Boosted Top Resonance Searches at CMS

Boosted Top Resonance Searches at CMS Justin Pilot, UC Davis on behalf of the CMS Collaboration Northwest Terascale Workshop, Using Jet Substructure University of Oregon 5 April 013

Introduction Many new physics models have large couplings to 3rd generation particles X top quark pairs Search for resonances in the top quark pair invariant mass distribution ) Fraction / (0 GeV/c CMS Simulation, s = 8 TeV 0.5 Z' (1% Width) 0.4 1 TeV 1.5 TeV 0.3 TeV 0. 3 TeV 0.1 0 0 500 00 1500 000 500 3000 3500 4000 4500 5000 tt Invariant Mass (GeV/c ) With the high collision energy, we enter the boosted regime Partially / fully merged top quark decay products Special (substructure) techniques required ) (GeV/c m jet 350 300 50 00 Top Tagging CMS Simulation Z` t t s = 7 TeV PAS JME--013 0.1 0.1 0.08 150 0.06 Recent new results from CMS to discuss today! 0 0.04 50 0.0 0 300 400 500 600 700 800 900 p jet T (GeV/c) 0

Recent CMS Results Two new ttbar resonance search results recently approved! Lepton + Jets (BG-1-006) Combination of standard, boosted techniques Top kinematic reconstruction Boosted top ID using jet mass All-Hadronic (BG-1-005) Top, W-tagging algorithms used Boosted top ID with substructure Data-driven background est. 3

Lepton + Jets Analysis 4

Lepton + Jets Analysis Combination of two techniques Threshold analysis Standard top reconstruction at low pt 4 or more jets, pt > 70, 50, 30, 30 GeV Isolated high pt electron/muon Missing ET > 0 GeV >= 1 b-tagged jet Boosted Analysis Merged top decay products or more jets pt > 150, 50 GeV jets fully merged top 3 jets partially merged top Both analyses use anti-kt R=0.5 jets 5

Lepton + Jets Analysis Combination of two techniques Threshold analysis Standard top reconstruction at low pt 4 or more jets, pt > 70, 50, 30, 30 GeV Isolated high pt electron/muon Missing ET > 0 GeV >= 1 b-tagged jet event fraction 0.1 0.08 0.06 CMS generator study s = 7TeV QCD t t Z, M=1TeV/c Z, M=3TeV/c Boosted Analysis Merged top decay products or more jets pt > 150, 50 GeV No isolation requirement on electron/muon 0 or 1 b-tagged jets HT > 150 GeV Missing ET > 50 GeV 0.04 Leptonic top 0.0 0 0 0.5 1 1.5.5 3 R(b-lep,lep) 6

Event Reconstruction With one lepton and several jets, how to find the right combinations to form top candidates? Choose assignment for each jet Hadronic hemisphere Leptonic hemisphere Nothing Form all combinations, compute χ function Cut to enhance sensitivity 7

- tt Invariant Mass Final discriminating distributions used to set limits 0 and 1 b-tag channels, muon+jets selection 0 b-tag >0 b-tag 8

Lepton + Jets Results Transition between threshold and boosted analysis determined by expected limit values Narrow Z exclusion limit. TeV x B [pb] Upper Limit σ Z' 1-1 -1 CMS, L = 19.6 fb, s = 8 TeV Threshold Boosted Z' with 1.% Decay Width Expected (95% CL) Observed (95% CL) Z' 1.% width Expected ± 1 s.d. Expected ± s.d. - 0.5 1 1.5.5 3 M tt [TeV] 9

Lepton + Jets Results Transition between threshold and boosted analysis determined by expected limit values Wide Z exclusion limit.68 TeV RS gluon exclusion limit.54 TeV -1 CMS, L = 19.6 fb, s = 8 TeV Z' with % Decay Width -1 CMS, L = 19.6 fb, s = 8 TeV KK Gluon x B [pb] Upper Limit σ Z' 1-1 Expected (95% CL) Observed (95% CL) Z'.0% width Expected ± 1 s.d. Expected ± s.d. x B [pb] Upper Limit σ g KK 1-1 Expected (95% CL) Observed (95% CL) KK Gluon Expected ± 1 s.d. Expected ± s.d. - 0.5 1 1.5.5 3 M tt [TeV] - 1 1.5.5 3 M tt [TeV]

All-Hadronic Analysis 11

Analysis Overview Search for all-hadronic top quark pairs in the boosted regime Two possible types of boosted hadronic top decays Type 1 -- decay products fully merged into a single jet Top Tagging algorithm Type -- b-jet escapes from the jet; W decays remain merged Signal region consists only of events in the type 1+1 topology 1+ events used for validation ) (GeV/c m jet 350 300 50 00 150 Top Tagging CMS Simulation Z` t t s = 7 TeV PAS JME--013 0.1 0.1 0.08 0.06 0 0.04 50 0.0 0 300 400 500 600 700 800 900 p jet T (GeV/c) 0 1

Top Tagging Algorithm Based on JHU top tagging algorithm Kaplan, Rehermann, Schwartz, Tweedie, PRL 1/14001 (008) ORIGINAL JET The algorithm uses jets with distance parameter R = 0.8, clustered with Cambridge-Aachen Uses cuts based on jet substructure information Acquired by reversing the jet clustering algorithm Step back in the pairwise sequence to find substructure Can find a maximum of 4 subjets if all decomposition criteria are met Optimized in simulation First reverse iteration Component A Component B Decomposition criteria: R AB > 0.4 p cluster T SUBJET A 0.004 p jet T > 0.05 p jet T SUBJET B?? 1 3 4 13

Event Selection We use Cambridge-Aachen R=0.8 PF jets We divide events into hemispheres based on the position of the leading jet Separation of Δφ >.1 All jets have η <.4 TYPE 1+1 Selection ( jets) Both jets pt > 400 GeV Both required to pass top tag algorithm 3 or 4 subjets Jet mass in [140, 50] GeV window Minimum pairwise subjet mass > 50 GeV 14

Enhancing Sensitivity A requirement on the rapidity difference, Δy, between the two jets imposed to enhance sensitivity High-mass particles generally give lower Δy Choice of cut is optimized using expected limit We use Δy < 1.0 Keep ~50% of signal, remove ~95% of background at high-mass 15

Algorithm Validation Select a muon+jets ttbar sample Isolated muon pt > 40 GeV At least one b-tagged jet found in the event Look for boosted hadronic top decays in the hemisphere opposite the muon Type -- partially merged top decay Type 1 -- fully merged top decay In 8 TeV analysis, enough statistics to start looking at the type 1 sample! 16

Subjet JES Identifying the type candidates in this sample, we compare the W mass for data and MC to measure the JES scale factor Fit to sum of two Gaussian distributions Results in a SF of 1.007 ± 0.004 Added in quadrature to standard jet energy correction 17

Substructure Efficiency We also measure a top tagging scale factor using the substructure quantities Use type 1 candidates Mass cut efficiency MC 90.8 ± 1.1 % Data 91.9 ± 1. % Minimum pairwise mass cut efficiency MC 67.0 ± 1.5 % Data 60.7 ± 1.7 % Combined Scale Factor is 0.96 ± 0.039 Applied to all MC events; once per top tag Total substructure uncertainty ~8.5% 18

Background Composition Two background processes in this analysis SM top pair production QCD multijet QCD multijet contribution determined from a data-derived method The mistag rate is measured using Type 1+1 events Similar kinematics to signal region Require the 1+1 event selection Jet pt > 400, 400 GeV Δy < 1.0 Form a sideband region Enriched in QCD dijet events Invert pairwise mass cut on one jet Anti-top tagged jet Probe jet 19

Background Composition Two background processes in this analysis SM top pair production QCD multijet QCD multijet contribution determined from a data-derived method Require the 1+1 event selection Jet pt > 400, 400 GeV Δy < 1.0 Form a sideband region Enriched in QCD dijet events Invert pairwise mass cut on one jet The mistag rate is measured using Type 1+1 events Measure tag rate of second jet Subtract ttbar contribution Anti-top tagged jet Probe jet 0

Forming the QCD Estimate Select events of type 1+X Apply the mistag rate to the X jet to model type 1+1 events Mass of the weighted jet not representative of the signal selection Biases shape of mtt distribution Jet mass of weighted jet set by hand to follow correct distribution Mass-modified procedure Top Tagged Jet No Requirements Weighted by Mistag Rate 1

Forming the QCD Estimate Select events of type 1+X Apply the mistag rate to the X jet to model type 1+1 events Mass of the weighted jet not representative of the signal selection Biases shape of mtt distribution Jet mass of weighted jet set by hand to follow correct distribution Mass-modified procedure Closure test performed to ensure procedure correctly estimates background If a signal were present in the data, the size of the recovered excess would be degraded by a small fraction Correct for this in limit setting also Top Tagged Jet No Requirements Weighted by Mistag Rate mtt

Event Yields Top candidate pair invariant mass used as signal discriminant Sub-plot shows two pieces of information Standard deviations of observation relative to bin uncertainty (black) Fractional error of observation vs. prediction (red) Good agreement between data and total expectation ) Events / (50 GeV/c N sigma 00 800 600 400 00 0 500 00 1500 000 500 3000 tt Invariant Mass (GeV/c ) 3 1 0-1 - -3 CMS Preliminary, -1 s = 8 TeV, 19.6 fb Data Non-Top Multijet SM tt 1 TeV RS KK gluon TeV RS KK gluon 3 TeV RS KK gluon 500 00 1500 000 500 3000 1.5 1 0.5 0-0.5-1 -1.5 - Data - Predicted Predicted ) Events / (50 GeV/c 3 Data Non-Top Multijet SM tt CMS Preliminary, -1 s = 8 TeV, 19.6 fb 1 TeV RS KK gluon TeV RS KK gluon 3 TeV RS KK gluon 1 500 00 1500 000 500 3000 tt Invariant Mass (GeV/c ) 3

Signal Models We use three signal hypothesis models with masses in the range 1-3 TeV Narrow (1%) Z Wide (%) Z RS KK gluon (width ~M/6) Template morphing used to form intermediate mass templates 4

Systematics We apply several shape and rate systematics to the samples Table shows normalization difference with/ without systematics Shape systematics (also affects rate) Jet energy scale Jet energy resolution QCD determination Rate systematics Luminosity (± 4.4%) t-tbar normalization (± 50%) Top-tag SF Trigger Efficiency 5

All-Hadronic Results Narrow Z exclusion limit < 1.65 TeV Wide Z exclusion limit <.35 TeV RS KK gluon exclusion < 1.8 TeV High mass cross section limits significantly improved due to Δy criteria 6

Enhancement Analysis To produce a limit on this general enhancement we use a simple counting experiment with events having mtt > 1 TeV Result is limit on the enhancement ratio arxiv:13.765 New result -- set limit of S < 1.79 (expect S <.9) at 95% CL Includes Δy cut ) Events / (50 GeV/c 00 800 600 400 00 CMS Preliminary, -1 s = 8 TeV, 19.6 fb Data Non-Top Multijet SM tt 1 TeV RS KK gluon TeV RS KK gluon 3 TeV RS KK gluon 0 500 00 1500 000 500 3000 tt Invariant Mass (GeV/c ) N sigma 3 1 0-1 - -3 500 00 1500 000 500 3000 1.5 1 0.5 0-0.5-1 -1.5 - Data - Predicted Predicted 7

Comparison of Results Comparable sensitivity in both the lepton+jets and all-hadronic analyses Fluctuations in opposite directions cause larger differences in the observed limits Observed Limits All-Hadronic Lepton+Jets Narrow Z 1.65 TeV. TeV Wide Z.35 TeV.68 TeV Combination of two results currently in preparation RS Gluon 1.8 TeV.54 TeV Expected Limits All-Hadronic Lepton+Jets Narrow Z 1.7 TeV.0 TeV Wide Z.5 TeV.6 TeV RS Gluon.15 TeV. TeV 8

Summary New boosted top resonance searches performed with the 8 TeV dataset at CMS Significant improvement in exclusion limits for new physics models Not final word for the 01 running Many new substructure techniques under study N-subjettiness Qjets b-tagging (on subjets) especially important to improve QCD multijet background rejection Stay tuned for updates Planning many new studies and analyses for Boost 013 Thanks for your attention! 9