Reconstruction and Identification of Boosted Tau Pair Topologies at ATLAS. David Kirchmeier TU Dresden

Transcription

1 Reconstruction and Identification of Boosted au Pair opologies at ALAS David Kirchmeier U Dresden

2 Motivation τ had τ had h A Search for BSM particles, e.g. heavy Higgs Bosons With higher energies in Run II: Higher masses reachable Highly boosted tau topologies more likely Highly boosted taus tend to end up in same jet Z l l DPG agung //5 David Kirchmeier (U Dresden)

3 au Lepton Decay Hadronic mode with 65% BR more collimated than QCD Jets David Kirchmeier (U Dresden) DPG agung //5

4 au Reconstruction and Identification Cells Clusters Anti-Kt-4-Jets au Candidates aus David Kirchmeier (U Dresden) DPG agung //5 4

5 au Reconstruction Results of current tau reconstruction (w/o identification step) Signal: simulated with Pythia, A Z h, gga, m A = ev Di-au Reconstruction Efficiency Di-au Reconstruction Efficiency ALAS work in progress Di-au p [GeV] Di-au p [GeV] David Kirchmeier (U Dresden) DPG agung //5 5

6 au Reconstruction R.4 R <.4 Di-au Reconstruction Efficiency reconstruction efficiency ττ R.4 ALAS work in progress R Reconstruction of just one anti-kt-4 jet highly boosted taus merge to one tau candidate Additional reconstruction method has to be developed David Kirchmeier (U Dresden) DPG agung //5 6

7 Di-au Reconstruction and Identification Cells Clusters Anti-Kt-4-Jets au Candidates aus Cells Clusters Anti-Kt--Jets Di-au Candidates Di-aus David Kirchmeier (U Dresden) DPG agung //5 7

8 Di-au Reconstruction - truth tau cells ditau candidate subjet Di-au ALAS work in progress - Z e e QCD Jet truth tau cells ditau candidate subjet R<.: inner tau cone.< R<.: outer tau cone.< R: isolation region ALAS work in progress anti-kt--jet David Kirchmeier (U Dresden) DPG agung //5 8

9 Di-au Reconstruction Signal Efficiency (Reconstruction) ALAS work in progress. di-tau reco, truth matched single tau reco, truth matched Signal Efficiency (Reconstruction) ALAS work in progress di-tau reco, truth matched and R <. single tau reco, truth matched R ditau p [GeV] Now di-taus with R <.4 and p > 5 GeV can be reconstructed Reconstruction efficiencies of up to 9% David Kirchmeier (U Dresden) DPG agung //5 9

10 Background Rejection Di-tau Signal.5.5 truth tau cells ditau candidate subjet ALAS work in progress ALAS work in progress truth tau cells ditau candidate subjet QCD Background normalized ALAS work in progress Di-au Signal High p Jet Data f_corelead f core = P R<. cells P R<. cells p cell p cell David Kirchmeier (U Dresden) DPG agung //5

11 Di-au Identification Signal: Pythia, A Z h ll + ττ, m A = GeV Background: high-p -Di-Jet data rained Boosted Decision rees with 9 variable Di-au Signal High-p -Di-Jet Background ALAS work in progress inverse Background Efficiency /eff 5 4 ALAS work in progress BD score Signal Efficiency David Kirchmeier (U Dresden) DPG agung //5

12 Conclusion and Outlook New di-tau reconstruction and identification Reconstruction efficiencies of up to 9% High background rejection Application in A Z h ττ + ll and H h h ττ + bb DPG agung //5 David Kirchmeier (U Dresden)

13 HANK YOU FOR YOUR AENION

14 BACKUP

15 Di-au Reconstruction Signal Efficiency (ID) working point: eff sig=.9 working point: eff sig=.8 working point: eff sig=.7 ALAS work in progress di-tau p [GeV] Background Efficiency (ID) ALAS work in progress working point: eff sig=.9 working point: eff sig=.8 working point: eff sig= di-tau p [GeV] David Kirchmeier (U Dresden) DPG agung //5 5

16 Di-au Reconstruction + ID Signal Efficiency (Reconstruction + ID) ALAS work in progress working point: eff sig=.9 working point: eff sig=.8 working point: eff sig= di-tau p [GeV] David Kirchmeier (U Dresden) DPG agung //5 6

17 Di-au Reconstruction Cuts anti-kt- jet n(subjets) n(subjet_tracks) 4 p (subjet) > 5 GeV p (track) > GeV n(pixel hits) n(pixel hits) + n(sc hits) 7 d.mm z sinθ.5mm DPG agung //5 David Kirchmeier (U Dresden) 7

18 Software Implementation r9 AHENA package with xaod input and Nuple output FastJet for jet reconstruction Python.7 classes to wrap reconstruction output into BD training and testing trees MVA for multivariate separation methods ROO 5.4 DPG agung //5 8 David Kirchmeier (U Dresden)

19 DPG agung //5 9 David Kirchmeier (U Dresden)

20 Di-au Identification normalized f lead/sublead core = Di-au Signal High p Jet Data f_corelead P R<. cells P R<. cells normalized p cell p cell Di-au Signal High p Jet Data f_coresub DPG agung //5 David Kirchmeier (U Dresden)

21 Di-au Identification fsubjet lead = pleading subjet p jet f sublead subjet = psubleading subjet p jet normalized.5 Di-au Signal High p Jet Data. normalized.5 Di-au Signal High p Jet Data f_leadpt_jetpt f_subleadpt_jetpt DPG agung //5 David Kirchmeier (U Dresden)

22 Di-au Identification normalized f lead track = p p leadtrk leading subjet.8 Di-au Signal.6 High p Jet Data f_tracklead f sublead track = normalized p p leadtrk subleading subjet Di-au Signal High p Jet Data f_tracksub DPG agung //5 David Kirchmeier (U Dresden)

23 Di-au Identification R max : maximal distance of subjet track to subjet axis normalized.5 Di-au Signal. High p Jet Data DPG agung // David Kirchmeier (U Dresden) R_max

24 Di-au Identification n lead trks Number of tracks in outer cone of leading subjet n sublead trks Number of tracks in outer cone of subleading subjet normalized Di-au Signal High p Jet Data normalized Di-au Signal High p Jet Data n_tracks_in_outer_lead n_tracks_in_outer_sub DPG agung //5 David Kirchmeier (U Dresden) 4