22.3.8 / 6 Reconstruction and Identification of Semi-Leptonic Di-Tau Decays in Boosted Topologies at ATLAS
22.3.8 2 / 6 Tau Lepton Properties 8% 7% other 6% 4% 35% leptonic % 9% 65% hadronic 9% 25%
22.3.8 3 / 6 Motivation - analysis G hh bbττ - high mass resonance decaying into pair of Higgs bosons production of Higgs bosons with high momentum two tau leptons with low spatial separation ( Boosted Topology ) one decays hadronically other decays leptonically Semi-Leptonic - algorithms for full hadronic boosted di-taus already exist
22.3.8 4 / 6 Motivation m G = GeV m G =3 GeV =5 GeV m G =5 GeV m G Events /.3.3 5 - Monte Carlo samples for G hh ττττ process with different graviton masses.5..5 How do existing tau-, electron and muon algorithms perform for low ΔR?..3.4.5.6.7.8.9 R (τ had-vis, e/µ)
Reconstruction of Hadronic Tau Decays, Electrons and Muons at ATLAS electron muon tau - energy in sliding window exceed threshold - inner detector track matches to calorimeter energy deposition - inner detector track - muon spectrometer track - anti-kt(r=.4) jets build from calorimeter energy deposition - one or three classified tracks in jet region http://atlas.physicsmasterclasses.org/en/wpath_teilchenid3.htm 22.3.8 5 / 6 M. Cacciari, G. P. Salam, G. Soyez, The Anti-k(t) jet clustering algorithm JHEP 84 (28) 63
22.3.8 6 / 6 Had-Electron Channel
22.3.8 7 / 6 Standard Tau Reconstruction/Identification Standard Electron Reconstruction/Identification Reconstruction VeryLoose Loose Medium Tight Reconstruction Loose Tight VeryLoose Medium Reco./Id. Efficiency.8.6 p (τhad-vis ) > 2 GeV T Reco./Id. Efficiency.8.6 p (e ) > GeV T η(e ) < 2.5 had-vis τ.4 e.4..3.4.5.6 R (τ had-vis,e )..3.4.5.6 R (e,τhad-vis )
22.3.8 8 / 6 Main Problem for Existing Algorithm anti-kt(r=.4)-jet - tau reconstruction uses anti-kt(r=.4) seed jet energy depositions merge into one seed jet for ΔR smaller than.4 axis of reconstructed hadronic tau shifted
22.3.8 9 / 6 New Reconstruction for the HadEl Channel anti-kt(r=.)-jet - approach based on Di-Tau reconstruction for HadHad channel tau anti-kt(2)-subjet electron reco electron. find seed: large anti-kt(r=.)- jet with > 3 GeV 2. groom small anti-kt(r=) subjets from clusters in seed jet 3. search for reconstructed electrons in region 4. create HadEl candidate for each reconstructed electron-subjet combination with ΔR >.
22.3.8 / 6 Performance of New HadEl-Reconstruction Efficiency.8.6.4 p (τhad-vis T p (e T ) > 2 GeV ) > 2 GeV - reconstruction efficiency for true HadEl di-taus - some cuts applied on all HadEl di-taus beforehand - good efficiency p (diτvis ) > 3 GeV T ~9% for η(τ ) < 2.5 had-vis <ΔR<.4 η(e ) < 2.5 - drops at because of size of..3.4.5.6.7.8.9 anti-kt(r=)-jets R(τ had-vis,e )
22.3.8 / 6 Identification of HadEl Di-Taus VeryLoose (.95) Loose (.8) VeryLoose (.95) Loose (.8) Identification Efficiency.2.8.6.4 Medium (.65) Tight (.5) Background Rejection 5 4 3 2 Medium (.65) Tight (.5)..3.4.5.6.7.8.9 R (e, τ Reco ) Reco - tight/medium/loose/veryloose working points tuned to 5%/65%/ 8%/95% efficiency..3.4.5.6.7.8.9 R (e, τ Reco ) Reco - high background rejection for ΔR >
22.3.8 2 / 6 Had-Muon Channel
22.3.8 3 / 6 Standard Tau Reconstruction/Identification Standard Muon Reconstruction/Identification Reconstruction VeryLoose Loose Medium Tight Reconstruction Loose Tight VeryLoose Medium Reco./Id. Efficiency.8.6 p (τhad-vis ) > 2 GeV T Reco./Id. Efficiency.8.6 p (µ T. < η(µ ) > GeV ) < 2.5 had-vis τ.4 µ.4..3.4.5.6 R (τ had-vis,µ ) - hadronic tau and muon reconstruction independent - build candidates by matching taus and muons with ΔR<. - new boosted decision tree uses: - muon identification - tau identification variables corrected for muon tracks..3.4.5.6 R (µ,τhad-vis )
22.3.8 4 / 6 Identification of HadMu Di-Taus VeryLoose (.95) Loose (.8) VeryLoose (.95) Loose (.8) Identification Efficiency.2.8.6.4 Medium (.65) Tight (.5) Background Rejection 5 4 3 2 Medium (.65) Tight (.5)..3.4.5.6.7.8.9 R (µ, τ Reco ) Reco - tight/medium/loose/veryloose working points tuned to 5%/65%/ 8%/95% efficiency..3.4.5.6.7.8.9 R (µ, τ Reco ) Reco - high background rejection for ΔR >
22.3.8 5 / 6 Summary and Outlook - new reconstruction and identification algorithms for boosted di-taus in the HadEl and HadMu decay channel - good performance where existing algorithms cease to work - tools already implemented in official ATLAS Athena software framework - ready for evaluation in analysis: - G HH bbττ - H aa ττττ (ma ~ GeV) - ongoing studies on systematic uncertainties (talk by Fabian Petsch)
22.3.8 6 / 6 Backup
22.3.8 7 / 6 Identification of HadEl Di-Taus Signal: matched HadEl di-taus from MC sample with graviton masses between TeV and 5 TeV Background: HadEl di-tau candidates from data5 with fat jet triggers. Reweight background to achieve same 2D tau pt vs electron pt distribution as in signal sample 2. Train boosted decision tree (BDT) 3. Tune WP to reach flat signal efficiency in ΔR vs ditau pt (different BDTScore cut for each 2D (ΔR,ditau pt) bin) three sets of ID variables: variables using information from electron variables using information from tau subjet variables using information of whole anti-kt() area
22.3.8 8 / 6 Example of HadEl Di-Tau Identification Variables signal background signal background Events /.5.35.3 5 Events.45.4.35.3 5.5..5.5..5..3.4.5.6.7.8.9 hadel-seed E frac 5 5 2 25 3 35 4 n tracks (diτ)
22.3.8 9 / 6 Example of HadEl Di-Tau Identification Variables signal background signal background Events /.5.35.3 5 Events /..4.35.3 5.5..5..5.5..3.4.5.6.7.8.9 f core (τ).2.4.6.8..2.4.6.8 R max (τ)
22.3.8 2 / 6 New Reconstruction for the HadMu Channel - use standard muon reconstruction and standard tau reconstruction - use muon as seed - create new Di-Tau- HadMu for every reconstructed tau in ΔR<. vicinity
Performance of New HadMu-Reconstruction Efficiency.8.6.4 p (τhad-vis ) > 2 GeV T p (µ ) > GeV T η(τ had-vis η(µ ) < 2.5 ) < 2.5 - reconstruction efficiency > ~9% - slight decrease for very low ΔR due to decreasing muon reconstruction efficiency..3.4.5.6.7.8.9 R(τ had-vis,µ ) 22.3.8 2 / 6
22.3.8 22 / 6 Example of HadMu Di-Tau Identification Variables signal background signal background Events Events.8.8.6.6.4.4 tight medium loose reco very loose.4.6.8.2.4.6.8 2 MuonSelection MuonIsIsolated
22.3.8 23 / 6 Example of HadMu Di-Tau Identification Variables signal background signal background Events /.3 5 Events /.2.5.4.5.3..5...3.4.5.6.7.8.9.5..5 5.3.35.4 f cent R Max