A new way to identify top quarks Torben Schell Institute for Theoretical Physics, Heidelberg University IMPRS-PTFS May 8, 2014 T. Schell (ITP – U Heidelberg) HEPTopTagger IMPRS-PTFS May 8, 2014 1 / 15 Outline 1 Top quarks? 2 From a hard process to a LHC event and back 3 HEPTopTagger T. Schell (ITP – U Heidelberg) HEPTopTagger IMPRS-PTFS May 8, 2014 2 / 15 Top quarks? Standard Model of particle physics [http://en.wikipedia.org/wiki/File:Standard Model of Elementary Particles.svg] T. Schell (ITP – U Heidelberg) HEPTopTagger IMPRS-PTFS May 8, 2014 3 / 15 Top quarks? Top history and basics 1995 discovery at the Tevatron 5 2 1 0 -1 160 170 180 190 2 Top Mass (GeV/c ) 4 2 Events/(10 GeV/c ) 1973 postulated by Kobayashi and Maskawa to allow for CP violation in the Standard Model ∆ln(likelihood) 6 3 2 1 mass mt ≈ 173 GeV 0 80 100 120 140 160 180 200 220 240 2 260 280 Reconstructed Mass (GeV/c ) [CDF, Phys. Rev. Lett. 74, 2626 (1995)] Why are we interested in top quarks? decay before hadronization weak scale mass → largest coupling to the Higgs boson → perfect laboratory to study electroweak symmetry breaking mediate Higgs production and decay from/to massless particles physics beyond the Standard Model T. Schell (ITP – U Heidelberg) HEPTopTagger IMPRS-PTFS May 8, 2014 4 / 15 From a hard process to a LHC event and back Production and decay on parton level ... top quark pair production at leading order top quark decay b t q W q¯ T. Schell (ITP – U Heidelberg) HEPTopTagger IMPRS-PTFS May 8, 2014 5 / 15 From a hard process to a LHC event and back ... and an event simulation for the LHC parton density functions parton shower hard final and initial state radiation underlying event hadronization (pile–up) [SHERPA, arXiv:0811.4622] T. Schell (ITP – U Heidelberg) HEPTopTagger IMPRS-PTFS May 8, 2014 6 / 15 From a hard process to a LHC event and back Jet clustering Reconstruction of parton level gluons and quarks from calorimeter data 1 Find the minimal distance of all objects dmin = min(dij , diB ). 2 If dmin ∈ {dij }, join the two corresponding objects. If dmin ∈ {diB }, remove object i → jet. 3 Iterate until no objects are left. kT : dij = min(pT ,i , pT ,j ) diB = pT ,i C/A: dij = ∆Rij R ∆Rij R diB = 1 anti-kT : ∆R dij = min( pT1,i , pT1,j ) R ij diB = 1 pT ,i [Cacciari, Salam, Soyez, arXiv:0802.1189] T. Schell (ITP – U Heidelberg) HEPTopTagger IMPRS-PTFS May 8, 2014 7 / 15 From a hard process to a LHC event and back Jet filtering Remove impurities from underlying event by reclustering the jet with an optimized cone size → reduced area. start from the calorimeter data that ended up in jet recluster with a reduced cone size Rfilt keep only the Nfilt hardest objects recluster to one object → filtered jet T. Schell (ITP – U Heidelberg) HEPTopTagger IMPRS-PTFS May 8, 2014 8 / 15 actual analysis, the numbers do not yet includ Higgs [30]. HEPTopTagger How to detect top quarks? The algorithm proceeds in the following ste ΔRbjj 1. define a fat jet using the C/A algorithm problem: tops decay products will decay into all directions → can not be distinguished from background 3 103 2 solution: boosted top quarks → fat jets 102 1 10 use moderately boosted tops 0 0 200 400 600 PT[GeV] 1 [Plehn et al. arXiv:1006.2833] HEPTopTagger T. Schell (ITP – U Heidelberg) Figure 2: Left: partonic Rbjj vs pT distributio only for tagged top quarks and based on the reco HEPTopTagger IMPRS-PTFS May 8, 2014 9 / 15 HEPTopTagger HEPTopTagger – Steps I [arXiv:1006.2833] b 1 t construction of fat jets: C/A algorithm with R = 1.5 require pT > 200 GeV q W q¯ 2 search for hard substructures: undo last clustering step: j → j1 j2 mass drop criterion: neglect j2 if mj1 > 0.8mj iterate until mi < msub = 30 GeV → hard substructures T. Schell (ITP – U Heidelberg) HEPTopTagger IMPRS-PTFS May 8, 2014 10 / 15 HEPTopTagger HEPTopTagger – Steps II 3 filtering: filter a triple of hard substructures to reduce contamination from underlying event → 3 jets (j1 , j2 , j3 ). 4 mass range cut: reject the top candidate if its mass is not inside a mass window around mt : 150 GeV < m123 < 200 GeV T. Schell (ITP – U Heidelberg) HEPTopTagger IMPRS-PTFS May 8, 2014 11 / 15 HEPTopTagger HEPTopTagger – Steps III 5 mass plane cuts: ask for 0.85 mmWt < mij m123 < 1.15 mmWt [Plehn et al. arXiv:1006.2833] additional cuts to reduce background: 13 if m23 ≈ mW 0.2 < arctan m m12 < 1.3; else 6 (tag) pT -cut: Finally, require pT T. Schell (ITP – U Heidelberg) m23 m123 > 0.35 > 200 GeV HEPTopTagger IMPRS-PTFS May 8, 2014 12 / 15 HEPTopTagger It is actually used close collaboration with ATLAS group of Prof. Sch¨oning 350 [ATLAS, CERN-PH-EP-2012-291] Data 2011 Z' (1 TeV) σ = 1.3 pb 300 tt 250 Multijet 200 ATLAS 150 ∫ L dt = 4.7 fb σ × BR(Z'→ tt) [pb] Events / 100 GeV used in ATLAS analyses Obs. 95% CL upper limit 102 Exp. 95% CL upper limit Exp. 1 σ uncertainty Exp. 2 σ uncertainty 10 Leptophobic Z' (LOx1.3) ATLAS HEPTopTagger -1 1 s = 7 TeV 100 HEPTopTagger 10-1 50 0 500 s = 7 TeV ∫ L dt = 4.7 fb -1 1000 1500 2000 2500 3000 0.6 0.8 1 1.2 tt Mass [GeV] [ATLAS, CERN-PH-EP-2012-291] (a) T. Schell (ITP – U Heidelberg) tt Multijet HEPTopTagger 1.8 2 2 Obs. 95% CL upper limit Exp. 95% CL upper limit Exp. 1 σ uncertainty KK BR(g → tt) [pb] vents / 100 GeV 25 1.6 (a) 35 10 searches for flavor violation in the top-Higgs sector Data 2011 30 g (1.6 TeV) σ = 0.35 pb [Greljo, Kamenik, Kopp, arXiv:1404.1278] KK 1.4 Z' Boson Mass [TeV] 10 Exp. 2 σ uncertainty KK gluon (LO) IMPRS-PTFS May 8, 2014 ATLAS 13 / 15 HEPTopTagger Recent developments extensions and improvements (cut order, distance measure, angular correlations, N–Subjettiness, low transverse momenta, . . . ) [Plehn et al. arXiv:1111.5034 & arXiv:1312.1504] reconstruction of heavy resonances [in preparation] dσ fb GeV dmrec t semilept. tt s = 8 TeV 150 dσ fb GeV dmrec t lept. W+jets s = 8 TeV 1000 1 - εB next step: full–hadronic decay of t ¯t H 1 0.98 old default p T,fat > 300 GeV 0.96 100 old default new default djsum new default djsum 500 50 0.94 pT,fat = [200,250] GeV 0.92 0.9 p T,fat = [250,300] GeV 0.88 0.86 0 0 100 200 300 [GeV] mrec t T. Schell (ITP – U Heidelberg) 0 0 100 200 300 [GeV] mrec t HEPTopTagger 0 R = 1.8, s = 13 TeV 0.2 0.4 0.6 0.8 1 εS IMPRS-PTFS May 8, 2014 14 / 15 Summary Summary there are many reasons to study top quarks the HEPTopTagger allows to reconstruct hadronically decaying top quarks in a moderately boosted regime based on jet substructure close collaboration with experimentalists which use the HEPTopTagger in ATLAS analyses T. Schell (ITP – U Heidelberg) HEPTopTagger IMPRS-PTFS May 8, 2014 15 / 15