Search for High Mass Top Pair Resonances in the Boosted All Hadronic Channel using the Template Overlap Method


An effort to identify highly boosted top quark pairs resulting from the production of a massive resonance and decaying into two "fat jets" is summarized here. This is an analysis falling under the Top Cross Section Working Group as well as the 4GT working group. This analysis is one of two fully hadronic top resonance searches, its focus is on reconstructing highly boosted tops from events containing at least one jet with pT > 500 GeV. Background rejection will be achieved through the use of a new method to tag boosted top jets, the top templating method (TopTemplateing). The alternative approach being pursued uses the HepTopTagger to select top quarks. A combined result is being sought between these two analyses.

This page will summarize all relevant information for the TopTemplating analysis: strategy, object definition, event selection, MC and Data samples, analysis parameters (e.g. top-templating settings), signal definitions, and links to relevant external information. In addition, a list of the analysis tasks and the people responsible is also provided.


  • Contact Person: Pekka Sinervo
  • Analysis Group: Trisha Farooque, Alex Martyniuk, Gilad Perez, Michele Petteni, Francesco Spano, Bernd Stelzer, Oliver Stelzer-Chilton, Stephen Swedish
  • Editors: Michele Petteni, Pekka Sinervo
  • Analysis Contacts: Trisha Farooque, Stephen Swedish
  • Mailing list: Pekka can provide you with a list


  • Selection of AKT10 jets
    • Trisha Farooque, Alex Martyniuk

  • Jet mass correction and incorporation of JES corrections
    • Trisha Farooque

  • b-tagging implementation
    • Jacob Groth-Jensen, Alex Martyniuk, Stephen Swedish

  • top-templating
    • Michele Petteni, Francesco Spano, (Stephen Swedish)

  • MC calculations
    • Stephen Swedish

  • Systematic uncertainties
    • Pekka Sinervo coordinating this effort

  • Limit calculations
    • Michele Petteni, Bernd Stelzer, Oliver Stelzer-Chilton, Stephen Swedish

  • Event Selection optimization
    • ???

Overall Strategy

  • Select events using the requirement of at least on high pT jet reconstructed with R=1.0 and with a pT_jet[0]>500 GeV. The jet should be in the central region defined by |eta|<2.0, though one could argue that a narrower "central" interval could be defined.
  • Require a recoil jet with pT>pTjet2, where pTjet2 is a fraction of pTjet1. This could notionally be around 50% of pTjet1, though the analysis is not too sensitive to this (either for background or signal). Also require this jet to be reasonable central |eta|<2.0. Need to do some optimization of this cut on the signal MC as a first step.
  • Require the events to have no evidence of significant missing Et by making a MetSignificance cut of around 5. The key here is to make this sample efficient for fully hadronic decays and fully complementary to the lepton+jets final state.
  • At this point, we'll be dominated by QCD dijet events -- "QCD jets".
  • We perform the pileup corrections for mass, which helps improve the S/B between QCD jets and top quark jets.
  • There are then at least three different strategies that should be investigated:
    1. B-tagging -- we look at the antikT 0.4 jets that are associated with the two leading jets, and then look at the MV1 results for each of these. We'll be able to use all the standard tools for this, though we'll need to work on understanding the efficiency and mistagging rates once we have a sizeable contribution of jets with pT>200 GeV. I've poked an idea in the ears of the b-tagging conveners as to how to get some constraints on these efficiencies.
    2. A template approach proposed by Almeida et al. (see this article for top jets and this article for an extension to Higgs jets).
    3. A conventional jet substructure requirement like using the jet mass and planar flow (the interesting thing here is that planar flow is substantially affected by pile-up, but Almeida et al. argue that the planar flow from the best template is more effective).
  • Do the b-tagging and top templating on both the leading jet and the non-leading jet, separately. In principle, these are independent methods, so using one against the other will allow us to measure fake rates.
  • Estimates of t-tbar acceptance would be done via MC for SM ttbar, KKG of various masses and Zprimes of various masses.
  • Background estimates would be done by looking at anti-tagged jet samples. We've used an ABCD technique in the past, and that may be the way to go here.
  • Optimize the data selection for two purposes: i) set an upper limit on a resonance decaying to t-tbar and ii) measure the top quark cross section above a top quark pT above some threshold. This is work that Sergei and Ben at Argonne are pioneering. The resonance search doesn't require seeing a top signal per se, so long as we understand the backgrounds.

The Template Overlap Method

A new method for identifying top quarks has been proposed that quantifies the agreement between the energy flow of reconstructed jets and that of boosted top partonic decays as a way of distinguishing boosted hadronic top quark decays from the QCD di-jet background (G. Perez et al, A template is defined as a unique partonic decay configuration corresponding to a specify top decay.

Sets of templates, fine grained in partonic phase space, are generated. Agreement between the top-jet candidate and each template is quantified using an overlap function. The maximal functional overlap is referred to as the template overlap, OV. The template method is generalizable to other massive particles, in the context of top quark selection the template overlap value is denoted OV3, where 3 indicates that candidate jets are being compared to 3-prong partonic decays. OV3 is an infrared safe jet observable that provides powerful discriminating power between boosted top-jets and the QCD background. For a detailed description see:

Event Selection (to be optimized)

  • Preselection
    • A good primary vertex with at least 4 tracks (check to make this consistent with other 2011 analyses)
    • At least 2 fat jets with |eta|<2.0:
      • leading jet: pT_j1>500 GeV
      • subleading jet: pT_j2 > 300 GeV
    • SMET < 5 (this is the primary anti-lepton+jet selection)

  • Control Region
    • mass_j2 < 100 GeV
  • Signal Region (to be optimized)
    • >1 b-tag ( or > 2 b-tags)
      • b-tag : Identify all AKT4 jets with R=1.0 of each fat jet and with pT>20 GeV and MV1 > 0.5
    • 130 < mass_j1 , mass_j2 < 210 GeV
    • OV3 >0.5 ( or OV2, planar flow (Pf) cut?)

Object Definition

Fat Jets

  • Selection
    • R=1.0 : AKTLC10
  • Corrections
    • Apply Jet mass correction for NPV, and JES correction, to jets

Flavor Tagging

    • We are attempting to follow the guidelines outlined here, Analysis17. Care must be taken to undertand the definition of amd jet.eta() in the D3PDs

Background Estimation

  • QCD
    • estimated using ABCD technique with correction calculated from QCD MC
  • SM ttbar
    • MC samples? Yet to be defined


  • Z-prime
  • KKGluon

Statistical Method

Summary of Mtt analysis and limit setting will go here.


Summary of the relevant systematics, estimation strategies, and the individuals responsible will go here.


Outstanding Issues

  • Currently, for data, no single D3PD type is simultaneously hadronically skimmed, and contains the required jet substructure information, b-tagging information, and lepton track information. Here is how the D3PD types break down

    • NTUP_TOPBOOST: contains all required variables, but D3PDs produced for data have TOP_El skim applied
    • Groom Jet : missing required lepton track information, (might have b-tagging information)
    • Others...

Analysis Tools


  • RootCore: package that helps developers build packages that work standalone (outside Athena).
    • atlasoff/PhysicsAnalysis/D3PDTools/RootCore

  • TopRootCore: RootCore-based package specific for the top group
    • atlasoff/PhysicsAnalysis/TopPhys/TopRootCoreRelease

Dataset Meta Data Access

Power User Status

  • Power user status will be requested to help us meet our timeline. Details will be added here as they are available.

Important Links

Relevant Published or Released papers

Relevant Internal Notes

Other related links/presentations

  • Overview of Analysis at 12 Apr 2012 4GT Meeting: Slides

Major updates:
-- PekkaSinervo - 12-Mar-2012

Edit | Attach | Watch | Print version | History: r1 | Backlinks | Raw View | WYSIWYG | More topic actions
Topic revision: r1 - 2012-04-24 - Swedish
    • Cern Search Icon Cern Search
    • TWiki Search Icon TWiki Search
    • Google Search Icon Google Search

    Sandbox All webs login

This site is powered by the TWiki collaboration platform Powered by PerlCopyright & 2008-2020 by the contributing authors. All material on this collaboration platform is the property of the contributing authors.
or Ideas, requests, problems regarding TWiki? use Discourse or Send feedback