Search for a Standard Model Higgs boson in the H->ZZ->llnunu decay channel with the ATLAS detector

A search for a heavy Standard Model Higgs boson decaying via H->ZZ->llnunu, where l=e,mu, is presented. It is based on proton-proton collision data at sqrt(s)=7 TeV, collected by the ATLAS experiment at the LHC in the first half of 2011 and corresponding to an integrated luminosity of 1.04 fb^-1. The data are compared to the expected Standard Model backgrounds. The data and the combined background expectations are found to be in agreement and upper limits are placed on the Higgs boson production cross section over the entire mass window considered; in particular the production of a Standard Model Higgs boson is excluded in the region 340

The search for the Standard Model (SM) Higgs boson [1][2][3] is one of the most important aspects of the Large Hadron Collider (LHC) physics program.Direct searches at the CERN LEP e + e − collider have set a lower limit of 114. 4 GeV on the Higgs boson mass, m H , at 95% confidence level [4].Searches by the CDF and D0 experiments at the Fermilab Tevatron pp collider have explored the mass range up to 200 GeV and exclude the additional region 156 < m H < 177 GeV [5].For m H greater than twice the Z boson mass, m Z , a significant fraction of Higgs bosons decay to two Z bosons.The ZZ → ℓ + ℓ − ν ν decay channel offers a substantial branching fraction in combination with a good separation from potential background processes owing to the high transverse momentum, p T , of the electron or muon pair from the leptonic Z decay and the high missing transverse momentum, E miss T , from the Z decaying to neutrinos.The first cross section limits for a SM Higgs boson in the mass region 200 < m H < 600 GeV were set by the ATLAS and CMS collaborations in Refs.[6,7].This letter extends the H → ZZ → ℓ + ℓ − ν ν results therein, with a 30-fold increase in the integrated luminosity, as well as a significant improvement in the event reconstruction and background rejection.
The data sample considered in this search was recorded by the ATLAS experiment during the first half of the 2011 LHC run at a center-of-mass energy √ s = 7 TeV.The integrated luminosity of the data sample, considering only data-taking periods where all relevant detector subsystems were operational, is 1.04 fb −1 .
The ATLAS detector has been described elsewhere [8].Simulated signal and background event samples are produced with Monte Carlo (MC) event generators, passed through a full GEANT4 [9] simulation of the ATLAS detector [10] and reconstructed with the same reconstruction software as the data.
H → ZZ → ℓ + ℓ − ν ν (ℓ = e, µ, τ ) events are mod-elled using the POWHEG [11,12] event generator, which includes matrix elements for the gluon fusion and the vector-boson fusion production mechanisms of the Higgs boson up to next-to-leading order.POWHEG is interfaced to PYTHIA [13] for the modelling of parton showers.The Higgs boson p T spectrum is reweighted to the calculation of Ref. [14], which provides QCD corrections up to next-to-leading order and QCD soft-gluon resummations up to next-to-next-to-leading logarithms.
An alternative sample of signal events is produced using the PYTHIA event generator, which includes only leading order matrix elements.In both cases PHOTOS [15] is used to model final-state radiation and TAUOLA [16] for the simulation of τ decays.
H → ZZ → ℓ + ℓ − ℓ + ℓ − and H → ZZ → ℓ + ℓ − q q samples are also simulated using the same generators as for the H → ZZ → ℓ + ℓ − ν ν samples, whilst H → W + W − → ℓ + νℓ − ν events are produced using the MC@NLO generator [17], interfaced to HERWIG [18] and JIMMY [19] in the gluon fusion channel and the SHERPA [20] generator in the vector-boson fusion channel.These channels contribute to the signal yield and are considered as part of the signal.In particular, H → W + W − → ℓ + νℓ − ν decays contribute as much as 77% to the signal expectation after the full selection for m H = 200 GeV decreasing to 13% at m H = 300 GeV.Independence of the analysis with respect to other ATLAS Higgs boson searches [21-23] is ensured through mutually exclusive selection requirements on the dilepton invariant mass, the number of leptons or the event missing transverse momentum.
The cross sections for Higgs boson production, the associated branching fractions [24], as well as their uncertainties, are compiled in Ref. [25].They correspond to next-to-next-to-leading order in QCD for the gluon fusion [26][27][28][29][30][31] and the vector boson fusion [32] processes.In addition, QCD soft-gluon resummations up to next-to-next-to-leading logarithms are available for the gluon fusion process [33], while next-to-leading order electroweak corrections are applied to both the gluon fusion [34,35] and the vector boson fusion [36,37] processes.These cross section calculations do not account for the width of the Higgs boson, which is implemented through an ad-hoc Breit-Wigner line shape applied at the event generator level.Recent studies [25,38] have indicated that effects due to off-shell Higgs boson production and interference with other SM processes may become sizeable at the highest masses (m H > 400 GeV) considered in this search.In the absence of a full calculation, a conservative estimate of the possible size of such effects was made and the impact on the obtained limits in this channel was found to be less than 2% for m H = 400 GeV growing to about 25% at m H = 600 GeV.
Different event generators are chosen to model a range of important background processes.The ALPGEN generator [39] interfaced with HERWIG for parton showers and hadronisation is used to simulate W/Z+jets backgrounds.MC@NLO, interfaced to HERWIG and JIMMY, is used for the production of top-pair, single top and diboson (W W , W Z and ZZ) backgrounds.PYTHIA is used to simulate b b and cc samples as well as alternative samples for the Z and ZZ backgrounds.All simulated background samples are scaled to the highest available precision calculations for the relevant process.An overview of the used predictions and their uncertainties is given in Ref. [40].
Data used for the search in the electron and muon channels were collected primarily using single lepton triggers with p T thresholds of 20 and 18 GeV respectively.The expected trigger efficiency is close to 100% in the electron channel and about 95% in the muon channel for signal events passing all the selection criteria described below Electron candidates are reconstructed from electromagnetic calorimeter clusters, with shapes consistent with those expected from electromagnetic showers, matched to tracks reconstructed in the inner detector.Details of the electron reconstruction and identification can be found in Ref. [41].The electron candidates are required to pass the standard ATLAS "medium" selection criteria and have p T > 20 GeV and pseudorapidity |η| < 2.47.
Muons are identified by reconstructing tracks in the muon spectrometer.These tracks are then extrapolated back to the beam line to find a matching inner detector track.Details of muon reconstruction and identification can be found in Ref. [41].Only muons with p T > 20 GeV and |η| < 2.5 are considered.
Jets are used in this analysis to reject backgrounds from events with heavy quark decays or from events with fake E miss T due to mis-measured jets.For this purpose jets are reconstructed from clusters of energy deposits in the calorimeters using the anti-k t algorithm [42] with a radius parameter R = 0.4.Only jets with p T > 25 GeV and |η| < 2.5 are considered.
To remove leptons associated with jets, such as those originating from semi-leptonic decays of b hadrons, leptons are not considered in the analysis if the sum of inner detector track momenta in a cone ∆R < 0.2 around the lepton direction is greater than 10% of the p T of the lepton itself or if the lepton is within a distance ∆R < 0.4 of the nearest jet.
The missing transverse momentum is measured as the (negative) vectorial sum of the transverse momenta of all clusters in the calorimeters within |η| < 4.5 and all selected muons in the event.Calorimeter deposits associated with muons are subtracted to avoid double counting.
Events are required to contain a reconstructed primary vertex formed from at least 3 tracks and exactly two oppositely charged electrons or muons, consistent with originating from the primary vertex.The dilepton mass distribution is shown in Fig. 1.Inclusive Z boson production is the dominant background at this stage of the analysis.To suppress backgrounds from top, W , and QCD multijet production, the dilepton invariant mass, m ℓℓ , is required to satisfy |m Z − m ℓℓ | < 15 GeV.FIG.1: The dilepton invariant mass distribution for events with exactly two oppositely charged electrons or muons.The inset at the bottom of the figure shows the ratio between the data and the combined background expectations as well as a band corresponding to the combined systematic uncertainties of the analysis.
To reduce the background from events with fake E miss T due to mis-measured jets, events are rejected if the azimuthal angle between the missing transverse momentum vector, p miss T , and the leading jet in the event satisfies ∆φ( p miss T , p jet T ) < 0.3.To reduce the background from top quark production, events with one or more btagged jets are rejected, where the b-tagging is based on a single discriminant combining information from both the impact parameter with respect to the primary vertex of tracks associated to the jet and the presence of displaced secondary vertices associated to the jet's tracks.FIG.2: The E miss T (left) and ∆φ(ℓ, ℓ) (right) distributions for events with exactly two oppositely charged electrons or muons inside the Z mass window.The insets at the bottom show the ratio between data and the combined background expectations as well as a band corresponding to the combined systematic uncertainties of the analysis.
The chosen cut achieves an efficiency of about 70% for identifying real b-jets, with a light-quark jet rejection of about 80 [43].
To exploit the mass dependent kinematic features of H → ZZ → ℓ + ℓ − ν ν production, the search is subdivided into a low Higgs boson mass (m H < 280 GeV) and a high Higgs boson mass (m H ≥ 280 GeV) search region, where dedicated cuts are applied to two important discriminating variables used to reduce the background contributions: E miss T and the azimuthal angle between the two leptons, ∆φ(ℓ, ℓ). Figure 2 shows the distributions of these variables after the application of the m ℓℓ window cut.Since inclusive Z production gives rise to a steeply falling E miss T distribution, systematic uncertainties on the E miss T reconstruction are particularly important to estimate this background correctly.The dominant contributions to the E miss T uncertainty come from the knowledge of the jet energy scale and the modelling of inclusive Z production.Figure 2 shows that a good agreement within systematic uncertainties is observed between data and the combined background expectation.In the low m H region, events are required to satisfy E miss T > 66 GeV, whilst in the high m H region the requirement is E miss T > 82 GeV.These cuts reduce significantly the backgrounds from processes with no or modest genuine missing transverse momentum originating from unobserved neutrinos.
The boost of the Z bosons originating from a Higgs boson decay increases with m H , thus reducing the expected opening angle between the leptons.In the low m H region this boost is expected to be modest and a cut 1 < ∆φ(ℓ, ℓ) < 2.64 is applied.In the high m H region an upper limit ∆φ(ℓ, ℓ) < 2.25 is required.
Finally, in the high m H region, events are also rejected if the azimuthal angle between the missing transverse momentum vector and the direction of the Z → ℓℓ boson candidate is ∆φ( p miss T , p ℓℓ T ) < 1.The efficiency of the event selection is very similar in the electron and muon channels, ranging from 3% for m H = 200 GeV to about 48% for m H = 600 GeV.SM pair production of Z bosons has a final state identical to the signal, and is therefore expected to survive most of the applied selection criteria and form a continuum in the transverse mass distribution (defined below).The normalisation for this background is obtained from a calculation including next-to-leading order terms [44] with an additional 6% term to account for missing quarkbox diagrams (gg → ZZ) [45].A 11% normalisation uncertainty is assigned to this background, estimated from scale, PDF and model uncertainties.W W and W Z backgrounds are normalised in a similar way.
The background from inclusive Z production is derived from MC, after checking that the simulation describes well the data in samples selected by requiring the presence of a lepton pair.The background from top events is also taken from the MC prediction.This prediction is verified to agree with data, within systematic uncertainties, in two independent control samples: the first one requires at least one identified b-jet, while the second selects events containing electron-muon pairs.
Additional backgrounds can arise from QCD multijet events or inclusive W production due to heavy flavour decays or jets faking leptons.The normalisation of the W background is obtained from the ratio between data and MC in control samples of like-sign electron-electron and electron-muon events with high E miss T .The QCD multijet background in the electron channel is determined using a data sample based on a loosened electron selection, thus dominated by jets; this sample is scaled to describe the tails of the m ℓℓ distribution.In the muon channel, the background from heavy flavour decays is studied using simulation, whereas other muon sources from multijet events are constrained using a sample of like-sign muon pairs in data.In both cases the background is found to be negligible.
The signal efficiencies and overall background expectations are similar in the electron and the muon channels, therefore only combined results are presented.The numbers of candidate H → ZZ → ℓ + ℓ − ν ν events selected in data and the expected yields from signal and background processes are shown in Table I.The systematic uncertainties include experimental uncertainties related to the selection and calibration of electrons, muons, jets and b-jets, which are also explicitly propagated to the E miss T calculation.Shape uncertainties for the signal and for the single Z and ZZ backgrounds are estimated using PYTHIA as an alternative MC generator.
Normalisation uncertainties for signal (gluon fusion +14% −10% and VBF 4%) and diboson backgrounds (11%) are obtained from theory [25]; uncertainties for the inclusive Z boson production (2.5%), top quark production (9%), inclusive W boson production (100%) and QCD multijet production in the electron channel (50%) are estimated from data.A 3.7% luminosity uncertainty [46] is included for those processes for which the normalisation is not obtained from the data.The dominant systematic uncertainties in the analysis are the E miss T uncertainties for the Z background, the b-tagging uncertainty for the top background and the normalisation uncertainties for the signal and the W and diboson backgrounds.
After the event selection, the Higgs boson search is performed by looking for an excess of data over the SM background expectation in the transverse mass distribution of the selected eeνν and µµνν events.The transverse mass is calculated from the lepton pair and the p miss T vector as: Figure 3 shows the m T distribution in the high m H search region.Signal to background ratios for different m H values, determined in a m T window defined to enclose 95% of the corresponding signal events, are listed in Table I.The number and distribution of candidate H → ZZ → ℓ + ℓ − ν ν events observed in the data agree with the expected backgrounds within the uncertainties, with no indication of an excess.Upper limits are set on the Higgs boson production cross section relative to its predicted SM value as a function of m H .The limits are extracted from a maximum likelihood fit to the m T distribution following the CL s modified frequentist formalism with the profile likelihood test statistic [47,48].All systematic uncertainties are taken into account.
Figure 4 shows the expected and observed limits at the 95% confidence level.The expected limit is lowest around m H = 380 GeV where it is 1.1 times the SM Higgs boson cross section.Fluctuations in the background can lead to better or worse expected limits.Over the entire mass range the observed limits agree with the expectations within the ±2σ band.A SM Higgs boson in the range 340 GeV < m H < 450 GeV is excluded at the 95% confidence level In summary, results of a search for a heavy SM Higgs boson with a mass in the range 200 < m H < 600 GeV decaying to ZZ → ℓ + ℓ − ν ν have been presented.These results are based on a data sample corresponding to an integrated luminosity of 1.04 fb −1 , recorded with the ATLAS detector at the LHC.No evidence for a signal is observed and cross section limits are placed over the entire mass range, excluding the production of a SM Higgs boson in the region 340 < m H < 450 GeV at the 95% confidence level.

FIG. 3 :
FIG.3:The transverse mass distribution of H → ZZ → ℓ + ℓ − ν ν candidates in the high mH search region for the data (dots), the expected backgrounds (histograms) and a Higgs boson of mass 380 GeV (filled histogram).The electron and muon channels are combined.

FIG. 4 :
FIG.4: Observed and expected 95% confidence level upper limits on the Higgs boson production cross section divided by the SM prediction.The green and yellow bands indicate the ±1σ and ±2σ fluctuations, respectively, around the median sensitivity.The limits are based on 1.04 fb −1 of data at √ s =7 TeV.

TABLE I :
The expected number of background and signal events for the Higgs boson search in the H → ZZ → ℓ + ℓ − ν ν channel, along with the observed numbers of candidates in data, for an integrated luminosity of 1.04 fb −1 .The quoted uncertainties are statistical and systematic respectively.Signal to background ratios are also given for various masses (see text).