PaperlayerView exact revision on arXiv

Source-linked AI summary

Search for massive resonances decaying into pairs of boosted bosons in semi-leptonic final states at sqrt(s) = 8 TeV

CMS Collaboration

arXiv:1405.3447v2hep-ex

TL;DR

The paper searches for new WW, ZZ, and WZ resonances in semileptonic final states, where boosted hadronic bosons can appear as single jets. It uses jet-substructure tagging and combines the semileptonic channels with an all-hadronic search, setting 95% CL bulk-graviton limits from 700 to 10 fb for masses from 600 to 2500 GeV.

  • Problem

    The search addresses whether new resonances predicted by extensions of the Standard Model can decay to pairs of massive vector bosons.

  • Method

    The analysis studies semileptonic WW, ZZ, and WZ events with boosted-boson jet substructure and combines them with a complementary all-hadronic search.

  • Results

    95% CL upper limits on bulk-graviton production range from 700 to 10 fb for resonance masses between 600 and 2500 GeV.

  • Takeaways & Limitations

    No excess with significance larger than two standard deviations is observed, and the combined analysis sets limits on bulk-graviton production.

  • Takeaways & Limitations

    The analysis is affected by disagreement between data and simulation in the τ21 distribution, attributed in part to parton-shower mismodeling.

Abstract

from arXiv · show

A search for new resonances decaying to WW, ZZ, or WZ is presented. Final states are considered in which one of the vector bosons decays leptonically and the other hadronically. Results are based on data corresponding to an integrated luminosity of 19.7 inverse femtobarns recorded in proton-proton collisions at sqrt(s) = 8 TeV with the CMS detector at the CERN LHC. Techniques aiming at identifying jet substructures are used to analyze signal events in which the hadronization products from the decay of highly boosted W or Z bosons are contained within a single reconstructed jet. Upper limits on the production of generic WW, ZZ, or WZ resonances are set as a function of the resonance mass and width. We increase the sensitivity of the analysis by statistically combining the results of this search with a complementary study of the all-hadronic final state. Upper limits at 95% confidence level are set on the bulk graviton production cross section in the range from 700 to 10 femtobarns for resonance masses between 600 and 2500 GeV, respectively. These limits on the bulk graviton model are the most stringent to date in the diboson final state.

1 Introduction

The paper searches for new resonances decaying to WW, ZZ, or WZ, motivated by extensions of the Standard Model and extra-dimensional models. It uses semileptonic final states, boosted-boson jet substructure, and a combination with an all-hadronic search.

  • Extensions of the Standard Model can predict new resonances coupling to pairs of massive vector bosons.
  • The bulk graviton model addresses the electroweak–gravitational scale difference through an extra-dimensional tower of Kaluza–Klein excitations.
  • The search uses 19.7 fb^-1 of 8 TeV CMS data in semileptonic WW, ZZ, and WZ final states.The considered topologies are ℓνqq(′) and ℓℓqq(′).
  • At high resonance masses, hadronic W or Z decays form a single massive V jet, so jet substructure is used for V tagging and background suppression.Resolved two-jet decays contribute negligibly in the mass range considered.
  • The semileptonic results are combined with a complementary all-hadronic VV search to improve sensitivity.The combined analysis uses the same V-tagging techniques in the all-hadronic channel.

2 The CMS detector

The CMS detector is built around a 3.8 T superconducting solenoid containing tracking and calorimetry, with muon detection outside the solenoid.

  • The CMS detector's central component is a 3.8 T superconducting solenoid with a 6 m internal diameter.The solenoid contains the silicon tracker, ECAL, and HCAL; the muon system lies outside it.

3 Simulated samples

Simulated samples model the principal backgrounds and benchmark bulk-graviton signals, with pileup and data–simulation efficiency corrections applied.

  • MADGRAPH, POWHEG, and PYTHIA generate the W+jets, Z+jets, top-quark, and diboson background samples.Parton showering and hadronization use PYTHIA with the Z2* tune.
  • The bulk graviton benchmark is generated for masses from 600 to 2500 GeV and decays to WW and ZZ.The vector bosons are longitudinally polarized in more than 99% of cases.
  • At 1 TeV and k/MPl = 0.5, the bulk-graviton production cross section is 15.1 fb, with WW and ZZ branching fractions of 18.7% and 9.5%.
  • Simulated events include supplementary minimum-bias interactions to reproduce pileup and are corrected for observed data–simulation efficiency differences.

4 Reconstruction and selection of events

The analysis reconstructs leptonic W or Z candidates and boosted hadronic V jets in semileptonic final states, using jet mass and substructure to identify merged boson decays. Event selections include channel-specific triggers, lepton requirements, missing transverse momentum, and mass constraints.

  • Leptonic W selection: The ℓν+V-jet channel uses single-lepton triggers, one selected lepton, missing transverse momentum, and a W-mass constraint to reconstruct the neutrino longitudinal momentum.The missing-transverse-momentum threshold is 40 GeV for muons and 80 GeV for electrons.
  • Leptonic Z selection: The ℓℓ+V-jet channel reconstructs a leptonic Z from two oppositely charged same-flavor leptons with invariant mass between 70 and 110 GeV.The selection accommodates collimated boosted-Z muons by requiring two tracker muons, with at least one also identified as a global muon.
  • Object reconstruction: The analysis uses particle-flow jets, requires CA8 and AK5 jets to be separated from identified leptons, and retains jets with pT > 30 GeV and |η| < 2.4.The missing transverse energy is computed from the vector sum of reconstructed particle-flow transverse momenta and corrected for jet energy-scale changes.
  • Hadronic V-jet reconstruction: Boosted W and Z decays are reconstructed as single CA8 V jets, whose mass and two-prong substructure discriminate them from multijet backgrounds.Jet grooming improves mass resolution and reduces pileup effects before applying the V-tagging requirements.
  • Selection validation: The τ21 distributions show data-simulation disagreement after the W- and Z-jet selections, so the analysis is designed to remain robust against such modeling differences.The relevant jet-mass windows are 65 < mjet < 105 GeV for W candidates and 70 < mjet < 110 GeV for Z candidates.

5 W tagging in a top-quark enriched control sample

A top-quark enriched control sample is used to quantify discrepancies between data and simulation, derive background normalization corrections, and calibrate W-tagging efficiencies for the signal selection.

  • Control-sample motivation: The τ21 discrepancy can bias simulated signal efficiencies, motivating corrections derived from a signal-free top-quark enriched control sample.The same parton-shower modeling is used for the control study and signal simulation so the corrections can be applied consistently.
  • Background normalization: 0.97 ± 0.02 and 0.96 ± 0.03 are the high-purity tt and single-top data-to-simulation scale factors in the muon and electron channels.The corresponding low-purity factors are 1.31 ± 0.05 and 1.39 ± 0.08.
  • W-tagging calibration: 0.89 ± 0.08 and 1.28 ± 0.30 are the W-tagging scale factors for the high-purity and low-purity categories, respectively.They are obtained from the ratio of tagging efficiencies measured in data and simulation using simultaneous jet-mass fits.
  • Jet-mass validation: 83.4 ± 0.3 and 7.2 ± 0.4 GeV are the simulated W-jet mass peak position and resolution, compared with 84.7 ± 0.4 and 7.9 ± 0.6 GeV in data.The mass shift is attributed to residual pileup, underlying-event, initial-state-radiation, and nearby b-jet energy contributions.
  • Extension to Z tagging: The same corrections are applied to Z-jet candidates because their kinematic properties are expected to be very similar to those of W jets.This extrapolation is presented as an expectation about agreement between data and simulation.

6 Modeling of background and signal

The analysis estimates dominant V+jets backgrounds from data sidebands and models signal and background mass distributions for the semileptonic channels. Final spectra show agreement between observed data and predicted backgrounds, including a highest-mass WW event near 3200 GeV that is outside the statistical fit range.

  • Background estimation: The dominant V+jets background is estimated from data using fits to jet-mass sidebands, while minor backgrounds are taken from simulation.The signal-region normalization is obtained from lower and upper mjet sidebands; the background mass shape uses the low-mjet sideband and an extrapolation function.
  • Background estimation: The high-mjet sideband is excluded to avoid possible contamination from beyond-SM VH resonances in the model-independent interpretation.Such events could populate the high-mass sidebands when the Higgs boson is reconstructed as a jet.
  • Background estimation: The V+jets mass-shape extrapolation uses αMC(mVV), whose validity is tested with a data closure test.The closure test successfully predicts the normalization and shape of the V+jets background in an upper sideband using lower-sideband data.
  • Background estimation: 700–3000 GeV is the fitted mVV range for the ℓν+V-jet analysis, while the ℓℓ+V-jet ranges are 500–2800 GeV for HP and 650–2800 GeV for LP.These ranges are chosen to provide smoothly falling spectra and stable background fits.
  • Results: The observed mWW spectrum agrees with the predicted background, while the highest-mass event has mWW ≈3200 GeV and is excluded from the fit performed up to 3 TeV.Its impact is negligible for limits on narrow bulk gravitons up to 2.5 TeV, and it is compatible with the background prediction above 2.5 TeV.
  • Signal modeling: The signal mass shape is modeled with bulk-graviton simulation at k/MPl = 0.2, whose intrinsic relative width is about 0.2%.For k/MPl ≲ 0.5, the natural width is sufficiently small to neglect relative to detector resolution.

7 Systematic uncertainties

Systematic uncertainties are evaluated for both signal efficiency and mass-shape modeling, while background uncertainties arise from sideband statistics, simulation, and extrapolation modeling. The dominant quoted signal-shape effects include jet-related uncertainties and unclustered energy in the semileptonic channel.

  • Background uncertainties: Background uncertainties affect both the normalization and shape of the mVV distribution and are driven mainly by sideband and simulation statistics.Shape uncertainties include the covariance matrix of the sideband fit and uncertainties in αMC(mVV).
  • Signal uncertainties: Signal systematic uncertainties affect both selection efficiency and the mVV shape through lepton, jet, and unclustered-energy variations.The signal efficiency is recalculated after varying reconstructed four-momenta by one standard deviation for each source.
  • Lepton uncertainties: Muon trigger and identification-plus-isolation uncertainties are 3% and 4%, respectively, while the total electron trigger, identification, and isolation uncertainty is 3%.These values account for the limited number of data events in the boosted regime.
  • Signal-shape uncertainties: Jet energy scale and resolution contribute about 3% relative uncertainty to the signal width in ℓν+V-jet and 2% in ℓℓ+V-jet.Unclustered-energy scale contributes a 1–3% uncertainty to the ℓν+V-jet signal width, larger at low resonance masses.

8 Statistical interpretation

The analysis sets limits using combined semileptonic and all-hadronic channels, with model-independent interpretations spanning resonance mass and width. A simplified efficiency-based procedure extends reinterpretation to generic signals while quantifying its systematic uncertainty.

  • Narrow-width bulk graviton limits: 95% CL exclusion limits are derived for a narrow bulk graviton by comparing observed and expected limits with predicted cross section times branching fraction.The interpretation assumes a natural width negligible relative to the experimental resolution and compares couplings k/MPl = 0.2 and 0.5.
  • Narrow-width bulk graviton limits: The semileptonic and all-hadronic analyses are statistically combined, with channel contributions varying across resonance-mass regions.The dilepton-plus-V-jet channel alone contributes below 800 GeV; the lepton-neutrino-plus-V-jet channel dominates from 800 to 2500 GeV, while the dijet channel contributes significantly above 1300 GeV.
  • Narrow-width bulk graviton limits: 15–20%: the combination makes expected cross-section limits more stringent than the individual analyses.
  • Model-independent limits: Generic-resonance limits are provided as functions of both resonance mass MX and natural width ΓX using a Breit–Wigner line shape convolved with detector resolution.
  • Model-independent limits: The model-independent reinterpretation supplies kinematic-dependent efficiencies and event limits without combining the two semileptonic analyses.The analyses remain separate to avoid assumptions about branching fractions into WW and ZZ channels.
  • Model-independent limits: 15%: an additional systematic uncertainty is assigned to total signal efficiency to cover imperfections in the efficiency parametrization.Parametrized efficiencies agree with direct simulation within 10% of their value for bulk and RS1 gravitons.

9 Summary

The search finds no evidence for new WW, ZZ, or WZ resonances in semileptonic final states. Combined limits constrain bulk graviton production across resonance masses from 600 to 2500 GeV.

  • No evidence for a signal is found in semileptonic WW, ZZ, or WZ final states.
  • The analysis reconstructs a leptonic W or Z candidate recoiling against a jet compatible with a hadronic vector boson.Jet substructure information is used to reduce multijet backgrounds.
  • 95% CL bulk graviton cross-section limits range from 700 to 10 fb for resonance masses from 600 to 2500 GeV, respectively.The limits combine the two semileptonic channels with a complementary fully hadronic search.
  • The bulk graviton limits are the most stringent to date in these final states.

A Detailed instructions and additional material for generic interpretation of the results

The generic interpretation procedure converts simulated resonance samples into approximate total efficiencies using generator-level filtering, parametrized reconstruction efficiencies, and veto factors. The resulting visible-event predictions are compared with mass- and width-dependent limits to assess exclusion power.

  • Efficiency parametrization: The efficiency tables provide reconstruction and identification efficiencies for leptonic and hadronic W and Z decays.They are parametrized by the generated boson transverse momentum and pseudorapidity.
  • Efficiency parametrization: Generic-model interpretation begins by generating resonance events at a specified mass and width, including relevant tau decays.
  • Efficiency parametrization: Generated events are filtered using WW or ZZ requirements, with corresponding hadronic-boson substitutions for WZ and ZW decays.
  • Efficiency parametrization: Passing events are weighted by the product of leptonic and hadronic efficiencies, including veto efficiencies of 90% for WW and 81% for WZ.
  • Efficiency parametrization: The sum of event weights divided by the generated event count approximates the total efficiency for the model.
  • Limits: The resulting visible event count is compared with observed limits to assess the experiment’s exclusion power.The limits depend on resonance mass and normalized width for generic WV and ZV resonances.
Loading 1405.3447v2…