New developments (JetFitter, optimization) allow to maintain performances despite increase of material • Still many developments/studies going on  CSC notes • Now shifting priorities towards first data • Impact of alignment a.s.a.p. • Commissioning b-tagging with simple/robust taggers • Strategies for cross-calibrating taggers and measuring performances/p.d.f. in data • Several methods for extracting b-efficiency: <10 % relative error achievable with 100 pb-1 • Work starting for estimating mistag rates Atelier Physique Atlas France 2007 - Seignosse

  23. Back-up slides

  24. More information • Talks from Marseille workshop (last May): • recent results covering the whole spectrum (~40 talks) • In CDS agenda: Physics/B-tagging/workshop • http://indico.cern.ch/conferenceDisplay.py?confId=14475 • HyperNews forum • WiKi page: https://uimon.cern.ch/twiki/bin/view/Atlas/FlavourTagging • And soon (hopefuly) the CSC notes !!! Atelier Physique Atlas France 2007 - Seignosse

  25. The case of high-pT jets • Performances decrease for moderate/high jet-pT (150-250 GeV): • pattern-recognition becomes more difficult in dense jets • b-tagging track selection even more critical  Optim. of cone size for track-jet assoc.: Fragmentation tracks selected: 15 < pT < 30 GeV/c Light jet rejection improvment w.r.t nominal +:for pT  250 GeV/c ~ 6% of b-jets have only tracks from fragmentation !! 200 < pT < 400 GeV/c • 20% (8%) of B in jets w/ pT>200 GeV decay beyond the beam-pipe (b-layer) • track selection (|d0|<1mm, NBLayer>1) too tight for high boosts • For very high pT (TeV) (e.g. little Higgs, Z’) even worse • requires dedicated tracking, tagging algorithms • but also simulation: propagation of B hadrons in matter in G4 ! Atelier Physique Atlas France 2007 - Seignosse

  26. Impact of (mis)alignment Random misalignments: IP3D tagger , ttH events, realistic conditions improving alignment up to almost perfect in one year, recovery of light jet rejection • will redo the exercise with misaligned detectors from simulation when available (more realistic than random misalignments) Atelier Physique Atlas France 2007 - Seignosse

  27. System8 on di-jet events Measurements system solvable analytically for , r and sample composition ! k: correlation between soft muon & track-based taggers ,: sample dependency Atelier Physique Atlas France 2007 - Seignosse

  28. L2 & HLT b-tagging • Aims:increase acceptance for multi- b-jets events, lower threshold X when used in combination with X, extract b-jet calibration samples • Inputs: L1 jet RoI (pT>15 GeV, ∆x∆=0.4x0.4), tracks from L2 (SiTrack/IDScan) • PV: in Z using histogramming • Taggers: impact parameter significance (LHR), secondary vertex in progress • Performances: • (NB: also an alternative à la CDF/SVT in the works: FTK) • Working on menus:e.g. 4j65/2b + 6j40/3b doubles efficiency for ttH(bb) w.r.t. 4j100+3j165, with 1.3kHz @ L1, 4 Hz @ EF (2.1033) • On-going improvements: • signed IP • speed of SV • PV search w/ multi RoI Atelier Physique Atlas France 2007 - Seignosse

More Related